Merge remote-tracking branch 'origin/master' into pr-local-plan

.github/actions/release/action.yml

@@ -0,0 +1,168 @@
+name: Release
+
+description: Makes patch releases and creates new release branch
+
+inputs:
+  ref:
+    description: 'Git reference (branch or commit sha) from which to create the release'
+    required: true
+    type: string
+  type:
+    description: 'The type of release: "new" for a new release or "patch" for a patch release'
+    required: true
+    type: choice
+    options:
+      - patch
+      - new
+  dry-run:
+    description: 'Dry run'
+    required: false
+    default: true
+    type: boolean
+  token:
+    required: true
+    type: string
+
+runs:
+  using: "composite"
+  steps:
+    - name: Prepare Release Info
+      shell: bash
+      run: |
+        python3 ./tests/ci/create_release.py --prepare-release-info \
+          --ref ${{ inputs.ref }} --release-type ${{ inputs.type }} \
+          ${{ inputs.dry-run && '--dry-run' || '' }}
+        echo "::group::Release Info"
+        python3 -m json.tool /tmp/release_info.json
+        echo "::endgroup::"
+        release_tag=$(jq -r '.release_tag' /tmp/release_info.json)
+        commit_sha=$(jq -r '.commit_sha' /tmp/release_info.json)
+        echo "Release Tag: $release_tag"
+        echo "RELEASE_TAG=$release_tag" >> "$GITHUB_ENV"
+        echo "COMMIT_SHA=$commit_sha" >> "$GITHUB_ENV"
+    - name: Download All Release Artifacts
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        python3 ./tests/ci/create_release.py --download-packages ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Push Git Tag for the Release
+      shell: bash
+      run: |
+        python3 ./tests/ci/create_release.py --push-release-tag ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Push New Release Branch
+      if: ${{ inputs.type == 'new' }}
+      shell: bash
+      run: |
+        python3 ./tests/ci/create_release.py --push-new-release-branch ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Bump CH Version and Update Contributors' List
+      shell: bash
+      run: |
+        python3 ./tests/ci/create_release.py --create-bump-version-pr ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Bump Docker versions, Changelog, Security
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        git checkout master
+        python3 ./tests/ci/create_release.py --set-progress-started --progress "update changelog, docker version, security"
+        echo "List versions"
+        ./utils/list-versions/list-versions.sh > ./utils/list-versions/version_date.tsv
+        echo "Update docker version"
+        ./utils/list-versions/update-docker-version.sh
+        echo "Generate ChangeLog"
+        export CI=1
+        docker run -u "${UID}:${GID}" -e PYTHONUNBUFFERED=1 -e CI=1 --network=host \
+          --volume=".:/ClickHouse" clickhouse/style-test \
+          /ClickHouse/tests/ci/changelog.py -v --debug-helpers \
+          --gh-user-or-token=${{ inputs.token }} --jobs=5 \
+          --output="/ClickHouse/docs/changelogs/${{ env.RELEASE_TAG }}.md" ${{ env.RELEASE_TAG }}
+        git add ./docs/changelogs/${{ env.RELEASE_TAG }}.md
+        echo "Generate Security"
+        python3 ./utils/security-generator/generate_security.py > SECURITY.md
+        git diff HEAD
+    - name: Create ChangeLog PR
+      if: ${{ inputs.type == 'patch' && ! inputs.dry-run }}
+      uses: peter-evans/create-pull-request@v6
+      with:
+        author: "robot-clickhouse <robot-clickhouse@users.noreply.github.com>"
+        token: ${{ inputs.token }}
+        committer: "robot-clickhouse <robot-clickhouse@users.noreply.github.com>"
+        commit-message: Update version_date.tsv and changelogs after ${{ env.RELEASE_TAG }}
+        branch: auto/${{ env.RELEASE_TAG }}
+        assignees: ${{ github.event.sender.login }}  # assign the PR to the tag pusher
+        delete-branch: true
+        title: Update version_date.tsv and changelog after ${{ env.RELEASE_TAG }}
+        labels: do not test
+        body: |
+          Update version_date.tsv and changelogs after ${{ env.RELEASE_TAG }}
+          ### Changelog category (leave one):
+          - Not for changelog (changelog entry is not required)
+    - name: Complete previous steps and Restore git state
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        python3 ./tests/ci/create_release.py --set-progress-completed
+        git reset --hard HEAD
+        git checkout "$GITHUB_REF_NAME"
+    - name: Create GH Release
+      shell: bash
+      if: ${{ inputs.type == 'patch' }}
+      run: |
+        python3 ./tests/ci/create_release.py --create-gh-release ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Export TGZ Packages
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        python3 ./tests/ci/artifactory.py --export-tgz ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Test TGZ Packages
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        python3 ./tests/ci/artifactory.py --test-tgz ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Export RPM Packages
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        python3 ./tests/ci/artifactory.py --export-rpm ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Test RPM Packages
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        python3 ./tests/ci/artifactory.py --test-rpm ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Export Debian Packages
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        python3 ./tests/ci/artifactory.py --export-debian ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Test Debian Packages
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        python3 ./tests/ci/artifactory.py --test-debian ${{ inputs.dry-run && '--dry-run' || '' }}
+    - name: Docker clickhouse/clickhouse-server building
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        cd "./tests/ci"
+        python3 ./create_release.py --set-progress-started --progress "docker server release"
+        export CHECK_NAME="Docker server image"
+        python3 docker_server.py --release-type auto --version ${{ env.RELEASE_TAG }} --check-name "$CHECK_NAME" --sha ${{ env.COMMIT_SHA }} ${{ ! inputs.dry-run && '--push' || '' }}
+        python3 ./create_release.py --set-progress-completed
+    - name: Docker clickhouse/clickhouse-keeper building
+      if: ${{ inputs.type == 'patch' }}
+      shell: bash
+      run: |
+        cd "./tests/ci"
+        python3 ./create_release.py --set-progress-started --progress "docker keeper release"
+        export CHECK_NAME="Docker keeper image"
+        python3 docker_server.py --release-type auto --version ${{ env.RELEASE_TAG }} --check-name "$CHECK_NAME" --sha ${{ env.COMMIT_SHA }} ${{ ! inputs.dry-run && '--push'  || '' }}
+        python3 ./create_release.py --set-progress-completed
+    - name: Set current Release progress to Completed with OK
+      shell: bash
+      run: |
+        python3 ./tests/ci/create_release.py --set-progress-started --progress "completed"
+        python3 ./tests/ci/create_release.py --set-progress-completed
+    - name: Post Slack Message
+      if: ${{ !cancelled() }}
+      shell: bash
+      run: |
+        python3 ./tests/ci/create_release.py --post-status ${{ inputs.dry-run && '--dry-run' || '' }}

.github/workflows/auto_release.yml

@@ -1,44 +1,110 @@
 name: AutoRelease
 
 env:
-  # Force the stdout and stderr streams to be unbuffered
   PYTHONUNBUFFERED: 1
+  DRY_RUN: true
 
 concurrency:
-  group: auto-release
+  group: release
 on: # yamllint disable-line rule:truthy
-  # schedule:
+  # Workflow uses a test bucket for packages and dry run mode (no real releases)
-  #   - cron: '0 10-16 * * 1-5'
+  schedule:
+    - cron: '0 9 * * *'
+    - cron: '0 15 * * *'
   workflow_dispatch:
+    inputs:
+      dry-run:
+        description: 'Dry run'
+        required: false
+        default: true
+        type: boolean
 
 jobs:
-  CherryPick:
+  AutoRelease:
-    runs-on: [self-hosted, style-checker-aarch64]
+    runs-on: [self-hosted, release-maker]
     steps:
+      - name: DebugInfo
+        uses: hmarr/debug-action@f7318c783045ac39ed9bb497e22ce835fdafbfe6
       - name: Set envs
-        # https://docs.github.com/en/actions/learn-github-actions/workflow-commands-for-github-actions#multiline-strings
         run: |
           cat >> "$GITHUB_ENV" << 'EOF'
-          TEMP_PATH=${{runner.temp}}/cherry_pick
           ROBOT_CLICKHOUSE_SSH_KEY<<RCSK
           ${{secrets.ROBOT_CLICKHOUSE_SSH_KEY}}
           RCSK
-          REPO_OWNER=ClickHouse
-          REPO_NAME=ClickHouse
-          REPO_TEAM=core
           EOF
+      - name: Set DRY_RUN for schedule
+        if: ${{ github.event_name == 'schedule' }}
+        run: echo "DRY_RUN=true" >> "$GITHUB_ENV"
+      - name: Set DRY_RUN for dispatch
+        if: ${{ github.event_name == 'workflow_dispatch' }}
+        run: echo "DRY_RUN=${{ github.event.inputs.dry-run }}" >> "$GITHUB_ENV"
       - name: Check out repository code
         uses: ClickHouse/checkout@v1
         with:
-          clear-repository: true
           token: ${{secrets.ROBOT_CLICKHOUSE_COMMIT_TOKEN}}
           fetch-depth: 0
-      - name: Auto-release
+      - name: Auto Release Prepare
         run: |
           cd "$GITHUB_WORKSPACE/tests/ci"
-          python3 auto_release.py --release-after-days=3
+          python3 auto_release.py --prepare
-      - name: Cleanup
+          echo "::group::Auto Release Info"
-        if: always()
+          python3 -m json.tool /tmp/autorelease_info.json
+          echo "::endgroup::"
+          {
+              echo 'AUTO_RELEASE_PARAMS<<EOF'
+              cat  /tmp/autorelease_info.json
+              echo 'EOF'
+          } >> "$GITHUB_ENV"
+      - name: Post Release Branch statuses
+        run: |
+          cd "$GITHUB_WORKSPACE/tests/ci"
+          python3 auto_release.py --post-status
+      - name: Release ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[0].release_branch }}
+        if: ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[0] && fromJson(env.AUTO_RELEASE_PARAMS).releases[0].ready }}
+        uses: ./.github/actions/release
+        with:
+          ref: ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[0].commit_sha }}
+          type: patch
+          dry-run: ${{ env.DRY_RUN }}
+          token: ${{secrets.ROBOT_CLICKHOUSE_COMMIT_TOKEN}}
+      - name: Release ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[1].release_branch }}
+        if: ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[0] && fromJson(env.AUTO_RELEASE_PARAMS).releases[1].ready }}
+        uses: ./.github/actions/release
+        with:
+          ref: ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[1].commit_sha }}
+          type: patch
+          dry-run: ${{ env.DRY_RUN }}
+          token: ${{secrets.ROBOT_CLICKHOUSE_COMMIT_TOKEN}}
+      - name: Release ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[2].release_branch }}
+        if: ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[2] && fromJson(env.AUTO_RELEASE_PARAMS).releases[2].ready }}
+        uses: ./.github/actions/release
+        with:
+          ref: ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[2].commit_sha }}
+          type: patch
+          dry-run: ${{ env.DRY_RUN }}
+          token: ${{secrets.ROBOT_CLICKHOUSE_COMMIT_TOKEN}}
+      - name: Release ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[3].release_branch }}
+        if: ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[3] && fromJson(env.AUTO_RELEASE_PARAMS).releases[3].ready }}
+        uses: ./.github/actions/release
+        with:
+          ref: ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[3].commit_sha }}
+          type: patch
+          dry-run: ${{ env.DRY_RUN }}
+          token: ${{secrets.ROBOT_CLICKHOUSE_COMMIT_TOKEN}}
+      - name: Release ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[4].release_branch }}
+        if: ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[4] && fromJson(env.AUTO_RELEASE_PARAMS).releases[4].ready }}
+        uses: ./.github/actions/release
+        with:
+          ref: ${{ fromJson(env.AUTO_RELEASE_PARAMS).releases[4].commit_sha }}
+          type: patch
+          dry-run: ${{ env.DRY_RUN }}
+          token: ${{secrets.ROBOT_CLICKHOUSE_COMMIT_TOKEN}}
+      - name: Post Slack Message
+        if: ${{ !cancelled() }}
+        run: |
+          cd "$GITHUB_WORKSPACE/tests/ci"
+          python3 auto_release.py --post-auto-release-complete --wf-status ${{ job.status }}
+      - name: Clean up
         run: |
           docker ps --quiet | xargs --no-run-if-empty docker kill ||:
           docker ps --all --quiet | xargs --no-run-if-empty docker rm -f ||:

.github/workflows/create_release.yml

@@ -2,7 +2,6 @@ name: CreateRelease
 
 concurrency:
   group: release
-
 'on':
   workflow_dispatch:
     inputs:
@@ -31,136 +30,15 @@ jobs:
     steps:
       - name: DebugInfo
         uses: hmarr/debug-action@f7318c783045ac39ed9bb497e22ce835fdafbfe6
-      - name: Set envs
-        # https://docs.github.com/en/actions/learn-github-actions/workflow-commands-for-github-actions#multiline-strings
-        run: |
-          cat >> "$GITHUB_ENV" << 'EOF'
-          ROBOT_CLICKHOUSE_SSH_KEY<<RCSK
-          ${{secrets.ROBOT_CLICKHOUSE_SSH_KEY}}
-          RCSK
-          RELEASE_INFO_FILE=${{ runner.temp }}/release_info.json
-          EOF
       - name: Check out repository code
         uses: ClickHouse/checkout@v1
         with:
           token: ${{secrets.ROBOT_CLICKHOUSE_COMMIT_TOKEN}}
           fetch-depth: 0
-      - name: Prepare Release Info
+      - name: Call Release Action
-        run: |
+        uses: ./.github/actions/release
-          python3 ./tests/ci/create_release.py --prepare-release-info \
-            --ref ${{ inputs.ref }} --release-type ${{ inputs.type }} \
-            --outfile ${{ env.RELEASE_INFO_FILE }} ${{ inputs.dry-run && '--dry-run' || '' }}
-          echo "::group::Release Info"
-          python3 -m json.tool "$RELEASE_INFO_FILE"
-          echo "::endgroup::"
-          release_tag=$(jq -r '.release_tag' "$RELEASE_INFO_FILE")
-          commit_sha=$(jq -r '.commit_sha' "$RELEASE_INFO_FILE")
-          echo "Release Tag: $release_tag"
-          echo "RELEASE_TAG=$release_tag" >> "$GITHUB_ENV"
-          echo "COMMIT_SHA=$commit_sha" >> "$GITHUB_ENV"
-      - name: Download All Release Artifacts
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-          python3 ./tests/ci/create_release.py --infile "$RELEASE_INFO_FILE" --download-packages ${{ inputs.dry-run && '--dry-run' || '' }}
-      - name: Push Git Tag for the Release
-        run: |
-          python3 ./tests/ci/create_release.py --push-release-tag --infile "$RELEASE_INFO_FILE" ${{ inputs.dry-run && '--dry-run' || '' }}
-      - name: Push New Release Branch
-        if: ${{ inputs.type == 'new' }}
-        run: |
-          python3 ./tests/ci/create_release.py --push-new-release-branch --infile "$RELEASE_INFO_FILE" ${{ inputs.dry-run && '--dry-run' || '' }}
-      - name: Bump CH Version and Update Contributors' List
-        run: |
-          python3 ./tests/ci/create_release.py --create-bump-version-pr --infile "$RELEASE_INFO_FILE"  ${{ inputs.dry-run && '--dry-run' || '' }}
-      - name: Checkout master
-        run: |
-          git checkout master
-      - name: Bump Docker versions, Changelog, Security
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-          [ "$(git branch --show-current)" != "master" ] && echo "not on the master" && exit 1
-          echo "List versions"
-          ./utils/list-versions/list-versions.sh > ./utils/list-versions/version_date.tsv
-          echo "Update docker version"
-          ./utils/list-versions/update-docker-version.sh
-          echo "Generate ChangeLog"
-          export CI=1
-          docker run -u "${UID}:${GID}" -e PYTHONUNBUFFERED=1 -e CI=1 --network=host \
-            --volume=".:/ClickHouse" clickhouse/style-test \
-            /ClickHouse/tests/ci/changelog.py -v --debug-helpers \
-            --gh-user-or-token="$GH_TOKEN" --jobs=5 \
-            --output="/ClickHouse/docs/changelogs/${{ env.RELEASE_TAG }}.md" ${{ env.RELEASE_TAG }}
-          git add ./docs/changelogs/${{ env.RELEASE_TAG }}.md
-          echo "Generate Security"
-          python3 ./utils/security-generator/generate_security.py > SECURITY.md
-          git diff HEAD
-      - name: Create ChangeLog PR
-        if: ${{ inputs.type == 'patch' && ! inputs.dry-run }}
-        uses: peter-evans/create-pull-request@v6
         with:
-          author: "robot-clickhouse <robot-clickhouse@users.noreply.github.com>"
+          ref: ${{ inputs.ref }}
-          token: ${{ secrets.ROBOT_CLICKHOUSE_COMMIT_TOKEN }}
+          type: ${{ inputs.type }}
-          committer: "robot-clickhouse <robot-clickhouse@users.noreply.github.com>"
+          dry-run: ${{ inputs.dry-run }}
-          commit-message: Update version_date.tsv and changelogs after ${{ env.RELEASE_TAG }}
+          token: ${{secrets.ROBOT_CLICKHOUSE_COMMIT_TOKEN}}
-          branch: auto/${{ env.RELEASE_TAG }}
-          assignees: ${{ github.event.sender.login }}  # assign the PR to the tag pusher
-          delete-branch: true
-          title: Update version_date.tsv and changelog after ${{ env.RELEASE_TAG }}
-          labels: do not test
-          body: |
-            Update version_date.tsv and changelogs after ${{ env.RELEASE_TAG }}
-            ### Changelog category (leave one):
-            - Not for changelog (changelog entry is not required)
-      - name: Reset changes if Dry-run
-        if: ${{ inputs.dry-run }}
-        run: |
-          git reset --hard HEAD
-      - name: Checkout back to GITHUB_REF
-        run: |
-          git checkout "$GITHUB_REF_NAME"
-      - name: Create GH Release
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-          python3 ./tests/ci/create_release.py --create-gh-release \
-            --infile ${{ env.RELEASE_INFO_FILE }} ${{ inputs.dry-run && '--dry-run' || '' }}
-
-      - name: Export TGZ Packages
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-          python3 ./tests/ci/artifactory.py --export-tgz --infile ${{ env.RELEASE_INFO_FILE }} ${{ inputs.dry-run && '--dry-run' || '' }}
-      - name: Test TGZ Packages
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-          python3 ./tests/ci/artifactory.py --test-tgz --infile ${{ env.RELEASE_INFO_FILE }} ${{ inputs.dry-run && '--dry-run' || '' }}
-      - name: Export RPM Packages
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-          python3 ./tests/ci/artifactory.py --export-rpm --infile ${{ env.RELEASE_INFO_FILE }} ${{ inputs.dry-run && '--dry-run' || '' }}
-      - name: Test RPM Packages
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-          python3 ./tests/ci/artifactory.py --test-rpm --infile ${{ env.RELEASE_INFO_FILE }} ${{ inputs.dry-run && '--dry-run' || '' }}
-      - name: Export Debian Packages
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-            python3 ./tests/ci/artifactory.py --export-debian --infile ${{ env.RELEASE_INFO_FILE }} ${{ inputs.dry-run && '--dry-run' || '' }}
-      - name: Test Debian Packages
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-          python3 ./tests/ci/artifactory.py --test-debian --infile ${{ env.RELEASE_INFO_FILE }} ${{ inputs.dry-run && '--dry-run' || '' }}
-      - name: Docker clickhouse/clickhouse-server building
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-          cd "./tests/ci"
-          export CHECK_NAME="Docker server image"
-          python3 docker_server.py --release-type auto --version ${{ env.RELEASE_TAG }} --check-name "$CHECK_NAME" --sha ${{ env.COMMIT_SHA }} ${{ ! inputs.dry-run && '--push' || '' }}
-      - name: Docker clickhouse/clickhouse-keeper building
-        if: ${{ inputs.type == 'patch' }}
-        run: |
-          cd "./tests/ci"
-          export CHECK_NAME="Docker keeper image"
-          python3 docker_server.py --release-type auto --version ${{ env.RELEASE_TAG }} --check-name "$CHECK_NAME" --sha ${{ env.COMMIT_SHA }} ${{ ! inputs.dry-run && '--push'  || '' }}
-      - name: Post Slack Message
-        if: always()
-        run: |
-          echo Slack Message

docs/en/sql-reference/statements/select/join.md

@@ -297,7 +297,7 @@ Algorithm requires the special column in tables. This column:
 
 - Must contain an ordered sequence.
 - Can be one of the following types: [Int, UInt](../../../sql-reference/data-types/int-uint.md), [Float](../../../sql-reference/data-types/float.md), [Date](../../../sql-reference/data-types/date.md), [DateTime](../../../sql-reference/data-types/datetime.md), [Decimal](../../../sql-reference/data-types/decimal.md).
-- Can’t be the only column in the `JOIN` clause.
+- For `hash` join algorithm it can’t be the only column in the `JOIN` clause.
 
 Syntax `ASOF JOIN ... ON`:
 
@@ -337,7 +337,8 @@ For example, consider the following tables:
 `ASOF JOIN` can take the timestamp of a user event from `table_1` and find an event in `table_2` where the timestamp is closest to the timestamp of the event from `table_1` corresponding to the closest match condition. Equal timestamp values are the closest if available. Here, the `user_id` column can be used for joining on equality and the `ev_time` column can be used for joining on the closest match. In our example, `event_1_1` can be joined with `event_2_1` and `event_1_2` can be joined with `event_2_3`, but `event_2_2` can’t be joined.
 
 :::note
-`ASOF` join is **not** supported in the [Join](../../../engines/table-engines/special/join.md) table engine.
+`ASOF JOIN` is supported only by `hash` and `full_sorting_merge` join algorithms.
+It's **not** supported in the [Join](../../../engines/table-engines/special/join.md) table engine.
 :::
 
 ## PASTE JOIN Usage

docs/ru/getting-started/playground.md

@@ -6,38 +6,38 @@ sidebar_label: Playground
 
 # ClickHouse Playground {#clickhouse-playground}
 
-[ClickHouse Playground](https://play.clickhouse.com/play?user=play) allows people to experiment with ClickHouse by running queries instantly, without setting up their server or cluster.
+[ClickHouse Playground](https://play.clickhouse.com/play?user=play) позволяет пользователям экспериментировать с ClickHouse, выполняя запросы мгновенно, без необходимости настройки сервера или кластера.
-Several example datasets are available in Playground.
+В Playground доступны несколько примеров наборов данных.
 
-You can make queries to Playground using any HTTP client, for example [curl](https://curl.haxx.se) or [wget](https://www.gnu.org/software/wget/), or set up a connection using [JDBC](../interfaces/jdbc.md) or [ODBC](../interfaces/odbc.md) drivers. More information about software products that support ClickHouse is available [here](../interfaces/index.md).
+Вы можете выполнять запросы к Playground, используя любой HTTP-клиент, например [curl](https://curl.haxx.se) или [wget](https://www.gnu.org/software/wget/), или настроить соединение, используя драйверы [JDBC](../interfaces/jdbc.md) или [ODBC](../interfaces/odbc.md). Дополнительную информацию о программных продуктах, поддерживающих ClickHouse, можно найти [здесь](../interfaces/index.md).
 
-## Credentials {#credentials}
+## Учетные данные {#credentials}
 
-| Parameter           | Value                              |
+| Параметр            | Значение                           |
 |:--------------------|:-----------------------------------|
-| HTTPS endpoint      | `https://play.clickhouse.com:443/` |
+| HTTPS-адрес         | `https://play.clickhouse.com:443/` |
-| Native TCP endpoint | `play.clickhouse.com:9440`         |
+| TCP-адрес           | `play.clickhouse.com:9440`         |
-| User                | `explorer` or `play`               |
+| Пользователь        | `explorer` или `play`              |
-| Password            | (empty)                            |
+| Пароль              | (пусто)                            |
 
-## Limitations {#limitations}
+## Ограничения {#limitations}
 
-The queries are executed as a read-only user. It implies some limitations:
+Запросы выполняются от имени пользователя с правами только на чтение. Это предполагает некоторые ограничения:
 
--   DDL queries are not allowed
+-   DDL-запросы не разрешены
--   INSERT queries are not allowed
+-   INSERT-запросы не разрешены
 
-The service also have quotas on its usage.
+Сервис также имеет квоты на использование.
 
-## Examples {#examples}
+## Примеры {#examples}
 
-HTTPS endpoint example with `curl`:
+Пример использования HTTPS-адреса с `curl`:
 
-``` bash
+```bash
 curl "https://play.clickhouse.com/?user=explorer" --data-binary "SELECT 'Play ClickHouse'"
 ```
 
-TCP endpoint example with [CLI](../interfaces/cli.md):
+Пример использования TCP-адреса с [CLI](../interfaces/cli.md):
 
 ``` bash
 clickhouse client --secure --host play.clickhouse.com --user explorer

src/Analyzer/Passes/LogicalExpressionOptimizerPass.cpp

@@ -68,6 +68,41 @@ QueryTreeNodePtr findEqualsFunction(const QueryTreeNodes & nodes)
     return nullptr;
 }
 
+/// Checks if the node is combination of isNull and notEquals functions of two the same arguments
+bool matchIsNullOfTwoArgs(const QueryTreeNodes & nodes, QueryTreeNodePtr & lhs, QueryTreeNodePtr & rhs)
+{
+    QueryTreeNodePtrWithHashSet all_arguments;
+    for (const auto & node : nodes)
+    {
+        const auto * func_node = node->as<FunctionNode>();
+        if (!func_node)
+            return false;
+
+        const auto & arguments = func_node->getArguments().getNodes();
+        if (func_node->getFunctionName() == "isNull" && arguments.size() == 1)
+            all_arguments.insert(QueryTreeNodePtrWithHash(arguments[0]));
+        else if (func_node->getFunctionName() == "notEquals" && arguments.size() == 2)
+        {
+            if (arguments[0]->isEqual(*arguments[1]))
+                return false;
+            all_arguments.insert(QueryTreeNodePtrWithHash(arguments[0]));
+            all_arguments.insert(QueryTreeNodePtrWithHash(arguments[1]));
+        }
+        else
+            return false;
+
+        if (all_arguments.size() > 2)
+            return false;
+    }
+
+    if (all_arguments.size() != 2)
+        return false;
+
+    lhs = all_arguments.begin()->node;
+    rhs = std::next(all_arguments.begin())->node;
+    return true;
+}
+
 bool isBooleanConstant(const QueryTreeNodePtr & node, bool expected_value)
 {
     const auto * constant_node = node->as<ConstantNode>();
@@ -213,11 +248,14 @@ private:
             else if (func_name == "and")
             {
                 const auto & and_arguments = argument_function->getArguments().getNodes();
-                bool all_are_is_null = and_arguments.size() == 2 && isNodeFunction(and_arguments[0], "isNull") && isNodeFunction(and_arguments[1], "isNull");
+
-                if (all_are_is_null)
+                QueryTreeNodePtr is_null_lhs_arg;
+                QueryTreeNodePtr is_null_rhs_arg;
+                if (matchIsNullOfTwoArgs(and_arguments, is_null_lhs_arg, is_null_rhs_arg))
                 {
-                    is_null_argument_to_indices[getFunctionArgument(and_arguments.front(), 0)].push_back(or_operands.size() - 1);
+                    is_null_argument_to_indices[is_null_lhs_arg].push_back(or_operands.size() - 1);
-                    is_null_argument_to_indices[getFunctionArgument(and_arguments.back(), 0)].push_back(or_operands.size() - 1);
+                    is_null_argument_to_indices[is_null_rhs_arg].push_back(or_operands.size() - 1);
+                    continue;
                 }
 
                 /// Expression `a = b AND (a IS NOT NULL) AND true AND (b IS NOT NULL)` we can be replaced with `a = b`

src/Analyzer/Resolve/QueryAnalyzer.cpp

@@ -4124,7 +4124,9 @@ void QueryAnalyzer::resolveInterpolateColumnsNodeList(QueryTreeNodePtr & interpo
 
         auto * column_to_interpolate = interpolate_node_typed.getExpression()->as<IdentifierNode>();
         if (!column_to_interpolate)
-            throw Exception(ErrorCodes::LOGICAL_ERROR, "INTERPOLATE can work only for indentifiers, but {} is found",
+            throw Exception(
+                ErrorCodes::LOGICAL_ERROR,
+                "INTERPOLATE can work only for identifiers, but {} is found",
                 interpolate_node_typed.getExpression()->formatASTForErrorMessage());
         auto column_to_interpolate_name = column_to_interpolate->getIdentifier().getFullName();
 

src/Columns/ColumnAggregateFunction.cpp

@@ -366,13 +366,10 @@ void ColumnAggregateFunction::updateHashWithValue(size_t n, SipHash & hash) cons
     hash.update(wbuf.str().c_str(), wbuf.str().size());
 }
 
-void ColumnAggregateFunction::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnAggregateFunction::getWeakHash32() const
 {
     auto s = data.size();
-    if (hash.getData().size() != data.size())
+    WeakHash32 hash(s);
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
-                        "column size is {}, hash size is {}", std::to_string(s), hash.getData().size());
-
     auto & hash_data = hash.getData();
 
     std::vector<UInt8> v;
@@ -383,6 +380,8 @@ void ColumnAggregateFunction::updateWeakHash32(WeakHash32 & hash) const
         wbuf.finalize();
         hash_data[i] = ::updateWeakHash32(v.data(), v.size(), hash_data[i]);
     }
+
+    return hash;
 }
 
 void ColumnAggregateFunction::updateHashFast(SipHash & hash) const

src/Columns/ColumnAggregateFunction.h

@@ -177,7 +177,7 @@ public:
 
     void updateHashWithValue(size_t n, SipHash & hash) const override;
 
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
 
     void updateHashFast(SipHash & hash) const override;
 

src/Columns/ColumnArray.cpp

@@ -271,15 +271,12 @@ void ColumnArray::updateHashWithValue(size_t n, SipHash & hash) const
         getData().updateHashWithValue(offset + i, hash);
 }
 
-void ColumnArray::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnArray::getWeakHash32() const
 {
     auto s = offsets->size();
-    if (hash.getData().size() != s)
+    WeakHash32 hash(s);
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
-                        "column size is {}, hash size is {}", s, hash.getData().size());
 
-    WeakHash32 internal_hash(data->size());
+    WeakHash32 internal_hash = data->getWeakHash32();
-    data->updateWeakHash32(internal_hash);
 
     Offset prev_offset = 0;
     const auto & offsets_data = getOffsets();
@@ -300,6 +297,8 @@ void ColumnArray::updateWeakHash32(WeakHash32 & hash) const
 
         prev_offset = offsets_data[i];
     }
+
+    return hash;
 }
 
 void ColumnArray::updateHashFast(SipHash & hash) const

src/Columns/ColumnArray.h

@@ -82,7 +82,7 @@ public:
     const char * deserializeAndInsertFromArena(const char * pos) override;
     const char * skipSerializedInArena(const char * pos) const override;
     void updateHashWithValue(size_t n, SipHash & hash) const override;
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
     void updateHashFast(SipHash & hash) const override;
 #if !defined(ABORT_ON_LOGICAL_ERROR)
     void insertRangeFrom(const IColumn & src, size_t start, size_t length) override;

src/Columns/ColumnCompressed.h

@@ -3,6 +3,7 @@
 #include <optional>
 #include <Core/Field.h>
 #include <Columns/IColumn.h>
+#include <Common/WeakHash.h>
 #include <IO/BufferWithOwnMemory.h>
 
 
@@ -98,7 +99,7 @@ public:
     const char * deserializeAndInsertFromArena(const char *) override { throwMustBeDecompressed(); }
     const char * skipSerializedInArena(const char *) const override { throwMustBeDecompressed(); }
     void updateHashWithValue(size_t, SipHash &) const override { throwMustBeDecompressed(); }
-    void updateWeakHash32(WeakHash32 &) const override { throwMustBeDecompressed(); }
+    WeakHash32 getWeakHash32() const override { throwMustBeDecompressed(); }
     void updateHashFast(SipHash &) const override { throwMustBeDecompressed(); }
     ColumnPtr filter(const Filter &, ssize_t) const override { throwMustBeDecompressed(); }
     void expand(const Filter &, bool) override { throwMustBeDecompressed(); }

src/Columns/ColumnConst.cpp

@@ -137,18 +137,10 @@ void ColumnConst::updatePermutation(PermutationSortDirection /*direction*/, Perm
 {
 }
 
-void ColumnConst::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnConst::getWeakHash32() const
 {
-    if (hash.getData().size() != s)
+    WeakHash32 element_hash = data->getWeakHash32();
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
+    return WeakHash32(s, element_hash.getData()[0]);
-                        "column size is {}, hash size is {}", std::to_string(s), std::to_string(hash.getData().size()));
-
-    WeakHash32 element_hash(1);
-    data->updateWeakHash32(element_hash);
-    size_t data_hash = element_hash.getData()[0];
-
-    for (auto & value : hash.getData())
-        value = static_cast<UInt32>(intHashCRC32(data_hash, value));
 }
 
 void ColumnConst::compareColumn(

src/Columns/ColumnConst.h

@@ -204,7 +204,7 @@ public:
         data->updateHashWithValue(0, hash);
     }
 
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
 
     void updateHashFast(SipHash & hash) const override
     {

src/Columns/ColumnDecimal.cpp

@@ -28,7 +28,6 @@ namespace ErrorCodes
     extern const int PARAMETER_OUT_OF_BOUND;
     extern const int SIZES_OF_COLUMNS_DOESNT_MATCH;
     extern const int NOT_IMPLEMENTED;
-    extern const int LOGICAL_ERROR;
 }
 
 template <is_decimal T>
@@ -76,13 +75,10 @@ void ColumnDecimal<T>::updateHashWithValue(size_t n, SipHash & hash) const
 }
 
 template <is_decimal T>
-void ColumnDecimal<T>::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnDecimal<T>::getWeakHash32() const
 {
     auto s = data.size();
-
+    WeakHash32 hash(s);
-    if (hash.getData().size() != s)
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
-                        "column size is {}, hash size is {}", std::to_string(s), std::to_string(hash.getData().size()));
 
     const T * begin = data.data();
     const T * end = begin + s;
@@ -94,6 +90,8 @@ void ColumnDecimal<T>::updateWeakHash32(WeakHash32 & hash) const
         ++begin;
         ++hash_data;
     }
+
+    return hash;
 }
 
 template <is_decimal T>

src/Columns/ColumnDecimal.h

@@ -102,7 +102,7 @@ public:
     const char * deserializeAndInsertFromArena(const char * pos) override;
     const char * skipSerializedInArena(const char * pos) const override;
     void updateHashWithValue(size_t n, SipHash & hash) const override;
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
     void updateHashFast(SipHash & hash) const override;
 #if !defined(ABORT_ON_LOGICAL_ERROR)
     int compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const override;

src/Columns/ColumnDynamic.h

@@ -4,6 +4,7 @@
 #include <Columns/ColumnVector.h>
 #include <Columns/ColumnVariant.h>
 #include <DataTypes/IDataType.h>
+#include <Common/WeakHash.h>
 
 
 namespace DB
@@ -174,9 +175,9 @@ public:
 
     void updateHashWithValue(size_t n, SipHash & hash) const override;
 
-    void updateWeakHash32(WeakHash32 & hash) const override
+    WeakHash32 getWeakHash32() const override
     {
-        variant_column->updateWeakHash32(hash);
+        return variant_column->getWeakHash32();
     }
 
     void updateHashFast(SipHash & hash) const override

src/Columns/ColumnFixedString.cpp

@@ -137,14 +137,10 @@ void ColumnFixedString::updateHashWithValue(size_t index, SipHash & hash) const
     hash.update(reinterpret_cast<const char *>(&chars[n * index]), n);
 }
 
-void ColumnFixedString::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnFixedString::getWeakHash32() const
 {
     auto s = size();
-
+    WeakHash32 hash(s);
-    if (hash.getData().size() != s)
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
-                        "column size is {}, "
-                        "hash size is {}", std::to_string(s), std::to_string(hash.getData().size()));
 
     const UInt8 * pos = chars.data();
     UInt32 * hash_data = hash.getData().data();
@@ -156,6 +152,8 @@ void ColumnFixedString::updateWeakHash32(WeakHash32 & hash) const
         pos += n;
         ++hash_data;
     }
+
+    return hash;
 }
 
 void ColumnFixedString::updateHashFast(SipHash & hash) const

src/Columns/ColumnFixedString.h

@@ -133,7 +133,7 @@ public:
 
     void updateHashWithValue(size_t index, SipHash & hash) const override;
 
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
 
     void updateHashFast(SipHash & hash) const override;
 

src/Columns/ColumnFunction.h

@@ -4,6 +4,7 @@
 #include <Core/NamesAndTypes.h>
 #include <Core/ColumnsWithTypeAndName.h>
 #include <Columns/IColumn.h>
+#include <Common/WeakHash.h>
 
 
 namespace DB
@@ -130,9 +131,9 @@ public:
         throw Exception(ErrorCodes::NOT_IMPLEMENTED, "updateHashWithValue is not implemented for {}", getName());
     }
 
-    void updateWeakHash32(WeakHash32 &) const override
+    WeakHash32 getWeakHash32() const override
     {
-        throw Exception(ErrorCodes::NOT_IMPLEMENTED, "updateWeakHash32 is not implemented for {}", getName());
+        throw Exception(ErrorCodes::NOT_IMPLEMENTED, "getWeakHash32 is not implemented for {}", getName());
     }
 
     void updateHashFast(SipHash &) const override

src/Columns/ColumnLowCardinality.cpp

@@ -7,8 +7,7 @@
 #include <Common/HashTable/HashMap.h>
 #include <Common/WeakHash.h>
 #include <Common/assert_cast.h>
-#include "Storages/IndicesDescription.h"
+#include <base/types.h>
-#include "base/types.h"
 #include <base/sort.h>
 #include <base/scope_guard.h>
 
@@ -320,19 +319,10 @@ const char * ColumnLowCardinality::skipSerializedInArena(const char * pos) const
     return getDictionary().skipSerializedInArena(pos);
 }
 
-void ColumnLowCardinality::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnLowCardinality::getWeakHash32() const
 {
-    auto s = size();
+    WeakHash32 dict_hash = getDictionary().getNestedColumn()->getWeakHash32();
-
+    return idx.getWeakHash(dict_hash);
-    if (hash.getData().size() != s)
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
-                        "column size is {}, hash size is {}", std::to_string(s), std::to_string(hash.getData().size()));
-
-    const auto & dict = getDictionary().getNestedColumn();
-    WeakHash32 dict_hash(dict->size());
-    dict->updateWeakHash32(dict_hash);
-
-    idx.updateWeakHash(hash, dict_hash);
 }
 
 void ColumnLowCardinality::updateHashFast(SipHash & hash) const
@@ -832,10 +822,11 @@ bool ColumnLowCardinality::Index::containsDefault() const
     return contains;
 }
 
-void ColumnLowCardinality::Index::updateWeakHash(WeakHash32 & hash, WeakHash32 & dict_hash) const
+WeakHash32 ColumnLowCardinality::Index::getWeakHash(const WeakHash32 & dict_hash) const
 {
+    WeakHash32 hash(positions->size());
     auto & hash_data = hash.getData();
-    auto & dict_hash_data = dict_hash.getData();
+    const auto & dict_hash_data = dict_hash.getData();
 
     auto update_weak_hash = [&](auto x)
     {
@@ -844,10 +835,11 @@ void ColumnLowCardinality::Index::updateWeakHash(WeakHash32 & hash, WeakHash32 &
         auto size = data.size();
 
         for (size_t i = 0; i < size; ++i)
-            hash_data[i] = static_cast<UInt32>(intHashCRC32(dict_hash_data[data[i]], hash_data[i]));
+            hash_data[i] = dict_hash_data[data[i]];
     };
 
     callForType(std::move(update_weak_hash), size_of_type);
+    return hash;
 }
 
 void ColumnLowCardinality::Index::collectSerializedValueSizes(

src/Columns/ColumnLowCardinality.h

@@ -111,7 +111,7 @@ public:
         getDictionary().updateHashWithValue(getIndexes().getUInt(n), hash);
     }
 
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
 
     void updateHashFast(SipHash &) const override;
 
@@ -325,7 +325,7 @@ public:
 
         bool containsDefault() const;
 
-        void updateWeakHash(WeakHash32 & hash, WeakHash32 & dict_hash) const;
+        WeakHash32 getWeakHash(const WeakHash32 & dict_hash) const;
 
         void collectSerializedValueSizes(PaddedPODArray<UInt64> & sizes, const PaddedPODArray<UInt64> & dict_sizes) const;
 

src/Columns/ColumnMap.cpp

@@ -143,9 +143,9 @@ void ColumnMap::updateHashWithValue(size_t n, SipHash & hash) const
     nested->updateHashWithValue(n, hash);
 }
 
-void ColumnMap::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnMap::getWeakHash32() const
 {
-    nested->updateWeakHash32(hash);
+    return nested->getWeakHash32();
 }
 
 void ColumnMap::updateHashFast(SipHash & hash) const

src/Columns/ColumnMap.h

@@ -64,7 +64,7 @@ public:
     const char * deserializeAndInsertFromArena(const char * pos) override;
     const char * skipSerializedInArena(const char * pos) const override;
     void updateHashWithValue(size_t n, SipHash & hash) const override;
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
     void updateHashFast(SipHash & hash) const override;
 
 #if !defined(ABORT_ON_LOGICAL_ERROR)

src/Columns/ColumnNullable.cpp

@@ -56,25 +56,21 @@ void ColumnNullable::updateHashWithValue(size_t n, SipHash & hash) const
         getNestedColumn().updateHashWithValue(n, hash);
 }
 
-void ColumnNullable::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnNullable::getWeakHash32() const
 {
     auto s = size();
 
-    if (hash.getData().size() != s)
+    WeakHash32 hash = nested_column->getWeakHash32();
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
-                        "column size is {}, hash size is {}", std::to_string(s), std::to_string(hash.getData().size()));
-
-    WeakHash32 old_hash = hash;
-    nested_column->updateWeakHash32(hash);
 
     const auto & null_map_data = getNullMapData();
     auto & hash_data = hash.getData();
-    auto & old_hash_data = old_hash.getData();
 
-    /// Use old data for nulls.
+    /// Use default for nulls.
     for (size_t row = 0; row < s; ++row)
         if (null_map_data[row])
-            hash_data[row] = old_hash_data[row];
+            hash_data[row] = WeakHash32::kDefaultInitialValue;
+
+    return hash;
 }
 
 void ColumnNullable::updateHashFast(SipHash & hash) const

src/Columns/ColumnNullable.h

@@ -133,7 +133,7 @@ public:
     void protect() override;
     ColumnPtr replicate(const Offsets & replicate_offsets) const override;
     void updateHashWithValue(size_t n, SipHash & hash) const override;
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
     void updateHashFast(SipHash & hash) const override;
     void getExtremes(Field & min, Field & max) const override;
     // Special function for nullable minmax index

src/Columns/ColumnObject.h

@@ -5,6 +5,7 @@
 #include <Core/Names.h>
 #include <DataTypes/Serializations/SubcolumnsTree.h>
 #include <Common/PODArray.h>
+#include <Common/WeakHash.h>
 
 #include <DataTypes/IDataType.h>
 
@@ -252,7 +253,7 @@ public:
     const char * deserializeAndInsertFromArena(const char *) override { throwMustBeConcrete(); }
     const char * skipSerializedInArena(const char *) const override { throwMustBeConcrete(); }
     void updateHashWithValue(size_t, SipHash &) const override { throwMustBeConcrete(); }
-    void updateWeakHash32(WeakHash32 &) const override { throwMustBeConcrete(); }
+    WeakHash32 getWeakHash32() const override { throwMustBeConcrete(); }
     void updateHashFast(SipHash & hash) const override;
     void expand(const Filter &, bool) override { throwMustBeConcrete(); }
     bool hasEqualValues() const override { throwMustBeConcrete(); }

src/Columns/ColumnSparse.cpp

@@ -678,20 +678,22 @@ void ColumnSparse::updateHashWithValue(size_t n, SipHash & hash) const
     values->updateHashWithValue(getValueIndex(n), hash);
 }
 
-void ColumnSparse::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnSparse::getWeakHash32() const
 {
-    if (hash.getData().size() != _size)
+    WeakHash32 values_hash = values->getWeakHash32();
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
+    WeakHash32 hash(size());
-        "column size is {}, hash size is {}", _size, hash.getData().size());
+
+    auto & hash_data = hash.getData();
+    auto & values_hash_data = values_hash.getData();
 
     auto offset_it = begin();
-    auto & hash_data = hash.getData();
     for (size_t i = 0; i < _size; ++i, ++offset_it)
     {
         size_t value_index = offset_it.getValueIndex();
-        auto data_ref = values->getDataAt(value_index);
+        hash_data[i] = values_hash_data[value_index];
-        hash_data[i] = ::updateWeakHash32(reinterpret_cast<const UInt8 *>(data_ref.data), data_ref.size, hash_data[i]);
     }
+
+    return hash;
 }
 
 void ColumnSparse::updateHashFast(SipHash & hash) const

src/Columns/ColumnSparse.h

@@ -139,7 +139,7 @@ public:
     void protect() override;
     ColumnPtr replicate(const Offsets & replicate_offsets) const override;
     void updateHashWithValue(size_t n, SipHash & hash) const override;
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
     void updateHashFast(SipHash & hash) const override;
     void getExtremes(Field & min, Field & max) const override;
 

src/Columns/ColumnString.cpp

@@ -108,13 +108,10 @@ MutableColumnPtr ColumnString::cloneResized(size_t to_size) const
     return res;
 }
 
-void ColumnString::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnString::getWeakHash32() const
 {
     auto s = offsets.size();
-
+    WeakHash32 hash(s);
-    if (hash.getData().size() != s)
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
-                        "column size is {}, hash size is {}", std::to_string(s), std::to_string(hash.getData().size()));
 
     const UInt8 * pos = chars.data();
     UInt32 * hash_data = hash.getData().data();
@@ -130,6 +127,8 @@ void ColumnString::updateWeakHash32(WeakHash32 & hash) const
         prev_offset = offset;
         ++hash_data;
     }
+
+    return hash;
 }
 
 

src/Columns/ColumnString.h

@@ -212,7 +212,7 @@ public:
         hash.update(reinterpret_cast<const char *>(&chars[offset]), string_size);
     }
 
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
 
     void updateHashFast(SipHash & hash) const override
     {

src/Columns/ColumnTuple.cpp

@@ -310,16 +310,15 @@ void ColumnTuple::updateHashWithValue(size_t n, SipHash & hash) const
         column->updateHashWithValue(n, hash);
 }
 
-void ColumnTuple::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnTuple::getWeakHash32() const
 {
     auto s = size();
-
+    WeakHash32 hash(s);
-    if (hash.getData().size() != s)
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
-                        "column size is {}, hash size is {}", std::to_string(s), std::to_string(hash.getData().size()));
 
     for (const auto & column : columns)
-        column->updateWeakHash32(hash);
+        hash.update(column->getWeakHash32());
+
+    return hash;
 }
 
 void ColumnTuple::updateHashFast(SipHash & hash) const

src/Columns/ColumnTuple.h

@@ -81,7 +81,7 @@ public:
     const char * deserializeAndInsertFromArena(const char * pos) override;
     const char * skipSerializedInArena(const char * pos) const override;
     void updateHashWithValue(size_t n, SipHash & hash) const override;
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
     void updateHashFast(SipHash & hash) const override;
 #if !defined(ABORT_ON_LOGICAL_ERROR)
     void insertRangeFrom(const IColumn & src, size_t start, size_t length) override;

src/Columns/ColumnVariant.cpp

@@ -789,36 +789,26 @@ void ColumnVariant::updateHashWithValue(size_t n, SipHash & hash) const
         variants[localDiscriminatorByGlobal(global_discr)]->updateHashWithValue(offsetAt(n), hash);
 }
 
-void ColumnVariant::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnVariant::getWeakHash32() const
 {
     auto s = size();
 
-    if (hash.getData().size() != s)
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
-                                                   "column size is {}, hash size is {}", std::to_string(s), std::to_string(hash.getData().size()));
-
     /// If we have only NULLs, keep hash unchanged.
     if (hasOnlyNulls())
-        return;
+        return WeakHash32(s);
 
     /// Optimization for case when there is only 1 non-empty variant and no NULLs.
     /// In this case we can just calculate weak hash for this variant.
     if (auto non_empty_local_discr = getLocalDiscriminatorOfOneNoneEmptyVariantNoNulls())
-    {
+        return variants[*non_empty_local_discr]->getWeakHash32();
-        variants[*non_empty_local_discr]->updateWeakHash32(hash);
-        return;
-    }
 
     /// Calculate weak hash for all variants.
     std::vector<WeakHash32> nested_hashes;
     for (const auto & variant : variants)
-    {
+        nested_hashes.emplace_back(variant->getWeakHash32());
-        WeakHash32 nested_hash(variant->size());
-        variant->updateWeakHash32(nested_hash);
-        nested_hashes.emplace_back(std::move(nested_hash));
-    }
 
     /// For each row hash is a hash of corresponding row from corresponding variant.
+    WeakHash32 hash(s);
     auto & hash_data = hash.getData();
     const auto & local_discriminators_data = getLocalDiscriminators();
     const auto & offsets_data = getOffsets();
@@ -827,11 +817,10 @@ void ColumnVariant::updateWeakHash32(WeakHash32 & hash) const
         Discriminator discr = local_discriminators_data[i];
         /// Update hash only for non-NULL values
         if (discr != NULL_DISCRIMINATOR)
-        {
+            hash_data[i] = nested_hashes[discr].getData()[offsets_data[i]];
-            auto nested_hash = nested_hashes[local_discriminators_data[i]].getData()[offsets_data[i]];
-            hash_data[i] = static_cast<UInt32>(hashCRC32(nested_hash, hash_data[i]));
-        }
     }
+
+    return hash;
 }
 
 void ColumnVariant::updateHashFast(SipHash & hash) const

src/Columns/ColumnVariant.h

@@ -213,7 +213,7 @@ public:
     const char * deserializeVariantAndInsertFromArena(Discriminator global_discr, const char * pos);
     const char * skipSerializedInArena(const char * pos) const override;
     void updateHashWithValue(size_t n, SipHash & hash) const override;
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
     void updateHashFast(SipHash & hash) const override;
     ColumnPtr filter(const Filter & filt, ssize_t result_size_hint) const override;
     void expand(const Filter & mask, bool inverted) override;

src/Columns/ColumnVector.cpp

@@ -73,13 +73,10 @@ void ColumnVector<T>::updateHashWithValue(size_t n, SipHash & hash) const
 }
 
 template <typename T>
-void ColumnVector<T>::updateWeakHash32(WeakHash32 & hash) const
+WeakHash32 ColumnVector<T>::getWeakHash32() const
 {
     auto s = data.size();
-
+    WeakHash32 hash(s);
-    if (hash.getData().size() != s)
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match size of column: "
-                        "column size is {}, hash size is {}", std::to_string(s), std::to_string(hash.getData().size()));
 
     const T * begin = data.data();
     const T * end = begin + s;
@@ -91,6 +88,8 @@ void ColumnVector<T>::updateWeakHash32(WeakHash32 & hash) const
         ++begin;
         ++hash_data;
     }
+
+    return hash;
 }
 
 template <typename T>

src/Columns/ColumnVector.h

@@ -114,7 +114,7 @@ public:
 
     void updateHashWithValue(size_t n, SipHash & hash) const override;
 
-    void updateWeakHash32(WeakHash32 & hash) const override;
+    WeakHash32 getWeakHash32() const override;
 
     void updateHashFast(SipHash & hash) const override;
 

src/Columns/IColumn.h

@@ -300,10 +300,10 @@ public:
     ///  passed bytes to hash must identify sequence of values unambiguously.
     virtual void updateHashWithValue(size_t n, SipHash & hash) const = 0;
 
-    /// Update hash function value. Hash is calculated for each element.
+    /// Get hash function value. Hash is calculated for each element.
     /// It's a fast weak hash function. Mainly need to scatter data between threads.
     /// WeakHash32 must have the same size as column.
-    virtual void updateWeakHash32(WeakHash32 & hash) const = 0;
+    virtual WeakHash32 getWeakHash32() const = 0;
 
     /// Update state of hash with all column.
     virtual void updateHashFast(SipHash & hash) const = 0;

src/Columns/IColumnDummy.h

@@ -1,6 +1,7 @@
 #pragma once
 
 #include <Columns/IColumn.h>
+#include <Common/WeakHash.h>
 
 
 namespace DB
@@ -63,8 +64,9 @@ public:
     {
     }
 
-    void updateWeakHash32(WeakHash32 & /*hash*/) const override
+    WeakHash32 getWeakHash32() const override
     {
+        return WeakHash32(s);
     }
 
     void updateHashFast(SipHash & /*hash*/) const override

src/Columns/IColumnUnique.h

@@ -1,6 +1,7 @@
 #pragma once
 #include <optional>
 #include <Columns/IColumn.h>
+#include <Common/WeakHash.h>
 
 namespace DB
 {
@@ -166,9 +167,9 @@ public:
         throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Method scatter is not supported for ColumnUnique.");
     }
 
-    void updateWeakHash32(WeakHash32 &) const override
+    WeakHash32 getWeakHash32() const override
     {
-        throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Method updateWeakHash32 is not supported for ColumnUnique.");
+        throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Method getWeakHash32 is not supported for ColumnUnique.");
     }
 
     void updateHashFast(SipHash &) const override

src/Columns/tests/gtest_weak_hash_32.cpp

@@ -60,8 +60,7 @@ TEST(WeakHash32, ColumnVectorU8)
             data.push_back(i);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -77,8 +76,7 @@ TEST(WeakHash32, ColumnVectorI8)
             data.push_back(i);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -94,8 +92,7 @@ TEST(WeakHash32, ColumnVectorU16)
             data.push_back(i);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -111,8 +108,7 @@ TEST(WeakHash32, ColumnVectorI16)
             data.push_back(i);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -128,8 +124,7 @@ TEST(WeakHash32, ColumnVectorU32)
             data.push_back(i << 16u);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -145,8 +140,7 @@ TEST(WeakHash32, ColumnVectorI32)
             data.push_back(i << 16);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -162,8 +156,7 @@ TEST(WeakHash32, ColumnVectorU64)
             data.push_back(i << 32u);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -179,8 +172,7 @@ TEST(WeakHash32, ColumnVectorI64)
             data.push_back(i << 32);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -204,8 +196,7 @@ TEST(WeakHash32, ColumnVectorU128)
         }
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), eq_data);
 }
@@ -221,8 +212,7 @@ TEST(WeakHash32, ColumnVectorI128)
             data.push_back(i << 32);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -238,8 +228,7 @@ TEST(WeakHash32, ColumnDecimal32)
             data.push_back(i << 16);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -255,8 +244,7 @@ TEST(WeakHash32, ColumnDecimal64)
             data.push_back(i << 32);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -272,8 +260,7 @@ TEST(WeakHash32, ColumnDecimal128)
             data.push_back(i << 32);
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), col->getData());
 }
@@ -294,8 +281,7 @@ TEST(WeakHash32, ColumnString1)
         }
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), data);
 }
@@ -331,8 +317,7 @@ TEST(WeakHash32, ColumnString2)
         }
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), data);
 }
@@ -369,8 +354,7 @@ TEST(WeakHash32, ColumnString3)
         }
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), data);
 }
@@ -397,8 +381,7 @@ TEST(WeakHash32, ColumnFixedString)
         }
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), data);
 }
@@ -444,8 +427,7 @@ TEST(WeakHash32, ColumnArray)
 
     auto col_arr = ColumnArray::create(std::move(val), std::move(off));
 
-    WeakHash32 hash(col_arr->size());
+    WeakHash32 hash = col_arr->getWeakHash32();
-    col_arr->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), eq_data);
 }
@@ -479,8 +461,7 @@ TEST(WeakHash32, ColumnArray2)
 
     auto col_arr = ColumnArray::create(std::move(val), std::move(off));
 
-    WeakHash32 hash(col_arr->size());
+    WeakHash32 hash = col_arr->getWeakHash32();
-    col_arr->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), eq_data);
 }
@@ -536,8 +517,7 @@ TEST(WeakHash32, ColumnArrayArray)
     auto col_arr = ColumnArray::create(std::move(val), std::move(off));
     auto col_arr_arr = ColumnArray::create(std::move(col_arr), std::move(off2));
 
-    WeakHash32 hash(col_arr_arr->size());
+    WeakHash32 hash = col_arr_arr->getWeakHash32();
-    col_arr_arr->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), eq_data);
 }
@@ -555,8 +535,7 @@ TEST(WeakHash32, ColumnConst)
 
     auto col_const = ColumnConst::create(std::move(inner_col), 256);
 
-    WeakHash32 hash(col_const->size());
+    WeakHash32 hash = col_const->getWeakHash32();
-    col_const->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), data);
 }
@@ -576,8 +555,7 @@ TEST(WeakHash32, ColumnLowcardinality)
         }
     }
 
-    WeakHash32 hash(col->size());
+    WeakHash32 hash = col->getWeakHash32();
-    col->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), data);
 }
@@ -602,8 +580,7 @@ TEST(WeakHash32, ColumnNullable)
 
     auto col_null = ColumnNullable::create(std::move(col), std::move(mask));
 
-    WeakHash32 hash(col_null->size());
+    WeakHash32 hash = col_null->getWeakHash32();
-    col_null->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), eq);
 }
@@ -633,8 +610,7 @@ TEST(WeakHash32, ColumnTupleUInt64UInt64)
     columns.emplace_back(std::move(col2));
     auto col_tuple = ColumnTuple::create(std::move(columns));
 
-    WeakHash32 hash(col_tuple->size());
+    WeakHash32 hash = col_tuple->getWeakHash32();
-    col_tuple->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), eq);
 }
@@ -671,8 +647,7 @@ TEST(WeakHash32, ColumnTupleUInt64String)
     columns.emplace_back(std::move(col2));
     auto col_tuple = ColumnTuple::create(std::move(columns));
 
-    WeakHash32 hash(col_tuple->size());
+    WeakHash32 hash = col_tuple->getWeakHash32();
-    col_tuple->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), eq);
 }
@@ -709,8 +684,7 @@ TEST(WeakHash32, ColumnTupleUInt64FixedString)
     columns.emplace_back(std::move(col2));
     auto col_tuple = ColumnTuple::create(std::move(columns));
 
-    WeakHash32 hash(col_tuple->size());
+    WeakHash32 hash = col_tuple->getWeakHash32();
-    col_tuple->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), eq);
 }
@@ -756,8 +730,7 @@ TEST(WeakHash32, ColumnTupleUInt64Array)
     columns.emplace_back(ColumnArray::create(std::move(val), std::move(off)));
     auto col_tuple = ColumnTuple::create(std::move(columns));
 
-    WeakHash32 hash(col_tuple->size());
+    WeakHash32 hash = col_tuple->getWeakHash32();
-    col_tuple->updateWeakHash32(hash);
 
     checkColumn(hash.getData(), eq_data);
 }

src/Common/Allocator.cpp

@@ -206,7 +206,7 @@ void * Allocator<clear_memory_, populate>::realloc(void * buf, size_t old_size,
         }
         else
         {
-            [[maybe_unused]] auto trace_free = CurrentMemoryTracker::free(old_size);
+            [[maybe_unused]] auto trace_free = CurrentMemoryTracker::free(new_size);
             ProfileEvents::increment(ProfileEvents::GWPAsanAllocateFailed);
         }
     }
@@ -239,7 +239,7 @@ void * Allocator<clear_memory_, populate>::realloc(void * buf, size_t old_size,
         void * new_buf = ::realloc(buf, new_size);
         if (nullptr == new_buf)
         {
-            [[maybe_unused]] auto trace_free = CurrentMemoryTracker::free(old_size);
+            [[maybe_unused]] auto trace_free = CurrentMemoryTracker::free(new_size);
             throw DB::ErrnoException(
                 DB::ErrorCodes::CANNOT_ALLOCATE_MEMORY,
                 "Allocator: Cannot realloc from {} to {}",

src/Common/Exception.h

@@ -244,6 +244,15 @@ private:
     const char * className() const noexcept override { return "DB::ErrnoException"; }
 };
 
+/// An exception to use in unit tests to test interfaces.
+/// It is distinguished from others, so it does not have to be logged.
+class TestException : public Exception
+{
+public:
+    using Exception::Exception;
+};
+
+
 using Exceptions = std::vector<std::exception_ptr>;
 
 /** Try to write an exception to the log (and forget about it).

src/Common/TraceSender.cpp

@@ -23,8 +23,20 @@ namespace DB
 
 LazyPipeFDs TraceSender::pipe;
 
+static thread_local bool inside_send = false;
 void TraceSender::send(TraceType trace_type, const StackTrace & stack_trace, Extras extras)
 {
+    /** The method shouldn't be called recursively or throw exceptions.
+      * There are several reasons:
+      * - avoid infinite recursion when some of subsequent functions invoke tracing;
+      * - avoid inconsistent writes if the method was interrupted by a signal handler in the middle of writing,
+      *   and then another tracing is invoked (e.g., from query profiler).
+      */
+    if (unlikely(inside_send))
+        return;
+    inside_send = true;
+    DENY_ALLOCATIONS_IN_SCOPE;
+
     constexpr size_t buf_size = sizeof(char) /// TraceCollector stop flag
         + sizeof(UInt8)                      /// String size
         + QUERY_ID_MAX_LEN                   /// Maximum query_id length
@@ -80,6 +92,8 @@ void TraceSender::send(TraceType trace_type, const StackTrace & stack_trace, Ext
     writePODBinary(extras.increment, out);
 
     out.next();
+
+    inside_send = false;
 }
 
 }

src/Common/WeakHash.cpp

@@ -1,2 +1,24 @@
 #include <Common/WeakHash.h>
+#include <Common/Exception.h>
+#include <Common/HashTable/Hash.h>
 
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int LOGICAL_ERROR;
+}
+
+void WeakHash32::update(const WeakHash32 & other)
+{
+    size_t size = data.size();
+    if (size != other.data.size())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Size of WeakHash32 does not match:"
+                        "left size is {}, right size is {}", size, other.data.size());
+
+    for (size_t i = 0; i < size; ++i)
+        data[i] = static_cast<UInt32>(intHashCRC32(other.data[i], data[i]));
+}
+
+}

src/Common/WeakHash.h

@@ -11,9 +11,8 @@ namespace DB
 /// The main purpose why this class needed is to support data initialization. Initially, every bit is 1.
 class WeakHash32
 {
-    static constexpr UInt32 kDefaultInitialValue = ~UInt32(0);
-
 public:
+    static constexpr UInt32 kDefaultInitialValue = ~UInt32(0);
 
     using Container = PaddedPODArray<UInt32>;
 
@@ -22,6 +21,8 @@ public:
 
     void reset(size_t size, UInt32 initial_value = kDefaultInitialValue) { data.assign(size, initial_value); }
 
+    void update(const WeakHash32 & other);
+
     const Container & getData() const { return data; }
     Container & getData() { return data; }
 

src/Common/getRandomASCIIString.cpp

@@ -6,12 +6,17 @@ namespace DB
 {
 
 String getRandomASCIIString(size_t length)
+{
+    return getRandomASCIIString(length, thread_local_rng);
+}
+
+String getRandomASCIIString(size_t length, pcg64 & rng)
 {
     std::uniform_int_distribution<int> distribution('a', 'z');
     String res;
     res.resize(length);
     for (auto & c : res)
-        c = distribution(thread_local_rng);
+        c = distribution(rng);
     return res;
 }
 

src/Common/getRandomASCIIString.h

@@ -2,11 +2,14 @@
 
 #include <Core/Types.h>
 
+#include <pcg_random.hpp>
+
 
 namespace DB
 {
 
 /// Slow random string. Useful for random names and things like this. Not for generating data.
 String getRandomASCIIString(size_t length);
+String getRandomASCIIString(size_t length, pcg64 & rng);
 
 }

src/Common/tests/gtest_shell_command.cpp

@@ -54,16 +54,3 @@ TEST(ShellCommand, ExecuteWithInput)
 
     EXPECT_EQ(res, "Hello, world!\n");
 }
-
-TEST(ShellCommand, AutoWait)
-{
-    // <defunct> hunting:
-    for (int i = 0; i < 1000; ++i)
-    {
-        auto command = ShellCommand::execute("echo " + std::to_string(i));
-        //command->wait(); // now automatic
-    }
-
-    // std::cerr << "inspect me: ps auxwwf\n";
-    // std::this_thread::sleep_for(std::chrono::seconds(100));
-}

src/Functions/IFunction.cpp

@@ -47,54 +47,85 @@ bool allArgumentsAreConstants(const ColumnsWithTypeAndName & args)
     return true;
 }
 
+/// Replaces single low cardinality column in a function call by its dictionary
+/// This can only happen after the arguments have been adapted in IFunctionOverloadResolver::getReturnType
+/// as it's only possible if there is one low cardinality column and, optionally, const columns
 ColumnPtr replaceLowCardinalityColumnsByNestedAndGetDictionaryIndexes(
     ColumnsWithTypeAndName & args, bool can_be_executed_on_default_arguments, size_t input_rows_count)
 {
-    size_t num_rows = input_rows_count;
+    /// We return the LC indexes so the LC can be reconstructed with the function result
     ColumnPtr indexes;
 
-    /// Find first LowCardinality column and replace it to nested dictionary.
+    size_t number_low_cardinality_columns = 0;
-    for (auto & column : args)
+    size_t last_low_cardinality = 0;
+    size_t number_const_columns = 0;
+    size_t number_full_columns = 0;
+
+    for (size_t i = 0; i < args.size(); i++)
     {
-        if (const auto * low_cardinality_column = checkAndGetColumn<ColumnLowCardinality>(column.column.get()))
+        auto const & arg = args[i];
+        if (checkAndGetColumn<ColumnLowCardinality>(arg.column.get()))
         {
-            /// Single LowCardinality column is supported now.
+            number_low_cardinality_columns++;
-            if (indexes)
+            last_low_cardinality = i;
-                throw Exception(ErrorCodes::LOGICAL_ERROR, "Expected single dictionary argument for function.");
-
-            const auto * low_cardinality_type = checkAndGetDataType<DataTypeLowCardinality>(column.type.get());
-
-            if (!low_cardinality_type)
-                throw Exception(ErrorCodes::LOGICAL_ERROR,
-                    "Incompatible type for LowCardinality column: {}",
-                    column.type->getName());
-
-            if (can_be_executed_on_default_arguments)
-            {
-                /// Normal case, when function can be executed on values' default.
-                column.column = low_cardinality_column->getDictionary().getNestedColumn();
-                indexes = low_cardinality_column->getIndexesPtr();
-            }
-            else
-            {
-                /// Special case when default value can't be used. Example: 1 % LowCardinality(Int).
-                /// LowCardinality always contains default, so 1 % 0 will throw exception in normal case.
-                auto dict_encoded = low_cardinality_column->getMinimalDictionaryEncodedColumn(0, low_cardinality_column->size());
-                column.column = dict_encoded.dictionary;
-                indexes = dict_encoded.indexes;
-            }
-
-            num_rows = column.column->size();
-            column.type = low_cardinality_type->getDictionaryType();
         }
+        else if (checkAndGetColumn<ColumnConst>(arg.column.get()))
+            number_const_columns++;
+        else
+            number_full_columns++;
     }
 
-    /// Change size of constants.
+    if (!number_low_cardinality_columns && !number_const_columns)
+        return nullptr;
+
+    if (number_full_columns > 0 || number_low_cardinality_columns > 1)
+    {
+        /// This should not be possible but currently there are multiple tests in CI failing because of it
+        /// TODO: Fix those cases, then enable this exception
+#if 0
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Unexpected low cardinality types found. Low cardinality: {}. Full {}. Const {}",
+                number_low_cardinality_columns, number_full_columns, number_const_columns);
+#else
+        return nullptr;
+#endif
+    }
+    else if (number_low_cardinality_columns == 1)
+    {
+        auto & lc_arg = args[last_low_cardinality];
+
+        const auto * low_cardinality_type = checkAndGetDataType<DataTypeLowCardinality>(lc_arg.type.get());
+        if (!low_cardinality_type)
+            throw Exception(ErrorCodes::LOGICAL_ERROR, "Incompatible type for LowCardinality column: {}", lc_arg.type->getName());
+
+        const auto * low_cardinality_column = checkAndGetColumn<ColumnLowCardinality>(lc_arg.column.get());
+        chassert(low_cardinality_column);
+
+        if (can_be_executed_on_default_arguments)
+        {
+            /// Normal case, when function can be executed on values' default.
+            lc_arg.column = low_cardinality_column->getDictionary().getNestedColumn();
+            indexes = low_cardinality_column->getIndexesPtr();
+        }
+        else
+        {
+            /// Special case when default value can't be used. Example: 1 % LowCardinality(Int).
+            /// LowCardinality always contains default, so 1 % 0 will throw exception in normal case.
+            auto dict_encoded = low_cardinality_column->getMinimalDictionaryEncodedColumn(0, low_cardinality_column->size());
+            lc_arg.column = dict_encoded.dictionary;
+            indexes = dict_encoded.indexes;
+        }
+
+        /// The new column will have a different number of rows, normally less but occasionally it might be more (NULL)
+        input_rows_count = lc_arg.column->size();
+        lc_arg.type = low_cardinality_type->getDictionaryType();
+    }
+
+    /// Change size of constants
     for (auto & column : args)
     {
         if (const auto * column_const = checkAndGetColumn<ColumnConst>(column.column.get()))
         {
-            column.column = ColumnConst::create(recursiveRemoveLowCardinality(column_const->getDataColumnPtr()), num_rows);
+            column.column = ColumnConst::create(recursiveRemoveLowCardinality(column_const->getDataColumnPtr()), input_rows_count);
             column.type = recursiveRemoveLowCardinality(column.type);
         }
     }
@@ -270,6 +301,8 @@ ColumnPtr IExecutableFunction::executeWithoutSparseColumns(const ColumnsWithType
             bool can_be_executed_on_default_arguments = canBeExecutedOnDefaultArguments();
 
             const auto & dictionary_type = res_low_cardinality_type->getDictionaryType();
+            /// The arguments should have been adapted in IFunctionOverloadResolver::getReturnType
+            /// So there is only one low cardinality column (and optionally some const columns) and no full column
             ColumnPtr indexes = replaceLowCardinalityColumnsByNestedAndGetDictionaryIndexes(
                     columns_without_low_cardinality, can_be_executed_on_default_arguments, input_rows_count);
 

src/Interpreters/ConcurrentHashJoin.cpp

@@ -310,7 +310,7 @@ IColumn::Selector ConcurrentHashJoin::selectDispatchBlock(const Strings & key_co
     {
         const auto & key_col = from_block.getByName(key_name).column->convertToFullColumnIfConst();
         const auto & key_col_no_lc = recursiveRemoveLowCardinality(recursiveRemoveSparse(key_col));
-        key_col_no_lc->updateWeakHash32(hash);
+        hash.update(key_col_no_lc->getWeakHash32());
     }
     return hashToSelector(hash, num_shards);
 }

src/Interpreters/InterpreterSelectQuery.cpp

@@ -1726,7 +1726,10 @@ void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, std::optional<P
                         const auto & join_clause = table_join.getOnlyClause();
 
                         auto join_kind = table_join.kind();
-                        bool kind_allows_filtering = isInner(join_kind) || isLeft(join_kind) || isRight(join_kind);
+                        auto join_strictness = table_join.strictness();
+
+                        bool join_type_allows_filtering = (join_strictness == JoinStrictness::All || join_strictness == JoinStrictness::Any)
+                                                       && (isInner(join_kind) || isLeft(join_kind) || isRight(join_kind));
 
                         auto has_non_const = [](const Block & block, const auto & keys)
                         {
@@ -1745,7 +1748,7 @@ void InterpreterSelectQuery::executeImpl(QueryPlan & query_plan, std::optional<P
                         bool has_non_const_keys = has_non_const(query_plan.getCurrentDataStream().header, join_clause.key_names_left)
                             && has_non_const(joined_plan->getCurrentDataStream().header, join_clause.key_names_right);
 
-                        if (settings.max_rows_in_set_to_optimize_join > 0 && kind_allows_filtering && has_non_const_keys)
+                        if (settings.max_rows_in_set_to_optimize_join > 0 && join_type_allows_filtering && has_non_const_keys)
                         {
                             auto * left_set = add_create_set(query_plan, join_clause.key_names_left, JoinTableSide::Left);
                             auto * right_set = add_create_set(*joined_plan, join_clause.key_names_right, JoinTableSide::Right);

src/Interpreters/JoinUtils.cpp

@@ -554,7 +554,7 @@ static Blocks scatterBlockByHashImpl(const Strings & key_columns_names, const Bl
     for (const auto & key_name : key_columns_names)
     {
         ColumnPtr key_col = materializeColumn(block, key_name);
-        key_col->updateWeakHash32(hash);
+        hash.update(key_col->getWeakHash32());
     }
     auto selector = hashToSelector(hash, sharder);
 

src/Parsers/ASTFunction.cpp

@@ -7,7 +7,6 @@
 #include <Common/FieldVisitorToString.h>
 #include <Common/KnownObjectNames.h>
 #include <Common/SipHash.h>
-#include <Common/typeid_cast.h>
 #include <IO/Operators.h>
 #include <IO/WriteBufferFromString.h>
 #include <IO/WriteHelpers.h>
@@ -19,9 +18,6 @@
 #include <Parsers/queryToString.h>
 #include <Parsers/ASTSetQuery.h>
 #include <Parsers/FunctionSecretArgumentsFinderAST.h>
-#include <Core/QualifiedTableName.h>
-
-#include <boost/algorithm/string.hpp>
 
 
 using namespace std::literals;
@@ -632,6 +628,7 @@ void ASTFunction::formatImplWithoutAlias(const FormatSettings & settings, Format
                     settings.ostr << ", ";
                 if (arguments->children[i]->as<ASTSetQuery>())
                     settings.ostr << "SETTINGS ";
+                nested_dont_need_parens.list_element_index = i;
                 arguments->children[i]->formatImpl(settings, state, nested_dont_need_parens);
             }
             settings.ostr << (settings.hilite ? hilite_operator : "") << ']' << (settings.hilite ? hilite_none : "");
@@ -642,12 +639,14 @@ void ASTFunction::formatImplWithoutAlias(const FormatSettings & settings, Format
         {
             settings.ostr << (settings.hilite ? hilite_operator : "") << ((frame.need_parens && !alias.empty()) ? "tuple" : "") << '('
                           << (settings.hilite ? hilite_none : "");
+
             for (size_t i = 0; i < arguments->children.size(); ++i)
             {
                 if (i != 0)
                     settings.ostr << ", ";
                 if (arguments->children[i]->as<ASTSetQuery>())
                     settings.ostr << "SETTINGS ";
+                nested_dont_need_parens.list_element_index = i;
                 arguments->children[i]->formatImpl(settings, state, nested_dont_need_parens);
             }
             settings.ostr << (settings.hilite ? hilite_operator : "") << ')' << (settings.hilite ? hilite_none : "");
@@ -663,6 +662,7 @@ void ASTFunction::formatImplWithoutAlias(const FormatSettings & settings, Format
                     settings.ostr << ", ";
                 if (arguments->children[i]->as<ASTSetQuery>())
                     settings.ostr << "SETTINGS ";
+                nested_dont_need_parens.list_element_index = i;
                 arguments->children[i]->formatImpl(settings, state, nested_dont_need_parens);
             }
             settings.ostr << (settings.hilite ? hilite_operator : "") << ')' << (settings.hilite ? hilite_none : "");

src/Planner/Planner.cpp

@@ -745,7 +745,12 @@ void addWithFillStepIfNeeded(QueryPlan & query_plan,
             {
                 auto & interpolate_node_typed = interpolate_node->as<InterpolateNode &>();
 
-                PlannerActionsVisitor planner_actions_visitor(planner_context);
+                PlannerActionsVisitor planner_actions_visitor(
+                    planner_context,
+                    /* use_column_identifier_as_action_node_name_, (default value)*/ true,
+                    /// Prefer the INPUT to CONSTANT nodes (actions must be non constant)
+                    /* always_use_const_column_for_constant_nodes */ false);
+
                 auto expression_to_interpolate_expression_nodes = planner_actions_visitor.visit(*interpolate_actions_dag,
                     interpolate_node_typed.getExpression());
                 if (expression_to_interpolate_expression_nodes.size() != 1)

src/Planner/PlannerActionsVisitor.cpp

@@ -487,16 +487,33 @@ public:
         return node;
     }
 
-    const ActionsDAG::Node * addConstantIfNecessary(const std::string & node_name, const ColumnWithTypeAndName & column)
+    [[nodiscard]] String addConstantIfNecessary(
+        const std::string & node_name, const ColumnWithTypeAndName & column, bool always_use_const_column_for_constant_nodes)
     {
+        chassert(column.column != nullptr);
         auto it = node_name_to_node.find(node_name);
+        if (it != node_name_to_node.end() && (!always_use_const_column_for_constant_nodes || it->second->column))
+            return {node_name};
+
         if (it != node_name_to_node.end())
-            return it->second;
+        {
+            /// There is a node with this name, but it doesn't have a column
+            /// This likely happens because we executed the query until WithMergeableState with a const node in the
+            /// WHERE clause and, as the results of headers are materialized, the column was removed
+            /// Let's add a new column and keep this
+            String dupped_name{node_name + "_dupped"};
+            if (node_name_to_node.find(dupped_name) != node_name_to_node.end())
+                return dupped_name;
+
+            const auto * node = &actions_dag.addColumn(column);
+            node_name_to_node[dupped_name] = node;
+            return dupped_name;
+        }
 
         const auto * node = &actions_dag.addColumn(column);
         node_name_to_node[node->result_name] = node;
 
-        return node;
+        return {node_name};
     }
 
     template <typename FunctionOrOverloadResolver>
@@ -525,7 +542,7 @@ public:
     }
 
 private:
-    std::unordered_map<std::string_view, const ActionsDAG::Node *> node_name_to_node;
+    std::unordered_map<String, const ActionsDAG::Node *> node_name_to_node;
     ActionsDAG & actions_dag;
     QueryTreeNodePtr scope_node;
 };
@@ -533,9 +550,11 @@ private:
 class PlannerActionsVisitorImpl
 {
 public:
-    PlannerActionsVisitorImpl(ActionsDAG & actions_dag,
+    PlannerActionsVisitorImpl(
+        ActionsDAG & actions_dag,
         const PlannerContextPtr & planner_context_,
-        bool use_column_identifier_as_action_node_name_);
+        bool use_column_identifier_as_action_node_name_,
+        bool always_use_const_column_for_constant_nodes_);
 
     ActionsDAG::NodeRawConstPtrs visit(QueryTreeNodePtr expression_node);
 
@@ -595,14 +614,18 @@ private:
     const PlannerContextPtr planner_context;
     ActionNodeNameHelper action_node_name_helper;
     bool use_column_identifier_as_action_node_name;
+    bool always_use_const_column_for_constant_nodes;
 };
 
-PlannerActionsVisitorImpl::PlannerActionsVisitorImpl(ActionsDAG & actions_dag,
+PlannerActionsVisitorImpl::PlannerActionsVisitorImpl(
+    ActionsDAG & actions_dag,
     const PlannerContextPtr & planner_context_,
-    bool use_column_identifier_as_action_node_name_)
+    bool use_column_identifier_as_action_node_name_,
+    bool always_use_const_column_for_constant_nodes_)
     : planner_context(planner_context_)
     , action_node_name_helper(node_to_node_name, *planner_context, use_column_identifier_as_action_node_name_)
     , use_column_identifier_as_action_node_name(use_column_identifier_as_action_node_name_)
+    , always_use_const_column_for_constant_nodes(always_use_const_column_for_constant_nodes_)
 {
     actions_stack.emplace_back(actions_dag, nullptr);
 }
@@ -725,17 +748,16 @@ PlannerActionsVisitorImpl::NodeNameAndNodeMinLevel PlannerActionsVisitorImpl::vi
     column.type = constant_type;
     column.column = column.type->createColumnConst(1, constant_literal);
 
-    actions_stack[0].addConstantIfNecessary(constant_node_name, column);
+    String final_name = actions_stack[0].addConstantIfNecessary(constant_node_name, column, always_use_const_column_for_constant_nodes);
 
     size_t actions_stack_size = actions_stack.size();
     for (size_t i = 1; i < actions_stack_size; ++i)
     {
         auto & actions_stack_node = actions_stack[i];
-        actions_stack_node.addInputConstantColumnIfNecessary(constant_node_name, column);
+        actions_stack_node.addInputConstantColumnIfNecessary(final_name, column);
     }
 
-    return {constant_node_name, Levels(0)};
+    return {final_name, Levels(0)};
-
 }
 
 PlannerActionsVisitorImpl::NodeNameAndNodeMinLevel PlannerActionsVisitorImpl::visitLambda(const QueryTreeNodePtr & node)
@@ -864,16 +886,16 @@ PlannerActionsVisitorImpl::NodeNameAndNodeMinLevel PlannerActionsVisitorImpl::ma
     else
         column.column = std::move(column_set);
 
-    actions_stack[0].addConstantIfNecessary(column.name, column);
+    String final_name = actions_stack[0].addConstantIfNecessary(column.name, column, always_use_const_column_for_constant_nodes);
 
     size_t actions_stack_size = actions_stack.size();
     for (size_t i = 1; i < actions_stack_size; ++i)
     {
         auto & actions_stack_node = actions_stack[i];
-        actions_stack_node.addInputConstantColumnIfNecessary(column.name, column);
+        actions_stack_node.addInputConstantColumnIfNecessary(final_name, column);
     }
 
-    return {column.name, Levels(0)};
+    return {final_name, Levels(0)};
 }
 
 PlannerActionsVisitorImpl::NodeNameAndNodeMinLevel PlannerActionsVisitorImpl::visitIndexHintFunction(const QueryTreeNodePtr & node)
@@ -1010,14 +1032,19 @@ PlannerActionsVisitorImpl::NodeNameAndNodeMinLevel PlannerActionsVisitorImpl::vi
 
 }
 
-PlannerActionsVisitor::PlannerActionsVisitor(const PlannerContextPtr & planner_context_, bool use_column_identifier_as_action_node_name_)
+PlannerActionsVisitor::PlannerActionsVisitor(
+    const PlannerContextPtr & planner_context_,
+    bool use_column_identifier_as_action_node_name_,
+    bool always_use_const_column_for_constant_nodes_)
     : planner_context(planner_context_)
     , use_column_identifier_as_action_node_name(use_column_identifier_as_action_node_name_)
+    , always_use_const_column_for_constant_nodes(always_use_const_column_for_constant_nodes_)
 {}
 
 ActionsDAG::NodeRawConstPtrs PlannerActionsVisitor::visit(ActionsDAG & actions_dag, QueryTreeNodePtr expression_node)
 {
-    PlannerActionsVisitorImpl actions_visitor_impl(actions_dag, planner_context, use_column_identifier_as_action_node_name);
+    PlannerActionsVisitorImpl actions_visitor_impl(
+        actions_dag, planner_context, use_column_identifier_as_action_node_name, always_use_const_column_for_constant_nodes);
     return actions_visitor_impl.visit(expression_node);
 }
 

src/Planner/PlannerActionsVisitor.h

@@ -27,11 +27,17 @@ using PlannerContextPtr = std::shared_ptr<PlannerContext>;
   * During actions build, there is special handling for following functions:
   * 1. Aggregate functions are added in actions dag as INPUT nodes. Aggregate functions arguments are not added.
   * 2. For function `in` and its variants, already collected sets from planner context are used.
+  * 3. When building actions that use CONSTANT nodes, by default we ignore pre-existing INPUTs if those don't have
+  * a column (a const column always has a column). This is for compatibility with previous headers. We disable this
+  * behaviour when we explicitly want to override CONSTANT nodes with the input (resolving InterpolateNode for example)
   */
 class PlannerActionsVisitor
 {
 public:
-    explicit PlannerActionsVisitor(const PlannerContextPtr & planner_context_, bool use_column_identifier_as_action_node_name_ = true);
+    explicit PlannerActionsVisitor(
+        const PlannerContextPtr & planner_context_,
+        bool use_column_identifier_as_action_node_name_ = true,
+        bool always_use_const_column_for_constant_nodes_ = true);
 
     /** Add actions necessary to calculate expression node into expression dag.
       * Necessary actions are not added in actions dag output.
@@ -42,6 +48,7 @@ public:
 private:
     const PlannerContextPtr planner_context;
     bool use_column_identifier_as_action_node_name = true;
+    bool always_use_const_column_for_constant_nodes = true;
 };
 
 /** Calculate query tree expression node action dag name and add them into node to name map.

src/Planner/PlannerJoinTree.cpp

@@ -77,7 +77,6 @@ namespace ErrorCodes
     extern const int INVALID_JOIN_ON_EXPRESSION;
     extern const int LOGICAL_ERROR;
     extern const int NOT_IMPLEMENTED;
-    extern const int SYNTAX_ERROR;
     extern const int ACCESS_DENIED;
     extern const int PARAMETER_OUT_OF_BOUND;
     extern const int TOO_MANY_COLUMNS;
@@ -1417,12 +1416,7 @@ JoinTreeQueryPlan buildQueryPlanForJoinNode(const QueryTreeNodePtr & join_table_
             {
                 if (!join_clause.hasASOF())
                     throw Exception(ErrorCodes::INVALID_JOIN_ON_EXPRESSION,
-                        "JOIN {} no inequality in ASOF JOIN ON section.",
+                        "JOIN {} no inequality in ASOF JOIN ON section",
-                        join_node.formatASTForErrorMessage());
-
-                if (table_join_clause.key_names_left.size() <= 1)
-                    throw Exception(ErrorCodes::SYNTAX_ERROR,
-                        "JOIN {} ASOF join needs at least one equi-join column",
                         join_node.formatASTForErrorMessage());
             }
 
@@ -1544,7 +1538,9 @@ JoinTreeQueryPlan buildQueryPlanForJoinNode(const QueryTreeNodePtr & join_table_
         {
             const auto & join_clause = table_join->getOnlyClause();
 
-            bool kind_allows_filtering = isInner(join_kind) || isLeft(join_kind) || isRight(join_kind);
+            bool join_type_allows_filtering = (join_strictness == JoinStrictness::All || join_strictness == JoinStrictness::Any)
+                                            && (isInner(join_kind) || isLeft(join_kind) || isRight(join_kind));
+
 
             auto has_non_const = [](const Block & block, const auto & keys)
             {
@@ -1564,7 +1560,7 @@ JoinTreeQueryPlan buildQueryPlanForJoinNode(const QueryTreeNodePtr & join_table_
             bool has_non_const_keys = has_non_const(left_plan.getCurrentDataStream().header, join_clause.key_names_left)
                 && has_non_const(right_plan.getCurrentDataStream().header, join_clause.key_names_right);
 
-            if (settings.max_rows_in_set_to_optimize_join > 0 && kind_allows_filtering && has_non_const_keys)
+            if (settings.max_rows_in_set_to_optimize_join > 0 && join_type_allows_filtering && has_non_const_keys)
             {
                 auto * left_set = add_create_set(left_plan, join_clause.key_names_left, JoinTableSide::Left);
                 auto * right_set = add_create_set(right_plan, join_clause.key_names_right, JoinTableSide::Right);

src/Processors/Transforms/MergeJoinTransform.cpp

@@ -34,13 +34,20 @@ namespace ErrorCodes
 namespace
 {
 
-FullMergeJoinCursorPtr createCursor(const Block & block, const Names & columns)
+FullMergeJoinCursorPtr createCursor(const Block & block, const Names & columns, JoinStrictness strictness)
 {
     SortDescription desc;
     desc.reserve(columns.size());
     for (const auto & name : columns)
         desc.emplace_back(name);
-    return std::make_unique<FullMergeJoinCursor>(block, desc);
+    return std::make_unique<FullMergeJoinCursor>(block, desc, strictness == JoinStrictness::Asof);
+}
+
+bool ALWAYS_INLINE isNullAt(const IColumn & column, size_t row)
+{
+    if (const auto * nullable_column = checkAndGetColumn<ColumnNullable>(&column))
+        return nullable_column->isNullAt(row);
+    return false;
 }
 
 template <bool has_left_nulls, bool has_right_nulls>
@@ -54,7 +61,7 @@ int nullableCompareAt(const IColumn & left_column, const IColumn & right_column,
         if (left_nullable && right_nullable)
         {
             int res = left_nullable->compareAt(lhs_pos, rhs_pos, right_column, null_direction_hint);
-            if (res)
+            if (res != 0)
                 return res;
 
             /// NULL != NULL case
@@ -90,9 +97,10 @@ int nullableCompareAt(const IColumn & left_column, const IColumn & right_column,
 
 int ALWAYS_INLINE compareCursors(const SortCursorImpl & lhs, size_t lpos,
                                  const SortCursorImpl & rhs, size_t rpos,
+                                 size_t key_length,
                                  int null_direction_hint)
 {
-    for (size_t i = 0; i < lhs.sort_columns_size; ++i)
+    for (size_t i = 0; i < key_length; ++i)
     {
         /// TODO(@vdimir): use nullableCompareAt only if there's nullable columns
         int cmp = nullableCompareAt<true, true>(*lhs.sort_columns[i], *rhs.sort_columns[i], lpos, rpos, null_direction_hint);
@@ -104,13 +112,18 @@ int ALWAYS_INLINE compareCursors(const SortCursorImpl & lhs, size_t lpos,
 
 int ALWAYS_INLINE compareCursors(const SortCursorImpl & lhs, const SortCursorImpl & rhs, int null_direction_hint)
 {
-    return compareCursors(lhs, lhs.getRow(), rhs, rhs.getRow(), null_direction_hint);
+    return compareCursors(lhs, lhs.getRow(), rhs, rhs.getRow(), lhs.sort_columns_size, null_direction_hint);
+}
+
+int compareAsofCursors(const FullMergeJoinCursor & lhs, const FullMergeJoinCursor & rhs, int null_direction_hint)
+{
+    return nullableCompareAt<true, true>(*lhs.getAsofColumn(), *rhs.getAsofColumn(), lhs->getRow(), rhs->getRow(), null_direction_hint);
 }
 
 bool ALWAYS_INLINE totallyLess(SortCursorImpl & lhs, SortCursorImpl & rhs, int null_direction_hint)
 {
     /// The last row of left cursor is less than the current row of the right cursor.
-    int cmp = compareCursors(lhs, lhs.rows - 1, rhs, rhs.getRow(), null_direction_hint);
+    int cmp = compareCursors(lhs, lhs.rows - 1, rhs, rhs.getRow(), lhs.sort_columns_size, null_direction_hint);
     return cmp < 0;
 }
 
@@ -222,25 +235,136 @@ Chunk getRowFromChunk(const Chunk & chunk, size_t pos)
     return result;
 }
 
-void inline addRange(PaddedPODArray<UInt64> & left_map, size_t start, size_t end)
+void inline addRange(PaddedPODArray<UInt64> & values, UInt64 start, UInt64 end)
 {
     assert(end > start);
-    for (size_t i = start; i < end; ++i)
+    for (UInt64 i = start; i < end; ++i)
-        left_map.push_back(i);
+        values.push_back(i);
 }
 
-void inline addMany(PaddedPODArray<UInt64> & left_or_right_map, size_t idx, size_t num)
+void inline addMany(PaddedPODArray<UInt64> & values, UInt64 value, size_t num)
 {
-    for (size_t i = 0; i < num; ++i)
+    values.resize_fill(values.size() + num, value);
-        left_or_right_map.push_back(idx);
 }
 }
 
-FullMergeJoinCursor::FullMergeJoinCursor(const Block & sample_block_, const SortDescription & description_)
+JoinKeyRow::JoinKeyRow(const FullMergeJoinCursor & cursor, size_t pos)
-    : sample_block(materializeBlock(sample_block_).cloneEmpty()), desc(description_)
 {
+    row.reserve(cursor->sort_columns.size());
+    for (const auto & col : cursor->sort_columns)
+    {
+        auto new_col = col->cloneEmpty();
+        new_col->insertFrom(*col, pos);
+        row.push_back(std::move(new_col));
+    }
+    if (const IColumn * asof_column = cursor.getAsofColumn())
+    {
+        if (const auto * nullable_asof_column = checkAndGetColumn<ColumnNullable>(asof_column))
+        {
+            /// We save matched column, and since NULL do not match anything, we can't use it as a key
+            chassert(!nullable_asof_column->isNullAt(pos));
+            asof_column = nullable_asof_column->getNestedColumnPtr().get();
+        }
+        auto new_col = asof_column->cloneEmpty();
+        new_col->insertFrom(*asof_column, pos);
+        row.push_back(std::move(new_col));
+    }
 }
 
+void JoinKeyRow::reset()
+{
+    row.clear();
+}
+
+bool JoinKeyRow::equals(const FullMergeJoinCursor & cursor) const
+{
+    if (row.empty())
+        return false;
+
+    for (size_t i = 0; i < cursor->sort_columns_size; ++i)
+    {
+        // int cmp = this->row[i]->compareAt(0, cursor->getRow(), *(cursor->sort_columns[i]), cursor->desc[i].nulls_direction);
+        int cmp = nullableCompareAt<true, true>(*this->row[i], *cursor->sort_columns[i], 0, cursor->getRow(), cursor->desc[i].nulls_direction);
+        if (cmp != 0)
+            return false;
+    }
+    return true;
+}
+
+bool JoinKeyRow::asofMatch(const FullMergeJoinCursor & cursor, ASOFJoinInequality asof_inequality) const
+{
+    chassert(this->row.size() == cursor->sort_columns_size + 1);
+    if (!equals(cursor))
+        return false;
+
+    const auto & asof_row = row.back();
+    if (isNullAt(*asof_row, 0) || isNullAt(*cursor.getAsofColumn(), cursor->getRow()))
+        return false;
+
+    int cmp = 0;
+    if (const auto * nullable_column = checkAndGetColumn<ColumnNullable>(cursor.getAsofColumn()))
+        cmp = nullable_column->getNestedColumn().compareAt(cursor->getRow(), 0, *asof_row, 1);
+    else
+        cmp = cursor.getAsofColumn()->compareAt(cursor->getRow(), 0, *asof_row, 1);
+
+    return (asof_inequality == ASOFJoinInequality::Less && cmp < 0)
+        || (asof_inequality == ASOFJoinInequality::LessOrEquals && cmp <= 0)
+        || (asof_inequality == ASOFJoinInequality::Greater && cmp > 0)
+        || (asof_inequality == ASOFJoinInequality::GreaterOrEquals && cmp >= 0);
+}
+
+void AnyJoinState::set(size_t source_num, const FullMergeJoinCursor & cursor)
+{
+    assert(cursor->rows);
+    keys[source_num] = JoinKeyRow(cursor, cursor->rows - 1);
+}
+
+void AnyJoinState::reset(size_t source_num)
+{
+    keys[source_num].reset();
+    value.clear();
+}
+
+void AnyJoinState::setValue(Chunk value_)
+{
+    value = std::move(value_);
+}
+
+bool AnyJoinState::empty() const { return keys[0].row.empty() && keys[1].row.empty(); }
+
+
+void AsofJoinState::set(const FullMergeJoinCursor & rcursor, size_t rpos)
+{
+    key = JoinKeyRow(rcursor, rpos);
+    value = rcursor.getCurrent().clone();
+    value_row = rpos;
+}
+
+void AsofJoinState::reset()
+{
+    key.reset();
+    value.clear();
+}
+
+FullMergeJoinCursor::FullMergeJoinCursor(const Block & sample_block_, const SortDescription & description_, bool is_asof)
+    : sample_block(materializeBlock(sample_block_).cloneEmpty())
+    , desc(description_)
+{
+    if (desc.empty())
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Got empty sort description for FullMergeJoinCursor");
+
+    if (is_asof)
+    {
+        /// For ASOF join prefix of sort description is used for equality comparison
+        /// and the last column is used for inequality comparison and is handled separately
+
+        auto asof_column_description = desc.back();
+        desc.pop_back();
+
+        chassert(asof_column_description.direction == 1 && asof_column_description.nulls_direction == 1);
+        asof_column_position = sample_block.getPositionByName(asof_column_description.column_name);
+    }
+}
 
 const Chunk & FullMergeJoinCursor::getCurrent() const
 {
@@ -278,48 +402,103 @@ bool FullMergeJoinCursor::fullyCompleted() const
     return !cursor.isValid() && recieved_all_blocks;
 }
 
+String FullMergeJoinCursor::dump() const
+{
+    Strings row_dump;
+    if (cursor.isValid())
+    {
+        Field val;
+        for (size_t i = 0; i < cursor.sort_columns_size; ++i)
+        {
+            cursor.sort_columns[i]->get(cursor.getRow(), val);
+            row_dump.push_back(val.dump());
+        }
+
+        if (const auto * asof_column = getAsofColumn())
+        {
+            asof_column->get(cursor.getRow(), val);
+            row_dump.push_back(val.dump());
+        }
+    }
+
+    return fmt::format("<{}/{}{}>[{}]",
+        cursor.getRow(), cursor.rows,
+        recieved_all_blocks ? "(finished)" : "",
+        fmt::join(row_dump, ", "));
+}
+
 MergeJoinAlgorithm::MergeJoinAlgorithm(
-    JoinPtr table_join_,
+    JoinKind kind_,
+    JoinStrictness strictness_,
+    const TableJoin::JoinOnClause & on_clause_,
     const Blocks & input_headers,
     size_t max_block_size_)
-    : table_join(table_join_)
+    : kind(kind_)
+    , strictness(strictness_)
     , max_block_size(max_block_size_)
     , log(getLogger("MergeJoinAlgorithm"))
 {
     if (input_headers.size() != 2)
         throw Exception(ErrorCodes::LOGICAL_ERROR, "MergeJoinAlgorithm requires exactly two inputs");
 
-    auto strictness = table_join->getTableJoin().strictness();
+    if (strictness != JoinStrictness::Any && strictness != JoinStrictness::All && strictness != JoinStrictness::Asof)
-    if (strictness != JoinStrictness::Any && strictness != JoinStrictness::All)
         throw Exception(ErrorCodes::NOT_IMPLEMENTED, "MergeJoinAlgorithm is not implemented for strictness {}", strictness);
 
-    auto kind = table_join->getTableJoin().kind();
+    if (strictness == JoinStrictness::Asof)
+    {
+        if (kind != JoinKind::Left && kind != JoinKind::Inner)
+            throw Exception(ErrorCodes::NOT_IMPLEMENTED, "MergeJoinAlgorithm does not implement ASOF {} join", kind);
+    }
+
     if (!isInner(kind) && !isLeft(kind) && !isRight(kind) && !isFull(kind))
         throw Exception(ErrorCodes::NOT_IMPLEMENTED, "MergeJoinAlgorithm is not implemented for kind {}", kind);
 
-    const auto & join_on = table_join->getTableJoin().getOnlyClause();
+    if (on_clause_.on_filter_condition_left || on_clause_.on_filter_condition_right)
-
-    if (join_on.on_filter_condition_left || join_on.on_filter_condition_right)
         throw Exception(ErrorCodes::NOT_IMPLEMENTED, "MergeJoinAlgorithm does not support ON filter conditions");
 
     cursors = {
-        createCursor(input_headers[0], join_on.key_names_left),
+        createCursor(input_headers[0], on_clause_.key_names_left, strictness),
-        createCursor(input_headers[1], join_on.key_names_right)
+        createCursor(input_headers[1], on_clause_.key_names_right, strictness),
     };
+}
 
-    for (const auto & [left_key, right_key] : table_join->getTableJoin().leftToRightKeyRemap())
+MergeJoinAlgorithm::MergeJoinAlgorithm(
+    JoinPtr join_ptr,
+    const Blocks & input_headers,
+    size_t max_block_size_)
+    : MergeJoinAlgorithm(
+        join_ptr->getTableJoin().kind(),
+        join_ptr->getTableJoin().strictness(),
+        join_ptr->getTableJoin().getOnlyClause(),
+        input_headers,
+        max_block_size_)
+{
+    for (const auto & [left_key, right_key] : join_ptr->getTableJoin().leftToRightKeyRemap())
     {
         size_t left_idx = input_headers[0].getPositionByName(left_key);
         size_t right_idx = input_headers[1].getPositionByName(right_key);
         left_to_right_key_remap[left_idx] = right_idx;
     }
 
-    const auto *smjPtr = typeid_cast<const FullSortingMergeJoin *>(table_join.get());
+    const auto *smjPtr = typeid_cast<const FullSortingMergeJoin *>(join_ptr.get());
     if (smjPtr)
     {
         null_direction_hint = smjPtr->getNullDirection();
     }
 
+    if (strictness == JoinStrictness::Asof)
+        setAsofInequality(join_ptr->getTableJoin().getAsofInequality());
+}
+
+void MergeJoinAlgorithm::setAsofInequality(ASOFJoinInequality asof_inequality_)
+{
+    if (strictness != JoinStrictness::Asof)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "setAsofInequality is only supported for ASOF joins");
+
+    if (asof_inequality_ == ASOFJoinInequality::None)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "ASOF inequality cannot be None");
+
+    asof_inequality = asof_inequality_;
 }
 
 void MergeJoinAlgorithm::logElapsed(double seconds)
@@ -407,7 +586,7 @@ struct AllJoinImpl
                 size_t lnum = nextDistinct(left_cursor.cursor);
                 size_t rnum = nextDistinct(right_cursor.cursor);
 
-                bool all_fit_in_block = std::max(left_map.size(), right_map.size()) + lnum * rnum <= max_block_size;
+                bool all_fit_in_block = !max_block_size || std::max(left_map.size(), right_map.size()) + lnum * rnum <= max_block_size;
                 bool have_all_ranges = left_cursor.cursor.isValid() && right_cursor.cursor.isValid();
                 if (all_fit_in_block && have_all_ranges)
                 {
@@ -421,7 +600,7 @@ struct AllJoinImpl
                 else
                 {
                     assert(state == nullptr);
-                    state = std::make_unique<AllJoinState>(left_cursor.cursor, lpos, right_cursor.cursor, rpos);
+                    state = std::make_unique<AllJoinState>(left_cursor, lpos, right_cursor, rpos);
                     state->addRange(0, left_cursor.getCurrent().clone(), lpos, lnum);
                     state->addRange(1, right_cursor.getCurrent().clone(), rpos, rnum);
                     return;
@@ -466,6 +645,17 @@ void dispatchKind(JoinKind kind, Args && ... args)
         throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Unsupported join kind: \"{}\"", kind);
 }
 
+MutableColumns MergeJoinAlgorithm::getEmptyResultColumns() const
+{
+    MutableColumns result_cols;
+    for (size_t i = 0; i < 2; ++i)
+    {
+        for (const auto & col : cursors[i]->sampleColumns())
+            result_cols.push_back(col->cloneEmpty());
+    }
+    return result_cols;
+}
+
 std::optional<MergeJoinAlgorithm::Status> MergeJoinAlgorithm::handleAllJoinState()
 {
     if (all_join_state && all_join_state->finished())
@@ -479,7 +669,7 @@ std::optional<MergeJoinAlgorithm::Status> MergeJoinAlgorithm::handleAllJoinState
         /// Accumulate blocks with same key in all_join_state
         for (size_t i = 0; i < 2; ++i)
         {
-            if (cursors[i]->cursor.isValid() && all_join_state->keys[i].equals(cursors[i]->cursor))
+            if (cursors[i]->cursor.isValid() && all_join_state->keys[i].equals(*cursors[i]))
             {
                 size_t pos = cursors[i]->cursor.getRow();
                 size_t num = nextDistinct(cursors[i]->cursor);
@@ -499,15 +689,10 @@ std::optional<MergeJoinAlgorithm::Status> MergeJoinAlgorithm::handleAllJoinState
         stat.max_blocks_loaded = std::max(stat.max_blocks_loaded, all_join_state->blocksStored());
 
         /// join all rows with current key
-        MutableColumns result_cols;
+        MutableColumns result_cols = getEmptyResultColumns();
-        for (size_t i = 0; i < 2; ++i)
-        {
-            for (const auto & col : cursors[i]->sampleColumns())
-                result_cols.push_back(col->cloneEmpty());
-        }
 
         size_t total_rows = 0;
-        while (total_rows < max_block_size)
+        while (!max_block_size || total_rows < max_block_size)
         {
             const auto & left_range = all_join_state->getLeft();
             const auto & right_range = all_join_state->getRight();
@@ -532,7 +717,52 @@ std::optional<MergeJoinAlgorithm::Status> MergeJoinAlgorithm::handleAllJoinState
     return {};
 }
 
-MergeJoinAlgorithm::Status MergeJoinAlgorithm::allJoin(JoinKind kind)
+std::optional<MergeJoinAlgorithm::Status> MergeJoinAlgorithm::handleAsofJoinState()
+{
+    if (strictness != JoinStrictness::Asof)
+        return {};
+
+    if (!cursors[1]->fullyCompleted())
+        return {};
+
+    auto & left_cursor = *cursors[0];
+    const auto & left_columns = left_cursor.getCurrent().getColumns();
+
+    MutableColumns result_cols = getEmptyResultColumns();
+
+    while (left_cursor->isValid() && asof_join_state.hasMatch(left_cursor, asof_inequality))
+    {
+        size_t i = 0;
+        for (const auto & col : left_columns)
+            result_cols[i++]->insertFrom(*col, left_cursor->getRow());
+        for (const auto & col : asof_join_state.value.getColumns())
+            result_cols[i++]->insertFrom(*col, asof_join_state.value_row);
+        chassert(i == result_cols.size());
+        left_cursor->next();
+    }
+
+    while (isLeft(kind) && left_cursor->isValid())
+    {
+        /// return row with default values at right side
+        size_t i = 0;
+        for (const auto & col : left_columns)
+            result_cols[i++]->insertFrom(*col, left_cursor->getRow());
+        for (; i < result_cols.size(); ++i)
+            result_cols[i]->insertDefault();
+        chassert(i == result_cols.size());
+
+        left_cursor->next();
+    }
+
+    size_t result_rows = result_cols.empty() ? 0 : result_cols.front()->size();
+    if (result_rows)
+        return Status(Chunk(std::move(result_cols), result_rows));
+
+    return {};
+}
+
+
+MergeJoinAlgorithm::Status MergeJoinAlgorithm::allJoin()
 {
     PaddedPODArray<UInt64> idx_map[2];
 
@@ -595,7 +825,7 @@ struct AnyJoinImpl
                      FullMergeJoinCursor & right_cursor,
                      PaddedPODArray<UInt64> & left_map,
                      PaddedPODArray<UInt64> & right_map,
-                     AnyJoinState & state,
+                     AnyJoinState & any_join_state,
                      int null_direction_hint)
     {
         assert(enabled);
@@ -656,21 +886,21 @@ struct AnyJoinImpl
             }
         }
 
-        /// Remember index of last joined row to propagate it to next block
+        /// Remember last joined row to propagate it to next block
 
-        state.setValue({});
+        any_join_state.setValue({});
         if (!left_cursor->isValid())
         {
-            state.set(0, left_cursor.cursor);
+            any_join_state.set(0, left_cursor);
             if (cmp == 0 && isLeft(kind))
-                state.setValue(getRowFromChunk(right_cursor.getCurrent(), rpos));
+                any_join_state.setValue(getRowFromChunk(right_cursor.getCurrent(), rpos));
         }
 
         if (!right_cursor->isValid())
         {
-            state.set(1, right_cursor.cursor);
+            any_join_state.set(1, right_cursor);
             if (cmp == 0 && isRight(kind))
-                state.setValue(getRowFromChunk(left_cursor.getCurrent(), lpos));
+                any_join_state.setValue(getRowFromChunk(left_cursor.getCurrent(), lpos));
         }
     }
 };
@@ -680,40 +910,34 @@ std::optional<MergeJoinAlgorithm::Status> MergeJoinAlgorithm::handleAnyJoinState
     if (any_join_state.empty())
         return {};
 
-    auto kind = table_join->getTableJoin().kind();
-
     Chunk result;
 
     for (size_t source_num = 0; source_num < 2; ++source_num)
     {
         auto & current = *cursors[source_num];
-        auto & state = any_join_state;
+        if (any_join_state.keys[source_num].equals(current))
-        if (any_join_state.keys[source_num].equals(current.cursor))
         {
             size_t start_pos = current->getRow();
             size_t length = nextDistinct(current.cursor);
 
             if (length && isLeft(kind) && source_num == 0)
             {
-                if (state.value)
+                if (any_join_state.value)
-                    result = copyChunkResized(current.getCurrent(), state.value, start_pos, length);
+                    result = copyChunkResized(current.getCurrent(), any_join_state.value, start_pos, length);
                 else
                     result = createBlockWithDefaults(source_num, start_pos, length);
             }
 
             if (length && isRight(kind) && source_num == 1)
             {
-                if (state.value)
+                if (any_join_state.value)
-                    result = copyChunkResized(state.value, current.getCurrent(), start_pos, length);
+                    result = copyChunkResized(any_join_state.value, current.getCurrent(), start_pos, length);
                 else
                     result = createBlockWithDefaults(source_num, start_pos, length);
             }
 
-            /// We've found row with other key, no need to skip more rows with current key
             if (current->isValid())
-            {
+                any_join_state.keys[source_num].reset();
-                state.keys[source_num].reset();
-            }
         }
         else
         {
@@ -726,7 +950,7 @@ std::optional<MergeJoinAlgorithm::Status> MergeJoinAlgorithm::handleAnyJoinState
     return {};
 }
 
-MergeJoinAlgorithm::Status MergeJoinAlgorithm::anyJoin(JoinKind kind)
+MergeJoinAlgorithm::Status MergeJoinAlgorithm::anyJoin()
 {
     if (auto result = handleAnyJoinState())
         return std::move(*result);
@@ -771,10 +995,151 @@ MergeJoinAlgorithm::Status MergeJoinAlgorithm::anyJoin(JoinKind kind)
     return Status(std::move(result));
 }
 
+
+MergeJoinAlgorithm::Status MergeJoinAlgorithm::asofJoin()
+{
+    auto & left_cursor = *cursors[0];
+    if (!left_cursor->isValid())
+        return Status(0);
+
+    auto & right_cursor = *cursors[1];
+    if (!right_cursor->isValid())
+        return Status(1);
+
+    const auto & left_columns = left_cursor.getCurrent().getColumns();
+    const auto & right_columns = right_cursor.getCurrent().getColumns();
+
+    MutableColumns result_cols = getEmptyResultColumns();
+
+    while (left_cursor->isValid() && right_cursor->isValid())
+    {
+        auto lpos = left_cursor->getRow();
+        auto rpos = right_cursor->getRow();
+        auto cmp = compareCursors(*left_cursor, *right_cursor, null_direction_hint);
+        if (cmp == 0)
+        {
+            if (isNullAt(*left_cursor.getAsofColumn(), lpos))
+                cmp = -1;
+            if (isNullAt(*right_cursor.getAsofColumn(), rpos))
+                cmp = 1;
+        }
+
+        if (cmp == 0)
+        {
+            auto asof_cmp = compareAsofCursors(left_cursor, right_cursor, null_direction_hint);
+
+            if ((asof_inequality == ASOFJoinInequality::Less && asof_cmp <= -1)
+             || (asof_inequality == ASOFJoinInequality::LessOrEquals && asof_cmp <= 0))
+            {
+                /// First row in right table that is greater (or equal) than current row in left table
+                /// matches asof join condition the best
+                size_t i = 0;
+                for (const auto & col : left_columns)
+                    result_cols[i++]->insertFrom(*col, lpos);
+                for (const auto & col : right_columns)
+                    result_cols[i++]->insertFrom(*col, rpos);
+                chassert(i == result_cols.size());
+
+                left_cursor->next();
+                continue;
+            }
+
+            if (asof_inequality == ASOFJoinInequality::Less || asof_inequality == ASOFJoinInequality::LessOrEquals)
+            {
+                /// Asof condition is not (yet) satisfied, skip row in right table
+                right_cursor->next();
+                continue;
+            }
+
+            if ((asof_inequality == ASOFJoinInequality::Greater && asof_cmp >= 1)
+             || (asof_inequality == ASOFJoinInequality::GreaterOrEquals && asof_cmp >= 0))
+            {
+                /// condition is satisfied, remember this row and move next to try to find better match
+                asof_join_state.set(right_cursor, rpos);
+                right_cursor->next();
+                continue;
+            }
+
+            if (asof_inequality == ASOFJoinInequality::Greater || asof_inequality == ASOFJoinInequality::GreaterOrEquals)
+            {
+                /// Asof condition is not satisfied anymore, use last matched row from right table
+                if (asof_join_state.hasMatch(left_cursor, asof_inequality))
+                {
+                    size_t i = 0;
+                    for (const auto & col : left_columns)
+                        result_cols[i++]->insertFrom(*col, lpos);
+                    for (const auto & col : asof_join_state.value.getColumns())
+                        result_cols[i++]->insertFrom(*col, asof_join_state.value_row);
+                    chassert(i == result_cols.size());
+                }
+                else
+                {
+                    asof_join_state.reset();
+                    if (isLeft(kind))
+                    {
+                        /// return row with default values at right side
+                        size_t i = 0;
+                        for (const auto & col : left_columns)
+                            result_cols[i++]->insertFrom(*col, lpos);
+                        for (; i < result_cols.size(); ++i)
+                            result_cols[i]->insertDefault();
+                        chassert(i == result_cols.size());
+                    }
+                }
+                left_cursor->next();
+                continue;
+            }
+
+            throw Exception(ErrorCodes::NOT_IMPLEMENTED, "TODO: implement ASOF equality join");
+        }
+        else if (cmp < 0)
+        {
+            if (asof_join_state.hasMatch(left_cursor, asof_inequality))
+            {
+                size_t i = 0;
+                for (const auto & col : left_columns)
+                    result_cols[i++]->insertFrom(*col, lpos);
+                for (const auto & col : asof_join_state.value.getColumns())
+                    result_cols[i++]->insertFrom(*col, asof_join_state.value_row);
+                chassert(i == result_cols.size());
+                left_cursor->next();
+                continue;
+            }
+            else
+            {
+                asof_join_state.reset();
+            }
+
+            /// no matches for rows in left table, just pass them through
+            size_t num = nextDistinct(*left_cursor);
+
+            if (isLeft(kind) && num)
+            {
+                /// return them with default values at right side
+                size_t i = 0;
+                for (const auto & col : left_columns)
+                    result_cols[i++]->insertRangeFrom(*col, lpos, num);
+                for (; i < result_cols.size(); ++i)
+                    result_cols[i]->insertManyDefaults(num);
+                chassert(i == result_cols.size());
+            }
+        }
+        else
+        {
+            /// skip rows in right table until we find match for current row in left table
+            nextDistinct(*right_cursor);
+        }
+    }
+    size_t num_rows = result_cols.empty() ? 0 : result_cols.front()->size();
+    return Status(Chunk(std::move(result_cols), num_rows));
+}
+
+
 /// if `source_num == 0` get data from left cursor and fill defaults at right
 /// otherwise - vice versa
 Chunk MergeJoinAlgorithm::createBlockWithDefaults(size_t source_num, size_t start, size_t num_rows) const
 {
+
     ColumnRawPtrs cols;
     {
         const auto & columns_left = source_num == 0 ? cursors[0]->getCurrent().getColumns() : cursors[0]->sampleColumns();
@@ -797,7 +1162,6 @@ Chunk MergeJoinAlgorithm::createBlockWithDefaults(size_t source_num, size_t star
             cols.push_back(col.get());
         }
     }
-
     Chunk result_chunk;
     copyColumnsResized(cols, start, num_rows, result_chunk);
     return result_chunk;
@@ -813,7 +1177,6 @@ Chunk MergeJoinAlgorithm::createBlockWithDefaults(size_t source_num)
 
 IMergingAlgorithm::Status MergeJoinAlgorithm::merge()
 {
-    auto kind = table_join->getTableJoin().kind();
 
     if (!cursors[0]->cursor.isValid() && !cursors[0]->fullyCompleted())
         return Status(0);
@@ -821,11 +1184,11 @@ IMergingAlgorithm::Status MergeJoinAlgorithm::merge()
     if (!cursors[1]->cursor.isValid() && !cursors[1]->fullyCompleted())
         return Status(1);
 
-
     if (auto result = handleAllJoinState())
-    {
         return std::move(*result);
-    }
+
+    if (auto result = handleAsofJoinState())
+        return std::move(*result);
 
     if (cursors[0]->fullyCompleted() || cursors[1]->fullyCompleted())
     {
@@ -839,7 +1202,7 @@ IMergingAlgorithm::Status MergeJoinAlgorithm::merge()
     }
 
     /// check if blocks are not intersecting at all
-    if (int cmp = totallyCompare(cursors[0]->cursor, cursors[1]->cursor, null_direction_hint); cmp != 0)
+    if (int cmp = totallyCompare(cursors[0]->cursor, cursors[1]->cursor, null_direction_hint); cmp != 0 && strictness != JoinStrictness::Asof)
     {
         if (cmp < 0)
         {
@@ -858,13 +1221,14 @@ IMergingAlgorithm::Status MergeJoinAlgorithm::merge()
         }
     }
 
-    auto strictness = table_join->getTableJoin().strictness();
-
     if (strictness == JoinStrictness::Any)
-        return anyJoin(kind);
+        return anyJoin();
 
     if (strictness == JoinStrictness::All)
-        return allJoin(kind);
+        return allJoin();
+
+    if (strictness == JoinStrictness::Asof)
+        return asofJoin();
 
     throw Exception(ErrorCodes::NOT_IMPLEMENTED, "Unsupported strictness '{}'", strictness);
 }
@@ -883,9 +1247,26 @@ MergeJoinTransform::MergeJoinTransform(
         /* always_read_till_end_= */ false,
         /* empty_chunk_on_finish_= */ true,
         table_join, input_headers, max_block_size)
-    , log(getLogger("MergeJoinTransform"))
 {
-    LOG_TRACE(log, "Use MergeJoinTransform");
+}
+
+MergeJoinTransform::MergeJoinTransform(
+        JoinKind kind_,
+        JoinStrictness strictness_,
+        const TableJoin::JoinOnClause & on_clause_,
+        const Blocks & input_headers,
+        const Block & output_header,
+        size_t max_block_size,
+        UInt64 limit_hint_)
+    : IMergingTransform<MergeJoinAlgorithm>(
+        input_headers,
+        output_header,
+        /* have_all_inputs_= */ true,
+        limit_hint_,
+        /* always_read_till_end_= */ false,
+        /* empty_chunk_on_finish_= */ true,
+        kind_, strictness_, on_clause_, input_headers, max_block_size)
+{
 }
 
 void MergeJoinTransform::onFinish()

src/Processors/Transforms/MergeJoinTransform.h

@@ -8,6 +8,7 @@
 #include <unordered_map>
 #include <utility>
 #include <vector>
+#include <array>
 
 #include <boost/core/noncopyable.hpp>
 
@@ -19,6 +20,7 @@
 #include <Processors/Chunk.h>
 #include <Processors/Merges/Algorithms/IMergingAlgorithm.h>
 #include <Processors/Merges/IMergingTransform.h>
+#include <Interpreters/TableJoin.h>
 
 namespace Poco { class Logger; }
 
@@ -35,57 +37,28 @@ using FullMergeJoinCursorPtr = std::unique_ptr<FullMergeJoinCursor>;
 /// Used instead of storing previous block
 struct JoinKeyRow
 {
-    std::vector<ColumnPtr> row;
-
     JoinKeyRow() = default;
 
-    explicit JoinKeyRow(const SortCursorImpl & impl_, size_t pos)
+    JoinKeyRow(const FullMergeJoinCursor & cursor, size_t pos);
-    {
-        row.reserve(impl_.sort_columns.size());
-        for (const auto & col : impl_.sort_columns)
-        {
-            auto new_col = col->cloneEmpty();
-            new_col->insertFrom(*col, pos);
-            row.push_back(std::move(new_col));
-        }
-    }
 
-    void reset()
+    bool equals(const FullMergeJoinCursor & cursor) const;
-    {
+    bool asofMatch(const FullMergeJoinCursor & cursor, ASOFJoinInequality asof_inequality) const;
-        row.clear();
-    }
 
-    bool equals(const SortCursorImpl & impl) const
+    void reset();
-    {
-        if (row.empty())
-            return false;
 
-        assert(this->row.size() == impl.sort_columns_size);
+    std::vector<ColumnPtr> row;
-        for (size_t i = 0; i < impl.sort_columns_size; ++i)
-        {
-            int cmp = this->row[i]->compareAt(0, impl.getRow(), *impl.sort_columns[i], impl.desc[i].nulls_direction);
-            if (cmp != 0)
-                return false;
-        }
-        return true;
-    }
 };
 
 /// Remembers previous key if it was joined in previous block
 class AnyJoinState : boost::noncopyable
 {
 public:
-    AnyJoinState() = default;
+    void set(size_t source_num, const FullMergeJoinCursor & cursor);
+    void setValue(Chunk value_);
 
-    void set(size_t source_num, const SortCursorImpl & cursor)
+    void reset(size_t source_num);
-    {
-        assert(cursor.rows);
-        keys[source_num] = JoinKeyRow(cursor, cursor.rows - 1);
-    }
 
-    void setValue(Chunk value_) { value = std::move(value_); }
+    bool empty() const;
-
-    bool empty() const { return keys[0].row.empty() && keys[1].row.empty(); }
 
     /// current keys
     JoinKeyRow keys[2];
@@ -118,8 +91,8 @@ public:
         Chunk chunk;
     };
 
-    AllJoinState(const SortCursorImpl & lcursor, size_t lpos,
+    AllJoinState(const FullMergeJoinCursor & lcursor, size_t lpos,
-                 const SortCursorImpl & rcursor, size_t rpos)
+                 const FullMergeJoinCursor & rcursor, size_t rpos)
         : keys{JoinKeyRow(lcursor, lpos), JoinKeyRow(rcursor, rpos)}
     {
     }
@@ -187,13 +160,32 @@ private:
     size_t ridx = 0;
 };
 
+
+class AsofJoinState : boost::noncopyable
+{
+public:
+    void set(const FullMergeJoinCursor & rcursor, size_t rpos);
+    void reset();
+
+    bool hasMatch(const FullMergeJoinCursor & cursor, ASOFJoinInequality asof_inequality) const
+    {
+        if (value.empty())
+            return false;
+        return key.asofMatch(cursor, asof_inequality);
+    }
+
+    JoinKeyRow key;
+    Chunk value;
+    size_t value_row = 0;
+};
+
 /*
  * Wrapper for SortCursorImpl
  */
 class FullMergeJoinCursor : boost::noncopyable
 {
 public:
-    explicit FullMergeJoinCursor(const Block & sample_block_, const SortDescription & description_);
+    explicit FullMergeJoinCursor(const Block & sample_block_, const SortDescription & description_, bool is_asof = false);
 
     bool fullyCompleted() const;
     void setChunk(Chunk && chunk);
@@ -203,17 +195,31 @@ public:
     SortCursorImpl * operator-> () { return &cursor; }
     const SortCursorImpl * operator-> () const { return &cursor; }
 
+    SortCursorImpl & operator* () { return cursor; }
+    const SortCursorImpl & operator* () const { return cursor; }
+
     SortCursorImpl cursor;
 
     const Block & sampleBlock() const { return sample_block; }
     Columns sampleColumns() const { return sample_block.getColumns(); }
 
+    const IColumn * getAsofColumn() const
+    {
+        if (!asof_column_position)
+            return nullptr;
+        return cursor.all_columns[*asof_column_position];
+    }
+
+    String dump() const;
+
 private:
     Block sample_block;
     SortDescription desc;
 
     Chunk current_chunk;
     bool recieved_all_blocks = false;
+
+    std::optional<size_t> asof_column_position;
 };
 
 /*
@@ -223,22 +229,33 @@ private:
 class MergeJoinAlgorithm final : public IMergingAlgorithm
 {
 public:
-    explicit MergeJoinAlgorithm(JoinPtr table_join, const Blocks & input_headers, size_t max_block_size_);
+    MergeJoinAlgorithm(JoinKind kind_,
+                       JoinStrictness strictness_,
+                       const TableJoin::JoinOnClause & on_clause_,
+                       const Blocks & input_headers,
+                       size_t max_block_size_);
+
+    MergeJoinAlgorithm(JoinPtr join_ptr, const Blocks & input_headers, size_t max_block_size_);
 
     const char * getName() const override { return "MergeJoinAlgorithm"; }
     void initialize(Inputs inputs) override;
     void consume(Input & input, size_t source_num) override;
     Status merge() override;
 
-    void logElapsed(double seconds);
+    void setAsofInequality(ASOFJoinInequality asof_inequality_);
 
+    void logElapsed(double seconds);
 private:
     std::optional<Status> handleAnyJoinState();
-    Status anyJoin(JoinKind kind);
+    Status anyJoin();
 
     std::optional<Status> handleAllJoinState();
-    Status allJoin(JoinKind kind);
+    Status allJoin();
 
+    std::optional<Status> handleAsofJoinState();
+    Status asofJoin();
+
+    MutableColumns getEmptyResultColumns() const;
     Chunk createBlockWithDefaults(size_t source_num);
     Chunk createBlockWithDefaults(size_t source_num, size_t start, size_t num_rows) const;
 
@@ -246,12 +263,15 @@ private:
     std::unordered_map<size_t, size_t> left_to_right_key_remap;
 
     std::array<FullMergeJoinCursorPtr, 2> cursors;
+    ASOFJoinInequality asof_inequality = ASOFJoinInequality::None;
 
-    /// Keep some state to make connection between data in different blocks
+    /// Keep some state to make handle data from different blocks
     AnyJoinState any_join_state;
     std::unique_ptr<AllJoinState> all_join_state;
+    AsofJoinState asof_join_state;
 
-    JoinPtr table_join;
+    JoinKind kind;
+    JoinStrictness strictness;
 
     size_t max_block_size;
     int null_direction_hint = 1;
@@ -281,12 +301,21 @@ public:
         size_t max_block_size,
         UInt64 limit_hint = 0);
 
+    MergeJoinTransform(
+        JoinKind kind_,
+        JoinStrictness strictness_,
+        const TableJoin::JoinOnClause & on_clause_,
+        const Blocks & input_headers,
+        const Block & output_header,
+        size_t max_block_size,
+        UInt64 limit_hint_ = 0);
+
     String getName() const override { return "MergeJoinTransform"; }
 
+    void setAsofInequality(ASOFJoinInequality asof_inequality_) { algorithm.setAsofInequality(asof_inequality_); }
+
 protected:
     void onFinish() override;
-
-    LoggerPtr log;
 };
 
 }

src/Processors/Transforms/ScatterByPartitionTransform.cpp

@@ -109,7 +109,7 @@ void ScatterByPartitionTransform::generateOutputChunks()
     hash.reset(num_rows);
 
     for (const auto & column_number : key_columns)
-        columns[column_number]->updateWeakHash32(hash);
+        hash.update(columns[column_number]->getWeakHash32());
 
     const auto & hash_data = hash.getData();
     IColumn::Selector selector(num_rows);

src/Processors/tests/gtest_full_sorting_join.cpp

@@ -0,0 +1,768 @@
+#include <gtest/gtest.h>
+
+#include <pcg_random.hpp>
+#include <random>
+#include <Poco/ConsoleChannel.h>
+#include <Poco/Logger.h>
+#include <Poco/AutoPtr.h>
+
+#include <Columns/ColumnsNumber.h>
+#include <Common/getRandomASCIIString.h>
+#include <Common/randomSeed.h>
+
+#include <DataTypes/DataTypesNumber.h>
+
+#include <Interpreters/TableJoin.h>
+
+#include <Processors/Executors/PipelineExecutor.h>
+#include <Processors/Executors/PullingPipelineExecutor.h>
+#include <Processors/Sinks/NullSink.h>
+#include <Processors/Sources/SourceFromChunks.h>
+#include <Processors/Sources/SourceFromSingleChunk.h>
+#include <Processors/Transforms/MergeJoinTransform.h>
+
+#include <Processors/Formats/Impl/PrettyCompactBlockOutputFormat.h>
+#include <Processors/Executors/CompletedPipelineExecutor.h>
+
+
+#include <QueryPipeline/QueryPipeline.h>
+
+using namespace DB;
+
+namespace
+{
+
+QueryPipeline buildJoinPipeline(
+    std::shared_ptr<ISource> left_source,
+    std::shared_ptr<ISource> right_source,
+    size_t key_length = 1,
+    JoinKind kind = JoinKind::Inner,
+    JoinStrictness strictness = JoinStrictness::All,
+    ASOFJoinInequality asof_inequality = ASOFJoinInequality::None)
+{
+    Blocks inputs;
+    inputs.emplace_back(left_source->getPort().getHeader());
+    inputs.emplace_back(right_source->getPort().getHeader());
+
+    Block out_header;
+    for (const auto & input : inputs)
+    {
+        for (ColumnWithTypeAndName column : input)
+        {
+            if (&input == &inputs.front())
+                column.name = "t1." + column.name;
+            else
+                column.name = "t2." + column.name;
+            out_header.insert(column);
+        }
+    }
+
+    TableJoin::JoinOnClause on_clause;
+    for (size_t i = 0; i < key_length; ++i)
+    {
+        on_clause.key_names_left.emplace_back(inputs[0].getByPosition(i).name);
+        on_clause.key_names_right.emplace_back(inputs[1].getByPosition(i).name);
+    }
+
+    auto joining = std::make_shared<MergeJoinTransform>(
+        kind,
+        strictness,
+        on_clause,
+        inputs, out_header, /* max_block_size = */ 0);
+
+    if (asof_inequality != ASOFJoinInequality::None)
+        joining->setAsofInequality(asof_inequality);
+
+    chassert(joining->getInputs().size() == 2);
+
+    connect(left_source->getPort(), joining->getInputs().front());
+    connect(right_source->getPort(), joining->getInputs().back());
+
+    auto * output_port = &joining->getOutputPort();
+
+    auto processors = std::make_shared<Processors>();
+    processors->emplace_back(std::move(left_source));
+    processors->emplace_back(std::move(right_source));
+    processors->emplace_back(std::move(joining));
+
+    QueryPipeline pipeline(QueryPlanResourceHolder{}, processors, output_port);
+    return pipeline;
+}
+
+
+std::shared_ptr<ISource> oneColumnSource(const std::vector<std::vector<UInt64>> & values)
+{
+    Block header = {
+        ColumnWithTypeAndName(std::make_shared<DataTypeUInt64>(), "key"),
+        ColumnWithTypeAndName(std::make_shared<DataTypeUInt64>(), "idx"),
+    };
+
+    UInt64 idx = 0;
+    Chunks chunks;
+    for (const auto & chunk_values : values)
+    {
+        auto key_column = ColumnUInt64::create();
+        auto idx_column = ColumnUInt64::create();
+
+        for (auto n : chunk_values)
+        {
+            key_column->insertValue(n);
+            idx_column->insertValue(idx);
+            ++idx;
+        }
+        chunks.emplace_back(Chunk(Columns{std::move(key_column), std::move(idx_column)}, chunk_values.size()));
+    }
+    return std::make_shared<SourceFromChunks>(header, std::move(chunks));
+}
+
+class SourceChunksBuilder
+{
+public:
+
+    explicit SourceChunksBuilder(const Block & header_)
+        : header(header_)
+    {
+        current_chunk = header.cloneEmptyColumns();
+        chassert(!current_chunk.empty());
+    }
+
+    void setBreakProbability(pcg64 & rng_)
+    {
+        /// random probability with possibility to have exact 0.0 and 1.0 values
+        break_prob = std::uniform_int_distribution<size_t>(0, 5)(rng_) / static_cast<double>(5);
+        rng = &rng_;
+    }
+
+    void addRow(const std::vector<Field> & row)
+    {
+        chassert(row.size() == current_chunk.size());
+        for (size_t i = 0; i < current_chunk.size(); ++i)
+            current_chunk[i]->insert(row[i]);
+
+        if (rng && std::uniform_real_distribution<>(0.0, 1.0)(*rng) < break_prob)
+            addChunk();
+    }
+
+    void addChunk()
+    {
+        if (current_chunk.front()->empty())
+            return;
+
+        size_t rows = current_chunk.front()->size();
+        chunks.emplace_back(std::move(current_chunk), rows);
+        current_chunk = header.cloneEmptyColumns();
+    }
+
+    std::shared_ptr<ISource> getSource()
+    {
+        addChunk();
+
+        /// copy chunk to allow reusing same builder
+        Chunks chunks_copy;
+        chunks_copy.reserve(chunks.size());
+        for (const auto & chunk : chunks)
+            chunks_copy.emplace_back(chunk.clone());
+        return std::make_shared<SourceFromChunks>(header, std::move(chunks_copy));
+    }
+
+private:
+    Block header;
+    Chunks chunks;
+    MutableColumns current_chunk;
+
+    pcg64 * rng = nullptr;
+    double break_prob = 0.0;
+};
+
+
+std::vector<std::vector<Field>> getValuesFromBlock(const Block & block, const Names & names)
+{
+    std::vector<std::vector<Field>> result;
+    for (size_t i = 0; i < block.rows(); ++i)
+    {
+        auto & row = result.emplace_back();
+        for (const auto & name : names)
+            block.getByName(name).column->get(i, row.emplace_back());
+    }
+    return result;
+}
+
+
+Block executePipeline(QueryPipeline && pipeline)
+{
+    PullingPipelineExecutor executor(pipeline);
+
+    Blocks result_blocks;
+    while (true)
+    {
+        Block block;
+        bool is_ok = executor.pull(block);
+        if (!is_ok)
+            break;
+        result_blocks.emplace_back(std::move(block));
+    }
+
+    return concatenateBlocks(result_blocks);
+}
+
+template <typename T>
+void assertColumnVectorEq(const typename ColumnVector<T>::Container & expected, const Block & block, const std::string & name)
+{
+    const auto * actual = typeid_cast<const ColumnVector<T> *>(block.getByName(name).column.get());
+    ASSERT_TRUE(actual) << "unexpected column type: " << block.getByName(name).column->dumpStructure() << "expected: " << typeid(ColumnVector<T>).name();
+
+    auto get_first_diff = [&]() -> String
+    {
+        const auto & actual_data = actual->getData();
+        size_t num_rows = std::min(expected.size(), actual_data.size());
+        for (size_t i = 0; i < num_rows; ++i)
+        {
+            if (expected[i] != actual_data[i])
+                return fmt::format(", expected: {}, actual: {} at row {}", expected[i], actual_data[i], i);
+        }
+        return "";
+    };
+
+    EXPECT_EQ(actual->getData().size(), expected.size());
+    ASSERT_EQ(actual->getData(), expected) << "column name: " << name << get_first_diff();
+}
+
+template <typename T>
+void assertColumnEq(const IColumn & expected, const Block & block, const std::string & name)
+{
+    const ColumnPtr & actual = block.getByName(name).column;
+    ASSERT_TRUE(checkColumn<T>(*actual));
+    ASSERT_TRUE(checkColumn<T>(expected));
+    EXPECT_EQ(actual->size(), expected.size());
+
+    auto dump_val = [](const IColumn & col, size_t i) -> String
+    {
+        Field value;
+        col.get(i, value);
+        return value.dump();
+    };
+
+    size_t num_rows = std::min(actual->size(), expected.size());
+    for (size_t i = 0; i < num_rows; ++i)
+        ASSERT_EQ(actual->compareAt(i, i, expected, 1), 0) << dump_val(*actual, i) << " != " << dump_val(expected, i) << " at row " << i;
+}
+
+template <typename T>
+T getRandomFrom(pcg64 & rng, const std::initializer_list<T> & opts)
+{
+    std::vector<T> options(opts.begin(), opts.end());
+    size_t idx = std::uniform_int_distribution<size_t>(0, options.size() - 1)(rng);
+    return options[idx];
+}
+
+void generateNextKey(pcg64 & rng, UInt64 & k1, String & k2)
+{
+    size_t str_len = std::uniform_int_distribution<>(1, 10)(rng);
+    String new_k2 = getRandomASCIIString(str_len, rng);
+    if (new_k2.compare(k2) <= 0)
+        ++k1;
+    k2 = new_k2;
+}
+
+bool isStrict(ASOFJoinInequality inequality)
+{
+    return inequality == ASOFJoinInequality::Less || inequality == ASOFJoinInequality::Greater;
+}
+
+}
+
+class FullSortingJoinTest : public ::testing::Test
+{
+public:
+    FullSortingJoinTest() = default;
+
+    void SetUp() override
+    {
+        Poco::AutoPtr<Poco::ConsoleChannel> channel(new Poco::ConsoleChannel(std::cerr));
+        Poco::Logger::root().setChannel(channel);
+        if (const char * test_log_level = std::getenv("TEST_LOG_LEVEL")) // NOLINT(concurrency-mt-unsafe)
+            Poco::Logger::root().setLevel(test_log_level);
+        else
+            Poco::Logger::root().setLevel("none");
+
+
+        UInt64 seed = randomSeed();
+        if (const char * random_seed = std::getenv("TEST_RANDOM_SEED")) // NOLINT(concurrency-mt-unsafe)
+            seed = std::stoull(random_seed);
+        std::cout << "TEST_RANDOM_SEED=" << seed << std::endl;
+        rng = pcg64(seed);
+    }
+
+    void TearDown() override
+    {
+    }
+
+    pcg64 rng;
+};
+
+TEST_F(FullSortingJoinTest, AllAnyOneKey)
+try
+{
+    {
+        SCOPED_TRACE("Inner All");
+        Block result = executePipeline(buildJoinPipeline(
+            oneColumnSource({ {1, 2, 3, 4, 5} }),
+            oneColumnSource({ {1}, {2}, {3}, {4}, {5} }),
+            1, JoinKind::Inner, JoinStrictness::All));
+
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({0, 1, 2, 3, 4}), result, "t1.idx");
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({0, 1, 2, 3, 4}), result, "t2.idx");
+    }
+    {
+        SCOPED_TRACE("Inner Any");
+        Block result = executePipeline(buildJoinPipeline(
+            oneColumnSource({ {1, 2, 3, 4, 5} }),
+            oneColumnSource({ {1}, {2}, {3}, {4}, {5} }),
+            1, JoinKind::Inner, JoinStrictness::Any));
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({0, 1, 2, 3, 4}), result, "t1.idx");
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({0, 1, 2, 3, 4}), result, "t2.idx");
+    }
+    {
+        SCOPED_TRACE("Inner All");
+        Block result = executePipeline(buildJoinPipeline(
+            oneColumnSource({ {2, 2, 2}, {2, 3}, {3, 5} }),
+            oneColumnSource({ {1, 1, 1}, {2, 2}, {3, 4} }),
+            1, JoinKind::Inner, JoinStrictness::All));
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({0, 1, 2, 0, 1, 2, 3, 3, 4, 5}), result, "t1.idx");
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({3, 3, 3, 4, 4, 4, 3, 4, 5, 5}), result, "t2.idx");
+    }
+    {
+        SCOPED_TRACE("Inner Any");
+        Block result = executePipeline(buildJoinPipeline(
+            oneColumnSource({ {2, 2, 2}, {2, 3}, {3, 5} }),
+            oneColumnSource({ {1, 1, 1}, {2, 2}, {3, 4} }),
+            1, JoinKind::Inner, JoinStrictness::Any));
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({0, 4}), result, "t1.idx");
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({3, 5}), result, "t2.idx");
+    }
+    {
+        SCOPED_TRACE("Inner Any");
+        Block result = executePipeline(buildJoinPipeline(
+            oneColumnSource({ {2, 2, 2, 2}, {3}, {3, 5} }),
+            oneColumnSource({ {1, 1, 1, 2}, {2}, {3, 4} }),
+            1, JoinKind::Inner, JoinStrictness::Any));
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({0, 4}), result, "t1.idx");
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({3, 5}), result, "t2.idx");
+    }
+    {
+
+        SCOPED_TRACE("Left Any");
+        Block result = executePipeline(buildJoinPipeline(
+            oneColumnSource({ {2, 2, 2}, {2, 3}, {3, 5} }),
+            oneColumnSource({ {1, 1, 1}, {2, 2}, {3, 4} }),
+            1, JoinKind::Left, JoinStrictness::Any));
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({0, 1, 2, 3, 4, 5, 6}), result, "t1.idx");
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({3, 3, 3, 3, 5, 5, 0}), result, "t2.idx");
+    }
+    {
+        SCOPED_TRACE("Left Any");
+        Block result = executePipeline(buildJoinPipeline(
+            oneColumnSource({ {2, 2, 2, 2}, {3}, {3, 5} }),
+            oneColumnSource({ {1, 1, 1, 2}, {2}, {3, 4} }),
+            1, JoinKind::Left, JoinStrictness::Any));
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({0, 1, 2, 3, 4, 5, 6}), result, "t1.idx");
+        assertColumnVectorEq<UInt64>(ColumnUInt64::Container({3, 3, 3, 3, 5, 5, 0}), result, "t2.idx");
+    }
+}
+catch (Exception & e)
+{
+    std::cout << e.getStackTraceString() << std::endl;
+    throw;
+}
+
+
+TEST_F(FullSortingJoinTest, AnySimple)
+try
+{
+    JoinKind kind = getRandomFrom(rng, {JoinKind::Inner, JoinKind::Left, JoinKind::Right});
+
+    SourceChunksBuilder left_source({
+        {std::make_shared<DataTypeUInt64>(), "k1"},
+        {std::make_shared<DataTypeString>(), "k2"},
+        {std::make_shared<DataTypeString>(), "attr"},
+    });
+
+    SourceChunksBuilder right_source({
+        {std::make_shared<DataTypeUInt64>(), "k1"},
+        {std::make_shared<DataTypeString>(), "k2"},
+        {std::make_shared<DataTypeString>(), "attr"},
+    });
+
+    left_source.setBreakProbability(rng);
+    right_source.setBreakProbability(rng);
+
+    size_t num_keys = std::uniform_int_distribution<>(100, 1000)(rng);
+
+    auto expected_left = ColumnString::create();
+    auto expected_right = ColumnString::create();
+
+    UInt64 k1 = 1;
+    String k2;
+
+    auto get_attr = [&](const String & side, size_t idx) -> String
+    {
+        return toString(k1) + "_" + k2 + "_" + side + "_" + toString(idx);
+    };
+
+    for (size_t i = 0; i < num_keys; ++i)
+    {
+        generateNextKey(rng, k1, k2);
+
+        /// Key is present in left, right or both tables. Both tables is more probable.
+        size_t key_presence = std::uniform_int_distribution<>(0, 10)(rng);
+
+        size_t num_rows_left = key_presence == 0 ? 0 : std::uniform_int_distribution<>(1, 10)(rng);
+        for (size_t j = 0; j < num_rows_left; ++j)
+            left_source.addRow({k1, k2, get_attr("left", j)});
+
+        size_t num_rows_right = key_presence == 1 ? 0 : std::uniform_int_distribution<>(1, 10)(rng);
+        for (size_t j = 0; j < num_rows_right; ++j)
+            right_source.addRow({k1, k2, get_attr("right", j)});
+
+        String left_attr = num_rows_left ? get_attr("left", 0) : "";
+        String right_attr = num_rows_right ? get_attr("right", 0) : "";
+
+        if (kind == JoinKind::Inner && num_rows_left && num_rows_right)
+        {
+            expected_left->insert(left_attr);
+            expected_right->insert(right_attr);
+        }
+        else if (kind == JoinKind::Left)
+        {
+            for (size_t j = 0; j < num_rows_left; ++j)
+            {
+                expected_left->insert(get_attr("left", j));
+                expected_right->insert(right_attr);
+            }
+        }
+        else if (kind == JoinKind::Right)
+        {
+            for (size_t j = 0; j < num_rows_right; ++j)
+            {
+                expected_left->insert(left_attr);
+                expected_right->insert(get_attr("right", j));
+            }
+        }
+    }
+
+    Block result_block = executePipeline(buildJoinPipeline(
+        left_source.getSource(), right_source.getSource(), /* key_length = */ 2,
+        kind, JoinStrictness::Any));
+    assertColumnEq<ColumnString>(*expected_left, result_block, "t1.attr");
+    assertColumnEq<ColumnString>(*expected_right, result_block, "t2.attr");
+}
+catch (Exception & e)
+{
+    std::cout << e.getStackTraceString() << std::endl;
+    throw;
+}
+
+TEST_F(FullSortingJoinTest, AsofSimple)
+try
+{
+    SourceChunksBuilder left_source({
+        {std::make_shared<DataTypeString>(), "key"},
+        {std::make_shared<DataTypeUInt64>(), "t"},
+    });
+    left_source.addRow({"AMZN", 3});
+    left_source.addRow({"AMZN", 4});
+    left_source.addRow({"AMZN", 6});
+    left_source.addRow({"SBUX", 10});
+
+    SourceChunksBuilder right_source({
+        {std::make_shared<DataTypeString>(), "key"},
+        {std::make_shared<DataTypeUInt64>(), "t"},
+        {std::make_shared<DataTypeUInt64>(), "value"},
+    });
+    right_source.addRow({"AAPL", 1, 97});
+    right_source.addChunk();
+    right_source.addRow({"AAPL", 2, 98});
+    right_source.addRow({"AAPL", 3, 99});
+    right_source.addRow({"AMZN", 1, 100});
+    right_source.addRow({"AMZN", 2, 110});
+    right_source.addChunk();
+    right_source.addRow({"AMZN", 2, 110});
+    right_source.addChunk();
+    right_source.addRow({"AMZN", 4, 130});
+    right_source.addRow({"AMZN", 5, 140});
+    right_source.addRow({"SBUX", 8, 180});
+    right_source.addChunk();
+    right_source.addRow({"SBUX", 9, 190});
+
+    {
+        Block result_block = executePipeline(buildJoinPipeline(
+            left_source.getSource(), right_source.getSource(), /* key_length = */ 2,
+            JoinKind::Inner, JoinStrictness::Asof, ASOFJoinInequality::LessOrEquals));
+        auto values = getValuesFromBlock(result_block, {"t1.key", "t1.t", "t2.t", "t2.value"});
+
+        ASSERT_EQ(values, (std::vector<std::vector<Field>>{
+            {"AMZN", 3u, 4u, 130u},
+            {"AMZN", 4u, 4u, 130u},
+        }));
+    }
+
+    {
+        Block result_block = executePipeline(buildJoinPipeline(
+            left_source.getSource(), right_source.getSource(), /* key_length = */ 2,
+            JoinKind::Inner, JoinStrictness::Asof, ASOFJoinInequality::GreaterOrEquals));
+        auto values = getValuesFromBlock(result_block, {"t1.key", "t1.t", "t2.t", "t2.value"});
+
+        ASSERT_EQ(values, (std::vector<std::vector<Field>>{
+            {"AMZN", 3u, 2u, 110u},
+            {"AMZN", 4u, 4u, 130u},
+            {"AMZN", 6u, 5u, 140u},
+            {"SBUX", 10u, 9u, 190u},
+        }));
+    }
+}
+catch (Exception & e)
+{
+    std::cout << e.getStackTraceString() << std::endl;
+    throw;
+}
+
+
+TEST_F(FullSortingJoinTest, AsofOnlyColumn)
+try
+{
+    auto left_source = oneColumnSource({ {3}, {3, 3, 3}, {3, 5, 5, 6}, {9, 9}, {10, 20} });
+
+    SourceChunksBuilder right_source_builder({
+        {std::make_shared<DataTypeUInt64>(), "t"},
+        {std::make_shared<DataTypeUInt64>(), "value"},
+    });
+
+    right_source_builder.setBreakProbability(rng);
+
+    for (const auto & row : std::vector<std::vector<Field>>{ {1, 101}, {2, 102}, {4, 104}, {5, 105}, {11, 111}, {15, 115} })
+        right_source_builder.addRow(row);
+
+    auto right_source = right_source_builder.getSource();
+
+    auto pipeline = buildJoinPipeline(
+        left_source, right_source, /* key_length = */ 1,
+        JoinKind::Inner, JoinStrictness::Asof, ASOFJoinInequality::LessOrEquals);
+
+    Block result_block = executePipeline(std::move(pipeline));
+
+    ASSERT_EQ(
+        assert_cast<const ColumnUInt64 *>(result_block.getByName("t1.key").column.get())->getData(),
+        (ColumnUInt64::Container{3, 3, 3, 3, 3, 5, 5, 6, 9, 9, 10})
+    );
+
+    ASSERT_EQ(
+        assert_cast<const ColumnUInt64 *>(result_block.getByName("t2.t").column.get())->getData(),
+        (ColumnUInt64::Container{4, 4, 4, 4, 4, 5, 5, 11, 11, 11, 11})
+    );
+
+    ASSERT_EQ(
+        assert_cast<const ColumnUInt64 *>(result_block.getByName("t2.value").column.get())->getData(),
+        (ColumnUInt64::Container{104, 104, 104, 104, 104, 105, 105, 111, 111, 111, 111})
+    );
+}
+catch (Exception & e)
+{
+    std::cout << e.getStackTraceString() << std::endl;
+    throw;
+}
+
+TEST_F(FullSortingJoinTest, AsofLessGeneratedTestData)
+try
+{
+    /// Generate data random and build expected result at the same time.
+
+    /// Test specific combinations of join kind and inequality per each run
+    auto join_kind = getRandomFrom(rng, { JoinKind::Inner, JoinKind::Left });
+    auto asof_inequality = getRandomFrom(rng, { ASOFJoinInequality::Less, ASOFJoinInequality::LessOrEquals });
+
+    SCOPED_TRACE(fmt::format("{} {}", join_kind, asof_inequality));
+
+    /// Key is complex, `k1, k2` for equality and `t` for asof
+    SourceChunksBuilder left_source_builder({
+        {std::make_shared<DataTypeUInt64>(), "k1"},
+        {std::make_shared<DataTypeString>(), "k2"},
+        {std::make_shared<DataTypeUInt64>(), "t"},
+        {std::make_shared<DataTypeInt64>(), "attr"},
+    });
+
+    SourceChunksBuilder right_source_builder({
+        {std::make_shared<DataTypeUInt64>(), "k1"},
+        {std::make_shared<DataTypeString>(), "k2"},
+        {std::make_shared<DataTypeUInt64>(), "t"},
+        {std::make_shared<DataTypeInt64>(), "attr"},
+    });
+
+    /// How small generated block should be
+    left_source_builder.setBreakProbability(rng);
+    right_source_builder.setBreakProbability(rng);
+
+    /// We are going to generate sorted data and remember expected result
+    ColumnInt64::Container expected;
+
+    UInt64 k1 = 1;
+    String k2;
+    auto key_num_total = std::uniform_int_distribution<>(1, 1000)(rng);
+    for (size_t key_num = 0; key_num < key_num_total; ++key_num)
+    {
+        /// Generate new key greater than previous
+        generateNextKey(rng, k1, k2);
+
+        Int64 left_t = 0;
+        /// Generate several rows for the key
+        size_t num_left_rows = std::uniform_int_distribution<>(1, 100)(rng);
+        for (size_t i = 0; i < num_left_rows; ++i)
+        {
+            /// t is strictly greater than previous
+            left_t += std::uniform_int_distribution<>(1, 10)(rng);
+
+            auto left_arrtibute_value = 10 * left_t;
+            left_source_builder.addRow({k1, k2, left_t, left_arrtibute_value});
+            expected.push_back(left_arrtibute_value);
+
+            auto num_matches = 1 + std::poisson_distribution<>(4)(rng);
+            /// Generate several matches in the right table
+            auto right_t = left_t;
+            for (size_t j = 0; j < num_matches; ++j)
+            {
+                int min_step = isStrict(asof_inequality) ? 1 : 0;
+                right_t += std::uniform_int_distribution<>(min_step, 3)(rng);
+
+                /// First row should match
+                bool is_match = j == 0;
+                right_source_builder.addRow({k1, k2, right_t, is_match ? 10 * left_arrtibute_value : -1});
+            }
+            /// Next left_t should be greater than right_t not to match with previous rows
+            left_t = right_t;
+        }
+
+        /// generate some rows with greater left_t to check that they are not matched
+        num_left_rows = std::bernoulli_distribution(0.5)(rng) ? std::uniform_int_distribution<>(1, 100)(rng) : 0;
+        for (size_t i = 0; i < num_left_rows; ++i)
+        {
+            left_t += std::uniform_int_distribution<>(1, 10)(rng);
+            left_source_builder.addRow({k1, k2, left_t, -10 * left_t});
+
+            if (join_kind == JoinKind::Left)
+                expected.push_back(-10 * left_t);
+        }
+    }
+
+    Block result_block = executePipeline(buildJoinPipeline(
+        left_source_builder.getSource(), right_source_builder.getSource(),
+        /* key_length = */ 3,
+        join_kind, JoinStrictness::Asof, asof_inequality));
+
+    assertColumnVectorEq<Int64>(expected, result_block, "t1.attr");
+
+    for (auto & e : expected)
+        /// Non matched rows from left table have negative attr
+        /// Value if attribute in right table is 10 times greater than in left table
+        e = e < 0 ? 0 : 10 * e;
+
+    assertColumnVectorEq<Int64>(expected, result_block, "t2.attr");
+}
+catch (Exception & e)
+{
+    std::cout << e.getStackTraceString() << std::endl;
+    throw;
+}
+
+TEST_F(FullSortingJoinTest, AsofGreaterGeneratedTestData)
+try
+{
+    /// Generate data random and build expected result at the same time.
+
+    /// Test specific combinations of join kind and inequality per each run
+    auto join_kind = getRandomFrom(rng, { JoinKind::Inner, JoinKind::Left });
+    auto asof_inequality = getRandomFrom(rng, { ASOFJoinInequality::Greater, ASOFJoinInequality::GreaterOrEquals });
+
+    SCOPED_TRACE(fmt::format("{} {}", join_kind, asof_inequality));
+
+    SourceChunksBuilder left_source_builder({
+        {std::make_shared<DataTypeUInt64>(), "k1"},
+        {std::make_shared<DataTypeString>(), "k2"},
+        {std::make_shared<DataTypeUInt64>(), "t"},
+        {std::make_shared<DataTypeInt64>(), "attr"},
+    });
+
+    SourceChunksBuilder right_source_builder({
+        {std::make_shared<DataTypeUInt64>(), "k1"},
+        {std::make_shared<DataTypeString>(), "k2"},
+        {std::make_shared<DataTypeUInt64>(), "t"},
+        {std::make_shared<DataTypeInt64>(), "attr"},
+    });
+
+    left_source_builder.setBreakProbability(rng);
+    right_source_builder.setBreakProbability(rng);
+
+    ColumnInt64::Container expected;
+
+    UInt64 k1 = 1;
+    String k2;
+    UInt64 left_t = 0;
+
+    auto key_num_total = std::uniform_int_distribution<>(1, 1000)(rng);
+    for (size_t key_num = 0; key_num < key_num_total; ++key_num)
+    {
+        /// Generate new key greater than previous
+        generateNextKey(rng, k1, k2);
+
+        /// Generate some rows with smaller left_t to check that they are not matched
+        size_t num_left_rows = std::bernoulli_distribution(0.5)(rng) ? std::uniform_int_distribution<>(1, 100)(rng) : 0;
+        for (size_t i = 0; i < num_left_rows; ++i)
+        {
+            left_t += std::uniform_int_distribution<>(1, 10)(rng);
+            left_source_builder.addRow({k1, k2, left_t, -10 * left_t});
+
+            if (join_kind == JoinKind::Left)
+                expected.push_back(-10 * left_t);
+        }
+
+        if (std::bernoulli_distribution(0.1)(rng))
+            continue;
+
+        size_t num_right_matches = std::uniform_int_distribution<>(1, 10)(rng);
+        auto right_t = left_t + std::uniform_int_distribution<>(isStrict(asof_inequality) ? 0 : 1, 10)(rng);
+        auto attribute_value = 10 * right_t;
+        for (size_t j = 0; j < num_right_matches; ++j)
+        {
+            right_t += std::uniform_int_distribution<>(0, 3)(rng);
+            bool is_match = j == num_right_matches - 1;
+            right_source_builder.addRow({k1, k2, right_t, is_match ? 10 * attribute_value : -1});
+        }
+
+        /// Next left_t should be greater than (or equals) right_t to match with previous rows
+        left_t = right_t + std::uniform_int_distribution<>(isStrict(asof_inequality) ? 1 : 0, 100)(rng);
+        size_t num_left_matches = std::uniform_int_distribution<>(1, 100)(rng);
+        for (size_t j = 0; j < num_left_matches; ++j)
+        {
+            left_t += std::uniform_int_distribution<>(0, 3)(rng);
+            left_source_builder.addRow({k1, k2, left_t, attribute_value});
+            expected.push_back(attribute_value);
+        }
+    }
+
+    Block result_block = executePipeline(buildJoinPipeline(
+        left_source_builder.getSource(), right_source_builder.getSource(),
+        /* key_length = */ 3,
+        join_kind, JoinStrictness::Asof, asof_inequality));
+
+    assertColumnVectorEq<Int64>(expected, result_block, "t1.attr");
+
+    for (auto & e : expected)
+        /// Non matched rows from left table have negative attr
+        /// Value if attribute in right table is 10 times greater than in left table
+        e = e < 0 ? 0 : 10 * e;
+
+    assertColumnVectorEq<Int64>(expected, result_block, "t2.attr");
+}
+catch (Exception & e)
+{
+    std::cout << e.getStackTraceString() << std::endl;
+    throw;
+}

src/Storages/MergeTree/MergeTreeBackgroundExecutor.cpp

@@ -155,6 +155,10 @@ void printExceptionWithRespectToAbort(LoggerPtr log, const String & query_id)
     {
         std::rethrow_exception(ex);
     }
+    catch (const TestException &) // NOLINT
+    {
+        /// Exception from a unit test, ignore it.
+    }
     catch (const Exception & e)
     {
         NOEXCEPT_SCOPE({

src/Storages/MergeTree/tests/gtest_executor.cpp

@@ -34,7 +34,7 @@ public:
 
         auto choice = distribution(generator);
         if (choice == 0)
-            throw std::runtime_error("Unlucky...");
+            throw TestException();
 
         return false;
     }
@@ -48,7 +48,7 @@ public:
     {
         auto choice = distribution(generator);
         if (choice == 0)
-            throw std::runtime_error("Unlucky...");
+            throw TestException();
     }
 
     Priority getPriority() const override { return {}; }

tests/ci/artifactory.py

@@ -3,8 +3,13 @@ import time
 from pathlib import Path
 from typing import Optional
 from shutil import copy2
-from create_release import PackageDownloader, ReleaseInfo, ShellRunner
+from create_release import (
-from ci_utils import WithIter
+    PackageDownloader,
+    ReleaseInfo,
+    ReleaseContextManager,
+    ReleaseProgress,
+)
+from ci_utils import WithIter, Shell
 
 
 class MountPointApp(metaclass=WithIter):
@@ -38,7 +43,6 @@ class R2MountPoint:
             self.bucket_name = self._PROD_BUCKET_NAME
 
         self.aux_mount_options = ""
-        self.async_mount = False
         if self.app == MountPointApp.S3FS:
             self.cache_dir = "/home/ubuntu/s3fs_cache"
             # self.aux_mount_options += "-o nomodtime " if self.NOMODTIME else "" not for s3fs
@@ -52,7 +56,6 @@ class R2MountPoint:
             self.mount_cmd = f"s3fs {self.bucket_name} {self.MOUNT_POINT} -o url={self.API_ENDPOINT} -o use_path_request_style -o umask=0000 -o nomultipart -o logfile={self.LOG_FILE} {self.aux_mount_options}"
         elif self.app == MountPointApp.RCLONE:
             # run rclone mount process asynchronously, otherwise subprocess.run(daemonized command) will not return
-            self.async_mount = True
             self.cache_dir = "/home/ubuntu/rclone_cache"
             self.aux_mount_options += "--no-modtime " if self.NOMODTIME else ""
             self.aux_mount_options += "-v " if self.DEBUG else ""  # -vv too verbose
@@ -76,19 +79,22 @@ class R2MountPoint:
         )
 
         _TEST_MOUNT_CMD = f"mount | grep -q {self.MOUNT_POINT}"
-        ShellRunner.run(_CLEAN_LOG_FILE_CMD)
+        Shell.run(_CLEAN_LOG_FILE_CMD)
-        ShellRunner.run(_UNMOUNT_CMD)
+        Shell.run(_UNMOUNT_CMD)
-        ShellRunner.run(_MKDIR_CMD)
+        Shell.run(_MKDIR_CMD)
-        ShellRunner.run(_MKDIR_FOR_CACHE)
+        Shell.run(_MKDIR_FOR_CACHE)
-        ShellRunner.run(self.mount_cmd, async_=self.async_mount)
+        if self.app == MountPointApp.S3FS:
-        if self.async_mount:
+            Shell.run(self.mount_cmd, check=True)
-            time.sleep(3)
+        else:
-        ShellRunner.run(_TEST_MOUNT_CMD)
+            # didn't manage to use simple run() and without blocking or failure
+            Shell.run_as_daemon(self.mount_cmd)
+        time.sleep(3)
+        Shell.run(_TEST_MOUNT_CMD, check=True)
 
     @classmethod
     def teardown(cls):
         print(f"Unmount [{cls.MOUNT_POINT}]")
-        ShellRunner.run(f"umount {cls.MOUNT_POINT}")
+        Shell.run(f"umount {cls.MOUNT_POINT}")
 
 
 class RepoCodenames(metaclass=WithIter):
@@ -101,6 +107,7 @@ class DebianArtifactory:
     _PROD_REPO_URL = "https://packages.clickhouse.com/deb"
 
     def __init__(self, release_info: ReleaseInfo, dry_run: bool):
+        self.release_info = release_info
         self.codename = release_info.codename
         self.version = release_info.version
         if dry_run:
@@ -124,8 +131,8 @@ class DebianArtifactory:
         cmd = f"{REPREPRO_CMD_PREFIX} includedeb {self.codename} {' '.join(paths)}"
         print("Running export command:")
         print(f"  {cmd}")
-        ShellRunner.run(cmd)
+        Shell.run(cmd, check=True)
-        ShellRunner.run("sync")
+        Shell.run("sync")
 
         if self.codename == RepoCodenames.LTS:
             packages_with_version = [
@@ -137,16 +144,19 @@ class DebianArtifactory:
             cmd = f"{REPREPRO_CMD_PREFIX} copy {RepoCodenames.STABLE} {RepoCodenames.LTS} {' '.join(packages_with_version)}"
             print("Running copy command:")
             print(f"  {cmd}")
-            ShellRunner.run(cmd)
+            Shell.run(cmd, check=True)
-            ShellRunner.run("sync")
+            Shell.run("sync")
 
     def test_packages(self):
-        ShellRunner.run("docker pull ubuntu:latest")
+        Shell.run("docker pull ubuntu:latest")
         print(f"Test packages installation, version [{self.version}]")
-        cmd = f"docker run --rm ubuntu:latest bash -c \"apt update -y; apt install -y sudo gnupg ca-certificates; apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv 8919F6BD2B48D754; echo 'deb {self.repo_url} stable main' | tee /etc/apt/sources.list.d/clickhouse.list; apt update -y; apt-get install -y clickhouse-client={self.version}\""
+        debian_command = f"echo 'deb {self.repo_url} stable main' | tee /etc/apt/sources.list.d/clickhouse.list; apt update -y; apt-get install -y clickhouse-common-static={self.version} clickhouse-client={self.version}"
+        cmd = f'docker run --rm ubuntu:latest bash -c "apt update -y; apt install -y sudo gnupg ca-certificates; apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv 8919F6BD2B48D754; {debian_command}"'
         print("Running test command:")
         print(f"  {cmd}")
-        ShellRunner.run(cmd)
+        Shell.run(cmd, check=True)
+        self.release_info.debian_command = debian_command
+        self.release_info.dump()
 
 
 def _copy_if_not_exists(src: Path, dst: Path) -> Path:
@@ -167,6 +177,7 @@ class RpmArtifactory:
     _SIGN_KEY = "885E2BDCF96B0B45ABF058453E4AD4719DDE9A38"
 
     def __init__(self, release_info: ReleaseInfo, dry_run: bool):
+        self.release_info = release_info
         self.codename = release_info.codename
         self.version = release_info.version
         if dry_run:
@@ -202,23 +213,26 @@ class RpmArtifactory:
         for command in commands:
             print("Running command:")
             print(f"    {command}")
-            ShellRunner.run(command)
+            Shell.run(command, check=True)
 
         update_public_key = f"gpg --armor --export {self._SIGN_KEY}"
         pub_key_path = dest_dir / "repodata" / "repomd.xml.key"
         print("Updating repomd.xml.key")
-        pub_key_path.write_text(ShellRunner.run(update_public_key)[1])
+        pub_key_path.write_text(Shell.run(update_public_key, check=True))
         if codename == RepoCodenames.LTS:
             self.export_packages(RepoCodenames.STABLE)
-        ShellRunner.run("sync")
+        Shell.run("sync")
 
     def test_packages(self):
-        ShellRunner.run("docker pull fedora:latest")
+        Shell.run("docker pull fedora:latest")
         print(f"Test package installation, version [{self.version}]")
-        cmd = f'docker run --rm fedora:latest /bin/bash -c "dnf -y install dnf-plugins-core && dnf config-manager --add-repo={self.repo_url} && dnf makecache && dnf -y install clickhouse-client-{self.version}-1"'
+        rpm_command = f"dnf config-manager --add-repo={self.repo_url} && dnf makecache && dnf -y install clickhouse-client-{self.version}-1"
+        cmd = f'docker run --rm fedora:latest /bin/bash -c "dnf -y install dnf-plugins-core && dnf config-manager --add-repo={self.repo_url} && {rpm_command}"'
         print("Running test command:")
         print(f"  {cmd}")
-        ShellRunner.run(cmd)
+        Shell.run(cmd, check=True)
+        self.release_info.rpm_command = rpm_command
+        self.release_info.dump()
 
 
 class TgzArtifactory:
@@ -226,6 +240,7 @@ class TgzArtifactory:
     _PROD_REPO_URL = "https://packages.clickhouse.com/tgz"
 
     def __init__(self, release_info: ReleaseInfo, dry_run: bool):
+        self.release_info = release_info
         self.codename = release_info.codename
         self.version = release_info.version
         if dry_run:
@@ -256,23 +271,28 @@ class TgzArtifactory:
 
         if codename == RepoCodenames.LTS:
             self.export_packages(RepoCodenames.STABLE)
-        ShellRunner.run("sync")
+        Shell.run("sync")
 
     def test_packages(self):
         tgz_file = "/tmp/tmp.tgz"
         tgz_sha_file = "/tmp/tmp.tgz.sha512"
-        ShellRunner.run(
+        cmd = f"curl -o {tgz_file} -f0 {self.repo_url}/stable/clickhouse-client-{self.version}-arm64.tgz"
-            f"curl -o {tgz_file} -f0 {self.repo_url}/stable/clickhouse-client-{self.version}-arm64.tgz"
+        Shell.run(
+            cmd,
+            check=True,
         )
-        ShellRunner.run(
+        Shell.run(
-            f"curl -o {tgz_sha_file} -f0 {self.repo_url}/stable/clickhouse-client-{self.version}-arm64.tgz.sha512"
+            f"curl -o {tgz_sha_file} -f0 {self.repo_url}/stable/clickhouse-client-{self.version}-arm64.tgz.sha512",
+            check=True,
         )
-        expected_checksum = ShellRunner.run(f"cut -d ' ' -f 1 {tgz_sha_file}")
+        expected_checksum = Shell.run(f"cut -d ' ' -f 1 {tgz_sha_file}", check=True)
-        actual_checksum = ShellRunner.run(f"sha512sum {tgz_file} | cut -d ' ' -f 1")
+        actual_checksum = Shell.run(f"sha512sum {tgz_file} | cut -d ' ' -f 1")
         assert (
             expected_checksum == actual_checksum
         ), f"[{actual_checksum} != {expected_checksum}]"
-        ShellRunner.run("rm /tmp/tmp.tgz*")
+        Shell.run("rm /tmp/tmp.tgz*")
+        self.release_info.tgz_command = cmd
+        self.release_info.dump()
 
 
 def parse_args() -> argparse.Namespace:
@@ -280,12 +300,6 @@ def parse_args() -> argparse.Namespace:
         formatter_class=argparse.ArgumentDefaultsHelpFormatter,
         description="Adds release packages to the repository",
     )
-    parser.add_argument(
-        "--infile",
-        type=str,
-        required=True,
-        help="input file with release info",
-    )
     parser.add_argument(
         "--export-debian",
         action="store_true",
@@ -326,9 +340,7 @@ def parse_args() -> argparse.Namespace:
 
 if __name__ == "__main__":
     args = parse_args()
-    assert args.dry_run
 
-    release_info = ReleaseInfo.from_file(args.infile)
     """
     Use S3FS. RCLONE has some errors with r2 remote which I didn't figure out how to resolve:
            ERROR : IO error: NotImplemented: versionId not implemented
@@ -336,20 +348,38 @@ if __name__ == "__main__":
     """
     mp = R2MountPoint(MountPointApp.S3FS, dry_run=args.dry_run)
     if args.export_debian:
-        mp.init()
+        with ReleaseContextManager(
-        DebianArtifactory(release_info, dry_run=args.dry_run).export_packages()
+            release_progress=ReleaseProgress.EXPORT_DEB
-        mp.teardown()
+        ) as release_info:
+            mp.init()
+            DebianArtifactory(release_info, dry_run=args.dry_run).export_packages()
+            mp.teardown()
     if args.export_rpm:
-        mp.init()
+        with ReleaseContextManager(
-        RpmArtifactory(release_info, dry_run=args.dry_run).export_packages()
+            release_progress=ReleaseProgress.EXPORT_RPM
-        mp.teardown()
+        ) as release_info:
+            mp.init()
+            RpmArtifactory(release_info, dry_run=args.dry_run).export_packages()
+            mp.teardown()
     if args.export_tgz:
-        mp.init()
+        with ReleaseContextManager(
-        TgzArtifactory(release_info, dry_run=args.dry_run).export_packages()
+            release_progress=ReleaseProgress.EXPORT_TGZ
-        mp.teardown()
+        ) as release_info:
+            mp.init()
+            TgzArtifactory(release_info, dry_run=args.dry_run).export_packages()
+            mp.teardown()
     if args.test_debian:
-        DebianArtifactory(release_info, dry_run=args.dry_run).test_packages()
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.TEST_DEB
+        ) as release_info:
+            DebianArtifactory(release_info, dry_run=args.dry_run).test_packages()
     if args.test_tgz:
-        TgzArtifactory(release_info, dry_run=args.dry_run).test_packages()
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.TEST_TGZ
+        ) as release_info:
+            TgzArtifactory(release_info, dry_run=args.dry_run).test_packages()
     if args.test_rpm:
-        RpmArtifactory(release_info, dry_run=args.dry_run).test_packages()
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.TEST_RPM
+        ) as release_info:
+            RpmArtifactory(release_info, dry_run=args.dry_run).test_packages()

tests/ci/auto_release.py

@@ -1,17 +1,17 @@
 import argparse
-from datetime import timedelta, datetime
+import dataclasses
-import logging
+import json
 import os
-from commit_status_helper import get_commit_filtered_statuses
+import sys
+from typing import List
+
 from get_robot_token import get_best_robot_token
 from github_helper import GitHub
-from release import Release, Repo as ReleaseRepo, RELEASE_READY_STATUS
+from ci_utils import Shell
+from env_helper import GITHUB_REPOSITORY
 from report import SUCCESS
-from ssh import SSHKey
+from ci_buddy import CIBuddy
-
+from ci_config import CI
-LOGGER_NAME = __name__
-HELPER_LOGGERS = ["github_helper", LOGGER_NAME]
-logger = logging.getLogger(LOGGER_NAME)
 
 
 def parse_args():
@@ -21,120 +21,198 @@ def parse_args():
     )
     parser.add_argument("--token", help="GitHub token, if not set, used from smm")
     parser.add_argument(
-        "--repo", default="ClickHouse/ClickHouse", help="Repo owner/name"
+        "--post-status",
-    )
-    parser.add_argument("--dry-run", action="store_true", help="Do not create anything")
-    parser.add_argument(
-        "--release-after-days",
-        type=int,
-        default=3,
-        help="Do automatic release on the latest green commit after the latest "
-        "release if the newest release is older than the specified days",
-    )
-    parser.add_argument(
-        "--debug-helpers",
         action="store_true",
-        help="Add debug logging for this script and github_helper",
+        help="Post release branch statuses",
     )
     parser.add_argument(
-        "--remote-protocol",
+        "--post-auto-release-complete",
-        "-p",
+        action="store_true",
-        default="ssh",
+        help="Post autorelease completion status",
-        choices=ReleaseRepo.VALID,
-        help="repo protocol for git commands remote, 'origin' is a special case and "
-        "uses 'origin' as a remote",
     )
+    parser.add_argument(
+        "--prepare",
+        action="store_true",
+        help="Prepare autorelease info",
+    )
+    parser.add_argument(
+        "--wf-status",
+        type=str,
+        default="",
+        help="overall workflow status [success|failure]",
+    )
+    return parser.parse_args(), parser
 
-    return parser.parse_args()
+
+MAX_NUMBER_OF_COMMITS_TO_CONSIDER_FOR_RELEASE = 5
+AUTORELEASE_INFO_FILE = "/tmp/autorelease_info.json"
+
+
+@dataclasses.dataclass
+class ReleaseParams:
+    ready: bool
+    ci_status: str
+    num_patches: int
+    release_branch: str
+    commit_sha: str
+    commits_to_branch_head: int
+    latest: bool
+
+    def to_dict(self):
+        return dataclasses.asdict(self)
+
+
+@dataclasses.dataclass
+class AutoReleaseInfo:
+    releases: List[ReleaseParams]
+
+    def add_release(self, release_params: ReleaseParams) -> None:
+        self.releases.append(release_params)
+
+    def dump(self):
+        print(f"Dump release info into [{AUTORELEASE_INFO_FILE}]")
+        with open(AUTORELEASE_INFO_FILE, "w", encoding="utf-8") as f:
+            print(json.dumps(dataclasses.asdict(self), indent=2), file=f)
+
+    @staticmethod
+    def from_file() -> "AutoReleaseInfo":
+        with open(AUTORELEASE_INFO_FILE, "r", encoding="utf-8") as json_file:
+            res = json.load(json_file)
+        releases = [ReleaseParams(**release) for release in res["releases"]]
+        return AutoReleaseInfo(releases=releases)
+
+
+def _prepare(token):
+    assert len(token) > 10
+    os.environ["GH_TOKEN"] = token
+    Shell.run("gh auth status", check=True)
+
+    gh = GitHub(token)
+    prs = gh.get_release_pulls(GITHUB_REPOSITORY)
+    prs.sort(key=lambda x: x.head.ref)
+    branch_names = [pr.head.ref for pr in prs]
+    print(f"Found release branches [{branch_names}]")
+
+    repo = gh.get_repo(GITHUB_REPOSITORY)
+    autoRelease_info = AutoReleaseInfo(releases=[])
+
+    for pr in prs:
+        print(f"\nChecking PR [{pr.head.ref}]")
+
+        refs = list(repo.get_git_matching_refs(f"tags/v{pr.head.ref}"))
+        assert refs
+
+        refs.sort(key=lambda ref: ref.ref)
+        latest_release_tag_ref = refs[-1]
+        latest_release_tag = repo.get_git_tag(latest_release_tag_ref.object.sha)
+
+        commits = Shell.run(
+            f"git rev-list --first-parent {latest_release_tag.tag}..origin/{pr.head.ref}",
+            check=True,
+        ).split("\n")
+        commit_num = len(commits)
+        print(
+            f"Previous release [{latest_release_tag.tag}] was [{commit_num}] commits ago, date [{latest_release_tag.tagger.date}]"
+        )
+
+        commits_to_check = commits[:-1]  # Exclude the version bump commit
+        commit_sha = ""
+        commit_ci_status = ""
+        commits_to_branch_head = 0
+
+        for idx, commit in enumerate(
+            commits_to_check[:MAX_NUMBER_OF_COMMITS_TO_CONSIDER_FOR_RELEASE]
+        ):
+            print(
+                f"Check commit [{commit}] [{pr.head.ref}~{idx+1}] as release candidate"
+            )
+            commit_num -= 1
+
+            is_completed = CI.GHActions.check_wf_completed(
+                token=token, commit_sha=commit
+            )
+            if not is_completed:
+                print(f"CI is in progress for [{commit}] - check previous commit")
+                commits_to_branch_head += 1
+                continue
+
+            commit_ci_status = CI.GHActions.get_commit_status_by_name(
+                token=token,
+                commit_sha=commit,
+                status_name=(CI.JobNames.BUILD_CHECK, "ClickHouse build check"),
+            )
+            commit_sha = commit
+            if commit_ci_status == SUCCESS:
+                break
+
+            print(f"CI status [{commit_ci_status}] - skip")
+            commits_to_branch_head += 1
+
+        ready = False
+        if commit_ci_status == SUCCESS and commit_sha:
+            print(
+                f"Add release ready info for commit [{commit_sha}] and release branch [{pr.head.ref}]"
+            )
+            ready = True
+        else:
+            print(f"WARNING: No ready commits found for release branch [{pr.head.ref}]")
+
+        autoRelease_info.add_release(
+            ReleaseParams(
+                release_branch=pr.head.ref,
+                commit_sha=commit_sha,
+                ready=ready,
+                ci_status=commit_ci_status,
+                num_patches=commit_num,
+                commits_to_branch_head=commits_to_branch_head,
+                latest=False,
+            )
+        )
+
+    if autoRelease_info.releases:
+        autoRelease_info.releases[-1].latest = True
+
+    autoRelease_info.dump()
 
 
 def main():
-    args = parse_args()
+    args, parser = parse_args()
-    logging.basicConfig(level=logging.INFO)
-    if args.debug_helpers:
-        for logger_name in HELPER_LOGGERS:
-            logging.getLogger(logger_name).setLevel(logging.DEBUG)
 
-    token = args.token or get_best_robot_token()
+    if args.post_status:
-    days_as_timedelta = timedelta(days=args.release_after_days)
+        info = AutoReleaseInfo.from_file()
-    now = datetime.now()
+        for release_info in info.releases:
-
+            if release_info.ready:
-    gh = GitHub(token)
+                CIBuddy(dry_run=False).post_info(
-    prs = gh.get_release_pulls(args.repo)
+                    title=f"Auto Release Status for {release_info.release_branch}",
-    branch_names = [pr.head.ref for pr in prs]
+                    body=release_info.to_dict(),
-
+                )
-    logger.info("Found release branches: %s\n ", " \n".join(branch_names))
+            else:
-    repo = gh.get_repo(args.repo)
+                CIBuddy(dry_run=False).post_warning(
-
+                    title=f"Auto Release Status for {release_info.release_branch}",
-    # In general there is no guarantee on which order the refs/commits are
+                    body=release_info.to_dict(),
-    # returned from the API, so we have to order them.
+                )
-    for pr in prs:
+    elif args.post_auto_release_complete:
-        logger.info("Checking PR %s", pr.head.ref)
+        assert args.wf_status, "--wf-status Required with --post-auto-release-complete"
-
+        if args.wf_status != SUCCESS:
-        refs = list(repo.get_git_matching_refs(f"tags/v{pr.head.ref}"))
+            CIBuddy(dry_run=False).post_job_error(
-        refs.sort(key=lambda ref: ref.ref)
+                error_description="Autorelease workflow failed",
-
+                job_name="Autorelease",
-        latest_release_tag_ref = refs[-1]
+                with_instance_info=False,
-        latest_release_tag = repo.get_git_tag(latest_release_tag_ref.object.sha)
+                with_wf_link=True,
-        logger.info("That last release was done at %s", latest_release_tag.tagger.date)
+                critical=True,
-
-        if latest_release_tag.tagger.date + days_as_timedelta > now:
-            logger.info(
-                "Not enough days since the last release %s,"
-                " no automatic release can be done",
-                latest_release_tag.tag,
             )
-            continue
+        else:
-
+            CIBuddy(dry_run=False).post_info(
-        unreleased_commits = list(
+                title=f"Autorelease completed",
-            repo.get_commits(sha=pr.head.ref, since=latest_release_tag.tagger.date)
+                body="",
-        )
+                with_wf_link=True,
-        unreleased_commits.sort(
-            key=lambda commit: commit.commit.committer.date, reverse=True
-        )
-
-        for commit in unreleased_commits:
-            logger.info("Checking statuses of commit %s", commit.sha)
-            statuses = get_commit_filtered_statuses(commit)
-            all_success = all(st.state == SUCCESS for st in statuses)
-            passed_ready_for_release_check = any(
-                st.context == RELEASE_READY_STATUS and st.state == SUCCESS
-                for st in statuses
             )
-            if not (all_success and passed_ready_for_release_check):
+    elif args.prepare:
-                logger.info("Commit is not green, thus not suitable for release")
+        _prepare(token=args.token or get_best_robot_token())
-                continue
+    else:
-
+        parser.print_help()
-            logger.info("Commit is ready for release, let's release!")
+        sys.exit(2)
-
-            release = Release(
-                ReleaseRepo(args.repo, args.remote_protocol),
-                commit.sha,
-                "patch",
-                args.dry_run,
-                True,
-            )
-            try:
-                release.do(True, True, True)
-            except:
-                if release.has_rollback:
-                    logging.error(
-                        "!!The release process finished with error, read the output carefully!!"
-                    )
-                    logging.error(
-                        "Probably, rollback finished with error. "
-                        "If you don't see any of the following commands in the output, "
-                        "execute them manually:"
-                    )
-                    release.log_rollback()
-                raise
-            logging.info("New release is done!")
-            break
 
 
 if __name__ == "__main__":
-    if os.getenv("ROBOT_CLICKHOUSE_SSH_KEY", ""):
+    main()
-        with SSHKey("ROBOT_CLICKHOUSE_SSH_KEY"):
-            main()
-    else:
-        main()

tests/ci/ci.py

@@ -1110,13 +1110,14 @@ def main() -> int:
             ci_cache.print_status()
 
         if IS_CI and not pr_info.is_merge_queue:
-            # wait for pending jobs to be finished, await_jobs is a long blocking call
-            ci_cache.await_pending_jobs(pr_info.is_release)
 
             if pr_info.is_release:
                 print("Release/master: CI Cache add pending records for all todo jobs")
                 ci_cache.push_pending_all(pr_info.is_release)
 
+            # wait for pending jobs to be finished, await_jobs is a long blocking call
+            ci_cache.await_pending_jobs(pr_info.is_release)
+
         # conclude results
         result["git_ref"] = git_ref
         result["version"] = version
@@ -1292,10 +1293,11 @@ def main() -> int:
                     pass
             if Utils.is_killed_with_oom():
                 print("WARNING: OOM while job execution")
+                print(subprocess.run("sudo dmesg -T", check=False))
                 error_description = f"Out Of Memory, exit_code {job_report.exit_code}"
             else:
                 error_description = f"Unknown, exit_code {job_report.exit_code}"
-            CIBuddy().post_error(
+            CIBuddy().post_job_error(
                 error_description + f" after {int(job_report.duration)}s",
                 job_name=_get_ext_check_name(args.job_name),
             )

tests/ci/ci_buddy.py

@@ -1,5 +1,6 @@
 import json
 import os
+from typing import Union, Dict
 
 import boto3
 import requests
@@ -60,14 +61,64 @@ class CIBuddy:
         except Exception as e:
             print(f"ERROR: Failed to post message, ex {e}")
 
-    def post_error(self, error_description, job_name="", with_instance_info=True):
+    def _post_formatted(
+        self, title: str, body: Union[Dict, str], with_wf_link: bool
+    ) -> None:
+        message = title
+        if isinstance(body, dict):
+            for name, value in body.items():
+                if "commit_sha" in name:
+                    value = (
+                        f"<https://github.com/{self.repo}/commit/{value}|{value[:8]}>"
+                    )
+                message += f"      *{name}*:    {value}\n"
+        else:
+            message += body + "\n"
+        run_id = os.getenv("GITHUB_RUN_ID", "")
+        if with_wf_link and run_id:
+            message += f"      *workflow*: <https://github.com/{self.repo}/actions/runs/{run_id}|{run_id}>\n"
+        self.post(message)
+
+    def post_info(
+        self, title: str, body: Union[Dict, str], with_wf_link: bool = True
+    ) -> None:
+        title_extended = f":white_circle:    *{title}*\n\n"
+        self._post_formatted(title_extended, body, with_wf_link)
+
+    def post_done(
+        self, title: str, body: Union[Dict, str], with_wf_link: bool = True
+    ) -> None:
+        title_extended = f":white_check_mark:    *{title}*\n\n"
+        self._post_formatted(title_extended, body, with_wf_link)
+
+    def post_warning(
+        self, title: str, body: Union[Dict, str], with_wf_link: bool = True
+    ) -> None:
+        title_extended = f":warning:    *{title}*\n\n"
+        self._post_formatted(title_extended, body, with_wf_link)
+
+    def post_critical(
+        self, title: str, body: Union[Dict, str], with_wf_link: bool = True
+    ) -> None:
+        title_extended = f":black_circle:    *{title}*\n\n"
+        self._post_formatted(title_extended, body, with_wf_link)
+
+    def post_job_error(
+        self,
+        error_description: str,
+        job_name: str = "",
+        with_instance_info: bool = True,
+        with_wf_link: bool = True,
+        critical: bool = False,
+    ) -> None:
         instance_id, instance_type = "unknown", "unknown"
         if with_instance_info:
             instance_id = Shell.run("ec2metadata --instance-id") or instance_id
             instance_type = Shell.run("ec2metadata --instance-type") or instance_type
         if not job_name:
             job_name = os.getenv("CHECK_NAME", "unknown")
-        line_err = f":red_circle:    *Error: {error_description}*\n\n"
+        sign = ":red_circle:" if not critical else ":black_circle:"
+        line_err = f"{sign}    *Error: {error_description}*\n\n"
         line_ghr = f"   *Runner:*    `{instance_type}`, `{instance_id}`\n"
         line_job = f"   *Job:*          `{job_name}`\n"
         line_pr_ = f"   *PR:*           <https://github.com/{self.repo}/pull/{self.pr_number}|#{self.pr_number}>, <{self.commit_url}|{self.sha}>\n"
@@ -82,10 +133,13 @@ class CIBuddy:
             message += line_pr_
         else:
             message += line_br_
+        run_id = os.getenv("GITHUB_RUN_ID", "")
+        if with_wf_link and run_id:
+            message += f"      *workflow*: <https://github.com/{self.repo}/actions/runs/{run_id}|{run_id}>\n"
         self.post(message)
 
 
 if __name__ == "__main__":
     # test
     buddy = CIBuddy(dry_run=True)
-    buddy.post_error("TEst")
+    buddy.post_job_error("TEst")

tests/ci/ci_cache.py

@@ -638,7 +638,14 @@ class CiCache:
         pushes pending records for all jobs that supposed to be run
         """
         for job, job_config in self.jobs_to_do.items():
-            if not job_config.has_digest():
+            if (
+                job in self.jobs_to_wait
+                or not job_config.has_digest()
+                or job_config.disable_await
+            ):
+                # 1. "job in self.jobs_to_wait" - this job already has a pending record in cache
+                # 2. "not job_config.has_digest()" - cache is not used for these jobs
+                # 3. "job_config.disable_await" - await is explicitly disabled
                 continue
             pending_state = PendingState(time.time(), run_url=GITHUB_RUN_URL)
             assert job_config.batches
@@ -708,7 +715,7 @@ class CiCache:
         Filter is to be applied in PRs to remove jobs that are not affected by the change
         :return:
         """
-        remove_from_to_do = []
+        remove_from_workflow = []
         required_builds = []
         has_test_jobs_to_skip = False
         for job_name, job_config in self.jobs_to_do.items():
@@ -723,26 +730,41 @@ class CiCache:
                     job=reference_name,
                     job_config=reference_config,
                 ):
-                    remove_from_to_do.append(job_name)
+                    remove_from_workflow.append(job_name)
                     has_test_jobs_to_skip = True
                 else:
                     required_builds += (
                         job_config.required_builds if job_config.required_builds else []
                     )
         if has_test_jobs_to_skip:
-            # If there are tests to skip, it means build digest has not been changed.
+            # If there are tests to skip, it means builds are not affected as well.
             # No need to test builds. Let's keep all builds required for test jobs and skip the others
             for job_name, job_config in self.jobs_to_do.items():
                 if CI.is_build_job(job_name):
                     if job_name not in required_builds:
-                        remove_from_to_do.append(job_name)
+                        remove_from_workflow.append(job_name)
 
-        for job in remove_from_to_do:
+        for job in remove_from_workflow:
             print(f"Filter job [{job}] - not affected by the change")
             if job in self.jobs_to_do:
                 del self.jobs_to_do[job]
             if job in self.jobs_to_wait:
                 del self.jobs_to_wait[job]
+            if job in self.jobs_to_skip:
+                self.jobs_to_skip.remove(job)
+
+        # special handling for the special job: BUILD_CHECK
+        has_builds = False
+        for job in list(self.jobs_to_do) + self.jobs_to_skip:
+            if CI.is_build_job(job):
+                has_builds = True
+                break
+        if not has_builds:
+            if CI.JobNames.BUILD_CHECK in self.jobs_to_do:
+                print(
+                    f"Filter job [{CI.JobNames.BUILD_CHECK}] - no builds are required in the workflow"
+                )
+                del self.jobs_to_do[CI.JobNames.BUILD_CHECK]
 
     def await_pending_jobs(self, is_release: bool, dry_run: bool = False) -> None:
         """
@@ -884,3 +906,87 @@ class CiCache:
                 self.jobs_to_wait[job] = job_config
 
         return self
+
+
+if __name__ == "__main__":
+    # for testing
+    job_digest = {
+        "package_release": "bbbd3519d1",
+        "package_aarch64": "bbbd3519d1",
+        "package_asan": "bbbd3519d1",
+        "package_ubsan": "bbbd3519d1",
+        "package_tsan": "bbbd3519d1",
+        "package_msan": "bbbd3519d1",
+        "package_debug": "bbbd3519d1",
+        "package_release_coverage": "bbbd3519d1",
+        "binary_release": "bbbd3519d1",
+        "binary_tidy": "bbbd3519d1",
+        "binary_darwin": "bbbd3519d1",
+        "binary_aarch64": "bbbd3519d1",
+        "binary_aarch64_v80compat": "bbbd3519d1",
+        "binary_freebsd": "bbbd3519d1",
+        "binary_darwin_aarch64": "bbbd3519d1",
+        "binary_ppc64le": "bbbd3519d1",
+        "binary_amd64_compat": "bbbd3519d1",
+        "binary_amd64_musl": "bbbd3519d1",
+        "binary_riscv64": "bbbd3519d1",
+        "binary_s390x": "bbbd3519d1",
+        "binary_loongarch64": "bbbd3519d1",
+        "Builds": "f5dffeecb8",
+        "Install packages (release)": "ba0c89660e",
+        "Install packages (aarch64)": "ba0c89660e",
+        "Stateful tests (asan)": "32a9a1aba9",
+        "Stateful tests (tsan)": "32a9a1aba9",
+        "Stateful tests (msan)": "32a9a1aba9",
+        "Stateful tests (ubsan)": "32a9a1aba9",
+        "Stateful tests (debug)": "32a9a1aba9",
+        "Stateful tests (release)": "32a9a1aba9",
+        "Stateful tests (coverage)": "32a9a1aba9",
+        "Stateful tests (aarch64)": "32a9a1aba9",
+        "Stateful tests (release, ParallelReplicas)": "32a9a1aba9",
+        "Stateful tests (debug, ParallelReplicas)": "32a9a1aba9",
+        "Stateless tests (asan)": "deb6778b88",
+        "Stateless tests (tsan)": "deb6778b88",
+        "Stateless tests (msan)": "deb6778b88",
+        "Stateless tests (ubsan)": "deb6778b88",
+        "Stateless tests (debug)": "deb6778b88",
+        "Stateless tests (release)": "deb6778b88",
+        "Stateless tests (coverage)": "deb6778b88",
+        "Stateless tests (aarch64)": "deb6778b88",
+        "Stateless tests (release, old analyzer, s3, DatabaseReplicated)": "deb6778b88",
+        "Stateless tests (debug, s3 storage)": "deb6778b88",
+        "Stateless tests (tsan, s3 storage)": "deb6778b88",
+        "Stress test (debug)": "aa298abf10",
+        "Stress test (tsan)": "aa298abf10",
+        "Upgrade check (debug)": "5ce4d3ee02",
+        "Integration tests (asan, old analyzer)": "42e58be3aa",
+        "Integration tests (tsan)": "42e58be3aa",
+        "Integration tests (aarch64)": "42e58be3aa",
+        "Integration tests flaky check (asan)": "42e58be3aa",
+        "Compatibility check (release)": "ecb69d8c4b",
+        "Compatibility check (aarch64)": "ecb69d8c4b",
+        "Unit tests (release)": "09d00b702e",
+        "Unit tests (asan)": "09d00b702e",
+        "Unit tests (msan)": "09d00b702e",
+        "Unit tests (tsan)": "09d00b702e",
+        "Unit tests (ubsan)": "09d00b702e",
+        "AST fuzzer (debug)": "c38ebf947f",
+        "AST fuzzer (asan)": "c38ebf947f",
+        "AST fuzzer (msan)": "c38ebf947f",
+        "AST fuzzer (tsan)": "c38ebf947f",
+        "AST fuzzer (ubsan)": "c38ebf947f",
+        "Stateless tests flaky check (asan)": "deb6778b88",
+        "Performance Comparison (release)": "a8a7179258",
+        "ClickBench (release)": "45c07c4aa6",
+        "ClickBench (aarch64)": "45c07c4aa6",
+        "Docker server image": "6a24d5b187",
+        "Docker keeper image": "6a24d5b187",
+        "Docs check": "4764154c62",
+        "Fast test": "cb269133f2",
+        "Style check": "ffffffffff",
+        "Stateful tests (ubsan, ParallelReplicas)": "32a9a1aba9",
+        "Stress test (msan)": "aa298abf10",
+        "Upgrade check (asan)": "5ce4d3ee02",
+    }
+    ci_cache = CiCache(job_digests=job_digest, cache_enabled=True, s3=S3Helper())
+    ci_cache.update()

tests/ci/ci_config.py

@@ -32,6 +32,9 @@ class CI:
     from ci_definitions import MQ_JOBS as MQ_JOBS
     from ci_definitions import WorkflowStages as WorkflowStages
     from ci_definitions import Runners as Runners
+    from ci_utils import Envs as Envs
+    from ci_utils import Utils as Utils
+    from ci_utils import GHActions as GHActions
     from ci_definitions import Labels as Labels
     from ci_definitions import TRUSTED_CONTRIBUTORS as TRUSTED_CONTRIBUTORS
     from ci_utils import CATEGORY_TO_LABEL as CATEGORY_TO_LABEL

tests/ci/ci_definitions.py

@@ -351,6 +351,8 @@ class JobConfig:
     run_by_label: str = ""
     # to run always regardless of the job digest or/and label
     run_always: bool = False
+    # disables CI await for a given job
+    disable_await: bool = False
     # if the job needs to be run on the release branch, including master (building packages, docker server).
     # NOTE: Subsequent runs on the same branch with the similar digest are still considered skip-able.
     required_on_release_branch: bool = False
@@ -395,6 +397,7 @@ class CommonJobConfigs:
             ],
         ),
         runner_type=Runners.STYLE_CHECKER_ARM,
+        disable_await=True,
     )
     COMPATIBILITY_TEST = JobConfig(
         job_name_keyword="compatibility",

tests/ci/ci_utils.py

@@ -1,9 +1,16 @@
 import os
 import re
 import subprocess
+import time
 from contextlib import contextmanager
 from pathlib import Path
-from typing import Any, Iterator, List, Union, Optional, Tuple
+from typing import Any, Iterator, List, Union, Optional, Sequence, Tuple
+
+import requests
+
+
+class Envs:
+    GITHUB_REPOSITORY = os.getenv("GITHUB_REPOSITORY", "ClickHouse/ClickHouse")
 
 
 LABEL_CATEGORIES = {
@@ -80,6 +87,71 @@ class GHActions:
             print(line)
         print("::endgroup::")
 
+    @staticmethod
+    def get_commit_status_by_name(
+        token: str, commit_sha: str, status_name: Union[str, Sequence]
+    ) -> str:
+        assert len(token) == 40
+        assert len(commit_sha) == 40
+        assert is_hex(commit_sha)
+        assert not is_hex(token)
+        url = f"https://api.github.com/repos/{Envs.GITHUB_REPOSITORY}/commits/{commit_sha}/statuses?per_page={200}"
+        headers = {
+            "Authorization": f"token {token}",
+            "Accept": "application/vnd.github.v3+json",
+        }
+        response = requests.get(url, headers=headers, timeout=5)
+
+        if isinstance(status_name, str):
+            status_name = (status_name,)
+        if response.status_code == 200:
+            assert "next" not in response.links, "Response truncated"
+            statuses = response.json()
+            for status in statuses:
+                if status["context"] in status_name:
+                    return status["state"]  # type: ignore
+        return ""
+
+    @staticmethod
+    def check_wf_completed(token: str, commit_sha: str) -> bool:
+        headers = {
+            "Authorization": f"token {token}",
+            "Accept": "application/vnd.github.v3+json",
+        }
+        url = f"https://api.github.com/repos/{Envs.GITHUB_REPOSITORY}/commits/{commit_sha}/check-runs?per_page={100}"
+
+        for i in range(3):
+            try:
+                response = requests.get(url, headers=headers, timeout=5)
+                response.raise_for_status()
+                # assert "next" not in response.links, "Response truncated"
+
+                data = response.json()
+                assert data["check_runs"], "?"
+
+                for check in data["check_runs"]:
+                    if check["status"] != "completed":
+                        print(
+                            f"   Check workflow status: Check not completed [{check['name']}]"
+                        )
+                        return False
+                return True
+            except Exception as e:
+                print(f"ERROR: exception after attempt [{i}]: {e}")
+                time.sleep(1)
+
+        return False
+
+    @staticmethod
+    def get_pr_url_by_branch(repo, branch):
+        get_url_cmd = (
+            f"gh pr list --repo {repo} --head {branch} --json url --jq '.[0].url'"
+        )
+        url = Shell.run(get_url_cmd)
+        if not url:
+            print(f"ERROR: PR nor found, branch [{branch}]")
+        return url
+
 
 class Shell:
     @classmethod
@@ -95,7 +167,11 @@ class Shell:
         return res.stdout.strip()
 
     @classmethod
-    def run(cls, command):
+    def run(cls, command, check=False, dry_run=False):
+        if dry_run:
+            print(f"Dry-ryn. Would run command [{command}]")
+            return ""
+        print(f"Run command [{command}]")
         res = ""
         result = subprocess.run(
             command,
@@ -106,13 +182,26 @@ class Shell:
             check=False,
         )
         if result.returncode == 0:
+            print(f"stdout: {result.stdout.strip()}")
             res = result.stdout
+        else:
+            print(
+                f"ERROR: stdout: {result.stdout.strip()}, stderr: {result.stderr.strip()}"
+            )
+            if check:
+                assert result.returncode == 0
         return res.strip()
 
+    @classmethod
+    def run_as_daemon(cls, command):
+        print(f"Run daemon command [{command}]")
+        subprocess.Popen(command.split(" "))  # pylint:disable=consider-using-with
+        return 0, ""
+
     @classmethod
     def check(cls, command):
         result = subprocess.run(
-            command + " 2>&1",
+            command,
             shell=True,
             stdout=subprocess.PIPE,
             stderr=subprocess.PIPE,

tests/ci/create_release.py

@@ -2,7 +2,6 @@ import argparse
 import dataclasses
 import json
 import os
-import subprocess
 
 from contextlib import contextmanager
 from copy import copy
@@ -13,7 +12,8 @@ from git_helper import Git, GIT_PREFIX
 from ssh import SSHAgent
 from env_helper import GITHUB_REPOSITORY, S3_BUILDS_BUCKET
 from s3_helper import S3Helper
-from ci_utils import Shell
+from ci_utils import Shell, GHActions
+from ci_buddy import CIBuddy
 from version_helper import (
     FILE_WITH_VERSION_PATH,
     GENERATED_CONTRIBUTORS,
@@ -27,34 +27,66 @@ from ci_config import CI
 
 CMAKE_PATH = get_abs_path(FILE_WITH_VERSION_PATH)
 CONTRIBUTORS_PATH = get_abs_path(GENERATED_CONTRIBUTORS)
+RELEASE_INFO_FILE = "/tmp/release_info.json"
 
 
-class ShellRunner:
+class ReleaseProgress:
+    STARTED = "started"
+    DOWNLOAD_PACKAGES = "download packages"
+    PUSH_RELEASE_TAG = "push release tag"
+    PUSH_NEW_RELEASE_BRANCH = "push new release branch"
+    BUMP_VERSION = "bump version"
+    CREATE_GH_RELEASE = "create GH release"
+    EXPORT_TGZ = "export TGZ packages"
+    EXPORT_RPM = "export RPM packages"
+    EXPORT_DEB = "export DEB packages"
+    TEST_TGZ = "test TGZ packages"
+    TEST_RPM = "test RPM packages"
+    TEST_DEB = "test DEB packages"
+    COMPLETED = "completed"
 
-    @classmethod
+
-    def run(
+class ReleaseProgressDescription:
-        cls, command, check_retcode=True, print_output=True, async_=False, dry_run=False
+    OK = "OK"
-    ):
+    FAILED = "FAILED"
-        if dry_run:
+
-            print(f"Dry-run: Would run shell command: [{command}]")
+
-            return 0, ""
+class ReleaseContextManager:
-        print(f"Running shell command: [{command}]")
+    def __init__(self, release_progress):
-        if async_:
+        self.release_progress = release_progress
-            subprocess.Popen(command.split(" "))  # pylint:disable=consider-using-with
+        self.release_info = None
-            return 0, ""
+
-        result = subprocess.run(
+    def __enter__(self):
-            command + " 2>&1",
+        if self.release_progress == ReleaseProgress.STARTED:
-            shell=True,
+            # create initial release info
-            stdout=subprocess.PIPE,
+            self.release_info = ReleaseInfo(
-            stderr=subprocess.PIPE,
+                release_branch="NA",
-            text=True,
+                commit_sha=args.ref,
-            check=True,
+                release_tag="NA",
-        )
+                version="NA",
-        if print_output:
+                codename="NA",
-            print(result.stdout)
+                previous_release_tag="NA",
-        if check_retcode:
+                previous_release_sha="NA",
-            assert result.returncode == 0, f"Return code [{result.returncode}]"
+                release_progress=ReleaseProgress.STARTED,
-        return result.returncode, result.stdout
+            ).dump()
+        else:
+            # fetch release info from fs and update
+            self.release_info = ReleaseInfo.from_file()
+            assert self.release_info
+            assert (
+                self.release_info.progress_description == ReleaseProgressDescription.OK
+            ), "Must be OK on the start of new context"
+            self.release_info.release_progress = self.release_progress
+            self.release_info.dump()
+        return self.release_info
+
+    def __exit__(self, exc_type, exc_value, traceback):
+        assert self.release_info
+        if exc_type is not None:
+            self.release_info.progress_description = ReleaseProgressDescription.FAILED
+        else:
+            self.release_info.progress_description = ReleaseProgressDescription.OK
+        self.release_info.dump()
 
 
 @dataclasses.dataclass
@@ -67,31 +99,50 @@ class ReleaseInfo:
     codename: str
     previous_release_tag: str
     previous_release_sha: str
+    changelog_pr: str = ""
+    version_bump_pr: str = ""
+    release_url: str = ""
+    debian_command: str = ""
+    rpm_command: str = ""
+    tgz_command: str = ""
+    docker_command: str = ""
+    release_progress: str = ""
+    progress_description: str = ""
+
+    def is_patch(self):
+        return self.release_branch != "master"
+
+    def is_new_release_branch(self):
+        return self.release_branch == "master"
 
     @staticmethod
-    def from_file(file_path: str) -> "ReleaseInfo":
+    def from_file() -> "ReleaseInfo":
-        with open(file_path, "r", encoding="utf-8") as json_file:
+        with open(RELEASE_INFO_FILE, "r", encoding="utf-8") as json_file:
             res = json.load(json_file)
         return ReleaseInfo(**res)
 
-    @staticmethod
+    def dump(self):
-    def prepare(commit_ref: str, release_type: str, outfile: str) -> None:
+        print(f"Dump release info into [{RELEASE_INFO_FILE}]")
-        Path(outfile).parent.mkdir(parents=True, exist_ok=True)
+        with open(RELEASE_INFO_FILE, "w", encoding="utf-8") as f:
-        Path(outfile).unlink(missing_ok=True)
+            print(json.dumps(dataclasses.asdict(self), indent=2), file=f)
+        return self
+
+    def prepare(self, commit_ref: str, release_type: str) -> "ReleaseInfo":
         version = None
         release_branch = None
         release_tag = None
         previous_release_tag = None
         previous_release_sha = None
-        codename = None
+        codename = ""
         assert release_type in ("patch", "new")
         if release_type == "new":
             # check commit_ref is right and on a right branch
-            ShellRunner.run(
+            Shell.run(
-                f"git merge-base --is-ancestor origin/{commit_ref} origin/master"
+                f"git merge-base --is-ancestor {commit_ref} origin/master",
+                check=True,
             )
             with checkout(commit_ref):
-                _, commit_sha = ShellRunner.run(f"git rev-parse {commit_ref}")
+                commit_sha = Shell.run(f"git rev-parse {commit_ref}", check=True)
                 # Git() must be inside "with checkout" contextmanager
                 git = Git()
                 version = get_version_from_repo(git=git)
@@ -102,9 +153,6 @@ class ReleaseInfo:
                     git.latest_tag == expected_prev_tag
                 ), f"BUG: latest tag [{git.latest_tag}], expected [{expected_prev_tag}]"
                 release_tag = version.describe
-                codename = (
-                    VersionType.STABLE
-                )  # dummy value (artifactory won't be updated for new release)
                 previous_release_tag = expected_prev_tag
                 previous_release_sha = Shell.run_strict(
                     f"git rev-parse {previous_release_tag}"
@@ -112,7 +160,7 @@ class ReleaseInfo:
                 assert previous_release_sha
         if release_type == "patch":
             with checkout(commit_ref):
-                _, commit_sha = ShellRunner.run(f"git rev-parse {commit_ref}")
+                commit_sha = Shell.run(f"git rev-parse {commit_ref}", check=True)
                 # Git() must be inside "with checkout" contextmanager
                 git = Git()
                 version = get_version_from_repo(git=git)
@@ -120,10 +168,11 @@ class ReleaseInfo:
                 version.with_description(codename)
                 release_branch = f"{version.major}.{version.minor}"
                 release_tag = version.describe
-            ShellRunner.run(f"{GIT_PREFIX} fetch origin {release_branch} --tags")
+            Shell.run(f"{GIT_PREFIX} fetch origin {release_branch} --tags", check=True)
             # check commit is right and on a right branch
-            ShellRunner.run(
+            Shell.run(
-                f"git merge-base --is-ancestor {commit_ref} origin/{release_branch}"
+                f"git merge-base --is-ancestor {commit_ref} origin/{release_branch}",
+                check=True,
             )
             if version.patch == 1:
                 expected_version = copy(version)
@@ -160,24 +209,24 @@ class ReleaseInfo:
             and commit_sha
             and release_tag
             and version
-            and codename in ("lts", "stable")
+            and (codename in ("lts", "stable") or release_type == "new")
         )
-        res = ReleaseInfo(
+
-            release_branch=release_branch,
+        self.release_branch = release_branch
-            commit_sha=commit_sha,
+        self.commit_sha = commit_sha
-            release_tag=release_tag,
+        self.release_tag = release_tag
-            version=version.string,
+        self.version = version.string
-            codename=codename,
+        self.codename = codename
-            previous_release_tag=previous_release_tag,
+        self.previous_release_tag = previous_release_tag
-            previous_release_sha=previous_release_sha,
+        self.previous_release_sha = previous_release_sha
-        )
+        self.release_progress = ReleaseProgress.STARTED
-        with open(outfile, "w", encoding="utf-8") as f:
+        self.progress_description = ReleaseProgressDescription.OK
-            print(json.dumps(dataclasses.asdict(res), indent=2), file=f)
+        return self
 
     def push_release_tag(self, dry_run: bool) -> None:
         if dry_run:
             # remove locally created tag from prev run
-            ShellRunner.run(
+            Shell.run(
                 f"{GIT_PREFIX} tag -l | grep -q {self.release_tag} && git tag -d {self.release_tag} ||:"
             )
         # Create release tag
@@ -185,16 +234,17 @@ class ReleaseInfo:
             f"Create and push release tag [{self.release_tag}], commit [{self.commit_sha}]"
         )
         tag_message = f"Release {self.release_tag}"
-        ShellRunner.run(
+        Shell.run(
-            f"{GIT_PREFIX} tag -a -m '{tag_message}' {self.release_tag} {self.commit_sha}"
+            f"{GIT_PREFIX} tag -a -m '{tag_message}' {self.release_tag} {self.commit_sha}",
+            check=True,
         )
         cmd_push_tag = f"{GIT_PREFIX} push origin {self.release_tag}:{self.release_tag}"
-        ShellRunner.run(cmd_push_tag, dry_run=dry_run)
+        Shell.run(cmd_push_tag, dry_run=dry_run, check=True)
 
     @staticmethod
     def _create_gh_label(label: str, color_hex: str, dry_run: bool) -> None:
         cmd = f"gh api repos/{GITHUB_REPOSITORY}/labels -f name={label} -f color={color_hex}"
-        ShellRunner.run(cmd, dry_run=dry_run)
+        Shell.run(cmd, dry_run=dry_run, check=True)
 
     def push_new_release_branch(self, dry_run: bool) -> None:
         assert (
@@ -211,8 +261,8 @@ class ReleaseInfo:
         ), f"Unexpected current version in git, must precede [{self.version}] by one step, actual [{version.string}]"
         if dry_run:
             # remove locally created branch from prev run
-            ShellRunner.run(
+            Shell.run(
-                f"{GIT_PREFIX} branch -l | grep -q {new_release_branch} && git branch -d {new_release_branch} ||:"
+                f"{GIT_PREFIX} branch -l | grep -q {new_release_branch} && git branch -d {new_release_branch}"
             )
         print(
             f"Create and push new release branch [{new_release_branch}], commit [{self.commit_sha}]"
@@ -225,7 +275,7 @@ class ReleaseInfo:
                 cmd_push_branch = (
                     f"{GIT_PREFIX} push --set-upstream origin {new_release_branch}"
                 )
-                ShellRunner.run(cmd_push_branch, dry_run=dry_run)
+                Shell.run(cmd_push_branch, dry_run=dry_run, check=True)
 
         print("Create and push backport tags for new release branch")
         ReleaseInfo._create_gh_label(
@@ -234,12 +284,13 @@ class ReleaseInfo:
         ReleaseInfo._create_gh_label(
             f"v{new_release_branch}-affected", "c2bfff", dry_run=dry_run
         )
-        ShellRunner.run(
+        Shell.run(
             f"""gh pr create --repo {GITHUB_REPOSITORY} --title 'Release pull request for branch {new_release_branch}'
             --head {new_release_branch} {pr_labels}
             --body 'This PullRequest is a part of ClickHouse release cycle. It is used by CI system only. Do not perform any changes with it.'
             """,
             dry_run=dry_run,
+            check=True,
         )
 
     def update_version_and_contributors_list(self, dry_run: bool) -> None:
@@ -265,32 +316,55 @@ class ReleaseInfo:
                 body_file = get_abs_path(".github/PULL_REQUEST_TEMPLATE.md")
                 actor = os.getenv("GITHUB_ACTOR", "") or "me"
                 cmd_create_pr = f"gh pr create --repo {GITHUB_REPOSITORY} --title 'Update version after release' --head {branch_upd_version_contributors} --base {self.release_branch} --body-file '{body_file} --label 'do not test' --assignee @{actor}"
-                ShellRunner.run(cmd_commit_version_upd, dry_run=dry_run)
+                Shell.run(cmd_commit_version_upd, check=True, dry_run=dry_run)
-                ShellRunner.run(cmd_push_branch, dry_run=dry_run)
+                Shell.run(cmd_push_branch, check=True, dry_run=dry_run)
-                ShellRunner.run(cmd_create_pr, dry_run=dry_run)
+                Shell.run(cmd_create_pr, check=True, dry_run=dry_run)
                 if dry_run:
-                    ShellRunner.run(
+                    Shell.run(f"{GIT_PREFIX} diff '{CMAKE_PATH}' '{CONTRIBUTORS_PATH}'")
-                        f"{GIT_PREFIX} diff '{CMAKE_PATH}' '{CONTRIBUTORS_PATH}'"
+                    Shell.run(
-                    )
-                    ShellRunner.run(
                         f"{GIT_PREFIX} checkout '{CMAKE_PATH}' '{CONTRIBUTORS_PATH}'"
                     )
+                    self.version_bump_pr = "dry-run"
+                else:
+                    self.version_bump_pr = GHActions.get_pr_url_by_branch(
+                        repo=GITHUB_REPOSITORY, branch=branch_upd_version_contributors
+                    )
+
+    def update_release_info(self, dry_run: bool) -> "ReleaseInfo":
+        if self.release_branch != "master":
+            branch = f"auto/{release_info.release_tag}"
+            if not dry_run:
+                url = GHActions.get_pr_url_by_branch(
+                    repo=GITHUB_REPOSITORY, branch=branch
+                )
+            else:
+                url = "dry-run"
+            print(f"ChangeLog PR url [{url}]")
+            self.changelog_pr = url
+            print(f"Release url [{url}]")
+            self.release_url = f"https://github.com/{GITHUB_REPOSITORY}/releases/tag/{self.release_tag}"
+            if self.release_progress == ReleaseProgress.COMPLETED:
+                self.docker_command = f"docker run --rm clickhouse/clickhouse:{self.version} clickhouse --version"
+        self.dump()
+        return self
 
     def create_gh_release(self, packages_files: List[str], dry_run: bool) -> None:
         repo = os.getenv("GITHUB_REPOSITORY")
         assert repo
-        cmds = []
+        cmds = [
-        cmds.append(
             f"gh release create --repo {repo} --title 'Release {self.release_tag}' {self.release_tag}"
-        )
+        ]
         for file in packages_files:
             cmds.append(f"gh release upload {self.release_tag} {file}")
         if not dry_run:
             for cmd in cmds:
-                ShellRunner.run(cmd)
+                Shell.run(cmd, check=True)
+            self.release_url = f"https://github.com/{GITHUB_REPOSITORY}/releases/tag/{self.release_tag}"
         else:
             print("Dry-run, would run commands:")
             print("\n  * ".join(cmds))
+            self.release_url = f"dry-run"
+        self.dump()
 
 
 class RepoTypes:
@@ -350,7 +424,7 @@ class PackageDownloader:
         self.macos_package_files = ["clickhouse-macos", "clickhouse-macos-aarch64"]
         self.file_to_type = {}
 
-        ShellRunner.run(f"mkdir -p {self.LOCAL_DIR}")
+        Shell.run(f"mkdir -p {self.LOCAL_DIR}")
 
         for package_type in self.PACKAGE_TYPES:
             for package in self.package_names:
@@ -400,7 +474,7 @@ class PackageDownloader:
         return res
 
     def run(self):
-        ShellRunner.run(f"rm -rf {self.LOCAL_DIR}/*")
+        Shell.run(f"rm -rf {self.LOCAL_DIR}/*")
         for package_file in (
             self.deb_package_files + self.rpm_package_files + self.tgz_package_files
         ):
@@ -473,6 +547,37 @@ class PackageDownloader:
         return True
 
 
+@contextmanager
+def checkout(ref: str) -> Iterator[None]:
+    orig_ref = Shell.run(f"{GIT_PREFIX} symbolic-ref --short HEAD", check=True)
+    rollback_cmd = f"{GIT_PREFIX} checkout {orig_ref}"
+    assert orig_ref
+    if ref not in (orig_ref,):
+        Shell.run(f"{GIT_PREFIX} checkout {ref}")
+    try:
+        yield
+    except (Exception, KeyboardInterrupt) as e:
+        print(f"ERROR: Exception [{e}]")
+        Shell.run(rollback_cmd)
+        raise
+    Shell.run(rollback_cmd)
+
+
+@contextmanager
+def checkout_new(ref: str) -> Iterator[None]:
+    orig_ref = Shell.run(f"{GIT_PREFIX} symbolic-ref --short HEAD", check=True)
+    rollback_cmd = f"{GIT_PREFIX} checkout {orig_ref}"
+    assert orig_ref
+    Shell.run(f"{GIT_PREFIX} checkout -b {ref}", check=True)
+    try:
+        yield
+    except (Exception, KeyboardInterrupt) as e:
+        print(f"ERROR: Exception [{e}]")
+        Shell.run(rollback_cmd)
+        raise
+    Shell.run(rollback_cmd)
+
+
 def parse_args() -> argparse.Namespace:
     parser = argparse.ArgumentParser(
         formatter_class=argparse.ArgumentDefaultsHelpFormatter,
@@ -508,6 +613,11 @@ def parse_args() -> argparse.Namespace:
         action="store_true",
         help="Create GH Release object and attach all packages",
     )
+    parser.add_argument(
+        "--post-status",
+        action="store_true",
+        help="Post release status into Slack",
+    )
     parser.add_argument(
         "--ref",
         type=str,
@@ -526,55 +636,25 @@ def parse_args() -> argparse.Namespace:
         help="do not make any actual changes in the repo, just show what will be done",
     )
     parser.add_argument(
-        "--outfile",
+        "--set-progress-started",
-        default="",
+        action="store_true",
-        type=str,
+        help="Set new progress step, --progress <PROGRESS STEP> must be set",
-        help="output file to write json result to, if not set - stdout",
     )
     parser.add_argument(
-        "--infile",
+        "--progress",
-        default="",
         type=str,
-        help="input file with release info",
+        help="Progress step name, see @ReleaseProgress",
+    )
+    parser.add_argument(
+        "--set-progress-completed",
+        action="store_true",
+        help="Set current progress step to OK (completed)",
     )
-
     return parser.parse_args()
 
 
-@contextmanager
-def checkout(ref: str) -> Iterator[None]:
-    _, orig_ref = ShellRunner.run(f"{GIT_PREFIX} symbolic-ref --short HEAD")
-    rollback_cmd = f"{GIT_PREFIX} checkout {orig_ref}"
-    assert orig_ref
-    if ref not in (orig_ref,):
-        ShellRunner.run(f"{GIT_PREFIX} checkout {ref}")
-    try:
-        yield
-    except (Exception, KeyboardInterrupt) as e:
-        print(f"ERROR: Exception [{e}]")
-        ShellRunner.run(rollback_cmd)
-        raise
-    ShellRunner.run(rollback_cmd)
-
-
-@contextmanager
-def checkout_new(ref: str) -> Iterator[None]:
-    _, orig_ref = ShellRunner.run(f"{GIT_PREFIX} symbolic-ref --short HEAD")
-    rollback_cmd = f"{GIT_PREFIX} checkout {orig_ref}"
-    assert orig_ref
-    ShellRunner.run(f"{GIT_PREFIX} checkout -b {ref}")
-    try:
-        yield
-    except (Exception, KeyboardInterrupt) as e:
-        print(f"ERROR: Exception [{e}]")
-        ShellRunner.run(rollback_cmd)
-        raise
-    ShellRunner.run(rollback_cmd)
-
-
 if __name__ == "__main__":
     args = parse_args()
-    assert args.dry_run
 
     # prepare ssh for git if needed
     _ssh_agent = None
@@ -586,43 +666,91 @@ if __name__ == "__main__":
         _ssh_agent.print_keys()
 
     if args.prepare_release_info:
-        assert (
+        with ReleaseContextManager(
-            args.ref and args.release_type and args.outfile
+            release_progress=ReleaseProgress.STARTED
-        ), "--ref, --release-type and --outfile must be provided with --prepare-release-info"
+        ) as release_info:
-        ReleaseInfo.prepare(
+            assert (
-            commit_ref=args.ref, release_type=args.release_type, outfile=args.outfile
+                args.ref and args.release_type
-        )
+            ), "--ref and --release-type must be provided with --prepare-release-info"
-    if args.push_release_tag:
+            release_info.prepare(commit_ref=args.ref, release_type=args.release_type)
-        assert args.infile, "--infile <release info file path> must be provided"
+
-        release_info = ReleaseInfo.from_file(args.infile)
-        release_info.push_release_tag(dry_run=args.dry_run)
-    if args.push_new_release_branch:
-        assert args.infile, "--infile <release info file path> must be provided"
-        release_info = ReleaseInfo.from_file(args.infile)
-        release_info.push_new_release_branch(dry_run=args.dry_run)
-    if args.create_bump_version_pr:
-        # TODO: store link to PR in release info
-        assert args.infile, "--infile <release info file path> must be provided"
-        release_info = ReleaseInfo.from_file(args.infile)
-        release_info.update_version_and_contributors_list(dry_run=args.dry_run)
     if args.download_packages:
-        assert args.infile, "--infile <release info file path> must be provided"
+        with ReleaseContextManager(
-        release_info = ReleaseInfo.from_file(args.infile)
+            release_progress=ReleaseProgress.DOWNLOAD_PACKAGES
-        p = PackageDownloader(
+        ) as release_info:
-            release=release_info.release_branch,
+            p = PackageDownloader(
-            commit_sha=release_info.commit_sha,
+                release=release_info.release_branch,
-            version=release_info.version,
+                commit_sha=release_info.commit_sha,
-        )
+                version=release_info.version,
-        p.run()
+            )
+            p.run()
+
+    if args.push_release_tag:
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.PUSH_RELEASE_TAG
+        ) as release_info:
+            release_info.push_release_tag(dry_run=args.dry_run)
+
+    if args.push_new_release_branch:
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.PUSH_NEW_RELEASE_BRANCH
+        ) as release_info:
+            release_info.push_new_release_branch(dry_run=args.dry_run)
+
+    if args.create_bump_version_pr:
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.BUMP_VERSION
+        ) as release_info:
+            release_info.update_version_and_contributors_list(dry_run=args.dry_run)
+
     if args.create_gh_release:
-        assert args.infile, "--infile <release info file path> must be provided"
+        with ReleaseContextManager(
-        release_info = ReleaseInfo.from_file(args.infile)
+            release_progress=ReleaseProgress.CREATE_GH_RELEASE
-        p = PackageDownloader(
+        ) as release_info:
-            release=release_info.release_branch,
+            p = PackageDownloader(
-            commit_sha=release_info.commit_sha,
+                release=release_info.release_branch,
-            version=release_info.version,
+                commit_sha=release_info.commit_sha,
-        )
+                version=release_info.version,
-        release_info.create_gh_release(p.get_all_packages_files(), args.dry_run)
+            )
+            release_info.create_gh_release(
+                packages_files=p.get_all_packages_files(), dry_run=args.dry_run
+            )
+
+    if args.post_status:
+        release_info = ReleaseInfo.from_file()
+        release_info.update_release_info(dry_run=args.dry_run)
+        if release_info.is_new_release_branch():
+            title = "New release branch"
+        else:
+            title = "New release"
+        if (
+            release_info.progress_description == ReleaseProgressDescription.OK
+            and release_info.release_progress == ReleaseProgress.COMPLETED
+        ):
+            title = "Completed: " + title
+            CIBuddy(dry_run=args.dry_run).post_done(
+                title, dataclasses.asdict(release_info)
+            )
+        else:
+            title = "Failed: " + title
+            CIBuddy(dry_run=args.dry_run).post_critical(
+                title, dataclasses.asdict(release_info)
+            )
+
+    if args.set_progress_started:
+        ri = ReleaseInfo.from_file()
+        ri.release_progress = args.progress
+        ri.progress_description = ReleaseProgressDescription.FAILED
+        ri.dump()
+        assert args.progress, "Progress step name must be provided"
+
+    if args.set_progress_completed:
+        ri = ReleaseInfo.from_file()
+        assert (
+            ri.progress_description == ReleaseProgressDescription.FAILED
+        ), "Must be FAILED before set to OK"
+        ri.progress_description = ReleaseProgressDescription.OK
+        ri.dump()
 
     # tear down ssh
     if _ssh_agent and _key_pub:

tests/ci/merge_pr.py

@@ -254,11 +254,14 @@ def main():
         statuses = get_commit_filtered_statuses(commit)
 
         has_failed_statuses = False
+        has_native_failed_status = False
         for status in statuses:
             print(f"Check status [{status.context}], [{status.state}]")
             if CI.is_required(status.context) and status.state != SUCCESS:
                 print(f"WARNING: Failed status [{status.context}], [{status.state}]")
                 has_failed_statuses = True
+                if status.context != CI.StatusNames.SYNC:
+                    has_native_failed_status = True
 
         if args.wf_status == SUCCESS or has_failed_statuses:
             # set Mergeable check if workflow is successful (green)
@@ -280,7 +283,7 @@ def main():
             print(
                 "Workflow failed but no failed statuses found (died runner?) - cannot set Mergeable Check status"
             )
-        if args.wf_status == SUCCESS and not has_failed_statuses:
+        if args.wf_status == SUCCESS and not has_native_failed_status:
             sys.exit(0)
         else:
             sys.exit(1)

tests/ci/pr_info.py

@@ -296,13 +296,16 @@ class PRInfo:
         else:
             if "schedule" in github_event:
                 self.event_type = EventType.SCHEDULE
-            else:
+            elif "inputs" in github_event:
                 # assume this is a dispatch
                 self.event_type = EventType.DISPATCH
-            logging.warning(
+                print("PR Info:")
-                "event.json does not match pull_request or push:\n%s",
+                print(self)
-                json.dumps(github_event, sort_keys=True, indent=4),
+            else:
-            )
+                logging.warning(
+                    "event.json does not match pull_request or push:\n%s",
+                    json.dumps(github_event, sort_keys=True, indent=4),
+                )
             self.sha = os.getenv(
                 "GITHUB_SHA", "0000000000000000000000000000000000000000"
             )

tests/ci/test_ci_config.py

@@ -587,11 +587,11 @@ class TestCIConfig(unittest.TestCase):
         for job, job_config in ci_cache.jobs_to_do.items():
             if job in MOCK_AFFECTED_JOBS:
                 MOCK_REQUIRED_BUILDS += job_config.required_builds
-            elif job not in MOCK_AFFECTED_JOBS:
+            elif job not in MOCK_AFFECTED_JOBS and not job_config.disable_await:
                 ci_cache.jobs_to_wait[job] = job_config
 
         for job, job_config in ci_cache.jobs_to_do.items():
-            if job_config.reference_job_name:
+            if job_config.reference_job_name or job_config.disable_await:
                 # jobs with reference_job_name in config are not supposed to have records in the cache - continue
                 continue
             if job in MOCK_AFFECTED_JOBS:
@@ -624,11 +624,76 @@ class TestCIConfig(unittest.TestCase):
             + MOCK_AFFECTED_JOBS
             + MOCK_REQUIRED_BUILDS
         )
+        self.assertTrue(
+            CI.JobNames.BUILD_CHECK not in ci_cache.jobs_to_wait,
+            "We must never await on Builds Report",
+        )
         self.assertCountEqual(
             list(ci_cache.jobs_to_wait),
-            [
+            MOCK_REQUIRED_BUILDS,
-                CI.JobNames.BUILD_CHECK,
+        )
-            ]
+        self.assertCountEqual(list(ci_cache.jobs_to_do), expected_to_do)
-            + MOCK_REQUIRED_BUILDS,
+
+    def test_ci_py_filters_not_affected_jobs_in_prs_no_builds(self):
+        """
+        checks ci.py filters not affected jobs in PRs, no builds required
+        """
+        settings = CiSettings()
+        settings.no_ci_cache = True
+        pr_info = PRInfo(github_event=_TEST_EVENT_JSON)
+        pr_info.event_type = EventType.PULL_REQUEST
+        pr_info.number = 123
+        assert pr_info.is_pr
+        ci_cache = CIPY._configure_jobs(
+            S3Helper(), pr_info, settings, skip_jobs=False, dry_run=True
+        )
+        self.assertTrue(not ci_cache.jobs_to_skip, "Must be no jobs in skip list")
+        assert not ci_cache.jobs_to_wait
+        assert not ci_cache.jobs_to_skip
+
+        MOCK_AFFECTED_JOBS = [
+            CI.JobNames.FAST_TEST,
+        ]
+        MOCK_REQUIRED_BUILDS = []
+
+        # pretend there are pending jobs that we need to wait
+        for job, job_config in ci_cache.jobs_to_do.items():
+            if job in MOCK_AFFECTED_JOBS:
+                if job_config.required_builds:
+                    MOCK_REQUIRED_BUILDS += job_config.required_builds
+            elif job not in MOCK_AFFECTED_JOBS and not job_config.disable_await:
+                ci_cache.jobs_to_wait[job] = job_config
+
+        for job, job_config in ci_cache.jobs_to_do.items():
+            if job_config.reference_job_name or job_config.disable_await:
+                # jobs with reference_job_name in config are not supposed to have records in the cache - continue
+                continue
+            if job in MOCK_AFFECTED_JOBS:
+                continue
+            for batch in range(job_config.num_batches):
+                # add any record into cache
+                record = CiCache.Record(
+                    record_type=random.choice(
+                        [
+                            CiCache.RecordType.FAILED,
+                            CiCache.RecordType.PENDING,
+                            CiCache.RecordType.SUCCESSFUL,
+                        ]
+                    ),
+                    job_name=job,
+                    job_digest=ci_cache.job_digests[job],
+                    batch=batch,
+                    num_batches=job_config.num_batches,
+                    release_branch=True,
+                )
+                for record_t_, records_ in ci_cache.records.items():
+                    if record_t_.value == CiCache.RecordType.FAILED.value:
+                        records_[record.to_str_key()] = record
+
+        ci_cache.filter_out_not_affected_jobs()
+        expected_to_do = MOCK_AFFECTED_JOBS + MOCK_REQUIRED_BUILDS
+        self.assertCountEqual(
+            list(ci_cache.jobs_to_wait),
+            MOCK_REQUIRED_BUILDS,
         )
         self.assertCountEqual(list(ci_cache.jobs_to_do), expected_to_do)

tests/ci/worker/init_runner.sh

@@ -50,7 +50,7 @@ set -uo pipefail
 # set accordingly to a runner role #
 ####################################
 
-echo "Running init v1"
+echo "Running init v1.1"
 export DEBIAN_FRONTEND=noninteractive
 export RUNNER_HOME=/home/ubuntu/actions-runner
 
@@ -66,6 +66,14 @@ bash /usr/local/share/scripts/init-network.sh
 RUNNER_TYPE=$(/usr/local/bin/aws ec2 describe-tags --filters "Name=resource-id,Values=$INSTANCE_ID" --query "Tags[?Key=='github:runner-type'].Value" --output text)
 LABELS="self-hosted,Linux,$(uname -m),$RUNNER_TYPE"
 export LABELS
+echo "Instance Labels: $LABELS"
+
+LIFE_CYCLE=$(curl -s --fail http://169.254.169.254/latest/meta-data/instance-life-cycle)
+export LIFE_CYCLE
+echo "Instance lifecycle: $LIFE_CYCLE"
+
+INSTANCE_TYPE=$(ec2metadata --instance-type)
+echo "Instance type: $INSTANCE_TYPE"
 
 # Refresh CloudWatch agent config
 aws ssm get-parameter --region us-east-1 --name AmazonCloudWatch-github-runners --query 'Parameter.Value' --output text > /opt/aws/amazon-cloudwatch-agent/etc/amazon-cloudwatch-agent.json
@@ -124,10 +132,6 @@ terminate_decrease_and_exit() {
 declare -f terminate_and_exit >> /tmp/actions-hooks/common.sh
 
 check_spot_instance_is_old() {
-    # This function should be executed ONLY BETWEEN runnings.
-    # It's unsafe to execute while the runner is working!
-    local LIFE_CYCLE
-    LIFE_CYCLE=$(curl -s --fail http://169.254.169.254/latest/meta-data/instance-life-cycle)
     if [ "$LIFE_CYCLE" == "spot" ]; then
         local UPTIME
         UPTIME=$(< /proc/uptime)

tests/integration/test_disk_configuration/test.py

@@ -208,13 +208,21 @@ def test_merge_tree_custom_disk_setting(start_cluster):
                 secret_access_key='minio123');
     """
     )
-    count = len(list(minio.list_objects(cluster.minio_bucket, "data/", recursive=True)))
+
+    list1 = list(minio.list_objects(cluster.minio_bucket, "data/", recursive=True))
+    count1 = len(list1)
+
     node1.query(f"INSERT INTO {TABLE_NAME}_3 SELECT number FROM numbers(100)")
     assert int(node1.query(f"SELECT count() FROM {TABLE_NAME}_3")) == 100
-    assert (
+
-        len(list(minio.list_objects(cluster.minio_bucket, "data/", recursive=True)))
+    list2 = list(minio.list_objects(cluster.minio_bucket, "data/", recursive=True))
-        == count
+    count2 = len(list2)
-    )
+
+    if count1 != count2:
+        print("list1: ", list1)
+        print("list2: ", list2)
+
+    assert count1 == count2
     assert (
         len(list(minio.list_objects(cluster.minio_bucket, "data2/", recursive=True)))
         > 0

tests/integration/test_ssl_cert_authentication/test.py

@@ -43,15 +43,10 @@ def started_cluster():
 config = """<clickhouse>
     <openSSL>
         <client>
-            <verificationMode>none</verificationMode>
+            <verificationMode>strict</verificationMode>
-
             <certificateFile>{certificateFile}</certificateFile>
             <privateKeyFile>{privateKeyFile}</privateKeyFile>
             <caConfig>{caConfig}</caConfig>
-
-            <invalidCertificateHandler>
-                <name>AcceptCertificateHandler</name>
-            </invalidCertificateHandler>
         </client>
     </openSSL>
 </clickhouse>"""

tests/integration/test_storage_rabbitmq/test.py

@@ -2220,13 +2220,11 @@ def test_rabbitmq_commit_on_block_write(rabbitmq_cluster):
 
 
 def test_rabbitmq_no_connection_at_startup_1(rabbitmq_cluster):
-    # no connection when table is initialized
+    error = instance.query_and_get_error(
-    rabbitmq_cluster.pause_container("rabbitmq1")
-    instance.query_and_get_error(
         """
         CREATE TABLE test.cs (key UInt64, value UInt64)
             ENGINE = RabbitMQ
-            SETTINGS rabbitmq_host_port = 'rabbitmq1:5672',
+            SETTINGS rabbitmq_host_port = 'no_connection_at_startup:5672',
                      rabbitmq_exchange_name = 'cs',
                      rabbitmq_format = 'JSONEachRow',
                      rabbitmq_flush_interval_ms=1000,
@@ -2234,7 +2232,7 @@ def test_rabbitmq_no_connection_at_startup_1(rabbitmq_cluster):
                      rabbitmq_row_delimiter = '\\n';
     """
     )
-    rabbitmq_cluster.unpause_container("rabbitmq1")
+    assert "CANNOT_CONNECT_RABBITMQ" in error
 
 
 def test_rabbitmq_no_connection_at_startup_2(rabbitmq_cluster):

tests/performance/final_big_column.xml

@@ -10,8 +10,8 @@
         PARTITION BY toYYYYMM(d) ORDER BY key
     </create_query>
 
-    <fill_query>INSERT INTO optimized_select_final SELECT toDate('2000-01-01'), 2*number, randomPrintableASCII(1000) FROM numbers(5000000)</fill_query>
+    <fill_query>INSERT INTO optimized_select_final SELECT toDate('2000-01-01'), 2*number, randomPrintableASCII(1000) FROM numbers(2500000)</fill_query>
-    <fill_query>INSERT INTO optimized_select_final SELECT toDate('2020-01-01'), 2*number+1, randomPrintableASCII(1000) FROM numbers(5000000)</fill_query>
+    <fill_query>INSERT INTO optimized_select_final SELECT toDate('2020-01-01'), 2*number+1, randomPrintableASCII(1000) FROM numbers(2500000)</fill_query>
 
     <query>SELECT * FROM optimized_select_final FINAL FORMAT Null SETTINGS max_threads = 8</query>
     <query>SELECT * FROM optimized_select_final FINAL WHERE key % 10 = 0 FORMAT Null</query>

tests/performance/function_tokens.xml

@@ -1,5 +1,5 @@
 <test>
     <query>with 'Many years later as he faced the firing squad, Colonel Aureliano Buendia was to remember that distant afternoon when his father took him to discover ice.' as s select splitByChar(' ', materialize(s)) as w from numbers(1000000)</query>
-    <query>with 'Many years later as he faced the firing squad, Colonel Aureliano Buendia was to remember that distant afternoon when his father took him to discover ice.' as s select splitByRegexp(' ', materialize(s)) as w from numbers(1000000)</query>
+    <query>with 'Many years later as he faced the firing squad, Colonel Aureliano Buendia was to remember that distant afternoon when his father took him to discover ice.' as s select splitByRegexp(' ', materialize(s)) as w from numbers(200000)</query>
-    <query>with 'Many years later as he faced the firing squad, Colonel Aureliano Buendia was to remember that distant afternoon when his father took him to discover ice.' as s select splitByRegexp('\s+', materialize(s)) as w from numbers(100000)</query>
+    <query>with 'Many years later as he faced the firing squad, Colonel Aureliano Buendia was to remember that distant afternoon when his father took him to discover ice.' as s select splitByRegexp('\s+', materialize(s)) as w from numbers(20000)</query>
 </test>

tests/performance/polymorphic_parts_s.xml

@@ -24,10 +24,10 @@
         <min_insert_block_size_rows>1</min_insert_block_size_rows>
     </settings>
 
-    <!-- 100 parts -->
+    <!-- 50 parts -->
-    <query>INSERT INTO hits_wide(UserID)    SELECT rand() FROM  numbers(100)</query>
+    <query>INSERT INTO hits_wide(UserID)    SELECT rand() FROM  numbers(50)</query>
-    <query>INSERT INTO hits_compact(UserID) SELECT rand() FROM numbers(1000)</query>
+    <query>INSERT INTO hits_compact(UserID) SELECT rand() FROM numbers(500)</query>
-    <query>INSERT INTO hits_buffer(UserID)  SELECT rand() FROM numbers(100)</query>
+    <query>INSERT INTO hits_buffer(UserID)  SELECT rand() FROM numbers(50)</query>
 
     <drop_query>DROP TABLE IF EXISTS hits_wide</drop_query>
     <drop_query>DROP TABLE IF EXISTS hits_compact</drop_query>

tests/performance/scripts/report.py

@@ -555,7 +555,7 @@ if args.report == "main":
             "Total client time for measured query runs, s",  # 2
             "Queries",  # 3
             "Longest query, total for measured runs, s",  # 4
-            "Wall clock time per query, s",  # 5
+            "Average query wall clock time, s",  # 5
             "Shortest query, total for measured runs, s",  # 6
             "",  # Runs                                               #7
         ]

tests/queries/0_stateless/00257_shard_no_aggregates_and_constant_keys.reference

@@ -8,13 +8,13 @@
 40
 41
 
-0
+41
 2	42
 
 2	42
 43
 
-0
+43
 11
 
 11

tests/queries/0_stateless/00927_asof_join_correct_bt.reference

@@ -1,13 +1,36 @@
+-- { echoOn }
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B1 B USING(k,t) ORDER BY (A.k, A.t);
 1	101	1	0	0	0
 1	102	2	2	102	1
 1	103	3	2	102	1
 1	104	4	4	104	1
 1	105	5	4	104	1
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B2 B USING(k,t) ORDER BY (A.k, A.t);
 1	101	1	0	0	0
 1	102	2	2	102	1
 1	103	3	2	102	1
 1	104	4	4	104	1
 1	105	5	4	104	1
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B3 B USING(k,t) ORDER BY (A.k, A.t);
+1	101	1	0	0	0
+1	102	2	2	102	1
+1	103	3	2	102	1
+1	104	4	4	104	1
+1	105	5	4	104	1
+SET join_algorithm = 'full_sorting_merge';
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B1 B USING(k,t) ORDER BY (A.k, A.t);
+1	101	1	0	0	0
+1	102	2	2	102	1
+1	103	3	2	102	1
+1	104	4	4	104	1
+1	105	5	4	104	1
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B2 B USING(k,t) ORDER BY (A.k, A.t);
+1	101	1	0	0	0
+1	102	2	2	102	1
+1	103	3	2	102	1
+1	104	4	4	104	1
+1	105	5	4	104	1
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B3 B USING(k,t) ORDER BY (A.k, A.t);
 1	101	1	0	0	0
 1	102	2	2	102	1
 1	103	3	2	102	1

tests/queries/0_stateless/00927_asof_join_correct_bt.sql

@@ -4,20 +4,29 @@ DROP TABLE IF EXISTS B;
 CREATE TABLE A(k UInt32, t UInt32, a UInt64) ENGINE = MergeTree() ORDER BY (k, t);
 INSERT INTO A(k,t,a) VALUES (1,101,1),(1,102,2),(1,103,3),(1,104,4),(1,105,5);
 
-CREATE TABLE B(k UInt32, t UInt32, b UInt64) ENGINE = MergeTree() ORDER BY (k, t);
+CREATE TABLE B1(k UInt32, t UInt32, b UInt64) ENGINE = MergeTree() ORDER BY (k, t);
-INSERT INTO B(k,t,b) VALUES (1,102,2), (1,104,4);
+INSERT INTO B1(k,t,b) VALUES (1,102,2), (1,104,4);
-SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B USING(k,t) ORDER BY (A.k, A.t);
-DROP TABLE B;
 
+CREATE TABLE B2(t UInt32, k UInt32, b UInt64) ENGINE = MergeTree() ORDER BY (k, t);
+INSERT INTO B2(k,t,b) VALUES (1,102,2), (1,104,4);
 
-CREATE TABLE B(t UInt32, k UInt32, b UInt64) ENGINE = MergeTree() ORDER BY (k, t);
+CREATE TABLE B3(k UInt32, b UInt64, t UInt32) ENGINE = MergeTree() ORDER BY (k, t);
-INSERT INTO B(k,t,b) VALUES (1,102,2), (1,104,4);
+INSERT INTO B3(k,t,b) VALUES (1,102,2), (1,104,4);
-SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B USING(k,t) ORDER BY (A.k, A.t);
-DROP TABLE B;
 
-CREATE TABLE B(k UInt32, b UInt64, t UInt32) ENGINE = MergeTree() ORDER BY (k, t);
+-- { echoOn }
-INSERT INTO B(k,t,b) VALUES (1,102,2), (1,104,4);
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B1 B USING(k,t) ORDER BY (A.k, A.t);
-SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B USING(k,t) ORDER BY (A.k, A.t);
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B2 B USING(k,t) ORDER BY (A.k, A.t);
-DROP TABLE B;
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B3 B USING(k,t) ORDER BY (A.k, A.t);
+
+SET join_algorithm = 'full_sorting_merge';
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B1 B USING(k,t) ORDER BY (A.k, A.t);
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B2 B USING(k,t) ORDER BY (A.k, A.t);
+SELECT A.k, A.t, A.a, B.b, B.t, B.k FROM A ASOF LEFT JOIN B3 B USING(k,t) ORDER BY (A.k, A.t);
+
+-- { echoOff }
+
+DROP TABLE B1;
+DROP TABLE B2;
+DROP TABLE B3;
 
 DROP TABLE A;

tests/queries/0_stateless/00927_asof_join_long.reference

@@ -1 +1,2 @@
 3000000
+3000000

tests/queries/0_stateless/00927_asof_join_long.sql

@@ -2,15 +2,28 @@
 
 DROP TABLE IF EXISTS tvs;
 
+-- to use different algorithms for in subquery
+SET allow_experimental_analyzer = 1;
+
 CREATE TABLE tvs(k UInt32, t UInt32, tv UInt64) ENGINE = Memory;
 INSERT INTO tvs(k,t,tv) SELECT k, t, t
 FROM (SELECT toUInt32(number) AS k FROM numbers(1000)) keys
-CROSS JOIN (SELECT toUInt32(number * 3) as t FROM numbers(10000)) tv_times;
+CROSS JOIN (SELECT toUInt32(number * 3) as t FROM numbers(10000)) tv_times
+SETTINGS join_algorithm = 'hash';
 
 SELECT SUM(trades.price - tvs.tv) FROM
 (SELECT k, t, t as price
     FROM (SELECT toUInt32(number) AS k FROM numbers(1000)) keys
-    CROSS JOIN (SELECT toUInt32(number * 10) AS t FROM numbers(3000)) trade_times) trades
+    CROSS JOIN (SELECT toUInt32(number * 10) AS t FROM numbers(3000)) trade_times
+    SETTINGS join_algorithm = 'hash') trades
 ASOF LEFT JOIN tvs USING(k,t);
 
+SELECT SUM(trades.price - tvs.tv) FROM
+(SELECT k, t, t as price
+    FROM (SELECT toUInt32(number) AS k FROM numbers(1000)) keys
+    CROSS JOIN (SELECT toUInt32(number * 10) AS t FROM numbers(3000)) trade_times
+    SETTINGS join_algorithm = 'hash') trades
+ASOF LEFT JOIN tvs USING(k,t)
+SETTINGS join_algorithm = 'full_sorting_merge';
+
 DROP TABLE tvs;

tests/queries/0_stateless/00927_asof_join_noninclusive.reference

@@ -27,3 +27,32 @@
 2	1970-01-01 00:00:03	3	3	1970-01-01 00:00:03	2
 2	1970-01-01 00:00:04	4	3	1970-01-01 00:00:03	2
 2	1970-01-01 00:00:05	5	3	1970-01-01 00:00:03	2
+1	1970-01-01 00:00:01	1	0	1970-01-01 00:00:00	0
+1	1970-01-01 00:00:02	2	2	1970-01-01 00:00:02	1
+1	1970-01-01 00:00:03	3	2	1970-01-01 00:00:02	1
+1	1970-01-01 00:00:04	4	4	1970-01-01 00:00:04	1
+1	1970-01-01 00:00:05	5	4	1970-01-01 00:00:04	1
+2	1970-01-01 00:00:01	1	0	1970-01-01 00:00:00	0
+2	1970-01-01 00:00:02	2	0	1970-01-01 00:00:00	0
+2	1970-01-01 00:00:03	3	3	1970-01-01 00:00:03	2
+2	1970-01-01 00:00:04	4	3	1970-01-01 00:00:03	2
+2	1970-01-01 00:00:05	5	3	1970-01-01 00:00:03	2
+3	1970-01-01 00:00:01	1	0	1970-01-01 00:00:00	0
+3	1970-01-01 00:00:02	2	0	1970-01-01 00:00:00	0
+3	1970-01-01 00:00:03	3	0	1970-01-01 00:00:00	0
+3	1970-01-01 00:00:04	4	0	1970-01-01 00:00:00	0
+3	1970-01-01 00:00:05	5	0	1970-01-01 00:00:00	0
+1	1970-01-01 00:00:02	2	2	1970-01-01 00:00:02	1
+1	1970-01-01 00:00:03	3	2	1970-01-01 00:00:02	1
+1	1970-01-01 00:00:04	4	4	1970-01-01 00:00:04	1
+1	1970-01-01 00:00:05	5	4	1970-01-01 00:00:04	1
+2	1970-01-01 00:00:03	3	3	1970-01-01 00:00:03	2
+2	1970-01-01 00:00:04	4	3	1970-01-01 00:00:03	2
+2	1970-01-01 00:00:05	5	3	1970-01-01 00:00:03	2
+1	1970-01-01 00:00:02	2	2	1970-01-01 00:00:02	1
+1	1970-01-01 00:00:03	3	2	1970-01-01 00:00:02	1
+1	1970-01-01 00:00:04	4	4	1970-01-01 00:00:04	1
+1	1970-01-01 00:00:05	5	4	1970-01-01 00:00:04	1
+2	1970-01-01 00:00:03	3	3	1970-01-01 00:00:03	2
+2	1970-01-01 00:00:04	4	3	1970-01-01 00:00:03	2
+2	1970-01-01 00:00:05	5	3	1970-01-01 00:00:03	2

tests/queries/0_stateless/00927_asof_join_noninclusive.sql

@@ -11,9 +11,12 @@ INSERT INTO B(k,t,b) VALUES (1,2,2),(1,4,4);
 INSERT INTO B(k,t,b) VALUES (2,3,3);
 
 SELECT A.k, toString(A.t, 'UTC'), A.a, B.b, toString(B.t, 'UTC'), B.k FROM A ASOF LEFT JOIN B USING(k,t) ORDER BY (A.k, A.t);
-
 SELECT A.k, toString(A.t, 'UTC'), A.a, B.b, toString(B.t, 'UTC'), B.k FROM A ASOF INNER JOIN B ON A.k == B.k AND A.t >= B.t ORDER BY (A.k, A.t);
+SELECT A.k, toString(A.t, 'UTC'), A.a, B.b, toString(B.t, 'UTC'), B.k FROM A ASOF JOIN B USING(k,t) ORDER BY (A.k, A.t);
 
+SET join_algorithm = 'full_sorting_merge';
+SELECT A.k, toString(A.t, 'UTC'), A.a, B.b, toString(B.t, 'UTC'), B.k FROM A ASOF LEFT JOIN B USING(k,t) ORDER BY (A.k, A.t);
+SELECT A.k, toString(A.t, 'UTC'), A.a, B.b, toString(B.t, 'UTC'), B.k FROM A ASOF INNER JOIN B ON A.k == B.k AND A.t >= B.t ORDER BY (A.k, A.t);
 SELECT A.k, toString(A.t, 'UTC'), A.a, B.b, toString(B.t, 'UTC'), B.k FROM A ASOF JOIN B USING(k,t) ORDER BY (A.k, A.t);
 
 DROP TABLE A;

tests/queries/0_stateless/00927_asof_join_other_types.reference

@@ -1,27 +1,72 @@
+-
 2	1	1	0
 2	3	3	3
 2	5	5	3
+-
 2	1	1	0
 2	3	3	3
 2	5	5	3
+-
 2	1	1	0
 2	3	3	3
 2	5	5	3
+-
 2	1	1	0
 2	3	3	3
 2	5	5	3
+-
+2	1	1	0
+2	3	3	3
+2	5	5	3
+-
+2	1	1	0
+2	3	3	3
+2	5	5	3
+-
+2	1	1	0
+2	3	3	3
+2	5	5	3
+-
+2	1	1	0
+2	3	3	3
+2	5	5	3
+-
 2	1970-01-01 02:00:01	1	0
 2	1970-01-01 02:00:03	3	3
 2	1970-01-01 02:00:05	5	3
+-
+2	1970-01-01 02:00:01	1	0
+2	1970-01-01 02:00:03	3	3
+2	1970-01-01 02:00:05	5	3
+-
 2	1	1	0
 2	3	3	3
 2	5	5	3
+-
 2	1	1	0
 2	3	3	3
 2	5	5	3
+-
 2	1	1	0
 2	3	3	3
 2	5	5	3
+-
+2	1	1	0
+2	3	3	3
+2	5	5	3
+-
+2	1	1	0
+2	3	3	3
+2	5	5	3
+-
+2	1	1	0
+2	3	3	3
+2	5	5	3
+-
+2	1970-01-01 02:00:00.001	1	0
+2	1970-01-01 02:00:00.003	3	3
+2	1970-01-01 02:00:00.005	5	3
+-
 2	1970-01-01 02:00:00.001	1	0
 2	1970-01-01 02:00:00.003	3	3
 2	1970-01-01 02:00:00.005	5	3

tests/queries/0_stateless/00927_asof_join_other_types.sh

@@ -1,27 +0,0 @@
-#!/usr/bin/env bash
-
-set -e
-
-CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
-# shellcheck source=../shell_config.sh
-. "$CURDIR"/../shell_config.sh
-
-for typename in "UInt32" "UInt64" "Float64" "Float32" "DateTime('Asia/Istanbul')" "Decimal32(5)" "Decimal64(5)" "Decimal128(5)" "DateTime64(3, 'Asia/Istanbul')"
-do
-    $CLICKHOUSE_CLIENT -mn <<EOF
-DROP TABLE IF EXISTS A;
-DROP TABLE IF EXISTS B;
-
-CREATE TABLE A(k UInt32, t ${typename}, a Float64) ENGINE = MergeTree() ORDER BY (k, t);
-INSERT INTO A(k,t,a) VALUES (2,1,1),(2,3,3),(2,5,5);
-
-CREATE TABLE B(k UInt32, t ${typename}, b Float64) ENGINE = MergeTree() ORDER BY (k, t);
-INSERT INTO B(k,t,b) VALUES (2,3,3);
-
-SELECT k, t, a, b FROM A ASOF LEFT JOIN B USING(k,t) ORDER BY (k,t);
-
-DROP TABLE A;
-DROP TABLE B;
-EOF
-
-done

tests/queries/0_stateless/00927_asof_join_other_types.sql.j2

@@ -0,0 +1,27 @@
+
+{% for typename in ["UInt32", "UInt64", "Float64", "Float32", "DateTime('Asia/Istanbul')", "Decimal32(5)", "Decimal64(5)", "Decimal128(5)", "DateTime64(3, 'Asia/Istanbul')"] -%}
+
+
+DROP TABLE IF EXISTS A;
+DROP TABLE IF EXISTS B;
+
+CREATE TABLE A(k UInt32, t {{ typename }}, a Float64) ENGINE = MergeTree() ORDER BY (k, t);
+INSERT INTO A(k,t,a) VALUES (2,1,1),(2,3,3),(2,5,5);
+
+CREATE TABLE B(k UInt32, t {{ typename }}, b Float64) ENGINE = MergeTree() ORDER BY (k, t);
+INSERT INTO B(k,t,b) VALUES (2,3,3);
+
+SELECT '-';
+
+SELECT k, t, a, b FROM A ASOF LEFT JOIN B USING(k,t) ORDER BY (k,t)
+SETTINGS join_algorithm = 'full_sorting_merge';
+
+SELECT '-';
+
+SELECT k, t, a, b FROM A ASOF LEFT JOIN B USING(k,t) ORDER BY (k,t)
+SETTINGS join_algorithm = 'hash';
+
+DROP TABLE A;
+DROP TABLE B;
+
+{% endfor %}

tests/queries/0_stateless/00927_asof_joins.reference

@@ -12,3 +12,18 @@
 2	1970-01-01 00:00:15	5	6.5	6
 2	1970-01-01 00:00:16	5	5.6	6
 2	1970-01-01 00:00:20	17	8.5	18
+-
+1	1970-01-01 00:00:05	1	1.5	2
+1	1970-01-01 00:00:06	1	1.51	2
+1	1970-01-01 00:00:10	11	11.5	12
+1	1970-01-01 00:00:11	11	11.51	12
+1	1970-01-01 00:00:15	5	5.5	6
+1	1970-01-01 00:00:16	5	5.6	6
+1	1970-01-01 00:00:20	7	7.5	8
+2	1970-01-01 00:00:05	11	2.5	12
+2	1970-01-01 00:00:06	11	2.51	12
+2	1970-01-01 00:00:10	21	12.5	22
+2	1970-01-01 00:00:11	21	12.51	22
+2	1970-01-01 00:00:15	5	6.5	6
+2	1970-01-01 00:00:16	5	5.6	6
+2	1970-01-01 00:00:20	17	8.5	18

tests/queries/0_stateless/00927_asof_joins.sql

@@ -9,7 +9,13 @@ CREATE TABLE tv(key UInt32, t DateTime, tv Float64) ENGINE = MergeTree() ORDER B
 INSERT INTO tv(key,t,tv) VALUES (1,5,1.5),(1,6,1.51),(1,10,11.5),(1,11,11.51),(1,15,5.5),(1,16,5.6),(1,20,7.5);
 INSERT INTO tv(key,t,tv) VALUES (2,5,2.5),(2,6,2.51),(2,10,12.5),(2,11,12.51),(2,15,6.5),(2,16,5.6),(2,20,8.5);
 
-SELECT tv.key, toString(tv.t, 'UTC'), md.bid, tv.tv, md.ask FROM tv ASOF LEFT JOIN md USING(key,t) ORDER BY (tv.key, tv.t);
+SELECT tv.key, toString(tv.t, 'UTC'), md.bid, tv.tv, md.ask FROM tv ASOF LEFT JOIN md USING(key,t) ORDER BY (tv.key, tv.t)
+;
+
+SELECT '-';
+
+SELECT tv.key, toString(tv.t, 'UTC'), md.bid, tv.tv, md.ask FROM tv ASOF LEFT JOIN md USING(key,t) ORDER BY (tv.key, tv.t)
+SETTINGS join_algorithm = 'full_sorting_merge';
 
 DROP TABLE md;
 DROP TABLE tv;

tests/queries/0_stateless/00976_asof_join_on.reference

@@ -1,3 +1,4 @@
+- default / join_use_nulls = 0 -
 1	1	0	0
 1	2	1	2
 1	3	1	2
@@ -34,3 +35,114 @@
 2	1	2	3
 2	2	2	3
 1	2	1	2
+- full_sorting_merge / join_use_nulls = 0 -
+1	1	0	0
+1	2	1	2
+1	3	1	2
+2	1	0	0
+2	2	0	0
+2	3	2	3
+3	1	0	0
+3	2	0	0
+3	3	0	0
+9
+1	2	1	2
+1	3	1	2
+2	3	2	3
+-
+1	1	1	2
+1	2	1	2
+1	3	1	4
+2	1	2	3
+2	2	2	3
+2	3	2	3
+-
+1	1	1	2
+1	2	1	2
+1	3	1	4
+2	1	2	3
+2	2	2	3
+2	3	2	3
+-
+1	3	1	2
+-
+1	1	1	2
+1	2	1	4
+1	3	1	4
+2	1	2	3
+2	2	2	3
+1	2	1	2
+- default / join_use_nulls = 1 -
+1	1	\N	\N
+1	2	1	2
+1	3	1	2
+2	1	\N	\N
+2	2	\N	\N
+2	3	2	3
+3	1	\N	\N
+3	2	\N	\N
+3	3	\N	\N
+9
+1	2	1	2
+1	3	1	2
+2	3	2	3
+-
+1	1	1	2
+1	2	1	2
+1	3	1	4
+2	1	2	3
+2	2	2	3
+2	3	2	3
+-
+1	1	1	2
+1	2	1	2
+1	3	1	4
+2	1	2	3
+2	2	2	3
+2	3	2	3
+-
+1	3	1	2
+-
+1	1	1	2
+1	2	1	4
+1	3	1	4
+2	1	2	3
+2	2	2	3
+1	2	1	2
+- full_sorting_merge / join_use_nulls = 1 -
+1	1	\N	\N
+1	2	1	2
+1	3	1	2
+2	1	\N	\N
+2	2	\N	\N
+2	3	2	3
+3	1	\N	\N
+3	2	\N	\N
+3	3	\N	\N
+9
+1	2	1	2
+1	3	1	2
+2	3	2	3
+-
+1	1	1	2
+1	2	1	2
+1	3	1	4
+2	1	2	3
+2	2	2	3
+2	3	2	3
+-
+1	1	1	2
+1	2	1	2
+1	3	1	4
+2	1	2	3
+2	2	2	3
+2	3	2	3
+-
+1	3	1	2
+-
+1	1	1	2
+1	2	1	4
+1	3	1	4
+2	1	2	3
+2	2	2	3
+1	2	1	2

tests/queries/0_stateless/00976_asof_join_on.sql.j2

@@ -7,6 +7,14 @@ CREATE TABLE B(b UInt32, t UInt32) ENGINE = Memory;
 INSERT INTO A (a,t) VALUES (1,1),(1,2),(1,3), (2,1),(2,2),(2,3), (3,1),(3,2),(3,3);
 INSERT INTO B (b,t) VALUES (1,2),(1,4),(2,3);
 
+{% for join_use_nulls in [0, 1] -%}
+{% for join_algorithm in ['default', 'full_sorting_merge'] -%}
+
+SET join_algorithm = '{{ join_algorithm }}';
+
+SELECT '- {{ join_algorithm }} / join_use_nulls = {{ join_use_nulls }} -';
+set join_use_nulls = {{ join_use_nulls }};
+
 SELECT A.a, A.t, B.b, B.t FROM A ASOF LEFT JOIN B ON A.a == B.b AND A.t >= B.t ORDER BY (A.a, A.t);
 SELECT count() FROM A ASOF LEFT JOIN B ON A.a == B.b AND B.t <= A.t;
 SELECT A.a, A.t, B.b, B.t FROM A ASOF INNER JOIN B ON B.t <= A.t AND A.a == B.b ORDER BY (A.a, A.t);
@@ -28,5 +36,8 @@ ASOF INNER JOIN (SELECT * FROM B UNION ALL SELECT 1, 3) AS B ON B.t <= A.t AND A
 WHERE B.t != 3 ORDER BY (A.a, A.t)
 ;
 
+{% endfor -%}
+{% endfor -%}
+
 DROP TABLE A;
 DROP TABLE B;

tests/queries/0_stateless/00997_set_index_array.sql

@@ -12,10 +12,10 @@ ORDER BY (primary_key);
 
 INSERT INTO set_array
 select
-  toString(intDiv(number, 1000000)) as primary_key,
+  toString(intDiv(number, 100000)) as primary_key,
   array(number) as index_array
 from system.numbers
-limit 10000000;
+limit 1000000;
 
 OPTIMIZE TABLE set_array FINAL;
 

tests/queries/0_stateless/01116_asof_join_dolbyzerr.reference

@@ -1,3 +1,6 @@
 v1	o1	['s2','s1']
 v1	o2	['s4']
 v2	o3	['s5','s3']
+v1	o1	['s2','s1']
+v1	o2	['s4']
+v2	o3	['s5','s3']