Merge branch 'master' into patch-1

2024-11-21 15:12:02 +00:00 · 2022-04-01 01:44:22 +00:00 · 2022-04-01 01:44:22 +00:00 · 1675c7f579
commit 1675c7f579
parent 902b402d20 5c6b84c398
330 changed files with 7473 additions and 2483 deletions
--- a/.clang-tidy
+++ b/.clang-tidy
@ -210,3 +210,6 @@ CheckOptions:
    value: false
  - key: performance-move-const-arg.CheckTriviallyCopyableMove
    value: false
+    # Workaround clang-tidy bug: https://github.com/llvm/llvm-project/issues/46097
+  - key: readability-identifier-naming.TypeTemplateParameterIgnoredRegexp
+    value: expr-type
--- a/.github/workflows/master.yml
+++ b/.github/workflows/master.yml
@ -149,7 +149,6 @@ jobs:
          sudo rm -fr "$TEMP_PATH"
  SplitBuildSmokeTest:
    needs: [BuilderDebSplitted]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
    runs-on: [self-hosted, style-checker]
    steps:
      - name: Set envs
@ -316,7 +315,6 @@ jobs:
          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
  BuilderBinRelease:
    needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
    runs-on: [self-hosted, builder]
    steps:
      - name: Set envs
@ -360,6 +358,51 @@ jobs:
          docker kill "$(docker ps -q)" ||:
          docker rm -f "$(docker ps -a -q)" ||:
          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
+  BuilderBinGCC:
+    needs: [DockerHubPush]
+    runs-on: [self-hosted, builder]
+    steps:
+      - name: Set envs
+        run: |
+          cat >> "$GITHUB_ENV" << 'EOF'
+          TEMP_PATH=${{runner.temp}}/build_check
+          IMAGES_PATH=${{runner.temp}}/images_path
+          REPO_COPY=${{runner.temp}}/build_check/ClickHouse
+          CACHES_PATH=${{runner.temp}}/../ccaches
+          CHECK_NAME=ClickHouse build check (actions)
+          BUILD_NAME=binary_gcc
+          EOF
+      - name: Download changed images
+        uses: actions/download-artifact@v2
+        with:
+          name: changed_images
+          path: ${{ env.IMAGES_PATH }}
+      - name: Clear repository
+        run: |
+          sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
+      - name: Check out repository code
+        uses: actions/checkout@v2
+        with:
+          submodules: 'true'
+          fetch-depth: 0 # otherwise we will have no info about contributors
+      - name: Build
+        run: |
+          sudo rm -fr "$TEMP_PATH"
+          mkdir -p "$TEMP_PATH"
+          cp -r "$GITHUB_WORKSPACE" "$TEMP_PATH"
+          cd "$REPO_COPY/tests/ci" && python3 build_check.py "$CHECK_NAME" "$BUILD_NAME"
+      - name: Upload build URLs to artifacts
+        if: ${{ success() || failure() }}
+        uses: actions/upload-artifact@v2
+        with:
+          name: ${{ env.BUILD_NAME }}
+          path: ${{ env.TEMP_PATH }}/${{ env.BUILD_NAME }}.json
+      - name: Cleanup
+        if: always()
+        run: |
+          docker kill "$(docker ps -q)" ||:
+          docker rm -f "$(docker ps -a -q)" ||:
+          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
  BuilderDebAsan:
    needs: [DockerHubPush]
    runs-on: [self-hosted, builder]
@ -590,7 +633,6 @@ jobs:
 ##########################################################################################
  BuilderDebSplitted:
    needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
    runs-on: [self-hosted, builder]
    steps:
      - name: Set envs
@ -636,7 +678,6 @@ jobs:
          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
  BuilderBinTidy:
    needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
    runs-on: [self-hosted, builder]
    steps:
      - name: Set envs
@ -682,7 +723,6 @@ jobs:
          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
  BuilderBinDarwin:
    needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
    runs-on: [self-hosted, builder]
    steps:
      - name: Set envs
@ -728,7 +768,6 @@ jobs:
          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
  BuilderBinAarch64:
    needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
    runs-on: [self-hosted, builder]
    steps:
      - name: Set envs
@ -774,7 +813,6 @@ jobs:
          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
  BuilderBinFreeBSD:
    needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
    runs-on: [self-hosted, builder]
    steps:
      - name: Set envs
@ -820,7 +858,6 @@ jobs:
          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
  BuilderBinDarwinAarch64:
    needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
    runs-on: [self-hosted, builder]
    steps:
      - name: Set envs
@ -866,7 +903,6 @@ jobs:
          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
  BuilderBinPPC64:
    needs: [DockerHubPush]
-    if: ${{ !contains(github.event.pull_request.labels.*.name, 'pr-documentation') && !contains(github.event.pull_request.labels.*.name, 'pr-doc-fix') }}
    runs-on: [self-hosted, builder]
    steps:
      - name: Set envs
@ -918,6 +954,7 @@ jobs:
      - BuilderDebRelease
      - BuilderDebAarch64
      - BuilderBinRelease
+      - BuilderBinGCC
      - BuilderDebAsan
      - BuilderDebTsan
      - BuilderDebUBsan
@ -2608,6 +2645,40 @@ jobs:
          docker kill "$(docker ps -q)" ||:
          docker rm -f "$(docker ps -a -q)" ||:
          sudo rm -fr "$TEMP_PATH"
+  UnitTestsReleaseGCC:
+    needs: [BuilderBinGCC]
+    runs-on: [self-hosted, fuzzer-unit-tester]
+    steps:
+      - name: Set envs
+        run: |
+          cat >> "$GITHUB_ENV" << 'EOF'
+          TEMP_PATH=${{runner.temp}}/unit_tests_asan
+          REPORTS_PATH=${{runner.temp}}/reports_dir
+          CHECK_NAME=Unit tests (release-gcc, actions)
+          REPO_COPY=${{runner.temp}}/unit_tests_asan/ClickHouse
+          EOF
+      - name: Download json reports
+        uses: actions/download-artifact@v2
+        with:
+          path: ${{ env.REPORTS_PATH }}
+      - name: Clear repository
+        run: |
+          sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
+      - name: Check out repository code
+        uses: actions/checkout@v2
+      - name: Unit test
+        run: |
+          sudo rm -fr "$TEMP_PATH"
+          mkdir -p "$TEMP_PATH"
+          cp -r "$GITHUB_WORKSPACE" "$TEMP_PATH"
+          cd "$REPO_COPY/tests/ci"
+          python3 unit_tests_check.py "$CHECK_NAME"
+      - name: Cleanup
+        if: always()
+        run: |
+          docker kill "$(docker ps -q)" ||:
+          docker rm -f "$(docker ps -a -q)" ||:
+          sudo rm -fr "$TEMP_PATH"
  UnitTestsTsan:
    needs: [BuilderDebTsan]
    runs-on: [self-hosted, fuzzer-unit-tester]
--- a/.github/workflows/pull_request.yml
+++ b/.github/workflows/pull_request.yml
@ -370,6 +370,48 @@ jobs:
          docker kill "$(docker ps -q)" ||:
          docker rm -f "$(docker ps -a -q)" ||:
          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
+  BuilderBinGCC:
+    needs: [DockerHubPush, FastTest]
+    runs-on: [self-hosted, builder]
+    steps:
+      - name: Set envs
+        run: |
+          cat >> "$GITHUB_ENV" << 'EOF'
+          TEMP_PATH=${{runner.temp}}/build_check
+          IMAGES_PATH=${{runner.temp}}/images_path
+          REPO_COPY=${{runner.temp}}/build_check/ClickHouse
+          CACHES_PATH=${{runner.temp}}/../ccaches
+          CHECK_NAME=ClickHouse build check (actions)
+          BUILD_NAME=binary_gcc
+          EOF
+      - name: Download changed images
+        uses: actions/download-artifact@v2
+        with:
+          name: changed_images
+          path: ${{ runner.temp }}/images_path
+      - name: Check out repository code
+        uses: actions/checkout@v2
+        with:
+          submodules: 'true'
+          fetch-depth: 0 # otherwise we will have no info about contributors
+      - name: Build
+        run: |
+          sudo rm -fr "$TEMP_PATH"
+          mkdir -p "$TEMP_PATH"
+          cp -r "$GITHUB_WORKSPACE" "$TEMP_PATH"
+          cd "$REPO_COPY/tests/ci" && python3 build_check.py "$CHECK_NAME" "$BUILD_NAME"
+      - name: Upload build URLs to artifacts
+        if: ${{ success() || failure() }}
+        uses: actions/upload-artifact@v2
+        with:
+          name: ${{ env.BUILD_NAME }}
+          path: ${{ runner.temp }}/build_check/${{ env.BUILD_NAME }}.json
+      - name: Cleanup
+        if: always()
+        run: |
+          docker kill "$(docker ps -q)" ||:
+          docker rm -f "$(docker ps -a -q)" ||:
+          sudo rm -fr "$TEMP_PATH" "$CACHES_PATH"
  BuilderDebAarch64:
    needs: [DockerHubPush, FastTest]
    runs-on: [self-hosted, builder]
@ -963,6 +1005,7 @@ jobs:
      - BuilderDebRelease
      - BuilderDebAarch64
      - BuilderBinRelease
+      - BuilderBinGCC
      - BuilderDebAsan
      - BuilderDebTsan
      - BuilderDebUBsan
@ -2808,6 +2851,40 @@ jobs:
          docker kill "$(docker ps -q)" ||:
          docker rm -f "$(docker ps -a -q)" ||:
          sudo rm -fr "$TEMP_PATH"
+  UnitTestsReleaseGCC:
+    needs: [BuilderBinGCC]
+    runs-on: [self-hosted, fuzzer-unit-tester]
+    steps:
+      - name: Set envs
+        run: |
+          cat >> "$GITHUB_ENV" << 'EOF'
+          TEMP_PATH=${{runner.temp}}/unit_tests_asan
+          REPORTS_PATH=${{runner.temp}}/reports_dir
+          CHECK_NAME=Unit tests (release-gcc, actions)
+          REPO_COPY=${{runner.temp}}/unit_tests_asan/ClickHouse
+          EOF
+      - name: Download json reports
+        uses: actions/download-artifact@v2
+        with:
+          path: ${{ env.REPORTS_PATH }}
+      - name: Clear repository
+        run: |
+          sudo rm -fr "$GITHUB_WORKSPACE" && mkdir "$GITHUB_WORKSPACE"
+      - name: Check out repository code
+        uses: actions/checkout@v2
+      - name: Unit test
+        run: |
+          sudo rm -fr "$TEMP_PATH"
+          mkdir -p "$TEMP_PATH"
+          cp -r "$GITHUB_WORKSPACE" "$TEMP_PATH"
+          cd "$REPO_COPY/tests/ci"
+          python3 unit_tests_check.py "$CHECK_NAME"
+      - name: Cleanup
+        if: always()
+        run: |
+          docker kill "$(docker ps -q)" ||:
+          docker rm -f "$(docker ps -a -q)" ||:
+          sudo rm -fr "$TEMP_PATH"
  UnitTestsTsan:
    needs: [BuilderDebTsan]
    runs-on: [self-hosted, fuzzer-unit-tester]
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@ -261,8 +261,8 @@ endif ()
 # Add a section with the hash of the compiled machine code for integrity checks.
 # Only for official builds, because adding a section can be time consuming (rewrite of several GB).
 # And cross compiled binaries are not supported (since you cannot execute clickhouse hash-binary)
-if (OBJCOPY_PATH AND YANDEX_OFFICIAL_BUILD AND (NOT CMAKE_TOOLCHAIN_FILE))
-    set (USE_BINARY_HASH 1)
+if (OBJCOPY_PATH AND CLICKHOUSE_OFFICIAL_BUILD AND (NOT CMAKE_TOOLCHAIN_FILE OR CMAKE_TOOLCHAIN_FILE MATCHES "linux/toolchain-x86_64.cmake$"))
+    set (USE_BINARY_HASH 1 CACHE STRING "Calculate binary hash and store it in the separate section")
 endif ()

 # Allows to build stripped binary in a separate directory
--- a/base/base/CMakeLists.txt
+++ b/base/base/CMakeLists.txt
@ -2,6 +2,7 @@ set (SRCS
    argsToConfig.cpp
    coverage.cpp
    demangle.cpp
+    getAvailableMemoryAmount.cpp
    getFQDNOrHostName.cpp
    getMemoryAmount.cpp
    getPageSize.cpp
--- a/base/base/getAvailableMemoryAmount.cpp
+++ b/base/base/getAvailableMemoryAmount.cpp
@ -0,0 +1,44 @@
+#include <stdexcept>
+#include <fstream>
+#include <base/getAvailableMemoryAmount.h>
+#include <base/getPageSize.h>
+
+#include <unistd.h>
+#include <sys/types.h>
+#include <sys/param.h>
+#if defined(BSD)
+#include <sys/sysctl.h>
+#include <sys/vmmeter.h>
+#endif
+
+
+uint64_t getAvailableMemoryAmountOrZero()
+{
+#if defined(_SC_AVPHYS_PAGES) // linux
+    return getPageSize() * sysconf(_SC_AVPHYS_PAGES);
+#elif defined(__FreeBSD__)
+    struct vmtotal vmt;
+    size_t vmt_size = sizeof(vmt);
+    if (sysctlbyname("vm.vmtotal", &vmt, &vmt_size, NULL, 0) == 0)
+        return getPageSize() * vmt.t_avm;
+    else
+        return 0;
+#else // darwin
+    unsigned int usermem;
+    size_t len = sizeof(usermem);
+    static int mib[2] = { CTL_HW, HW_USERMEM };
+    if (sysctl(mib, 2, &usermem, &len, nullptr, 0) == 0 && len == sizeof(usermem))
+        return usermem;
+    else
+        return 0;
+#endif
+}
+
+
+uint64_t getAvailableMemoryAmount()
+{
+    auto res = getAvailableMemoryAmountOrZero();
+    if (!res)
+        throw std::runtime_error("Cannot determine available memory amount");
+    return res;
+}
--- a/base/base/getAvailableMemoryAmount.h
+++ b/base/base/getAvailableMemoryAmount.h
@ -0,0 +1,12 @@
+#pragma once
+
+#include <cstdint>
+
+/** Returns the size of currently available physical memory (RAM) in bytes.
+  * Returns 0 on unsupported platform or if it cannot determine the size of physical memory.
+  */
+uint64_t getAvailableMemoryAmountOrZero();
+
+/** Throws exception if it cannot determine the size of physical memory.
+  */
+uint64_t getAvailableMemoryAmount();
--- a/base/glibc-compatibility/CMakeLists.txt
+++ b/base/glibc-compatibility/CMakeLists.txt
@ -51,6 +51,6 @@ if (GLIBC_COMPATIBILITY)

    message (STATUS "Some symbols from glibc will be replaced for compatibility")

-elseif (YANDEX_OFFICIAL_BUILD)
+elseif (CLICKHOUSE_OFFICIAL_BUILD)
    message (WARNING "Option GLIBC_COMPATIBILITY must be turned on for production builds.")
 endif ()
--- a/cmake/version.cmake
+++ b/cmake/version.cmake
@ -18,6 +18,6 @@ set (VERSION_STRING_SHORT "${VERSION_MAJOR}.${VERSION_MINOR}")

 math (EXPR VERSION_INTEGER "${VERSION_PATCH} + ${VERSION_MINOR}*1000 + ${VERSION_MAJOR}*1000000")

-if(YANDEX_OFFICIAL_BUILD)
+if(CLICKHOUSE_OFFICIAL_BUILD)
    set(VERSION_OFFICIAL " (official build)")
 endif()
--- a/contrib/avro-cmake/CMakeLists.txt
+++ b/contrib/avro-cmake/CMakeLists.txt
@ -69,9 +69,10 @@ endif ()
 target_compile_options(_avrocpp PRIVATE ${SUPPRESS_WARNINGS})

 # create a symlink to include headers with <avro/...>
+set(AVRO_INCLUDE_DIR "${CMAKE_CURRENT_BINARY_DIR}/include")
 ADD_CUSTOM_TARGET(avro_symlink_headers ALL
-    COMMAND ${CMAKE_COMMAND} -E make_directory "${AVROCPP_ROOT_DIR}/include"
-    COMMAND ${CMAKE_COMMAND} -E create_symlink "${AVROCPP_ROOT_DIR}/api" "${AVROCPP_ROOT_DIR}/include/avro"
+    COMMAND ${CMAKE_COMMAND} -E make_directory "${AVRO_INCLUDE_DIR}"
+    COMMAND ${CMAKE_COMMAND} -E create_symlink "${AVROCPP_ROOT_DIR}/api" "${AVRO_INCLUDE_DIR}/avro"
 )
 add_dependencies(_avrocpp avro_symlink_headers)
-target_include_directories(_avrocpp SYSTEM BEFORE PUBLIC "${AVROCPP_ROOT_DIR}/include")
+target_include_directories(_avrocpp SYSTEM BEFORE PUBLIC "${AVRO_INCLUDE_DIR}")
--- a/contrib/boost-cmake/CMakeLists.txt
+++ b/contrib/boost-cmake/CMakeLists.txt
@ -27,7 +27,11 @@ target_include_directories (_boost_headers_only SYSTEM BEFORE INTERFACE ${LIBRAR

 # asio

-target_compile_definitions (_boost_headers_only INTERFACE BOOST_ASIO_STANDALONE=1)
+target_compile_definitions (_boost_headers_only INTERFACE
+  BOOST_ASIO_STANDALONE=1
+  # Avoid using of deprecated in c++ > 17 std::result_of
+  BOOST_ASIO_HAS_STD_INVOKE_RESULT=1
+)

 # iostreams

--- a/contrib/hyperscan
+++ b/contrib/hyperscan
@ -1 +1 @@
-Subproject commit e9f08df0213fc637aac0a5bbde9beeaeba2fe9fa
+Subproject commit 5edc68c5ac68d2d4f876159e9ee84def6d3dc87c
--- a/contrib/libcxx
+++ b/contrib/libcxx
@ -1 +1 @@
-Subproject commit 61e60294b1de01483caa9f5d00f437c99b674de6
+Subproject commit 172b2ae074f6755145b91c53a95c8540c1468239
--- a/contrib/libcxx-cmake/CMakeLists.txt
+++ b/contrib/libcxx-cmake/CMakeLists.txt
@ -18,12 +18,14 @@ set(SRCS
 "${LIBCXX_SOURCE_DIR}/src/filesystem/directory_iterator.cpp"
 "${LIBCXX_SOURCE_DIR}/src/filesystem/int128_builtins.cpp"
 "${LIBCXX_SOURCE_DIR}/src/filesystem/operations.cpp"
+"${LIBCXX_SOURCE_DIR}/src/format.cpp"
 "${LIBCXX_SOURCE_DIR}/src/functional.cpp"
 "${LIBCXX_SOURCE_DIR}/src/future.cpp"
 "${LIBCXX_SOURCE_DIR}/src/hash.cpp"
 "${LIBCXX_SOURCE_DIR}/src/ios.cpp"
 "${LIBCXX_SOURCE_DIR}/src/ios.instantiations.cpp"
 "${LIBCXX_SOURCE_DIR}/src/iostream.cpp"
+"${LIBCXX_SOURCE_DIR}/src/legacy_pointer_safety.cpp"
 "${LIBCXX_SOURCE_DIR}/src/locale.cpp"
 "${LIBCXX_SOURCE_DIR}/src/memory.cpp"
 "${LIBCXX_SOURCE_DIR}/src/mutex.cpp"
@ -33,6 +35,9 @@ set(SRCS
 "${LIBCXX_SOURCE_DIR}/src/random.cpp"
 "${LIBCXX_SOURCE_DIR}/src/random_shuffle.cpp"
 "${LIBCXX_SOURCE_DIR}/src/regex.cpp"
+"${LIBCXX_SOURCE_DIR}/src/ryu/d2fixed.cpp"
+"${LIBCXX_SOURCE_DIR}/src/ryu/d2s.cpp"
+"${LIBCXX_SOURCE_DIR}/src/ryu/f2s.cpp"
 "${LIBCXX_SOURCE_DIR}/src/shared_mutex.cpp"
 "${LIBCXX_SOURCE_DIR}/src/stdexcept.cpp"
 "${LIBCXX_SOURCE_DIR}/src/string.cpp"
@ -49,7 +54,9 @@ set(SRCS
 add_library(cxx ${SRCS})
 set_target_properties(cxx PROPERTIES FOLDER "contrib/libcxx-cmake")

-target_include_directories(cxx SYSTEM BEFORE PUBLIC $<BUILD_INTERFACE:${LIBCXX_SOURCE_DIR}/include>)
+target_include_directories(cxx SYSTEM BEFORE PUBLIC
+        $<BUILD_INTERFACE:${LIBCXX_SOURCE_DIR}/include>
+        $<BUILD_INTERFACE:${LIBCXX_SOURCE_DIR}>/src)
 target_compile_definitions(cxx PRIVATE -D_LIBCPP_BUILDING_LIBRARY -DLIBCXX_BUILDING_LIBCXXABI)

 # Enable capturing stack traces for all exceptions.
--- a/contrib/libcxxabi
+++ b/contrib/libcxxabi
@ -1 +1 @@
-Subproject commit df8f1e727dbc9e2bedf2282096fa189dc3fe0076
+Subproject commit 6eb7cc7a7bdd779e6734d1b9fb451df2274462d7
--- a/contrib/libcxxabi-cmake/CMakeLists.txt
+++ b/contrib/libcxxabi-cmake/CMakeLists.txt
@ -1,24 +1,24 @@
 set(LIBCXXABI_SOURCE_DIR "${ClickHouse_SOURCE_DIR}/contrib/libcxxabi")

 set(SRCS
-"${LIBCXXABI_SOURCE_DIR}/src/stdlib_stdexcept.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/cxa_virtual.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/cxa_thread_atexit.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/fallback_malloc.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/cxa_guard.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/cxa_default_handlers.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/cxa_personality.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/stdlib_exception.cpp"
 "${LIBCXXABI_SOURCE_DIR}/src/abort_message.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/cxa_aux_runtime.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/cxa_default_handlers.cpp"
 "${LIBCXXABI_SOURCE_DIR}/src/cxa_demangle.cpp"
 "${LIBCXXABI_SOURCE_DIR}/src/cxa_exception.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/cxa_handlers.cpp"
 "${LIBCXXABI_SOURCE_DIR}/src/cxa_exception_storage.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/private_typeinfo.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/stdlib_typeinfo.cpp"
-"${LIBCXXABI_SOURCE_DIR}/src/cxa_aux_runtime.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/cxa_guard.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/cxa_handlers.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/cxa_personality.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/cxa_thread_atexit.cpp"
 "${LIBCXXABI_SOURCE_DIR}/src/cxa_vector.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/cxa_virtual.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/fallback_malloc.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/private_typeinfo.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/stdlib_exception.cpp"
 "${LIBCXXABI_SOURCE_DIR}/src/stdlib_new_delete.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/stdlib_stdexcept.cpp"
+"${LIBCXXABI_SOURCE_DIR}/src/stdlib_typeinfo.cpp"
 )

 add_library(cxxabi ${SRCS})
@ -30,6 +30,7 @@ target_compile_options(cxxabi PRIVATE -w)
 target_include_directories(cxxabi SYSTEM BEFORE
    PUBLIC $<BUILD_INTERFACE:${LIBCXXABI_SOURCE_DIR}/include>
    PRIVATE $<BUILD_INTERFACE:${LIBCXXABI_SOURCE_DIR}/../libcxx/include>
+    PRIVATE $<BUILD_INTERFACE:${LIBCXXABI_SOURCE_DIR}/../libcxx/src>
 )
 target_compile_definitions(cxxabi PRIVATE -D_LIBCPP_BUILDING_LIBRARY)
 target_compile_options(cxxabi PRIVATE -nostdinc++ -fno-sanitize=undefined -Wno-macro-redefined) # If we don't disable UBSan, infinite recursion happens in dynamic_cast.
--- a/contrib/replxx
+++ b/contrib/replxx
@ -1 +1 @@
-Subproject commit 9460e5e0fc10f78f460af26a6bd928798cac864d
+Subproject commit 6f0b6f151ae2a044625ae93acd19ca365fcea64d
--- a/docker/packager/packager
+++ b/docker/packager/packager
@ -163,6 +163,7 @@ def parse_env_variables(
        cmake_flags.append("-DCMAKE_INSTALL_PREFIX=/usr")
        cmake_flags.append("-DCMAKE_INSTALL_SYSCONFDIR=/etc")
        cmake_flags.append("-DCMAKE_INSTALL_LOCALSTATEDIR=/var")
+        cmake_flags.append("-DBUILD_STANDALONE_KEEPER=ON")
        if is_release_build(build_type, package_type, sanitizer, split_binary):
            cmake_flags.append("-DINSTALL_STRIPPED_BINARIES=ON")

@ -244,7 +245,7 @@ def parse_env_variables(
        result.append(f"AUTHOR='{author}'")

    if official:
-        cmake_flags.append("-DYANDEX_OFFICIAL_BUILD=1")
+        cmake_flags.append("-DCLICKHOUSE_OFFICIAL_BUILD=1")

    result.append('CMAKE_FLAGS="' + " ".join(cmake_flags) + '"')

--- a/docker/test/fasttest/run.sh
+++ b/docker/test/fasttest/run.sh
@ -267,6 +267,7 @@ function run_tests
    local test_opts=(
        --hung-check
        --fast-tests-only
+        --no-random-settings
        --no-long
        --testname
        --shard
--- a/docker/test/fuzzer/run-fuzzer.sh
+++ b/docker/test/fuzzer/run-fuzzer.sh
@ -13,7 +13,7 @@ script_dir="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
 echo "$script_dir"
 repo_dir=ch
 BINARY_TO_DOWNLOAD=${BINARY_TO_DOWNLOAD:="clang-13_debug_none_bundled_unsplitted_disable_False_binary"}
-BINARY_URL_TO_DOWNLOAD=${BINARY_URL_TO_DOWNLOAD:="https://clickhouse-builds.s3.yandex.net/$PR_TO_TEST/$SHA_TO_TEST/clickhouse_build_check/$BINARY_TO_DOWNLOAD/clickhouse"}
+BINARY_URL_TO_DOWNLOAD=${BINARY_URL_TO_DOWNLOAD:="https://clickhouse-builds.s3.amazonaws.com/$PR_TO_TEST/$SHA_TO_TEST/clickhouse_build_check/$BINARY_TO_DOWNLOAD/clickhouse"}

 function clone
 {
--- a/docker/test/keeper-jepsen/run.sh
+++ b/docker/test/keeper-jepsen/run.sh
@ -2,7 +2,7 @@
 set -euo pipefail


-CLICKHOUSE_PACKAGE=${CLICKHOUSE_PACKAGE:="https://clickhouse-builds.s3.yandex.net/$PR_TO_TEST/$SHA_TO_TEST/clickhouse_build_check/clang-13_relwithdebuginfo_none_bundled_unsplitted_disable_False_binary/clickhouse"}
+CLICKHOUSE_PACKAGE=${CLICKHOUSE_PACKAGE:="https://clickhouse-builds.s3.amazonaws.com/$PR_TO_TEST/$SHA_TO_TEST/clickhouse_build_check/clang-13_relwithdebuginfo_none_bundled_unsplitted_disable_False_binary/clickhouse"}
 CLICKHOUSE_REPO_PATH=${CLICKHOUSE_REPO_PATH:=""}


@ -10,7 +10,7 @@ if [ -z "$CLICKHOUSE_REPO_PATH" ]; then
    CLICKHOUSE_REPO_PATH=ch
    rm -rf ch ||:
    mkdir ch ||:
-    wget -nv -nd -c "https://clickhouse-test-reports.s3.yandex.net/$PR_TO_TEST/$SHA_TO_TEST/repo/clickhouse_no_subs.tar.gz"
+    wget -nv -nd -c "https://clickhouse-test-reports.s3.amazonaws.com/$PR_TO_TEST/$SHA_TO_TEST/repo/clickhouse_no_subs.tar.gz"
    tar -C ch --strip-components=1 -xf clickhouse_no_subs.tar.gz
    ls -lath ||:
 fi
--- a/docker/test/performance-comparison/compare.sh
+++ b/docker/test/performance-comparison/compare.sh
@ -1294,15 +1294,15 @@ create table ci_checks engine File(TSVWithNamesAndTypes, 'ci-checks.tsv')
        select '' test_name,
            '$(sed -n 's/.*<!--message: \(.*\)-->/\1/p' report.html)' test_status,
            0 test_duration_ms,
-            'https://clickhouse-test-reports.s3.yandex.net/$PR_TO_TEST/$SHA_TO_TEST/performance_comparison/report.html#fail1' report_url
+            'https://clickhouse-test-reports.s3.amazonaws.com/$PR_TO_TEST/$SHA_TO_TEST/performance_comparison/report.html#fail1' report_url
        union all
            select test || ' #' || toString(query_index), 'slower' test_status, 0 test_duration_ms,
-                'https://clickhouse-test-reports.s3.yandex.net/$PR_TO_TEST/$SHA_TO_TEST/performance_comparison/report.html#changes-in-performance.'
+                'https://clickhouse-test-reports.s3.amazonaws.com/$PR_TO_TEST/$SHA_TO_TEST/performance_comparison/report.html#changes-in-performance.'
                    || test || '.' || toString(query_index) report_url
            from queries where changed_fail != 0 and diff > 0
        union all
            select test || ' #' || toString(query_index), 'unstable' test_status, 0 test_duration_ms,
-                'https://clickhouse-test-reports.s3.yandex.net/$PR_TO_TEST/$SHA_TO_TEST/performance_comparison/report.html#unstable-queries.'
+                'https://clickhouse-test-reports.s3.amazonaws.com/$PR_TO_TEST/$SHA_TO_TEST/performance_comparison/report.html#unstable-queries.'
                    || test || '.' || toString(query_index) report_url
            from queries where unstable_fail != 0
    )
@ -1378,7 +1378,7 @@ $REF_SHA	$SHA_TO_TEST	$(numactl --hardware | sed -n 's/^available:[[:space:]]\+/
 EOF

    # Also insert some data about the check into the CI checks table.
-    "${client[@]}" --query "INSERT INTO "'"'"gh-data"'"'".checks FORMAT TSVWithNamesAndTypes" \
+    "${client[@]}" --query "INSERT INTO "'"'"default"'"'".checks FORMAT TSVWithNamesAndTypes" \
        < ci-checks.tsv

    set -x
--- a/docker/test/performance-comparison/download.sh
+++ b/docker/test/performance-comparison/download.sh
@ -16,26 +16,17 @@ right_sha=$4
 datasets=${CHPC_DATASETS-"hits1 hits10 hits100 values"}

 declare -A dataset_paths
-if [[ $S3_URL == *"s3.amazonaws.com"* ]]; then
-    dataset_paths["hits10"]="https://clickhouse-private-datasets.s3.amazonaws.com/hits_10m_single/partitions/hits_10m_single.tar"
-    dataset_paths["hits100"]="https://clickhouse-private-datasets.s3.amazonaws.com/hits_100m_single/partitions/hits_100m_single.tar"
-    dataset_paths["hits1"]="https://clickhouse-datasets.s3.amazonaws.com/hits/partitions/hits_v1.tar"
-    dataset_paths["values"]="https://clickhouse-datasets.s3.amazonaws.com/values_with_expressions/partitions/test_values.tar"
-else
-    dataset_paths["hits10"]="https://s3.mds.yandex.net/clickhouse-private-datasets/hits_10m_single/partitions/hits_10m_single.tar"
-    dataset_paths["hits100"]="https://s3.mds.yandex.net/clickhouse-private-datasets/hits_100m_single/partitions/hits_100m_single.tar"
-    dataset_paths["hits1"]="https://clickhouse-datasets.s3.yandex.net/hits/partitions/hits_v1.tar"
-    dataset_paths["values"]="https://clickhouse-datasets.s3.yandex.net/values_with_expressions/partitions/test_values.tar"
-fi
+dataset_paths["hits10"]="https://clickhouse-private-datasets.s3.amazonaws.com/hits_10m_single/partitions/hits_10m_single.tar"
+dataset_paths["hits100"]="https://clickhouse-private-datasets.s3.amazonaws.com/hits_100m_single/partitions/hits_100m_single.tar"
+dataset_paths["hits1"]="https://clickhouse-datasets.s3.amazonaws.com/hits/partitions/hits_v1.tar"
+dataset_paths["values"]="https://clickhouse-datasets.s3.amazonaws.com/values_with_expressions/partitions/test_values.tar"


 function download
 {
    # Historically there were various paths for the performance test package.
    # Test all of them.
-    declare -a urls_to_try=("https://s3.amazonaws.com/clickhouse-builds/$left_pr/$left_sha/performance/performance.tgz"
-                            "https://clickhouse-builds.s3.yandex.net/$left_pr/$left_sha/clickhouse_build_check/performance/performance.tgz"
-                           )
+    declare -a urls_to_try=("https://s3.amazonaws.com/clickhouse-builds/$left_pr/$left_sha/performance/performance.tgz")

    for path in "${urls_to_try[@]}"
    do
--- a/docker/test/performance-comparison/entrypoint.sh
+++ b/docker/test/performance-comparison/entrypoint.sh
@ -4,7 +4,7 @@ set -ex
 CHPC_CHECK_START_TIMESTAMP="$(date +%s)"
 export CHPC_CHECK_START_TIMESTAMP

-S3_URL=${S3_URL:="https://clickhouse-builds.s3.yandex.net"}
+S3_URL=${S3_URL:="https://clickhouse-builds.s3.amazonaws.com"}

 COMMON_BUILD_PREFIX="/clickhouse_build_check"
 if [[ $S3_URL == *"s3.amazonaws.com"* ]]; then
@ -64,9 +64,7 @@ function find_reference_sha
        # Historically there were various path for the performance test package,
        # test all of them.
        unset found
-        declare -a urls_to_try=("https://s3.amazonaws.com/clickhouse-builds/0/$REF_SHA/performance/performance.tgz"
-                                "https://clickhouse-builds.s3.yandex.net/0/$REF_SHA/clickhouse_build_check/performance/performance.tgz"
-                               )
+        declare -a urls_to_try=("https://s3.amazonaws.com/clickhouse-builds/0/$REF_SHA/performance/performance.tgz")
        for path in "${urls_to_try[@]}"
        do
            if curl_with_retry "$path"
--- a/docker/test/stateful/Dockerfile
+++ b/docker/test/stateful/Dockerfile
@ -11,7 +11,7 @@ RUN apt-get update -y \

 COPY s3downloader /s3downloader

-ENV S3_URL="https://clickhouse-datasets.s3.yandex.net"
+ENV S3_URL="https://clickhouse-datasets.s3.amazonaws.com"
 ENV DATASETS="hits visits"
 ENV EXPORT_S3_STORAGE_POLICIES=1

--- a/docker/test/stateful/run.sh
+++ b/docker/test/stateful/run.sh
@ -115,7 +115,7 @@ function run_tests()
    fi

    set +e
-    clickhouse-test --testname --shard --zookeeper --check-zookeeper-session --no-stateless --hung-check --print-time \
+    clickhouse-test -j 2 --testname --shard --zookeeper --check-zookeeper-session --no-stateless --hung-check --print-time \
        --skip 00168_parallel_processing_on_replicas "${ADDITIONAL_OPTIONS[@]}" \
        "$SKIP_TESTS_OPTION" 2>&1 | ts '%Y-%m-%d %H:%M:%S' | tee test_output/test_result.txt

--- a/docker/test/stateful/s3downloader
+++ b/docker/test/stateful/s3downloader
@ -10,7 +10,7 @@ import requests
 import tempfile


-DEFAULT_URL = 'https://clickhouse-datasets.s3.yandex.net'
+DEFAULT_URL = 'https://clickhouse-datasets.s3.amazonaws.com'

 AVAILABLE_DATASETS = {
    'hits': 'hits_v1.tar',
--- a/docker/test/stateless/run.sh
+++ b/docker/test/stateless/run.sh
@ -131,8 +131,23 @@ clickhouse-client -q "system flush logs" ||:

 grep -Fa "Fatal" /var/log/clickhouse-server/clickhouse-server.log ||:
 pigz < /var/log/clickhouse-server/clickhouse-server.log > /test_output/clickhouse-server.log.gz &
-clickhouse-client -q "select * from system.query_log format TSVWithNamesAndTypes" | pigz > /test_output/query-log.tsv.gz &
-clickhouse-client -q "select * from system.query_thread_log format TSVWithNamesAndTypes" | pigz > /test_output/query-thread-log.tsv.gz &
+
+# Compress tables.
+#
+# NOTE:
+# - that due to tests with s3 storage we cannot use /var/lib/clickhouse/data
+#   directly
+# - even though ci auto-compress some files (but not *.tsv) it does this only
+#   for files >64MB, we want this files to be compressed explicitly
+for table in query_log zookeeper_log trace_log
+do
+    clickhouse-client -q "select * from system.$table format TSVWithNamesAndTypes" | pigz > /test_output/$table.tsv.gz &
+    if [[ -n "$USE_DATABASE_REPLICATED" ]] && [[ "$USE_DATABASE_REPLICATED" -eq 1 ]]; then
+        clickhouse-client --port 19000 -q "select * from system.$table format TSVWithNamesAndTypes" | pigz > /test_output/$table.1.tsv.gz &
+        clickhouse-client --port 29000 -q "select * from system.$table format TSVWithNamesAndTypes" | pigz > /test_output/$table.2.tsv.gz &
+    fi
+done
+wait ||:

 # Also export trace log in flamegraph-friendly format.
 for trace_type in CPU Memory Real
@ -161,14 +176,6 @@ fi

 tar -chf /test_output/coordination.tar /var/lib/clickhouse/coordination ||:

-# Replace the engine with Ordinary to avoid extra symlinks stuff in artifacts.
-# (so that clickhouse-local --path can read it w/o extra care).
-sed -i -e "s/ATTACH DATABASE _ UUID '[^']*'/ATTACH DATABASE system/" -e "s/Atomic/Ordinary/" /var/lib/clickhouse/metadata/system.sql
-for table in text_log query_log zookeeper_log trace_log; do
-    sed -i "s/ATTACH TABLE _ UUID '[^']*'/ATTACH TABLE $table/" /var/lib/clickhouse/metadata/system/${table}.sql
-    tar -chf /test_output/${table}_dump.tar /var/lib/clickhouse/metadata/system.sql /var/lib/clickhouse/metadata/system/${table}.sql /var/lib/clickhouse/data/system/${table} ||:
-done
-
 if [[ -n "$USE_DATABASE_REPLICATED" ]] && [[ "$USE_DATABASE_REPLICATED" -eq 1 ]]; then
    grep -Fa "Fatal" /var/log/clickhouse-server/clickhouse-server1.log ||:
    grep -Fa "Fatal" /var/log/clickhouse-server/clickhouse-server2.log ||:
@ -179,8 +186,6 @@ if [[ -n "$USE_DATABASE_REPLICATED" ]] && [[ "$USE_DATABASE_REPLICATED" -eq 1 ]]
    rm /var/log/clickhouse-server/clickhouse-server2.log
    mv /var/log/clickhouse-server/stderr1.log /test_output/ ||:
    mv /var/log/clickhouse-server/stderr2.log /test_output/ ||:
-    tar -chf /test_output/zookeeper_log_dump1.tar /var/lib/clickhouse1/data/system/zookeeper_log ||:
-    tar -chf /test_output/zookeeper_log_dump2.tar /var/lib/clickhouse2/data/system/zookeeper_log ||:
    tar -chf /test_output/coordination1.tar /var/lib/clickhouse1/coordination ||:
    tar -chf /test_output/coordination2.tar /var/lib/clickhouse2/coordination ||:
 fi
--- a/docker/test/stateless/setup_minio.sh
+++ b/docker/test/stateless/setup_minio.sh
@ -41,6 +41,7 @@ sleep 5
 ./mc admin user add clickminio test testtest
 ./mc admin policy set clickminio readwrite user=test
 ./mc mb clickminio/test
+./mc policy set public clickminio/test


 # Upload data to Minio. By default after unpacking all tests will in
--- a/docker/test/stress/Dockerfile
+++ b/docker/test/stress/Dockerfile
@ -29,7 +29,7 @@ COPY ./download_previous_release /download_previous_release
 COPY run.sh /

 ENV DATASETS="hits visits"
-ENV S3_URL="https://clickhouse-datasets.s3.yandex.net"
+ENV S3_URL="https://clickhouse-datasets.s3.amazonaws.com"
 ENV EXPORT_S3_STORAGE_POLICIES=1

 CMD ["/bin/bash", "/run.sh"]
--- a/docs/en/engines/table-engines/mergetree-family/mergetree.md
+++ b/docs/en/engines/table-engines/mergetree-family/mergetree.md
@ -688,7 +688,7 @@ Tags:
 -   `volume_name_N` — Volume name. Volume names must be unique.
 -   `disk` — a disk within a volume.
 -   `max_data_part_size_bytes` — the maximum size of a part that can be stored on any of the volume’s disks. If the a size of a merged part estimated to be bigger than `max_data_part_size_bytes` then this part will be written to a next volume. Basically this feature allows to keep new/small parts on a hot (SSD) volume and move them to a cold (HDD) volume when they reach large size. Do not use this setting if your policy has only one volume.
-   `move_factor` — when the amount of available space gets lower than this factor, data automatically start to move on the next volume if any (by default, 0.1).
+-   `move_factor` — when the amount of available space gets lower than this factor, data automatically starts to move on the next volume if any (by default, 0.1). ClickHouse sorts existing parts by size from largest to smallest (in descending order) and selects parts with the total size that is sufficient to meet the `move_factor` condition. If the total size of all parts is insufficient, all parts will be moved. 
 -   `prefer_not_to_merge` — Disables merging of data parts on this volume. When this setting is enabled, merging data on this volume is not allowed. This allows controlling how ClickHouse works with slow disks.

 Cofiguration examples:
--- a/docs/en/operations/named-collections.md
+++ b/docs/en/operations/named-collections.md
@ -36,6 +36,7 @@ Example of configuration:
            <access_key_id>AKIAIOSFODNN7EXAMPLE</access_key_id>
            <secret_access_key> wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY</secret_access_key>
            <format>CSV</format>
+            <url>https://s3.us-east-1.amazonaws.com/yourbucket/mydata/</url>
        </s3_mydata>
    </named_collections>
 </clickhouse>
@ -44,12 +45,12 @@ Example of configuration:
 ### Example of using named connections with the s3 function

 ```sql
-INSERT INTO FUNCTION s3(s3_mydata, url = 'https://s3.us-east-1.amazonaws.com/yourbucket/mydata/test_file.tsv.gz',
+INSERT INTO FUNCTION s3(s3_mydata, filename = 'test_file.tsv.gz',
   format = 'TSV', structure = 'number UInt64', compression_method = 'gzip')
 SELECT * FROM numbers(10000);

 SELECT count()
-FROM s3(s3_mydata, url = 'https://s3.us-east-1.amazonaws.com/yourbucket/mydata/test_file.tsv.gz')
+FROM s3(s3_mydata, filename = 'test_file.tsv.gz')

 ┌─count()─┐
 │   10000 │
--- a/docs/en/operations/server-configuration-parameters/settings.md
+++ b/docs/en/operations/server-configuration-parameters/settings.md
@ -1616,3 +1616,14 @@ Possible values:

 Default value: `10000`.

+## global_memory_usage_overcommit_max_wait_microseconds {#global_memory_usage_overcommit_max_wait_microseconds}
+
+Sets maximum waiting time for global overcommit tracker.
+
+Possible values:
+
+-   Positive integer.
+
+Default value: `0`.
+
+
--- a/docs/en/operations/settings/memory-overcommit.md
+++ b/docs/en/operations/settings/memory-overcommit.md
@ -0,0 +1,31 @@
+# Memory overcommit
+
+Memory overcommit is an experimental technique intended to allow to set more flexible memory limits for queries.
+
+The idea of this technique is to introduce settings which can represent guaranteed amount of memory a query can use.
+When memory overcommit is enabled and the memory limit is reached ClickHouse will select the most overcommitted query and try to free memory by killing this query.
+
+When memory limit is reached any query will wait some time during atempt to allocate new memory.
+If timeout is passed and memory is freed, the query continues execution. Otherwise an exception will be thrown and the query is killed.
+
+Selection of query to stop or kill is performed by either global or user overcommit trackers depending on what memory limit is reached.
+
+## User overcommit tracker
+
+User overcommit tracker finds a query with the biggest overcommit ratio in the user's query list.
+Overcommit ratio is computed as number of allocated bytes divided by value of `max_guaranteed_memory_usage` setting.
+
+Waiting timeout is set by `memory_usage_overcommit_max_wait_microseconds` setting.
+
+**Example**
+
+```sql
+SELECT number FROM numbers(1000) GROUP BY number SETTINGS max_guaranteed_memory_usage=4000, memory_usage_overcommit_max_wait_microseconds=500
+```
+
+## Global overcommit tracker
+
+Global overcommit tracker finds a query with the biggest overcommit ratio in the list of all queries.
+In this case overcommit ratio is computed as number of allocated bytes divided by value of `max_guaranteed_memory_usage_for_user` setting.
+
+Waiting timeout is set by `global_memory_usage_overcommit_max_wait_microseconds` parameter in the configuration file.
--- a/docs/en/operations/settings/settings.md
+++ b/docs/en/operations/settings/settings.md
@ -4220,10 +4220,36 @@ Possible values:
 -   0 — Disabled.
 -   1 — Enabled. The wait time equal shutdown_wait_unfinished config.

-Default value: 0.
+Default value: `0`.

 ## shutdown_wait_unfinished

 The waiting time in seconds for currently handled connections when shutdown server.

-Default Value: 5.
+Default Value: `5`.
+
+## max_guaranteed_memory_usage
+
+Maximum guaranteed memory usage for processing of single query.
+It represents soft limit in case when hard limit is reached on user level.
+Zero means unlimited.
+Read more about [memory overcommit](memory-overcommit.md).
+
+Default value: `0`.
+
+## memory_usage_overcommit_max_wait_microseconds
+
+Maximum time thread will wait for memory to be freed in the case of memory overcommit on a user level.
+If the timeout is reached and memory is not freed, an exception is thrown.
+Read more about [memory overcommit](memory-overcommit.md).
+
+Default value: `0`.
+
+## max_guaranteed_memory_usage_for_user
+
+Maximum guaranteed memory usage for processing all concurrently running queries for the user.
+It represents soft limit in case when hard limit is reached on global level.
+Zero means unlimited.
+Read more about [memory overcommit](memory-overcommit.md).
+
+Default value: `0`.
--- a/docs/en/sql-reference/aggregate-functions/reference/grouparraysorted.md
+++ b/docs/en/sql-reference/aggregate-functions/reference/grouparraysorted.md
@ -0,0 +1,48 @@
+---
+toc_priority: 108
+---
+
+# groupArraySorted {#groupArraySorted}
+
+Returns an array with the first N items in ascending order.
+
+``` sql
+groupArraySorted(N)(column)
+```
+
+**Arguments**
+
+-   `N` – The number of elements to return.
+
+If the parameter is omitted, default value 10 is used.
+
+**Arguments**
+
+-   `column` – The value.
+-   `expr` — Optional. The field or expresion to sort by. If not set values are sorted by themselves.
+
+**Example**
+
+Gets the first 10 numbers:
+
+``` sql
+SELECT groupArraySorted(10)(number) FROM numbers(100)
+```
+
+``` text
+┌─groupArraySorted(10)(number)─┐
+│ [0,1,2,3,4,5,6,7,8,9]        │
+└──────────────────────────────┘
+```
+
+Or the last 10:
+
+``` sql
+SELECT groupArraySorted(10)(number, -number) FROM numbers(100)
+```
+
+``` text
+┌─groupArraySorted(10)(number, negate(number))─┐
+│ [99,98,97,96,95,94,93,92,91,90]              │
+└──────────────────────────────────────────────┘
+```
--- a/docs/en/sql-reference/aggregate-functions/reference/index.md
+++ b/docs/en/sql-reference/aggregate-functions/reference/index.md
@ -35,6 +35,7 @@ ClickHouse-specific aggregate functions:
 -   [groupArrayInsertAt](../../../sql-reference/aggregate-functions/reference/grouparrayinsertat.md)
 -   [groupArrayMovingAvg](../../../sql-reference/aggregate-functions/reference/grouparraymovingavg.md)
 -   [groupArrayMovingSum](../../../sql-reference/aggregate-functions/reference/grouparraymovingsum.md)
+-   [groupArraySorted](../../../sql-reference/aggregate-functions/reference/grouparraysorted.md)
 -   [groupBitAnd](../../../sql-reference/aggregate-functions/reference/groupbitand.md)
 -   [groupBitOr](../../../sql-reference/aggregate-functions/reference/groupbitor.md)
 -   [groupBitXor](../../../sql-reference/aggregate-functions/reference/groupbitxor.md)
--- a/docs/ja/development/developer-instruction.md
+++ b/docs/ja/development/developer-instruction.md
@ -273,7 +273,7 @@ GitHubのUIでforkリポジトリに移動します。 ブランチで開発し

 プル要求は、作業がまだ完了していない場合でも作成できます。 この場合、単語を入れてください “WIP” （進行中の作業）タイトルの先頭に、それは後で変更することができます。 これは、変更の協調的なレビューと議論、および利用可能なすべてのテストの実行に役立ちます。 変更の簡単な説明を提供することが重要です。

-Yandexの従業員がタグであなたのPRにラベルを付けるとすぐにテストが開始されます “can be tested”. The results of some first checks (e.g. code style) will come in within several minutes. Build check results will arrive within half an hour. And the main set of tests will report itself within an hour.
+ClickHouseの従業員がタグであなたのPRにラベルを付けるとすぐにテストが開始されます “can be tested”. The results of some first checks (e.g. code style) will come in within several minutes. Build check results will arrive within half an hour. And the main set of tests will report itself within an hour.

 システムは、プル要求用にClickHouseバイナリビルドを個別に準備します。 これらのビルドを取得するには “Details” 次のリンク “ClickHouse build check” 小切手のリストのエントリ。 そこには、ビルドへの直接リンクがあります。ClickHouseのdebパッケージは、本番サーバーにも展開できます（恐れがない場合）。

--- a/docs/ja/sql-reference/aggregate-functions/reference/grouparraysorted.md
+++ b/docs/ja/sql-reference/aggregate-functions/reference/grouparraysorted.md
--- a/docs/ru/development/developer-instruction.md
+++ b/docs/ru/development/developer-instruction.md
@ -72,11 +72,11 @@ ClickHouse не работает и не собирается на 32-битны

 Этот вариант не подходит для отправки изменений на сервер. Вы можете временно его использовать, а затем добавить ssh ключи и заменить адрес репозитория с помощью команды `git remote`.

-Вы можете также добавить для своего локального репозитория адрес оригинального репозитория Яндекса, чтобы притягивать оттуда обновления:
+Вы можете также добавить для своего локального репозитория адрес оригинального репозитория, чтобы притягивать оттуда обновления:

    git remote add upstream git@github.com:ClickHouse/ClickHouse.git

-После этого, вы сможете добавлять в свой репозиторий обновления из репозитория Яндекса с помощью команды `git pull upstream master`.
+После этого, вы сможете добавлять в свой репозиторий обновления из репозитория ClickHouse с помощью команды `git pull upstream master`.

 ### Работа с сабмодулями Git {#rabota-s-sabmoduliami-git}

@ -288,7 +288,7 @@ sudo ./llvm.sh 12

 Pull request можно создать, даже если работа над задачей ещё не завершена. В этом случае, добавьте в его название слово «WIP» (work in progress). Название можно будет изменить позже. Это полезно для совместного просмотра и обсуждения изменений, а также для запуска всех имеющихся тестов. Введите краткое описание изменений - впоследствии, оно будет использовано для релизных changelog.

-Тесты будут запущены, как только сотрудники Яндекса поставят для pull request тег «Can be tested». Результаты первых проверок (стиль кода) появятся уже через несколько минут. Результаты сборки появятся примерно через пол часа. Результаты основного набора тестов будут доступны в пределах часа.
+Тесты будут запущены, как только сотрудники ClickHouse поставят для pull request тег «Can be tested». Результаты первых проверок (стиль кода) появятся уже через несколько минут. Результаты сборки появятся примерно через пол часа. Результаты основного набора тестов будут доступны в пределах часа.

 Система подготовит сборки ClickHouse специально для вашего pull request. Для их получения, нажмите на ссылку «Details» у проверки «Clickhouse build check». Там вы сможете найти прямые ссылки на собранные .deb пакеты ClickHouse, которые, при желании, вы даже сможете установить на свои продакшен серверы (если не страшно).

--- a/docs/ru/engines/table-engines/mergetree-family/mergetree.md
+++ b/docs/ru/engines/table-engines/mergetree-family/mergetree.md
@ -678,7 +678,7 @@ TTL d + INTERVAL 1 MONTH GROUP BY k1, k2 SET x = max(x), y = min(y);
 -   `volume_name_N` — название тома. Названия томов должны быть уникальны.
 -   `disk` — диск, находящийся внутри тома.
 -   `max_data_part_size_bytes` — максимальный размер куска данных, который может находится на любом из дисков этого тома. Если в результате слияния размер куска ожидается больше, чем max_data_part_size_bytes, то этот кусок будет записан в следующий том. В основном эта функция позволяет хранить новые / мелкие куски на горячем (SSD) томе и перемещать их на холодный (HDD) том, когда они достигают большого размера. Не используйте этот параметр, если политика имеет только один том. 
-   `move_factor` — доля доступного свободного места на томе, если места становится меньше, то данные начнут перемещение на следующий том, если он есть (по умолчанию 0.1).
+-   `move_factor` — доля доступного свободного места на томе, если места становится меньше, то данные начнут перемещение на следующий том, если он есть (по умолчанию 0.1). Для перемещения куски сортируются по размеру от большего к меньшему (по убыванию) и выбираются куски, совокупный размер которых достаточен для соблюдения условия `move_factor`, если совокупный размер всех партов недостаточен, будут перемещены все парты.
 -   `prefer_not_to_merge` — Отключает слияние кусков данных, хранящихся на данном томе. Если данная настройка включена, то слияние данных, хранящихся на данном томе, не допускается. Это позволяет контролировать работу ClickHouse с медленными дисками.

 Примеры конфигураций:
--- a/docs/zh/development/developer-instruction.md
+++ b/docs/zh/development/developer-instruction.md
@ -259,7 +259,7 @@ ClickHouse的架构描述可以在此处查看：https://clickhouse.com/docs/en/

 即使工作尚未完成，也可以创建拉取请求。在这种情况下，请在标题的开头加上«WIP»（正在进行中），以便后续更改。这对于协同审查和讨论更改以及运行所有可用测试用例很有用。提供有关变更的简短描述很重要，这将在后续用于生成重新发布变更日志。

-Yandex成员一旦在您的拉取请求上贴上«可以测试»标签，就会开始测试。一些初始检查项（例如，代码类型）的结果会在几分钟内反馈。构建的检查结果将在半小时内完成。而主要的测试用例集结果将在一小时内报告给您。
+ClickHouse成员一旦在您的拉取请求上贴上«可以测试»标签，就会开始测试。一些初始检查项（例如，代码类型）的结果会在几分钟内反馈。构建的检查结果将在半小时内完成。而主要的测试用例集结果将在一小时内报告给您。

 系统将分别为您的拉取请求准备ClickHouse二进制版本。若要检索这些构建信息，请在检查列表中单击« ClickHouse构建检查»旁边的«详细信息»链接。在这里，您会找到指向ClickHouse的.deb软件包的直接链接，此外，甚至可以将其部署在生产服务器上（如果您不担心）。

--- a/docs/zh/engines/table-engines/mergetree-family/versionedcollapsingmergetree.md
+++ b/docs/zh/engines/table-engines/mergetree-family/versionedcollapsingmergetree.md
@ -8,7 +8,7 @@ toc_title: "版本折叠MergeTree"
 这个引擎:

 -   允许快速写入不断变化的对象状态。
-   删除后台中的旧对象状态。 这显着降低了存储体积。
+-   删除后台中的旧对象状态。 这显著降低了存储体积。

 请参阅部分 [崩溃](#table_engines_versionedcollapsingmergetree) 有关详细信息。

--- a/packages/clickhouse-keeper-dbg.yaml
+++ b/packages/clickhouse-keeper-dbg.yaml
@ -0,0 +1,28 @@
+# package sources should be placed in ${PWD}/root
+# nfpm should run from the same directory with a config
+name: "clickhouse-keeper-dbg"
+arch: "${DEB_ARCH}" # amd64, arm64
+platform: "linux"
+version: "${CLICKHOUSE_VERSION_STRING}"
+vendor: "ClickHouse Inc."
+homepage: "https://clickhouse.com"
+license: "Apache"
+section: "database"
+priority: "optional"
+maintainer: "ClickHouse Dev Team <packages+linux@clickhouse.com>"
+description: |
+  debugging symbols for clickhouse-keeper
+    This package contains the debugging symbols for clickhouse-keeper.
+
+contents:
+- src: root/usr/lib/debug/usr/bin/clickhouse-keeper.debug
+  dst: /usr/lib/debug/usr/bin/clickhouse-keeper.debug
+# docs
+- src: ../AUTHORS
+  dst: /usr/share/doc/clickhouse-keeper-dbg/AUTHORS
+- src: ../CHANGELOG.md
+  dst: /usr/share/doc/clickhouse-keeper-dbg/CHANGELOG.md
+- src: ../LICENSE
+  dst: /usr/share/doc/clickhouse-keeper-dbg/LICENSE
+- src: ../README.md
+  dst: /usr/share/doc/clickhouse-keeper-dbg/README.md
--- a/packages/clickhouse-keeper.yaml
+++ b/packages/clickhouse-keeper.yaml
@ -0,0 +1,40 @@
+# package sources should be placed in ${PWD}/root
+# nfpm should run from the same directory with a config
+name: "clickhouse-keeper"
+arch: "${DEB_ARCH}" # amd64, arm64
+platform: "linux"
+version: "${CLICKHOUSE_VERSION_STRING}"
+vendor: "ClickHouse Inc."
+homepage: "https://clickhouse.com"
+license: "Apache"
+section: "database"
+priority: "optional"
+
+conflicts:
+- clickhouse-server
+depends:
+- adduser
+suggests:
+- clickhouse-keeper-dbg
+
+maintainer: "ClickHouse Dev Team <packages+linux@clickhouse.com>"
+description: |
+  Static clickhouse-keeper binary
+    A stand-alone clickhouse-keeper package
+
+
+contents:
+- src: root/etc/clickhouse-keeper
+  dst: /etc/clickhouse-keeper
+  type: config
+- src: root/usr/bin/clickhouse-keeper
+  dst: /usr/bin/clickhouse-keeper
+# docs
+- src: ../AUTHORS
+  dst: /usr/share/doc/clickhouse-keeper/AUTHORS
+- src: ../CHANGELOG.md
+  dst: /usr/share/doc/clickhouse-keeper/CHANGELOG.md
+- src: ../LICENSE
+  dst: /usr/share/doc/clickhouse-keeper/LICENSE
+- src: ../README.md
+  dst: /usr/share/doc/clickhouse-keeper/README.md
--- a/programs/keeper/CMakeLists.txt
+++ b/programs/keeper/CMakeLists.txt
@ -71,17 +71,11 @@ if (BUILD_STANDALONE_KEEPER)
        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressedReadBuffer.cpp
        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressedReadBufferFromFile.cpp
        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressedWriteBuffer.cpp
-        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressionCodecDelta.cpp
-        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressionCodecDoubleDelta.cpp
-        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressionCodecEncrypted.cpp
-        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressionCodecGorilla.cpp
        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressionCodecLZ4.cpp
        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressionCodecMultiple.cpp
        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressionCodecNone.cpp
-        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressionCodecT64.cpp
        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressionCodecZSTD.cpp
        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/CompressionFactory.cpp
-        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/getCompressionCodecForFile.cpp
        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/ICompressionCodec.cpp
        ${CMAKE_CURRENT_SOURCE_DIR}/../../src/Compression/LZ4_decompress_faster.cpp

--- a/programs/local/LocalServer.cpp
+++ b/programs/local/LocalServer.cpp
@ -184,6 +184,11 @@ void LocalServer::tryInitPath()
    if (path.back() != '/')
        path += '/';

+    fs::create_directories(fs::path(path) / "user_defined/");
+    fs::create_directories(fs::path(path) / "data/");
+    fs::create_directories(fs::path(path) / "metadata/");
+    fs::create_directories(fs::path(path) / "metadata_dropped/");
+
    global_context->setPath(path);

    global_context->setTemporaryStorage(path + "tmp");
@ -565,7 +570,6 @@ void LocalServer::processConfig()
        /// Lock path directory before read
        status.emplace(fs::path(path) / "status", StatusFile::write_full_info);

-        fs::create_directories(fs::path(path) / "user_defined/");
        LOG_DEBUG(log, "Loading user defined objects from {}", path);
        Poco::File(path + "user_defined/").createDirectories();
        UserDefinedSQLObjectsLoader::instance().loadObjects(global_context);
@ -573,9 +577,6 @@ void LocalServer::processConfig()
        LOG_DEBUG(log, "Loaded user defined objects.");

        LOG_DEBUG(log, "Loading metadata from {}", path);
-        fs::create_directories(fs::path(path) / "data/");
-        fs::create_directories(fs::path(path) / "metadata/");
-
        loadMetadataSystem(global_context);
        attachSystemTablesLocal(global_context, *createMemoryDatabaseIfNotExists(global_context, DatabaseCatalog::SYSTEM_DATABASE));
        attachInformationSchema(global_context, *createMemoryDatabaseIfNotExists(global_context, DatabaseCatalog::INFORMATION_SCHEMA));
--- a/programs/server/Server.cpp
+++ b/programs/server/Server.cpp
@ -20,6 +20,7 @@
 #include <base/phdr_cache.h>
 #include <base/ErrorHandlers.h>
 #include <base/getMemoryAmount.h>
+#include <base/getAvailableMemoryAmount.h>
 #include <base/errnoToString.h>
 #include <base/coverage.h>
 #include <base/getFQDNOrHostName.h>
@ -45,6 +46,7 @@
 #include <Core/ServerUUID.h>
 #include <IO/HTTPCommon.h>
 #include <IO/ReadHelpers.h>
+#include <IO/ReadBufferFromFile.h>
 #include <IO/IOThreadPool.h>
 #include <IO/UseSSL.h>
 #include <Interpreters/AsynchronousMetrics.h>
@ -80,6 +82,7 @@
 #include <Common/SensitiveDataMasker.h>
 #include <Common/ThreadFuzzer.h>
 #include <Common/getHashOfLoadedBinary.h>
+#include <Common/filesystemHelpers.h>
 #include <Common/Elf.h>
 #include <Server/MySQLHandlerFactory.h>
 #include <Server/PostgreSQLHandlerFactory.h>
@ -505,6 +508,101 @@ void checkForUsersNotInMainConfig(
    }
 }

+/// Unused in other builds
+#if defined(OS_LINUX)
+static String readString(const String & path)
+{
+    ReadBufferFromFile in(path);
+    String contents;
+    readStringUntilEOF(contents, in);
+    return contents;
+}
+
+static int readNumber(const String & path)
+{
+    ReadBufferFromFile in(path);
+    int result;
+    readText(result, in);
+    return result;
+}
+
+#endif
+
+static void sanityChecks(Server * server)
+{
+    std::string data_path = getCanonicalPath(server->config().getString("path", DBMS_DEFAULT_PATH));
+    std::string logs_path = server->config().getString("logger.log", "");
+
+#if defined(OS_LINUX)
+    try
+    {
+        if (readString("/sys/devices/system/clocksource/clocksource0/current_clocksource").find("tsc") == std::string::npos)
+            server->context()->addWarningMessage("Linux is not using fast TSC clock source. Performance can be degraded.");
+    }
+    catch (...)
+    {
+    }
+
+    try
+    {
+        if (readNumber("/proc/sys/vm/overcommit_memory") == 2)
+            server->context()->addWarningMessage("Linux memory overcommit is disabled.");
+    }
+    catch (...)
+    {
+    }
+
+    try
+    {
+        if (readString("/sys/kernel/mm/transparent_hugepage/enabled").find("[always]") != std::string::npos)
+            server->context()->addWarningMessage("Linux transparent hugepage are set to \"always\".");
+    }
+    catch (...)
+    {
+    }
+
+    try
+    {
+        if (readNumber("/proc/sys/kernel/pid_max") < 30000)
+            server->context()->addWarningMessage("Linux max PID is too low.");
+    }
+    catch (...)
+    {
+    }
+
+    try
+    {
+        if (readNumber("/proc/sys/kernel/threads-max") < 30000)
+            server->context()->addWarningMessage("Linux threads max count is too low.");
+    }
+    catch (...)
+    {
+    }
+
+    std::string dev_id = getBlockDeviceId(data_path);
+    if (getBlockDeviceType(dev_id) == BlockDeviceType::ROT && getBlockDeviceReadAheadBytes(dev_id) == 0)
+        server->context()->addWarningMessage("Rotational disk with disabled readahead is in use. Performance can be degraded.");
+#endif
+
+    try
+    {
+        if (getAvailableMemoryAmount() < (2l << 30))
+            server->context()->addWarningMessage("Available memory at server startup is too low (2GiB).");
+
+        if (!enoughSpaceInDirectory(data_path, 1ull << 30))
+            server->context()->addWarningMessage("Available disk space at server startup is too low (1GiB).");
+
+        if (!logs_path.empty())
+        {
+            if (!enoughSpaceInDirectory(fs::path(logs_path).parent_path(), 1ull << 30))
+                server->context()->addWarningMessage("Available disk space at server startup is too low (1GiB).");
+        }
+    }
+    catch (...)
+    {
+    }
+}
+
 int Server::main(const std::vector<std::string> & /*args*/)
 {
    Poco::Logger * log = &logger();
@ -538,13 +636,14 @@ int Server::main(const std::vector<std::string> & /*args*/)
    global_context->addWarningMessage("Server was built in debug mode. It will work slowly.");
 #endif

-if (ThreadFuzzer::instance().isEffective())
-    global_context->addWarningMessage("ThreadFuzzer is enabled. Application will run slowly and unstable.");
+    if (ThreadFuzzer::instance().isEffective())
+        global_context->addWarningMessage("ThreadFuzzer is enabled. Application will run slowly and unstable.");

 #if defined(SANITIZER)
    global_context->addWarningMessage("Server was built with sanitizer. It will work slowly.");
 #endif

+    sanityChecks(this);

    // Initialize global thread pool. Do it before we fetch configs from zookeeper
    // nodes (`from_zk`), because ZooKeeper interface uses the pool. We will
@ -766,6 +865,38 @@ if (ThreadFuzzer::instance().isEffective())
        }
    }

+    /// Try to increase limit on number of threads.
+    {
+        rlimit rlim;
+        if (getrlimit(RLIMIT_NPROC, &rlim))
+            throw Poco::Exception("Cannot getrlimit");
+
+        if (rlim.rlim_cur == rlim.rlim_max)
+        {
+            LOG_DEBUG(log, "rlimit on number of threads is {}", rlim.rlim_cur);
+        }
+        else
+        {
+            rlim_t old = rlim.rlim_cur;
+            rlim.rlim_cur = rlim.rlim_max;
+            int rc = setrlimit(RLIMIT_NPROC, &rlim);
+            if (rc != 0)
+            {
+                LOG_WARNING(log, "Cannot set max number of threads to {}. error: {}", rlim.rlim_cur, strerror(errno));
+                rlim.rlim_cur = old;
+            }
+            else
+            {
+                LOG_DEBUG(log, "Set max number of threads to {} (was {}).", rlim.rlim_cur, old);
+            }
+        }
+
+        if (rlim.rlim_cur < 30000)
+        {
+            global_context->addWarningMessage("Maximum number of threads is lower than 30000. There could be problems with handling a lot of simultaneous queries.");
+        }
+    }
+
    static ServerErrorHandler error_handler;
    Poco::ErrorHandler::set(&error_handler);

@ -829,6 +960,36 @@ if (ThreadFuzzer::instance().isEffective())
        fs::create_directories(path / "metadata_dropped/");
    }

+#if USE_ROCKSDB
+    /// Initialize merge tree metadata cache
+    if (config().has("merge_tree_metadata_cache"))
+    {
+        fs::create_directories(path / "rocksdb/");
+        size_t size = config().getUInt64("merge_tree_metadata_cache.lru_cache_size", 256 << 20);
+        bool continue_if_corrupted = config().getBool("merge_tree_metadata_cache.continue_if_corrupted", false);
+        try
+        {
+            LOG_DEBUG(
+                log, "Initiailizing merge tree metadata cache lru_cache_size:{} continue_if_corrupted:{}", size, continue_if_corrupted);
+            global_context->initializeMergeTreeMetadataCache(path_str + "/" + "rocksdb", size);
+        }
+        catch (...)
+        {
+            if (continue_if_corrupted)
+            {
+                /// Rename rocksdb directory and reinitialize merge tree metadata cache
+                time_t now = time(nullptr);
+                fs::rename(path / "rocksdb", path / ("rocksdb.old." + std::to_string(now)));
+                global_context->initializeMergeTreeMetadataCache(path_str + "/" + "rocksdb", size);
+            }
+            else
+            {
+                throw;
+            }
+        }
+    }
+#endif
+
    if (config().has("interserver_http_port") && config().has("interserver_https_port"))
        throw Exception("Both http and https interserver ports are specified", ErrorCodes::EXCESSIVE_ELEMENT_IN_CONFIG);

--- a/programs/server/config.xml
+++ b/programs/server/config.xml
@ -1294,4 +1294,10 @@
        </tables>
    </rocksdb>
    -->
+
+    <!-- Uncomment if enable merge tree metadata cache -->
+    <merge_tree_metadata_cache>
+        <lru_cache_size>268435456</lru_cache_size>
+        <continue_if_corrupted>true</continue_if_corrupted>
+    </merge_tree_metadata_cache>
 </clickhouse>
--- a/src/Access/Common/QuotaDefs.h
+++ b/src/Access/Common/QuotaDefs.h
@ -13,7 +13,7 @@ enum class QuotaType
 {
    QUERIES,        /// Number of queries.
    QUERY_SELECTS,  /// Number of select queries.
-    QUERY_INSERTS,  /// Number of inserts queries.
+    QUERY_INSERTS,  /// Number of insert queries.
    ERRORS,         /// Number of queries with exceptions.
    RESULT_ROWS,    /// Number of rows returned as result.
    RESULT_BYTES,   /// Number of bytes returned as result.
--- a/src/AggregateFunctions/AggregateFunctionGroupArraySorted.cpp
+++ b/src/AggregateFunctions/AggregateFunctionGroupArraySorted.cpp
@ -0,0 +1,147 @@
+#include <AggregateFunctions/AggregateFunctionFactory.h>
+#include <AggregateFunctions/AggregateFunctionGroupArraySorted.h>
+#include <AggregateFunctions/FactoryHelpers.h>
+#include <AggregateFunctions/Helpers.h>
+#include <DataTypes/DataTypeDate.h>
+#include <DataTypes/DataTypeDateTime.h>
+#include <DataTypes/DataTypeString.h>
+#include <Common/FieldVisitorConvertToNumber.h>
+
+
+static inline constexpr UInt64 GROUP_SORTED_ARRAY_MAX_SIZE = 0xFFFFFF;
+static inline constexpr UInt64 GROUP_SORTED_ARRAY_DEFAULT_THRESHOLD = 10;
+
+
+namespace DB
+{
+struct Settings;
+
+namespace ErrorCodes
+{
+    extern const int ARGUMENT_OUT_OF_BOUND;
+    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
+    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
+}
+
+
+namespace
+{
+    template <typename T, bool expr_sorted, typename TColumnB, bool is_plain_b>
+    class AggregateFunctionGroupArraySortedNumeric : public AggregateFunctionGroupArraySorted<T, false, expr_sorted, TColumnB, is_plain_b>
+    {
+        using AggregateFunctionGroupArraySorted<T, false, expr_sorted, TColumnB, is_plain_b>::AggregateFunctionGroupArraySorted;
+    };
+
+    template <typename T, bool expr_sorted, typename TColumnB, bool is_plain_b>
+    class AggregateFunctionGroupArraySortedFieldType
+        : public AggregateFunctionGroupArraySorted<typename T::FieldType, false, expr_sorted, TColumnB, is_plain_b>
+    {
+        using AggregateFunctionGroupArraySorted<typename T::FieldType, false, expr_sorted, TColumnB, is_plain_b>::
+            AggregateFunctionGroupArraySorted;
+        DataTypePtr getReturnType() const override { return std::make_shared<DataTypeArray>(std::make_shared<T>()); }
+    };
+
+    template <template <typename, bool, typename, bool> class AggregateFunctionTemplate, typename TColumnA, bool expr_sorted, typename TColumnB, bool is_plain_b, typename... TArgs>
+    AggregateFunctionPtr
+    createAggregateFunctionGroupArraySortedTypedFinal(TArgs && ... args)
+    {
+        return AggregateFunctionPtr(new AggregateFunctionTemplate<TColumnA, expr_sorted, TColumnB, is_plain_b>(std::forward<TArgs>(args)...));
+    }
+
+    template <bool expr_sorted = false, typename TColumnB = UInt64, bool is_plain_b = false>
+    AggregateFunctionPtr
+    createAggregateFunctionGroupArraySortedTyped(const DataTypes & argument_types, const Array & params, UInt64 threshold)
+    {
+#define DISPATCH(A, C, B) \
+    if (which.idx == TypeIndex::A) \
+        return createAggregateFunctionGroupArraySortedTypedFinal<C, B, expr_sorted, TColumnB, is_plain_b>(threshold, argument_types, params);
+#define DISPATCH_NUMERIC(A) DISPATCH(A, AggregateFunctionGroupArraySortedNumeric, A)
+        WhichDataType which(argument_types[0]);
+        FOR_NUMERIC_TYPES(DISPATCH_NUMERIC)
+        DISPATCH(Enum8, AggregateFunctionGroupArraySortedNumeric, Int8)
+        DISPATCH(Enum16, AggregateFunctionGroupArraySortedNumeric, Int16)
+        DISPATCH(Date, AggregateFunctionGroupArraySortedFieldType, DataTypeDate)
+        DISPATCH(DateTime, AggregateFunctionGroupArraySortedFieldType, DataTypeDateTime)
+#undef DISPATCH
+#undef DISPATCH_NUMERIC
+
+        if (argument_types[0]->isValueUnambiguouslyRepresentedInContiguousMemoryRegion())
+        {
+            return AggregateFunctionPtr(new AggregateFunctionGroupArraySorted<StringRef, true, expr_sorted, TColumnB, is_plain_b>(
+                threshold, argument_types, params));
+        }
+        else
+        {
+            return AggregateFunctionPtr(new AggregateFunctionGroupArraySorted<StringRef, false, expr_sorted, TColumnB, is_plain_b>(
+                threshold, argument_types, params));
+        }
+    }
+
+
+    AggregateFunctionPtr createAggregateFunctionGroupArraySorted(
+        const std::string & name, const DataTypes & argument_types, const Array & params, const Settings *)
+    {
+        UInt64 threshold = GROUP_SORTED_ARRAY_DEFAULT_THRESHOLD;
+
+        if (params.size() == 1)
+        {
+            UInt64 k = applyVisitor(FieldVisitorConvertToNumber<UInt64>(), params[0]);
+
+            if (k > GROUP_SORTED_ARRAY_MAX_SIZE)
+                throw Exception(
+                    "Too large parameter(s) for aggregate function " + name + ". Maximum: " + toString(GROUP_SORTED_ARRAY_MAX_SIZE),
+                    ErrorCodes::ARGUMENT_OUT_OF_BOUND);
+
+            if (k == 0)
+                throw Exception("Parameter 0 is illegal for aggregate function " + name, ErrorCodes::ARGUMENT_OUT_OF_BOUND);
+
+            threshold = k;
+        }
+        else if (!params.empty())
+        {
+            throw Exception("Aggregate function " + name + " only supports 1 parameter.", ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+        }
+
+        if (argument_types.size() == 2)
+        {
+            if (isNumber(argument_types[1]))
+            {
+#define DISPATCH2(A, B) \
+    if (which.idx == TypeIndex::A) \
+        return createAggregateFunctionGroupArraySortedTyped<true, B>(argument_types, params, threshold);
+#define DISPATCH(A) DISPATCH2(A, A)
+                WhichDataType which(argument_types[1]);
+                FOR_NUMERIC_TYPES(DISPATCH)
+                DISPATCH2(Enum8, Int8)
+                DISPATCH2(Enum16, Int16)
+#undef DISPATCH
+#undef DISPATCH2
+                throw Exception("Invalid parameter type.", ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+            }
+            else if (argument_types[1]->isValueUnambiguouslyRepresentedInContiguousMemoryRegion())
+            {
+                return createAggregateFunctionGroupArraySortedTyped<true, StringRef, true>(argument_types, params, threshold);
+            }
+            else
+            {
+                return createAggregateFunctionGroupArraySortedTyped<true, StringRef, false>(argument_types, params, threshold);
+            }
+        }
+        else if (argument_types.size() == 1)
+        {
+            return createAggregateFunctionGroupArraySortedTyped<>(argument_types, params, threshold);
+        }
+        else
+        {
+            throw Exception(
+                "Aggregate function " + name + " requires one or two parameters.", ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+        }
+    }
+}
+
+void registerAggregateFunctionGroupArraySorted(AggregateFunctionFactory & factory)
+{
+    AggregateFunctionProperties properties = {.returns_default_when_only_null = false, .is_order_dependent = true};
+    factory.registerFunction("groupArraySorted", {createAggregateFunctionGroupArraySorted, properties});
+}
+}
--- a/src/AggregateFunctions/AggregateFunctionGroupArraySorted.h
+++ b/src/AggregateFunctions/AggregateFunctionGroupArraySorted.h
@ -0,0 +1,310 @@
+#pragma once
+
+#include <Columns/ColumnArray.h>
+#include <DataTypes/DataTypeArray.h>
+
+#include <AggregateFunctions/AggregateFunctionGroupArraySortedData.h>
+#include <AggregateFunctions/IAggregateFunction.h>
+
+namespace DB
+{
+template <typename TColumn, bool is_plain>
+inline TColumn readItem(const IColumn * column, Arena * arena, size_t row)
+{
+    if constexpr (std::is_same_v<TColumn, StringRef>)
+    {
+        if constexpr (is_plain)
+        {
+            StringRef str = column->getDataAt(row);
+            auto ptr = arena->alloc(str.size);
+            std::copy(str.data, str.data + str.size, ptr);
+            return StringRef(ptr, str.size);
+        }
+        else
+        {
+            const char * begin = nullptr;
+            return column->serializeValueIntoArena(row, *arena, begin);
+        }
+    }
+    else
+    {
+        if constexpr (std::is_same_v<TColumn, UInt64>)
+            return column->getUInt(row);
+        else
+            return column->getInt(row);
+    }
+}
+
+template <typename TColumn, typename TFilter = void>
+size_t
+getFirstNElements_low_threshold(const TColumn * data, int num_elements, int threshold, size_t * results, const TFilter * filter = nullptr)
+{
+    for (int i = 0; i < threshold; i++)
+    {
+        results[i] = 0;
+    }
+
+    threshold = std::min(num_elements, threshold);
+    int current_max = 0;
+    int cur;
+    int z;
+    for (int i = 0; i < num_elements; i++)
+    {
+        if constexpr (!std::is_same_v<TFilter, void>)
+        {
+            if (filter[i] == 0)
+                continue;
+        }
+
+        //Starting from the highest values and we look for the immediately lower than the given one
+        for (cur = current_max; cur > 0; cur--)
+        {
+            if (data[i] > data[results[cur - 1]])
+                break;
+        }
+
+        if (cur < threshold)
+        {
+            //Move all the higher values 1 position to the right
+            for (z = std::min(threshold - 1, current_max); z > cur; z--)
+                results[z] = results[z - 1];
+
+            if (current_max < threshold)
+                ++current_max;
+
+            //insert element into the given position
+            results[cur] = i;
+        }
+    }
+
+    return current_max;
+}
+
+template <typename T>
+struct SortableItem
+{
+    T a;
+    size_t b;
+    bool operator<(const SortableItem & other) const { return (this->a < other.a); }
+};
+
+template <typename TColumn, typename TFilter = void>
+size_t getFirstNElements_high_threshold(
+    const TColumn * data, size_t num_elements, size_t threshold, size_t * results, const TFilter * filter = nullptr)
+{
+    std::vector<SortableItem<TColumn>> dataIndexed(num_elements);
+    size_t num_elements_filtered = 0;
+
+    for (size_t i = 0; i < num_elements; i++)
+    {
+        if constexpr (!std::is_same_v<TFilter, void>)
+        {
+            if (filter[i] == 0)
+                continue;
+        }
+
+        dataIndexed.data()[num_elements_filtered].a = data[i];
+        dataIndexed.data()[num_elements_filtered].b = i;
+        num_elements_filtered++;
+    }
+
+    threshold = std::min(num_elements_filtered, threshold);
+
+    std::nth_element(dataIndexed.data(), dataIndexed.data() + threshold, dataIndexed.data() + num_elements_filtered);
+    std::sort(dataIndexed.data(), dataIndexed.data() + threshold);
+
+    for (size_t i = 0; i < threshold; i++)
+    {
+        results[i] = dataIndexed[i].b;
+    }
+
+    return threshold;
+}
+
+static const size_t THRESHOLD_MAX_CUSTOM_FUNCTION = 1000;
+
+template <typename TColumn>
+size_t getFirstNElements(const TColumn * data, size_t num_elements, size_t threshold, size_t * results, const UInt8 * filter = nullptr)
+{
+    if (threshold < THRESHOLD_MAX_CUSTOM_FUNCTION)
+    {
+        if (filter != nullptr)
+            return getFirstNElements_low_threshold(data, num_elements, threshold, results, filter);
+        else
+            return getFirstNElements_low_threshold(data, num_elements, threshold, results);
+    }
+    else
+    {
+        if (filter != nullptr)
+            return getFirstNElements_high_threshold(data, num_elements, threshold, results, filter);
+        else
+            return getFirstNElements_high_threshold(data, num_elements, threshold, results);
+    }
+}
+
+template <typename TColumnA, bool is_plain_a, bool use_column_b, typename TColumnB, bool is_plain_b>
+class AggregateFunctionGroupArraySorted : public IAggregateFunctionDataHelper<
+                                              AggregateFunctionGroupArraySortedData<TColumnA, use_column_b, TColumnB>,
+                                              AggregateFunctionGroupArraySorted<TColumnA, is_plain_a, use_column_b, TColumnB, is_plain_b>>
+{
+protected:
+    using State = AggregateFunctionGroupArraySortedData<TColumnA, use_column_b, TColumnB>;
+    using Base = IAggregateFunctionDataHelper<
+        AggregateFunctionGroupArraySortedData<TColumnA, use_column_b, TColumnB>,
+        AggregateFunctionGroupArraySorted>;
+
+    UInt64 threshold;
+    DataTypePtr & input_data_type;
+    mutable std::mutex mutex;
+
+    static void deserializeAndInsert(StringRef str, IColumn & data_to);
+
+public:
+    AggregateFunctionGroupArraySorted(UInt64 threshold_, const DataTypes & argument_types_, const Array & params)
+        : IAggregateFunctionDataHelper<
+            AggregateFunctionGroupArraySortedData<TColumnA, use_column_b, TColumnB>,
+            AggregateFunctionGroupArraySorted>(argument_types_, params)
+        , threshold(threshold_)
+        , input_data_type(this->argument_types[0])
+    {
+    }
+
+    void create(AggregateDataPtr place) const override
+    {
+        Base::create(place);
+        this->data(place).threshold = threshold;
+    }
+
+    String getName() const override { return "groupArraySorted"; }
+
+    DataTypePtr getReturnType() const override { return std::make_shared<DataTypeArray>(input_data_type); }
+
+    bool allocatesMemoryInArena() const override
+    {
+        if constexpr (std::is_same_v<TColumnA, StringRef>)
+            return true;
+        else
+            return false;
+    }
+
+    void add(AggregateDataPtr __restrict place, const IColumn ** columns, size_t row_num, Arena * arena) const override
+    {
+        State & data = this->data(place);
+        if constexpr (use_column_b)
+        {
+            data.add(
+                readItem<TColumnA, is_plain_a>(columns[0], arena, row_num), readItem<TColumnB, is_plain_b>(columns[1], arena, row_num));
+        }
+        else
+        {
+            data.add(readItem<TColumnA, is_plain_a>(columns[0], arena, row_num));
+        }
+    }
+
+    template <typename TColumn, bool is_plain, typename TFunc>
+    void
+    forFirstRows(size_t batch_size, const IColumn ** columns, size_t data_column, Arena * arena, ssize_t if_argument_pos, TFunc func) const
+    {
+        const TColumn * values = nullptr;
+        std::unique_ptr<std::vector<TColumn>> values_vector;
+        std::vector<size_t> best_rows(threshold);
+
+        if constexpr (std::is_same_v<TColumn, StringRef>)
+        {
+            values_vector.reset(new std::vector<TColumn>(batch_size));
+            for (size_t i = 0; i < batch_size; i++)
+                (*values_vector)[i] = readItem<TColumn, is_plain>(columns[data_column], arena, i);
+            values = (*values_vector).data();
+        }
+        else
+        {
+            const auto & column = assert_cast<const ColumnVector<TColumn> &>(*columns[data_column]);
+            values = column.getData().data();
+        }
+
+        const UInt8 * filter = nullptr;
+        StringRef refFilter;
+
+        if (if_argument_pos >= 0)
+        {
+            refFilter = columns[if_argument_pos]->getRawData();
+            filter = reinterpret_cast<const UInt8 *>(refFilter.data);
+        }
+
+        size_t num_elements = getFirstNElements(values, batch_size, threshold, best_rows.data(), filter);
+        for (size_t i = 0; i < num_elements; i++)
+        {
+            func(best_rows[i], values);
+        }
+    }
+
+    void addBatchSinglePlace(
+        size_t batch_size, AggregateDataPtr place, const IColumn ** columns, Arena * arena, ssize_t if_argument_pos) const override
+    {
+        State & data = this->data(place);
+
+        if constexpr (use_column_b)
+        {
+            forFirstRows<TColumnB, is_plain_b>(
+                batch_size, columns, 1, arena, if_argument_pos, [columns, &arena, &data](size_t row, const TColumnB * values)
+                {
+                    data.add(readItem<TColumnA, is_plain_a>(columns[0], arena, row), values[row]);
+                });
+        }
+        else
+        {
+            forFirstRows<TColumnA, is_plain_a>(
+                batch_size, columns, 0, arena, if_argument_pos, [&data](size_t row, const TColumnA * values)
+                {
+                    data.add(values[row]);
+                });
+        }
+    }
+
+    void merge(AggregateDataPtr __restrict place, ConstAggregateDataPtr rhs, Arena *) const override
+    {
+        this->data(place).merge(this->data(rhs));
+    }
+
+    void serialize(ConstAggregateDataPtr __restrict place, WriteBuffer & buf, std::optional<size_t> /* version */) const override
+    {
+        this->data(place).serialize(buf);
+    }
+
+    void
+    deserialize(AggregateDataPtr __restrict place, ReadBuffer & buf, std::optional<size_t> /* version  */, Arena * arena) const override
+    {
+        this->data(place).deserialize(buf, arena);
+    }
+
+    void insertResultInto(AggregateDataPtr __restrict place, IColumn & to, Arena * /*arena*/) const override
+    {
+        ColumnArray & arr_to = assert_cast<ColumnArray &>(to);
+        ColumnArray::Offsets & offsets_to = arr_to.getOffsets();
+
+        auto & values = this->data(place).values;
+        offsets_to.push_back(offsets_to.back() + values.size());
+
+        IColumn & data_to = arr_to.getData();
+        for (auto value : values)
+        {
+            if constexpr (std::is_same_v<TColumnA, StringRef>)
+            {
+                auto str = State::itemValue(value);
+                if constexpr (is_plain_a)
+                {
+                    data_to.insertData(str.data, str.size);
+                }
+                else
+                {
+                    data_to.deserializeAndInsertFromArena(str.data);
+                }
+            }
+            else
+            {
+                data_to.insert(State::itemValue(value));
+            }
+        }
+    }
+};
+}
--- a/src/AggregateFunctions/AggregateFunctionGroupArraySortedData.h
+++ b/src/AggregateFunctions/AggregateFunctionGroupArraySortedData.h
@ -0,0 +1,162 @@
+#pragma once
+
+#include <IO/ReadBuffer.h>
+#include <IO/ReadHelpers.h>
+#include <IO/VarInt.h>
+#include <IO/WriteBuffer.h>
+#include <IO/WriteHelpers.h>
+
+
+static inline constexpr UInt64 GROUP_SORTED_DEFAULT_THRESHOLD = 0xFFFFFF;
+
+namespace DB
+{
+template <typename T>
+static void writeOneItem(WriteBuffer & buf, T item)
+{
+    if constexpr (std::numeric_limits<T>::is_signed)
+    {
+        writeVarInt(item, buf);
+    }
+    else
+    {
+        writeVarUInt(item, buf);
+    }
+}
+
+static void writeOneItem(WriteBuffer & buf, const StringRef & item)
+{
+    writeBinary(item, buf);
+}
+
+template <typename T>
+static void readOneItem(ReadBuffer & buf, Arena * /*arena*/, T & item)
+{
+    if constexpr (std::numeric_limits<T>::is_signed)
+    {
+        DB::Int64 val;
+        readVarT(val, buf);
+        item = val;
+    }
+    else
+    {
+        DB::UInt64 val;
+        readVarT(val, buf);
+        item = val;
+    }
+}
+
+static void readOneItem(ReadBuffer & buf, Arena * arena, StringRef & item)
+{
+    item = readStringBinaryInto(*arena, buf);
+}
+
+template <typename Storage>
+struct AggregateFunctionGroupArraySortedDataBase
+{
+    typedef typename Storage::value_type ValueType;
+    AggregateFunctionGroupArraySortedDataBase(UInt64 threshold_ = GROUP_SORTED_DEFAULT_THRESHOLD) : threshold(threshold_) { }
+
+    virtual ~AggregateFunctionGroupArraySortedDataBase() { }
+    inline void narrowDown()
+    {
+        while (values.size() > threshold)
+            values.erase(--values.end());
+    }
+
+    void merge(const AggregateFunctionGroupArraySortedDataBase & other)
+    {
+        values.merge(Storage(other.values));
+        narrowDown();
+    }
+
+    void serialize(WriteBuffer & buf) const
+    {
+        writeOneItem(buf, UInt64(values.size()));
+        for (auto value : values)
+        {
+            serializeItem(buf, value);
+        }
+    }
+
+    virtual void serializeItem(WriteBuffer & buf, ValueType & val) const = 0;
+    virtual ValueType deserializeItem(ReadBuffer & buf, Arena * arena) const = 0;
+
+    void deserialize(ReadBuffer & buf, Arena * arena)
+    {
+        values.clear();
+        UInt64 length;
+        readOneItem(buf, nullptr, length);
+
+        while (length--)
+        {
+            values.insert(deserializeItem(buf, arena));
+        }
+
+        narrowDown();
+    }
+
+    UInt64 threshold;
+    Storage values;
+};
+
+template <typename T, bool expr_sorted, typename TIndex>
+struct AggregateFunctionGroupArraySortedData
+{
+};
+
+template <typename T, typename TIndex>
+struct AggregateFunctionGroupArraySortedData<T, true, TIndex> : public AggregateFunctionGroupArraySortedDataBase<std::multimap<TIndex, T>>
+{
+    using Base = AggregateFunctionGroupArraySortedDataBase<std::multimap<TIndex, T>>;
+    using Base::Base;
+
+    void add(T item, TIndex weight)
+    {
+        Base::values.insert({weight, item});
+        Base::narrowDown();
+    }
+
+    void serializeItem(WriteBuffer & buf, typename Base::ValueType & value) const override
+    {
+        writeOneItem(buf, value.first);
+        writeOneItem(buf, value.second);
+    }
+
+    virtual typename Base::ValueType deserializeItem(ReadBuffer & buf, Arena * arena) const override
+    {
+        TIndex first;
+        T second;
+        readOneItem(buf, arena, first);
+        readOneItem(buf, arena, second);
+
+        return {first, second};
+    }
+
+    static T itemValue(typename Base::ValueType & value) { return value.second; }
+};
+
+template <typename T, typename TIndex>
+struct AggregateFunctionGroupArraySortedData<T, false, TIndex> : public AggregateFunctionGroupArraySortedDataBase<std::multiset<T>>
+{
+    using Base = AggregateFunctionGroupArraySortedDataBase<std::multiset<T>>;
+    using Base::Base;
+
+    void add(T item)
+    {
+        Base::values.insert(item);
+        Base::narrowDown();
+    }
+
+    void serializeItem(WriteBuffer & buf, typename Base::ValueType & value) const override { writeOneItem(buf, value); }
+
+    typename Base::ValueType deserializeItem(ReadBuffer & buf, Arena * arena) const override
+    {
+        T value;
+        readOneItem(buf, arena, value);
+        return value;
+    }
+
+    static T itemValue(typename Base::ValueType & value) { return value; }
+};
+}
--- a/src/AggregateFunctions/AggregateFunctionSumMap.cpp
+++ b/src/AggregateFunctions/AggregateFunctionSumMap.cpp
@ -67,7 +67,7 @@ auto parseArguments(const std::string & name, const DataTypes & arguments)
        values_types.push_back(array_type->getNestedType());
    }

-    return std::tuple{std::move(keys_type), std::move(values_types), tuple_argument};
+    return std::tuple<DataTypePtr, DataTypes, bool>{std::move(keys_type), std::move(values_types), tuple_argument};
 }

 // This function instantiates a particular overload of the sumMap family of
--- a/src/AggregateFunctions/registerAggregateFunctions.cpp
+++ b/src/AggregateFunctions/registerAggregateFunctions.cpp
@ -59,6 +59,7 @@ void registerAggregateFunctionNothing(AggregateFunctionFactory &);
 void registerAggregateFunctionExponentialMovingAverage(AggregateFunctionFactory &);
 void registerAggregateFunctionSparkbar(AggregateFunctionFactory &);
 void registerAggregateFunctionIntervalLengthSum(AggregateFunctionFactory &);
+void registerAggregateFunctionGroupArraySorted(AggregateFunctionFactory & factory);

 class AggregateFunctionCombinatorFactory;
 void registerAggregateFunctionCombinatorIf(AggregateFunctionCombinatorFactory &);
@ -130,6 +131,7 @@ void registerAggregateFunctions()
        registerAggregateFunctionIntervalLengthSum(factory);
        registerAggregateFunctionExponentialMovingAverage(factory);
        registerAggregateFunctionSparkbar(factory);
+        registerAggregateFunctionGroupArraySorted(factory);

        registerWindowFunctions(factory);
    }
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@ -494,6 +494,11 @@ endif()

 target_link_libraries (clickhouse_common_io PUBLIC ch_contrib::fast_float)

+if (USE_ORC)
+    dbms_target_link_libraries(PUBLIC ${ORC_LIBRARIES})
+    dbms_target_include_directories(SYSTEM BEFORE PUBLIC ${ORC_INCLUDE_DIR} "${CMAKE_BINARY_DIR}/contrib/orc/c++/include")
+endif ()
+
 if (TARGET ch_contrib::rocksdb)
    dbms_target_link_libraries(PUBLIC ch_contrib::rocksdb)
 endif()
@ -573,10 +578,6 @@ if (ENABLE_TESTS)
        target_link_libraries(unit_tests_dbms PRIVATE ch_contrib::simdjson)
    endif()

-    if(TARGET ch_contrib::rapidjson)
-        target_include_directories(unit_tests_dbms PRIVATE ch_contrib::rapidjson)
-    endif()
-
    if (TARGET ch_contrib::yaml_cpp)
        target_link_libraries(unit_tests_dbms PRIVATE ch_contrib::yaml_cpp)
    endif()
--- a/src/Client/ClientBase.cpp
+++ b/src/Client/ClientBase.cpp
@ -220,7 +220,7 @@ static void incrementProfileEventsBlock(Block & dst, const Block & src)
 }


-std::atomic_flag exit_on_signal = ATOMIC_FLAG_INIT;
+std::atomic_flag exit_on_signal;

 class QueryInterruptHandler : private boost::noncopyable
 {
--- a/src/Columns/ColumnObject.cpp
+++ b/src/Columns/ColumnObject.cpp
@ -521,7 +521,7 @@ ColumnObject::ColumnObject(bool is_nullable_)
 {
 }

-ColumnObject::ColumnObject(SubcolumnsTree && subcolumns_, bool is_nullable_)
+ColumnObject::ColumnObject(Subcolumns && subcolumns_, bool is_nullable_)
    : is_nullable(is_nullable_)
    , subcolumns(std::move(subcolumns_))
    , num_rows(subcolumns.empty() ? 0 : (*subcolumns.begin())->data.size())
@ -696,7 +696,7 @@ const ColumnObject::Subcolumn & ColumnObject::getSubcolumn(const PathInData & ke
 ColumnObject::Subcolumn & ColumnObject::getSubcolumn(const PathInData & key)
 {
    if (const auto * node = subcolumns.findLeaf(key))
-        return const_cast<SubcolumnsTree::Node *>(node)->data;
+        return const_cast<Subcolumns::Node *>(node)->data;

    throw Exception(ErrorCodes::ILLEGAL_COLUMN, "There is no subcolumn {} in ColumnObject", key.getPath());
 }
@ -794,7 +794,7 @@ bool ColumnObject::isFinalized() const
 void ColumnObject::finalize()
 {
    size_t old_size = size();
-    SubcolumnsTree new_subcolumns;
+    Subcolumns new_subcolumns;
    for (auto && entry : subcolumns)
    {
        const auto & least_common_type = entry->data.getLeastCommonType();
--- a/src/Columns/ColumnObject.h
+++ b/src/Columns/ColumnObject.h
@ -138,20 +138,20 @@ public:
        size_t num_of_defaults_in_prefix = 0;
    };

-    using SubcolumnsTree = SubcolumnsTree<Subcolumn>;
+    using Subcolumns = SubcolumnsTree<Subcolumn>;

 private:
    /// If true then all subcolumns are nullable.
    const bool is_nullable;

-    SubcolumnsTree subcolumns;
+    Subcolumns subcolumns;
    size_t num_rows;

 public:
    static constexpr auto COLUMN_NAME_DUMMY = "_dummy";

    explicit ColumnObject(bool is_nullable_);
-    ColumnObject(SubcolumnsTree && subcolumns_, bool is_nullable_);
+    ColumnObject(Subcolumns && subcolumns_, bool is_nullable_);

    /// Checks that all subcolumns have consistent sizes.
    void checkConsistency() const;
@ -173,8 +173,8 @@ public:
    /// It cares about consistency of sizes of Nested arrays.
    void addNestedSubcolumn(const PathInData & key, const FieldInfo & field_info, size_t new_size);

-    const SubcolumnsTree & getSubcolumns() const { return subcolumns; }
-    SubcolumnsTree & getSubcolumns() { return subcolumns; }
+    const Subcolumns & getSubcolumns() const { return subcolumns; }
+    Subcolumns & getSubcolumns() { return subcolumns; }
    PathsInData getKeys() const;

    /// Finalizes all subcolumns.
--- a/src/Common/Exception.h
+++ b/src/Common/Exception.h
@ -35,10 +35,10 @@ public:
    {}

    // Format message with fmt::format, like the logging functions.
-    template <typename ...Args>
-    Exception(int code, const std::string & fmt, Args&&... args)
-        : Exception(fmt::format(fmt::runtime(fmt), std::forward<Args>(args)...), code)
-    {}
+    template <typename... Args>
+    Exception(int code, fmt::format_string<Args...> fmt, Args &&... args) : Exception(fmt::format(fmt, std::forward<Args>(args)...), code)
+    {
+    }

    struct CreateFromPocoTag {};
    struct CreateFromSTDTag {};
@ -52,10 +52,10 @@ public:
    const char * what() const throw() override { return message().data(); }

    /// Add something to the existing message.
-    template <typename ...Args>
-    void addMessage(const std::string& format, Args&&... args)
+    template <typename... Args>
+    void addMessage(fmt::format_string<Args...> format, Args &&... args)
    {
-        extendedMessage(fmt::format(fmt::runtime(format), std::forward<Args>(args)...));
+        extendedMessage(fmt::format(format, std::forward<Args>(args)...));
    }

    void addMessage(const std::string& message)
@ -117,10 +117,10 @@ public:
    ParsingException(int code, const std::string & message);

    // Format message with fmt::format, like the logging functions.
-    template <typename ...Args>
-    ParsingException(int code, const std::string & fmt, Args&&... args)
-        : Exception(fmt::format(fmt::runtime(fmt), std::forward<Args>(args)...), code)
-    {}
+    template <typename... Args>
+    ParsingException(int code, fmt::format_string<Args...> fmt, Args &&... args) : Exception(code, fmt, std::forward<Args>(args)...)
+    {
+    }


    std::string displayText() const
--- a/src/Common/FileSegment.cpp
+++ b/src/Common/FileSegment.cpp
@ -437,6 +437,7 @@ String FileSegment::stateToString(FileSegment::State state)
        case FileSegment::State::SKIP_CACHE:
            return "SKIP_CACHE";
    }
+    __builtin_unreachable();
 }

 String FileSegmentsHolder::toString()
--- a/src/Common/HashTable/FixedHashTable.h
+++ b/src/Common/HashTable/FixedHashTable.h
@ -67,6 +67,9 @@ struct FixedHashTableCalculatedSize
 {
    size_t getSize(const Cell * buf, const typename Cell::State & state, size_t num_cells) const
    {
+        if (!buf)
+            return 0;
+
        size_t res = 0;
        for (const Cell * end = buf + num_cells; buf != end; ++buf)
            if (!buf->isZero(state))
@ -76,6 +79,9 @@ struct FixedHashTableCalculatedSize

    bool isEmpty(const Cell * buf, const typename Cell::State & state, size_t num_cells) const
    {
+        if (!buf)
+            return true;
+
        for (const Cell * end = buf + num_cells; buf != end; ++buf)
            if (!buf->isZero(state))
                return false;
--- a/src/Common/HashTable/TwoLevelHashTable.h
+++ b/src/Common/HashTable/TwoLevelHashTable.h
@ -94,6 +94,12 @@ public:

    TwoLevelHashTable() = default;

+    explicit TwoLevelHashTable(size_t size_hint)
+    {
+        for (auto & impl : impls)
+            impl.reserve(size_hint / NUM_BUCKETS);
+    }
+
    /// Copy the data from another (normal) hash table. It should have the same hash function.
    template <typename Source>
    explicit TwoLevelHashTable(const Source & src)
--- a/src/Common/OvercommitTracker.cpp
+++ b/src/Common/OvercommitTracker.cpp
@ -23,6 +23,12 @@ void OvercommitTracker::setMaxWaitTime(UInt64 wait_time)

 bool OvercommitTracker::needToStopQuery(MemoryTracker * tracker)
 {
+    // NOTE: Do not change the order of locks
+    //
+    // global_mutex must be acquired before overcommit_m, because
+    // method OvercommitTracker::unsubscribe(MemoryTracker *) is
+    // always called with already acquired global_mutex in
+    // ProcessListEntry::~ProcessListEntry().
    std::unique_lock<std::mutex> global_lock(global_mutex);
    std::unique_lock<std::mutex> lk(overcommit_m);

@ -76,7 +82,7 @@ void UserOvercommitTracker::pickQueryToExcludeImpl()
    MemoryTracker * query_tracker = nullptr;
    OvercommitRatio current_ratio{0, 0};
    // At this moment query list must be read only.
-    // BlockQueryIfMemoryLimit is used in ProcessList to guarantee this.
+    // This is guaranteed by locking global_mutex in OvercommitTracker::needToStopQuery.
    auto & queries = user_process_list->queries;
    LOG_DEBUG(logger, "Trying to choose query to stop from {} queries", queries.size());
    for (auto const & query : queries)
@ -111,9 +117,9 @@ void GlobalOvercommitTracker::pickQueryToExcludeImpl()
    MemoryTracker * query_tracker = nullptr;
    OvercommitRatio current_ratio{0, 0};
    // At this moment query list must be read only.
-    // BlockQueryIfMemoryLimit is used in ProcessList to guarantee this.
-    LOG_DEBUG(logger, "Trying to choose query to stop");
-    process_list->processEachQueryStatus([&](DB::QueryStatus const & query)
+    // This is guaranteed by locking global_mutex in OvercommitTracker::needToStopQuery.
+    LOG_DEBUG(logger, "Trying to choose query to stop from {} queries", process_list->size());
+    for (auto const & query : process_list->processes)
    {
        if (query.isKilled())
            return;
@ -134,7 +140,7 @@ void GlobalOvercommitTracker::pickQueryToExcludeImpl()
            query_tracker = memory_tracker;
            current_ratio   = ratio;
        }
-    });
+    }
    LOG_DEBUG(logger, "Selected to stop query with overcommit ratio {}/{}",
        current_ratio.committed, current_ratio.soft_limit);
    picked_tracker = query_tracker;
--- a/src/Common/OvercommitTracker.h
+++ b/src/Common/OvercommitTracker.h
@ -43,8 +43,6 @@ class MemoryTracker;
 // is killed to free memory.
 struct OvercommitTracker : boost::noncopyable
 {
-    explicit OvercommitTracker(std::mutex & global_mutex_);
-
    void setMaxWaitTime(UInt64 wait_time);

    bool needToStopQuery(MemoryTracker * tracker);
@ -54,8 +52,12 @@ struct OvercommitTracker : boost::noncopyable
    virtual ~OvercommitTracker() = default;

 protected:
+    explicit OvercommitTracker(std::mutex & global_mutex_);
+
    virtual void pickQueryToExcludeImpl() = 0;

+    // This mutex is used to disallow concurrent access
+    // to picked_tracker and cancelation_state variables.
    mutable std::mutex overcommit_m;
    mutable std::condition_variable cv;

@ -87,6 +89,11 @@ private:
        }
    }

+    // Global mutex which is used in ProcessList to synchronize
+    // insertion and deletion of queries.
+    // OvercommitTracker::pickQueryToExcludeImpl() implementations
+    // require this mutex to be locked, because they read list (or sublist)
+    // of queries.
    std::mutex & global_mutex;
 };

--- a/src/Common/ProfileEvents.cpp
+++ b/src/Common/ProfileEvents.cpp
@ -9,6 +9,7 @@
    M(SelectQuery, "Same as Query, but only for SELECT queries.") \
    M(InsertQuery, "Same as Query, but only for INSERT queries.") \
    M(AsyncInsertQuery, "Same as InsertQuery, but only for asynchronous INSERT queries.") \
+    M(AsyncInsertBytes, "Data size in bytes of asynchronous INSERT queries.") \
    M(FailedQuery, "Number of failed queries.") \
    M(FailedSelectQuery, "Same as FailedQuery, but only for SELECT queries.") \
    M(FailedInsertQuery, "Same as FailedQuery, but only for INSERT queries.") \
@ -284,6 +285,16 @@
    \
    M(MainConfigLoads, "Number of times the main configuration was reloaded.") \
    \
+    M(AggregationPreallocatedElementsInHashTables, "How many elements were preallocated in hash tables for aggregation.") \
+    M(AggregationHashTablesInitializedAsTwoLevel, "How many hash tables were inited as two-level for aggregation.") \
+    \
+    M(MergeTreeMetadataCacheGet, "Number of rocksdb reads(used for merge tree metadata cache)") \
+    M(MergeTreeMetadataCachePut, "Number of rocksdb puts(used for merge tree metadata cache)") \
+    M(MergeTreeMetadataCacheDelete, "Number of rocksdb deletes(used for merge tree metadata cache)") \
+    M(MergeTreeMetadataCacheSeek, "Number of rocksdb seeks(used for merge tree metadata cache)") \
+    M(MergeTreeMetadataCacheHit, "Number of times the read of meta file was done from MergeTree metadata cache") \
+    M(MergeTreeMetadataCacheMiss, "Number of times the read of meta file was not done from MergeTree metadata cache") \
+    \
    M(ScalarSubqueriesGlobalCacheHit, "Number of times a read from a scalar subquery was done using the global cache") \
    M(ScalarSubqueriesLocalCacheHit, "Number of times a read from a scalar subquery was done using the local cache") \
    M(ScalarSubqueriesCacheMiss, "Number of times a read from a scalar subquery was not cached and had to be calculated completely")
--- a/src/Common/RangeGenerator.h
+++ b/src/Common/RangeGenerator.h
@ -0,0 +1,46 @@
+#pragma once
+
+#include <optional>
+#include <cmath>
+
+namespace DB
+{
+
+class RangeGenerator
+{
+public:
+    explicit RangeGenerator(size_t total_size_, size_t range_step_, size_t range_start = 0)
+        : from(range_start), range_step(range_step_), total_size(total_size_)
+    {
+    }
+
+    size_t totalRanges() const { return static_cast<size_t>(round(static_cast<float>(total_size - from) / range_step)); }
+
+    using Range = std::pair<size_t, size_t>;
+
+    // return upper exclusive range of values, i.e. [from_range, to_range>
+    std::optional<Range> nextRange()
+    {
+        if (from >= total_size)
+        {
+            return std::nullopt;
+        }
+
+        auto to = from + range_step;
+        if (to >= total_size)
+        {
+            to = total_size;
+        }
+
+        Range range{from, to};
+        from = to;
+        return range;
+    }
+
+private:
+    size_t from;
+    size_t range_step;
+    size_t total_size;
+};
+
+}
--- a/src/Common/filesystemHelpers.cpp
+++ b/src/Common/filesystemHelpers.cpp
@ -4,6 +4,8 @@
 #if defined(__linux__)
 #    include <cstdio>
 #    include <mntent.h>
+#    include <sys/stat.h>
+#    include <sys/sysmacros.h>
 #endif
 #include <cerrno>
 #include <Poco/Version.h>
@ -13,6 +15,9 @@
 #include <unistd.h>
 #include <sys/types.h>
 #include <utime.h>
+#include <IO/ReadBufferFromFile.h>
+#include <IO/Operators.h>
+#include <IO/WriteBufferFromString.h>

 namespace fs = std::filesystem;

@ -24,6 +29,7 @@ namespace ErrorCodes
    extern const int LOGICAL_ERROR;
    extern const int SYSTEM_ERROR;
    extern const int NOT_IMPLEMENTED;
+    extern const int CANNOT_STAT;
    extern const int CANNOT_STATVFS;
    extern const int PATH_ACCESS_DENIED;
    extern const int CANNOT_CREATE_FILE;
@ -57,6 +63,68 @@ std::unique_ptr<TemporaryFile> createTemporaryFile(const std::string & path)
    return std::make_unique<TemporaryFile>(path);
 }

+#if !defined(__linux__)
+[[noreturn]]
+#endif
+String getBlockDeviceId([[maybe_unused]] const String & path)
+{
+#if defined(__linux__)
+    struct stat sb;
+    if (lstat(path.c_str(), &sb))
+        throwFromErrnoWithPath("Cannot lstat " + path, path, ErrorCodes::CANNOT_STAT);
+    WriteBufferFromOwnString ss;
+    ss << major(sb.st_dev) << ":" << minor(sb.st_dev);
+    return ss.str();
+#else
+    throw DB::Exception("The function getDeviceId is supported on Linux only", ErrorCodes::NOT_IMPLEMENTED);
+#endif
+}
+
+#if !defined(__linux__)
+[[noreturn]]
+#endif
+BlockDeviceType getBlockDeviceType([[maybe_unused]] const String & device_id)
+{
+#if defined(__linux__)
+    try
+    {
+        ReadBufferFromFile in("/sys/dev/block/" + device_id + "/queue/rotational");
+        int rotational;
+        readText(rotational, in);
+        return rotational ? BlockDeviceType::ROT : BlockDeviceType::NONROT;
+    }
+    catch (...)
+    {
+        return BlockDeviceType::UNKNOWN;
+    }
+#else
+    throw DB::Exception("The function getDeviceType is supported on Linux only", ErrorCodes::NOT_IMPLEMENTED);
+#endif
+}
+
+#if !defined(__linux__)
+[[noreturn]]
+#endif
+UInt64 getBlockDeviceReadAheadBytes([[maybe_unused]] const String & device_id)
+{
+#if defined(__linux__)
+    try
+    {
+        ReadBufferFromFile in("/sys/dev/block/" + device_id + "/queue/read_ahead_kb");
+        int read_ahead_kb;
+        readText(read_ahead_kb, in);
+        return read_ahead_kb * 1024;
+    }
+    catch (...)
+    {
+        return static_cast<UInt64>(-1);
+    }
+#else
+    throw DB::Exception("The function getDeviceType is supported on Linux only", ErrorCodes::NOT_IMPLEMENTED);
+#endif
+}
+
+/// Returns name of filesystem mounted to mount_point
 std::filesystem::path getMountPoint(std::filesystem::path absolute_path)
 {
    if (absolute_path.is_relative())
--- a/src/Common/filesystemHelpers.h
+++ b/src/Common/filesystemHelpers.h
@ -18,6 +18,31 @@ using TemporaryFile = Poco::TemporaryFile;
 bool enoughSpaceInDirectory(const std::string & path, size_t data_size);
 std::unique_ptr<TemporaryFile> createTemporaryFile(const std::string & path);

+// Determine what block device is responsible for specified path
+#if !defined(__linux__)
+[[noreturn]]
+#endif
+String getBlockDeviceId([[maybe_unused]] const String & path);
+
+enum class BlockDeviceType
+{
+    UNKNOWN = 0, // we were unable to determine device type
+    NONROT = 1, // not a rotational device (SSD, NVME, etc)
+    ROT = 2 // rotational device (HDD)
+};
+
+// Try to determine block device type
+#if !defined(__linux__)
+[[noreturn]]
+#endif
+BlockDeviceType getBlockDeviceType([[maybe_unused]] const String & device_id);
+
+// Get size of read-ahead in bytes for specified block device
+#if !defined(__linux__)
+[[noreturn]]
+#endif
+UInt64 getBlockDeviceReadAheadBytes([[maybe_unused]] const String & device_id);
+
 /// Returns mount point of filesystem where absolute_path (must exist) is located
 std::filesystem::path getMountPoint(std::filesystem::path absolute_path);

--- a/src/Common/format.h
+++ b/src/Common/format.h
@ -0,0 +1,178 @@
+#pragma once
+
+#include <base/types.h>
+#include <Common/Exception.h>
+#include <Common/PODArray.h>
+#include <Common/StringUtils/StringUtils.h>
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int BAD_ARGUMENTS;
+}
+
+namespace Format
+{
+    using IndexPositions = PODArrayWithStackMemory<UInt64, 64>;
+
+    static inline void parseNumber(const String & description, UInt64 l, UInt64 r, UInt64 & res, UInt64 argument_number)
+    {
+        res = 0;
+        for (UInt64 pos = l; pos < r; ++pos)
+        {
+            if (!isNumericASCII(description[pos]))
+                throw Exception("Not a number in curly braces at position " + std::to_string(pos), ErrorCodes::BAD_ARGUMENTS);
+            res = res * 10 + description[pos] - '0';
+            if (res >= argument_number)
+                throw Exception(
+                    "Too big number for arguments, must be at most " + std::to_string(argument_number - 1), ErrorCodes::BAD_ARGUMENTS);
+        }
+    }
+
+    static inline void init(
+        const String & pattern,
+        size_t argument_number,
+        const std::vector<std::optional<String>> & constant_strings,
+        IndexPositions & index_positions,
+        std::vector<String> & substrings)
+    {
+        /// Is current position after open curly brace.
+        bool is_open_curly = false;
+        /// The position of last open token.
+        size_t last_open = -1;
+
+        /// Is formatting in a plain {} token.
+        std::optional<bool> is_plain_numbering;
+        UInt64 index_if_plain = 0;
+
+        /// Left position of adding substrings, just to the closed brace position or the start of the string.
+        /// Invariant --- the start of substring is in this position.
+        size_t start_pos = 0;
+
+        /// A flag to decide whether we should glue the constant strings.
+        bool glue_to_next = false;
+
+        /// Handling double braces (escaping).
+        auto double_brace_removal = [](String & str)
+        {
+            size_t i = 0;
+            bool should_delete = true;
+            str.erase(
+                std::remove_if(
+                    str.begin(),
+                    str.end(),
+                    [&i, &should_delete, &str](char)
+                    {
+                        bool is_double_brace = (str[i] == '{' && str[i + 1] == '{') || (str[i] == '}' && str[i + 1] == '}');
+                        ++i;
+                        if (is_double_brace && should_delete)
+                        {
+                            should_delete = false;
+                            return true;
+                        }
+                        should_delete = true;
+                        return false;
+                    }),
+                str.end());
+        };
+
+        index_positions.emplace_back();
+
+        for (size_t i = 0; i < pattern.size(); ++i)
+        {
+            if (pattern[i] == '{')
+            {
+                /// Escaping handling
+                /// It is safe to access because of null termination
+                if (pattern[i + 1] == '{')
+                {
+                    ++i;
+                    continue;
+                }
+
+                if (is_open_curly)
+                    throw Exception("Two open curly braces without close one at position " + std::to_string(i), ErrorCodes::BAD_ARGUMENTS);
+
+                String to_add = String(pattern.data() + start_pos, i - start_pos);
+                double_brace_removal(to_add);
+                if (!glue_to_next)
+                    substrings.emplace_back(to_add);
+                else
+                    substrings.back() += to_add;
+
+                glue_to_next = false;
+
+                is_open_curly = true;
+                last_open = i + 1;
+            }
+            else if (pattern[i] == '}')
+            {
+                if (pattern[i + 1] == '}')
+                {
+                    ++i;
+                    continue;
+                }
+
+                if (!is_open_curly)
+                    throw Exception("Closed curly brace without open one at position " + std::to_string(i), ErrorCodes::BAD_ARGUMENTS);
+
+                is_open_curly = false;
+
+                if (last_open == i)
+                {
+                    if (is_plain_numbering && !*is_plain_numbering)
+                        throw Exception(
+                            "Cannot switch from automatic field numbering to manual field specification", ErrorCodes::BAD_ARGUMENTS);
+                    is_plain_numbering = true;
+                    if (index_if_plain >= argument_number)
+                        throw Exception("Argument is too big for formatting", ErrorCodes::BAD_ARGUMENTS);
+                    index_positions.back() = index_if_plain++;
+                }
+                else
+                {
+                    if (is_plain_numbering && *is_plain_numbering)
+                        throw Exception(
+                            "Cannot switch from automatic field numbering to manual field specification", ErrorCodes::BAD_ARGUMENTS);
+                    is_plain_numbering = false;
+
+                    UInt64 arg;
+                    parseNumber(pattern, last_open, i, arg, argument_number);
+
+                    if (arg >= argument_number)
+                        throw Exception(
+                            "Argument is too big for formatting. Note that indexing starts from zero", ErrorCodes::BAD_ARGUMENTS);
+
+                    index_positions.back() = arg;
+                }
+
+                if (!constant_strings.empty() && constant_strings[index_positions.back()])
+                {
+                    /// The next string should be glued to last `A {} C`.format('B') -> `A B C`.
+                    glue_to_next = true;
+                    substrings.back() += *constant_strings[index_positions.back()];
+                }
+                else
+                    index_positions.emplace_back(); /// Otherwise we commit arg number and proceed.
+
+                start_pos = i + 1;
+            }
+        }
+
+        if (is_open_curly)
+            throw Exception("Last open curly brace is not closed", ErrorCodes::BAD_ARGUMENTS);
+
+        String to_add = String(pattern.data() + start_pos, pattern.size() - start_pos);
+        double_brace_removal(to_add);
+
+        if (!glue_to_next)
+            substrings.emplace_back(to_add);
+        else
+            substrings.back() += to_add;
+
+        index_positions.pop_back();
+    }
+}
+
+}
--- a/src/Common/getNumberOfPhysicalCPUCores.cpp
+++ b/src/Common/getNumberOfPhysicalCPUCores.cpp
@ -38,21 +38,7 @@ unsigned getCGroupLimitedCPUCores(unsigned default_cpu_count)
        quota_count = ceil(static_cast<float>(cgroup_quota) / static_cast<float>(cgroup_period));
    }

-    // Share number (typically a number relative to 1024) (2048 typically expresses 2 CPUs worth of processing)
-    // -1 for no share setup
-    int cgroup_share = read_from("/sys/fs/cgroup/cpu/cpu.shares", -1);
-    // Convert 1024 to no shares setup
-    if (cgroup_share == 1024)
-        cgroup_share = -1;
-
-#    define PER_CPU_SHARES 1024
-    unsigned share_count = default_cpu_count;
-    if (cgroup_share > -1)
-    {
-        share_count = ceil(static_cast<float>(cgroup_share) / static_cast<float>(PER_CPU_SHARES));
-    }
-
-    return std::min(default_cpu_count, std::min(share_count, quota_count));
+    return std::min(default_cpu_count, quota_count);
 }
 #endif // OS_LINUX

@ -91,6 +77,7 @@ unsigned getNumberOfPhysicalCPUCores()
            cpu_count = std::thread::hardware_concurrency();

 #if defined(OS_LINUX)
+        /// TODO: add a setting for disabling that, similar to UseContainerSupport in java
        cpu_count = getCGroupLimitedCPUCores(cpu_count);
 #endif // OS_LINUX
        return cpu_count;
--- a/src/Common/tests/gtest_global_context.cpp
+++ b/src/Common/tests/gtest_global_context.cpp
@ -1,7 +1,18 @@
 #include "gtest_global_context.h"

 const ContextHolder & getContext()
+{
+    return getMutableContext();
+}
+
+ContextHolder & getMutableContext()
 {
    static ContextHolder holder;
    return holder;
 }
+
+void destroyContext()
+{
+    auto & holder = getMutableContext();
+    return holder.destroy();
+}
--- a/src/Common/tests/gtest_global_context.h
+++ b/src/Common/tests/gtest_global_context.h
@ -16,6 +16,17 @@ struct ContextHolder
    }

    ContextHolder(ContextHolder &&) = default;
+
+    void destroy()
+    {
+        context->shutdown();
+        context.reset();
+        shared_context.reset();
+    }
 };

 const ContextHolder & getContext();
+
+ContextHolder & getMutableContext();
+
+void destroyContext();
--- a/src/Compression/CompressionFactory.cpp
+++ b/src/Compression/CompressionFactory.cpp
@ -165,25 +165,36 @@ void registerCodecNone(CompressionCodecFactory & factory);
 void registerCodecLZ4(CompressionCodecFactory & factory);
 void registerCodecLZ4HC(CompressionCodecFactory & factory);
 void registerCodecZSTD(CompressionCodecFactory & factory);
+void registerCodecMultiple(CompressionCodecFactory & factory);
+
+
+/// Keeper use only general-purpose codecs, so we don't need these special codecs
+/// in standalone build
+#ifndef KEEPER_STANDALONE_BUILD
+
 void registerCodecDelta(CompressionCodecFactory & factory);
 void registerCodecT64(CompressionCodecFactory & factory);
 void registerCodecDoubleDelta(CompressionCodecFactory & factory);
 void registerCodecGorilla(CompressionCodecFactory & factory);
 void registerCodecEncrypted(CompressionCodecFactory & factory);
-void registerCodecMultiple(CompressionCodecFactory & factory);
+
+#endif

 CompressionCodecFactory::CompressionCodecFactory()
 {
-    registerCodecLZ4(*this);
    registerCodecNone(*this);
+    registerCodecLZ4(*this);
    registerCodecZSTD(*this);
    registerCodecLZ4HC(*this);
+    registerCodecMultiple(*this);
+
+#ifndef KEEPER_STANDALONE_BUILD
    registerCodecDelta(*this);
    registerCodecT64(*this);
    registerCodecDoubleDelta(*this);
    registerCodecGorilla(*this);
    registerCodecEncrypted(*this);
-    registerCodecMultiple(*this);
+#endif

    default_codec = get("LZ4", {});
 }
--- a/src/Core/Settings.h
+++ b/src/Core/Settings.h
@ -500,6 +500,10 @@ class IColumn;
    M(Bool, optimize_rewrite_sum_if_to_count_if, true, "Rewrite sumIf() and sum(if()) function countIf() function when logically equivalent", 0) \
    M(UInt64, insert_shard_id, 0, "If non zero, when insert into a distributed table, the data will be inserted into the shard `insert_shard_id` synchronously. Possible values range from 1 to `shards_number` of corresponding distributed table", 0) \
    \
+    M(Bool, collect_hash_table_stats_during_aggregation, true, "Enable collecting hash table statistics to optimize memory allocation", 0) \
+    M(UInt64, max_entries_for_hash_table_stats, 10'000, "How many entries hash table statistics collected during aggregation is allowed to have", 0) \
+    M(UInt64, max_size_to_preallocate_for_aggregation, 10'000'000, "For how many elements it is allowed to preallocate space in all hash tables in total before aggregation", 0) \
+    \
    /** Experimental feature for moving data between shards. */ \
    \
    M(Bool, allow_experimental_query_deduplication, false, "Experimental data deduplication for SELECT queries based on part UUIDs", 0) \
--- a/src/Core/SettingsEnums.h
+++ b/src/Core/SettingsEnums.h
@ -187,5 +187,4 @@ DECLARE_SETTING_ENUM_WITH_RENAME(EnumComparingMode, FormatSettings::EnumComparin
 DECLARE_SETTING_ENUM_WITH_RENAME(EscapingRule, FormatSettings::EscapingRule)

 DECLARE_SETTING_ENUM_WITH_RENAME(MsgPackUUIDRepresentation, FormatSettings::MsgPackUUIDRepresentation)
-
 }
--- a/src/DataTypes/Serializations/SerializationObject.cpp
+++ b/src/DataTypes/Serializations/SerializationObject.cpp
@ -63,12 +63,12 @@ private:
    size_t num_dimensions_to_keep;
 };

-using Node = typename ColumnObject::SubcolumnsTree::Node;
+using Node = typename ColumnObject::Subcolumns::Node;

 /// Finds a subcolumn from the same Nested type as @entry and inserts
 /// an array with default values with consistent sizes as in Nested type.
 bool tryInsertDefaultFromNested(
-    const std::shared_ptr<Node> & entry, const ColumnObject::SubcolumnsTree & subcolumns)
+    const std::shared_ptr<Node> & entry, const ColumnObject::Subcolumns & subcolumns)
 {
    if (!entry->path.hasNested())
        return false;
@ -198,7 +198,7 @@ void SerializationObject<Parser>::deserializeWholeText(IColumn & column, ReadBuf
 template <typename Parser>
 void SerializationObject<Parser>::deserializeTextEscaped(IColumn & column, ReadBuffer & istr, const FormatSettings &) const
 {
-    deserializeTextImpl(column, [&](String & s) { readEscapedStringInto(s, istr); });
+    deserializeTextImpl(column, [&](String & s) { readEscapedString(s, istr); });
 }

 template <typename Parser>
--- a/src/Databases/DatabaseOnDisk.cpp
+++ b/src/Databases/DatabaseOnDisk.cpp
@ -6,6 +6,7 @@
 #include <IO/WriteHelpers.h>
 #include <Interpreters/Context.h>
 #include <Interpreters/InterpreterCreateQuery.h>
+#include <Interpreters/ApplyWithSubqueryVisitor.h>
 #include <Parsers/ASTCreateQuery.h>
 #include <Parsers/ASTFunction.h>
 #include <Parsers/ParserCreateQuery.h>
@ -55,6 +56,9 @@ std::pair<String, StoragePtr> createTableFromAST(
    ast_create_query.attach = true;
    ast_create_query.setDatabase(database_name);

+    if (ast_create_query.select && ast_create_query.isView())
+        ApplyWithSubqueryVisitor().visit(*ast_create_query.select);
+
    if (ast_create_query.as_table_function)
    {
        const auto & factory = TableFunctionFactory::instance();
--- a/src/Databases/DatabaseReplicatedWorker.cpp
+++ b/src/Databases/DatabaseReplicatedWorker.cpp
@ -179,8 +179,12 @@ String DatabaseReplicatedDDLWorker::tryEnqueueAndExecuteEntry(DDLLogEntry & entr

    if (!task->was_executed)
    {
-        throw Exception(ErrorCodes::LOGICAL_ERROR, "Entry {} was executed, but was not committed: code {}: {}",
-                        task->execution_status.code, task->execution_status.message);
+        throw Exception(
+            ErrorCodes::LOGICAL_ERROR,
+            "Entry {} was executed, but was not committed: code {}: {}",
+            task->entry_name,
+            task->execution_status.code,
+            task->execution_status.message);
    }

    try_node->setAlreadyRemoved();
--- a/src/Dictionaries/XDBCDictionarySource.cpp
+++ b/src/Dictionaries/XDBCDictionarySource.cpp
@ -50,7 +50,7 @@ namespace
        {
            if (!qualified_name.database.empty())
                throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
-                    "Dictionary source of type {} specifies a schema but schema is not supported by {}-driver",
+                    "Dictionary source specifies a schema but schema is not supported by {}-driver",
                    bridge_.getName());
        }

--- a/src/Disks/IO/CachedReadBufferFromRemoteFS.cpp
+++ b/src/Disks/IO/CachedReadBufferFromRemoteFS.cpp
@ -392,8 +392,13 @@ void CachedReadBufferFromRemoteFS::predownload(FileSegmentPtr & file_segment)
                if (bytes_to_predownload)
                    throw Exception(
                        ErrorCodes::LOGICAL_ERROR,
-                        "Failed to predownload remaining {} bytes. Current file segment: {}, current download offset: {}, expected: {}, eof: {}",
-                        file_segment->range().toString(), file_segment->getDownloadOffset(), file_offset_of_buffer_end, implementation_buffer->eof());
+                        "Failed to predownload remaining {} bytes. Current file segment: {}, current download offset: {}, expected: {}, "
+                        "eof: {}",
+                        bytes_to_predownload,
+                        file_segment->range().toString(),
+                        file_segment->getDownloadOffset(),
+                        file_offset_of_buffer_end,
+                        implementation_buffer->eof());

                auto result = implementation_buffer->hasPendingData();

--- a/src/Disks/IO/CachedReadBufferFromRemoteFS.h
+++ b/src/Disks/IO/CachedReadBufferFromRemoteFS.h
@ -96,6 +96,7 @@ private:
            case ReadType::REMOTE_FS_READ_AND_PUT_IN_CACHE:
                return "REMOTE_FS_READ_AND_PUT_IN_CACHE";
        }
+        __builtin_unreachable();
    }
    size_t first_offset = 0;
 };
--- a/src/Disks/IO/ReadBufferFromRemoteFSGather.cpp
+++ b/src/Disks/IO/ReadBufferFromRemoteFSGather.cpp
@ -44,7 +44,7 @@ SeekableReadBufferPtr ReadBufferFromS3Gather::createImplementationBuffer(const S
    {
        return std::make_unique<ReadBufferFromS3>(
            client_ptr, bucket, fs::path(metadata.remote_fs_root_path) / path, max_single_read_retries,
-            settings, /* use_external_buffer */true, read_until_position, /* restricted_seek */true);
+            settings, /* use_external_buffer */true, /* offset */ 0, read_until_position, /* restricted_seek */true);
    };

    if (with_cache)
--- a/src/Formats/FormatSchemaInfo.cpp
+++ b/src/Formats/FormatSchemaInfo.cpp
@ -85,9 +85,12 @@ FormatSchemaInfo::FormatSchemaInfo(const String & format_schema, const String &
    else if (path.has_parent_path() && !fs::weakly_canonical(default_schema_directory_path / path).string().starts_with(fs::weakly_canonical(default_schema_directory_path).string()))
    {
        if (is_server)
-            throw Exception(ErrorCodes::BAD_ARGUMENTS,
-                            "Path in the 'format_schema' setting shouldn't go outside the 'format_schema_path' directory: {} ({} not in {})",
-                            path.string());
+            throw Exception(
+                ErrorCodes::BAD_ARGUMENTS,
+                "Path in the 'format_schema' setting shouldn't go outside the 'format_schema_path' directory: {} ({} not in {})",
+                default_schema_directory(),
+                path.string(),
+                default_schema_directory());
        path = default_schema_directory_path / path;
        schema_path = path.filename();
        schema_directory = path.parent_path() / "";
--- a/src/Formats/JSONEachRowUtils.cpp
+++ b/src/Formats/JSONEachRowUtils.cpp
@ -9,6 +9,7 @@
 #include <DataTypes/DataTypeArray.h>
 #include <DataTypes/DataTypeTuple.h>
 #include <DataTypes/DataTypeMap.h>
+#include <DataTypes/DataTypeObject.h>
 #include <Common/JSONParsers/SimdJSONParser.h>
 #include <Common/JSONParsers/RapidJSONParser.h>
 #include <Common/JSONParsers/DummyJSONParser.h>
@ -158,22 +159,37 @@ DataTypePtr getDataTypeFromJSONFieldImpl(const Element & field)
    {
        auto object = field.getObject();
        DataTypePtr value_type;
+        bool is_object = false;
        for (const auto key_value_pair : object)
        {
            auto type = getDataTypeFromJSONFieldImpl(key_value_pair.second);
            if (!type)
-                return nullptr;
+                continue;

-            if (value_type && value_type->getName() != type->getName())
-                return nullptr;
+            if (isObject(type))
+            {
+                is_object = true;
+                break;
+            }

-            value_type = type;
+            if (!value_type)
+            {
+                value_type = type;
+            }
+            else if (!value_type->equals(*type))
+            {
+                is_object = true;
+                break;
+            }
        }

-        if (!value_type)
-            return nullptr;
+        if (is_object)
+            return std::make_shared<DataTypeObject>("json", false);

-        return std::make_shared<DataTypeMap>(std::make_shared<DataTypeString>(), value_type);
+        if (value_type)
+            return std::make_shared<DataTypeMap>(std::make_shared<DataTypeString>(), value_type);
+
+        return nullptr;
    }

    throw Exception{ErrorCodes::INCORRECT_DATA, "Unexpected JSON type"};
--- a/src/Formats/ReadSchemaUtils.cpp
+++ b/src/Formats/ReadSchemaUtils.cpp
@ -7,6 +7,8 @@
 #include <Formats/ReadSchemaUtils.h>
 #include <Processors/Formats/ISchemaReader.h>
 #include <Common/assert_cast.h>
+#include <Interpreters/Context.h>
+#include <Storages/IStorage.h>

 namespace DB
 {
@ -17,6 +19,28 @@ namespace ErrorCodes
    extern const int BAD_ARGUMENTS;
 }

+static std::optional<NamesAndTypesList> getOrderedColumnsList(
+    const NamesAndTypesList & columns_list, const Names & columns_order_hint)
+{
+    if (columns_list.size() != columns_order_hint.size())
+        return {};
+
+    std::unordered_map<String, DataTypePtr> available_columns;
+    for (const auto & [name, type] : columns_list)
+        available_columns.emplace(name, type);
+
+    NamesAndTypesList res;
+    for (const auto & name : columns_order_hint)
+    {
+        auto it = available_columns.find(name);
+        if (it == available_columns.end())
+            return {};
+
+        res.emplace_back(name, it->second);
+    }
+    return res;
+}
+
 ColumnsDescription readSchemaFromFormat(
    const String & format_name,
    const std::optional<FormatSettings> & format_settings,
@ -52,6 +76,22 @@ ColumnsDescription readSchemaFromFormat(
        {
            throw Exception(ErrorCodes::CANNOT_EXTRACT_TABLE_STRUCTURE, "Cannot extract table structure from {} format file. Error: {}", format_name, e.message());
        }
+
+        /// If we have "INSERT SELECT" query then try to order
+        /// columns as they are ordered in table schema for formats
+        /// without strict column order (like JSON and TSKV).
+        /// It will allow to execute simple data loading with query
+        /// "INSERT INTO table SELECT * FROM ..."
+        const auto & insertion_table = context->getInsertionTable();
+        if (!schema_reader->hasStrictOrderOfColumns() && !insertion_table.empty())
+        {
+            auto storage = DatabaseCatalog::instance().getTable(insertion_table, context);
+            auto metadata = storage->getInMemoryMetadataPtr();
+            auto names_in_storage = metadata->getColumns().getNamesOfPhysical();
+            auto ordered_list = getOrderedColumnsList(names_and_types, names_in_storage);
+            if (ordered_list)
+                names_and_types = *ordered_list;
+        }
    }
    else
        throw Exception(ErrorCodes::BAD_ARGUMENTS, "{} file format doesn't support schema inference", format_name);
--- a/src/Formats/registerFormats.cpp
+++ b/src/Formats/registerFormats.cpp
@ -13,6 +13,7 @@ void registerFileSegmentationEngineCSV(FormatFactory & factory);
 void registerFileSegmentationEngineJSONEachRow(FormatFactory & factory);
 void registerFileSegmentationEngineRegexp(FormatFactory & factory);
 void registerFileSegmentationEngineJSONAsString(FormatFactory & factory);
+void registerFileSegmentationEngineJSONAsObject(FormatFactory & factory);
 void registerFileSegmentationEngineJSONCompactEachRow(FormatFactory & factory);

 /// Formats for both input/output.
@ -103,6 +104,7 @@ void registerProtobufSchemaReader(FormatFactory & factory);
 void registerProtobufListSchemaReader(FormatFactory & factory);
 void registerLineAsStringSchemaReader(FormatFactory & factory);
 void registerJSONAsStringSchemaReader(FormatFactory & factory);
+void registerJSONAsObjectSchemaReader(FormatFactory & factory);
 void registerRawBLOBSchemaReader(FormatFactory & factory);
 void registerMsgPackSchemaReader(FormatFactory & factory);
 void registerCapnProtoSchemaReader(FormatFactory & factory);
@ -123,6 +125,7 @@ void registerFormats()
    registerFileSegmentationEngineJSONEachRow(factory);
    registerFileSegmentationEngineRegexp(factory);
    registerFileSegmentationEngineJSONAsString(factory);
+    registerFileSegmentationEngineJSONAsObject(factory);
    registerFileSegmentationEngineJSONCompactEachRow(factory);

    registerInputFormatNative(factory);
@ -207,6 +210,7 @@ void registerFormats()
    registerProtobufListSchemaReader(factory);
    registerLineAsStringSchemaReader(factory);
    registerJSONAsStringSchemaReader(factory);
+    registerJSONAsObjectSchemaReader(factory);
    registerRawBLOBSchemaReader(factory);
    registerMsgPackSchemaReader(factory);
    registerCapnProtoSchemaReader(factory);
--- a/src/Functions/FunctionsConversion.h
+++ b/src/Functions/FunctionsConversion.h
@ -53,6 +53,7 @@
 #include <DataTypes/DataTypeLowCardinality.h>
 #include <Columns/ColumnLowCardinality.h>
 #include <Interpreters/Context.h>
+#include <Common/HashTable/HashMap.h>


 namespace DB
@ -886,7 +887,7 @@ struct ConvertImplGenericToString
        const IColumn & col_from = *col_with_type_and_name.column;

        size_t size = col_from.size();
-        auto col_to = result_type->createColumn();
+        auto col_to = removeNullable(result_type)->createColumn();

        {
            ColumnStringHelpers::WriteHelper write_helper(
@ -3140,52 +3141,138 @@ private:
        }
    }

+    WrapperType createTupleToObjectWrapper(const DataTypeTuple & from_tuple, bool has_nullable_subcolumns) const
+    {
+        if (!from_tuple.haveExplicitNames())
+            throw Exception(ErrorCodes::TYPE_MISMATCH,
+            "Cast to Object can be performed only from flatten Named Tuple. Got: {}", from_tuple.getName());
+
+        PathsInData paths;
+        DataTypes from_types;
+
+        std::tie(paths, from_types) = flattenTuple(from_tuple.getPtr());
+        auto to_types = from_types;
+
+        for (auto & type : to_types)
+        {
+            if (isTuple(type) || isNested(type))
+                throw Exception(ErrorCodes::TYPE_MISMATCH,
+                    "Cast to Object can be performed only from flatten Named Tuple. Got: {}",
+                    from_tuple.getName());
+
+            type = recursiveRemoveLowCardinality(type);
+        }
+
+        return [element_wrappers = getElementWrappers(from_types, to_types),
+            has_nullable_subcolumns, from_types, to_types, paths]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t input_rows_count)
+        {
+            size_t tuple_size = to_types.size();
+            auto flattened_column = flattenTuple(arguments.front().column);
+            const auto & column_tuple = assert_cast<const ColumnTuple &>(*flattened_column);
+
+            if (tuple_size != column_tuple.getColumns().size())
+                throw Exception(ErrorCodes::TYPE_MISMATCH,
+                    "Expected tuple with {} subcolumn, but got {} subcolumns",
+                    tuple_size, column_tuple.getColumns().size());
+
+            auto res = ColumnObject::create(has_nullable_subcolumns);
+            for (size_t i = 0; i < tuple_size; ++i)
+            {
+                ColumnsWithTypeAndName element = {{column_tuple.getColumns()[i], from_types[i], "" }};
+                auto converted_column = element_wrappers[i](element, to_types[i], nullable_source, input_rows_count);
+                res->addSubcolumn(paths[i], converted_column->assumeMutable());
+            }
+
+            return res;
+        };
+    }
+
+    WrapperType createMapToObjectWrapper(const DataTypeMap & from_map, bool has_nullable_subcolumns) const
+    {
+        auto key_value_types = from_map.getKeyValueTypes();
+
+        if (!isStringOrFixedString(key_value_types[0]))
+            throw Exception(ErrorCodes::TYPE_MISMATCH,
+                "Cast to Object from Map can be performed only from Map "
+                "with String or FixedString key. Got: {}", from_map.getName());
+
+        const auto & value_type = key_value_types[1];
+        auto to_value_type = value_type;
+
+        if (!has_nullable_subcolumns && value_type->isNullable())
+            to_value_type = removeNullable(value_type);
+
+        if (has_nullable_subcolumns && !value_type->isNullable())
+            to_value_type = makeNullable(value_type);
+
+        DataTypes to_key_value_types{std::make_shared<DataTypeString>(), std::move(to_value_type)};
+        auto element_wrappers = getElementWrappers(key_value_types, to_key_value_types);
+
+        return [has_nullable_subcolumns, element_wrappers, key_value_types, to_key_value_types]
+            (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t) -> ColumnPtr
+        {
+            const auto & column_map = assert_cast<const ColumnMap &>(*arguments.front().column);
+            const auto & offsets = column_map.getNestedColumn().getOffsets();
+            auto key_value_columns = column_map.getNestedData().getColumnsCopy();
+
+            for (size_t i = 0; i < 2; ++i)
+            {
+                ColumnsWithTypeAndName element{{key_value_columns[i], key_value_types[i], ""}};
+                key_value_columns[i] = element_wrappers[i](element, to_key_value_types[i], nullable_source, key_value_columns[i]->size());
+            }
+
+            const auto & key_column_str = assert_cast<const ColumnString &>(*key_value_columns[0]);
+            const auto & value_column = *key_value_columns[1];
+
+            using SubcolumnsMap = HashMap<StringRef, MutableColumnPtr, StringRefHash>;
+            SubcolumnsMap subcolumns;
+
+            for (size_t row = 0; row < offsets.size(); ++row)
+            {
+                for (size_t i = offsets[static_cast<ssize_t>(row) - 1]; i < offsets[row]; ++i)
+                {
+                    auto ref = key_column_str.getDataAt(i);
+
+                    bool inserted;
+                    SubcolumnsMap::LookupResult it;
+                    subcolumns.emplace(ref, it, inserted);
+                    auto & subcolumn = it->getMapped();
+
+                    if (inserted)
+                        subcolumn = value_column.cloneEmpty()->cloneResized(row);
+
+                    /// Map can have duplicated keys. We insert only first one.
+                    if (subcolumn->size() == row)
+                        subcolumn->insertFrom(value_column, i);
+                }
+
+                /// Insert default values for keys missed in current row.
+                for (const auto & [_, subcolumn] : subcolumns)
+                    if (subcolumn->size() == row)
+                        subcolumn->insertDefault();
+            }
+
+            auto column_object = ColumnObject::create(has_nullable_subcolumns);
+            for (auto && [key, subcolumn] : subcolumns)
+            {
+                PathInData path(key.toView());
+                column_object->addSubcolumn(path, std::move(subcolumn));
+            }
+
+            return column_object;
+        };
+    }
+
    WrapperType createObjectWrapper(const DataTypePtr & from_type, const DataTypeObject * to_type) const
    {
        if (const auto * from_tuple = checkAndGetDataType<DataTypeTuple>(from_type.get()))
        {
-            if (!from_tuple->haveExplicitNames())
-                 throw Exception(ErrorCodes::TYPE_MISMATCH,
-                    "Cast to Object can be performed only from flatten Named Tuple. Got: {}", from_type->getName());
-
-            PathsInData paths;
-            DataTypes from_types;
-
-            std::tie(paths, from_types) = flattenTuple(from_type);
-            auto to_types = from_types;
-
-            for (auto & type : to_types)
-            {
-                if (isTuple(type) || isNested(type))
-                     throw Exception(ErrorCodes::TYPE_MISMATCH,
-                        "Cast to Object can be performed only from flatten Named Tuple. Got: {}", from_type->getName());
-
-                type = recursiveRemoveLowCardinality(type);
-            }
-
-            return [element_wrappers = getElementWrappers(from_types, to_types),
-                has_nullable_subcolumns = to_type->hasNullableSubcolumns(), from_types, to_types, paths]
-                (ColumnsWithTypeAndName & arguments, const DataTypePtr &, const ColumnNullable * nullable_source, size_t input_rows_count)
-            {
-                size_t tuple_size = to_types.size();
-                auto flattened_column = flattenTuple(arguments.front().column);
-                const auto & column_tuple = assert_cast<const ColumnTuple &>(*flattened_column);
-
-                if (tuple_size != column_tuple.getColumns().size())
-                    throw Exception(ErrorCodes::TYPE_MISMATCH,
-                        "Expected tuple with {} subcolumn, but got {} subcolumns",
-                        tuple_size, column_tuple.getColumns().size());
-
-                auto res = ColumnObject::create(has_nullable_subcolumns);
-                for (size_t i = 0; i < tuple_size; ++i)
-                {
-                    ColumnsWithTypeAndName element = {{column_tuple.getColumns()[i], from_types[i], "" }};
-                    auto converted_column = element_wrappers[i](element, to_types[i], nullable_source, input_rows_count);
-                    res->addSubcolumn(paths[i], converted_column->assumeMutable());
-                }
-
-                return res;
-            };
+            return createTupleToObjectWrapper(*from_tuple, to_type->hasNullableSubcolumns());
+        }
+        else if (const auto * from_map = checkAndGetDataType<DataTypeMap>(from_type.get()))
+        {
+            return createMapToObjectWrapper(*from_map, to_type->hasNullableSubcolumns());
        }
        else if (checkAndGetDataType<DataTypeString>(from_type.get()))
        {
@ -3199,7 +3286,7 @@ private:
        }

        throw Exception(ErrorCodes::TYPE_MISMATCH,
-            "Cast to Object can be performed only from flatten named tuple or string. Got: {}", from_type->getName());
+            "Cast to Object can be performed only from flatten named Tuple, Map or String. Got: {}", from_type->getName());
    }

    template <typename FieldType>
--- a/src/Functions/FunctionsStringArray.h
+++ b/src/Functions/FunctionsStringArray.h
@ -259,7 +259,7 @@ public:
            throw Exception(
                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH,
                "Function '{}' needs at least 2 arguments, at most 3 arguments; passed {}.",
-                arguments.size());
+                name, arguments.size());

        if (!isString(arguments[0]))
            throw Exception("Illegal type " + arguments[0]->getName() + " of first argument of function " + getName() + ". Must be String.",
--- a/src/Functions/IFunction.cpp
+++ b/src/Functions/IFunction.cpp
@ -181,9 +181,12 @@ ColumnPtr IExecutableFunction::defaultImplementationForNulls(
        // Default implementation for nulls returns null result for null arguments,
        // so the result type must be nullable.
        if (!result_type->isNullable())
-            throw Exception(ErrorCodes::LOGICAL_ERROR,
-                            "Function {} with Null argument and default implementation for Nulls "
-                            "is expected to return Nullable result, got {}", result_type->getName());
+            throw Exception(
+                ErrorCodes::LOGICAL_ERROR,
+                "Function {} with Null argument and default implementation for Nulls "
+                "is expected to return Nullable result, got {}",
+                getName(),
+                result_type->getName());

        return result_type->createColumnConstWithDefaultValue(input_rows_count);
    }
--- a/src/Functions/caseWithExpression.cpp
+++ b/src/Functions/caseWithExpression.cpp
@ -43,6 +43,9 @@ public:
        for (size_t i = 2; i < args.size() - 1; i += 2)
            dst_array_types.push_back(args[i]);

+        // Type of the ELSE branch
+        dst_array_types.push_back(args.back());
+
        return getLeastSupertype(dst_array_types);
    }

--- a/src/Functions/castOrDefault.cpp
+++ b/src/Functions/castOrDefault.cpp
@ -231,7 +231,7 @@ private:
            {
                throw Exception(ErrorCodes::BAD_ARGUMENTS,
                    "Function {} decimal scale should have native UInt type. Actual {}",
-                    scale_argument.type->getName());
+                    getName(), scale_argument.type->getName());
            }

            scale = arguments[additional_argument_index].column->getUInt(0);
--- a/src/Functions/concat.cpp
+++ b/src/Functions/concat.cpp
@ -52,23 +52,21 @@ public:
    {
        if (arguments.size() < 2)
            throw Exception(
-                "Number of arguments for function " + getName() + " doesn't match: passed " + toString(arguments.size())
-                    + ", should be at least 2.",
-                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
-
-        if (arguments.size() > FormatImpl::argument_threshold)
-            throw Exception(
-                "Number of arguments for function " + getName() + " doesn't match: passed " + toString(arguments.size())
-                    + ", should be at most " + std::to_string(FormatImpl::argument_threshold),
-                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH,
+                "Number of arguments for function {} doesn't match: passed {}, should be at least 2",
+                getName(),
+                arguments.size());

        for (const auto arg_idx : collections::range(0, arguments.size()))
        {
            const auto * arg = arguments[arg_idx].get();
            if (!isStringOrFixedString(arg))
-                throw Exception{"Illegal type " + arg->getName() + " of argument " + std::to_string(arg_idx + 1) + " of function "
-                                    + getName(),
-                                ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT};
+                throw Exception(
+                    ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                    "Illegal type {} of argument {} of function {}",
+                    arg->getName(),
+                    arg_idx + 1,
+                    getName());
        }

        return std::make_shared<DataTypeString>();
@ -125,7 +123,7 @@ private:
        std::vector<const ColumnString::Chars *> data(num_arguments);
        std::vector<const ColumnString::Offsets *> offsets(num_arguments);
        std::vector<size_t> fixed_string_sizes(num_arguments);
-        std::vector<String> constant_strings(num_arguments);
+        std::vector<std::optional<String>> constant_strings(num_arguments);
        bool has_column_string = false;
        bool has_column_fixed_string = false;
        for (size_t i = 0; i < num_arguments; ++i)
--- a/src/Functions/dateName.cpp
+++ b/src/Functions/dateName.cpp
@ -112,7 +112,7 @@ public:
            || (res = executeType<DataTypeDateTime64>(arguments, result_type))))
            throw Exception(
                ErrorCodes::ILLEGAL_COLUMN,
-                "Illegal column {} of function {], must be Date or DateTime.",
+                "Illegal column {} of function {}, must be Date or DateTime.",
                arguments[1].column->getName(),
                getName());

--- a/src/Functions/flattenTuple.cpp
+++ b/src/Functions/flattenTuple.cpp
@ -0,0 +1,68 @@
+#include <Functions/IFunction.h>
+#include <Functions/FunctionFactory.h>
+#include <Functions/FunctionHelpers.h>
+#include <DataTypes/DataTypeTuple.h>
+#include <DataTypes/ObjectUtils.h>
+#include <Columns/ColumnTuple.h>
+
+namespace DB
+{
+
+namespace ErrorCodes
+{
+    extern const int ILLEGAL_COLUMN;
+    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
+}
+
+namespace
+{
+
+class FunctionFlattenTuple : public IFunction
+{
+public:
+    static constexpr auto name = "flattenTuple";
+    static FunctionPtr create(ContextPtr) { return std::make_shared<FunctionFlattenTuple>(); }
+
+    String getName() const override { return name; }
+    size_t getNumberOfArguments() const override { return 1; }
+    bool isSuitableForShortCircuitArgumentsExecution(const DataTypesWithConstInfo &) const override { return true; }
+    bool useDefaultImplementationForConstants() const override { return true; }
+
+    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    {
+        const auto & type = arguments[0];
+        const auto * type_tuple = checkAndGetDataType<DataTypeTuple>(type.get());
+        if (!type_tuple || !type_tuple->haveExplicitNames())
+            throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                "Argument for function '{}' must be Named Tuple. Got '{}'",
+                getName(), type->getName());
+
+        auto [paths, types] = flattenTuple(type);
+        Names names;
+        names.reserve(paths.size());
+        for (const auto & path : paths)
+            names.push_back(path.getPath());
+
+        return std::make_shared<DataTypeTuple>(types, names);
+    }
+
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t /*input_rows_count*/) const override
+    {
+        auto column = arguments.at(0).column;
+        if (!checkAndGetColumn<ColumnTuple>(column.get()))
+            throw Exception(ErrorCodes::ILLEGAL_COLUMN,
+                "Illegal column {} of first argument of function {}. Expected ColumnTuple",
+                column->getName(), getName());
+
+        return flattenTuple(column);
+    }
+};
+
+}
+
+void registerFunctionFlattenTuple(FunctionFactory & factory)
+{
+    factory.registerFunction<FunctionFlattenTuple>();
+}
+
+}
--- a/src/Functions/formatString.cpp
+++ b/src/Functions/formatString.cpp
@ -45,25 +45,23 @@ public:

    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
    {
-        if (arguments.empty())
+        if (arguments.size() < 2)
            throw Exception(
-                "Number of arguments for function " + getName() + " doesn't match: passed " + toString(arguments.size())
-                    + ", should be at least 1",
-                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
-
-        if (arguments.size() > FormatImpl::argument_threshold)
-            throw Exception(
-                "Number of arguments for function " + getName() + " doesn't match: passed " + toString(arguments.size())
-                    + ", should be at most " + std::to_string(FormatImpl::argument_threshold),
-                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH);
+                ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH,
+                "Number of arguments for function {} doesn't match: passed {}, should be at least 2",
+                getName(),
+                arguments.size());

        for (const auto arg_idx : collections::range(0, arguments.size()))
        {
            const auto * arg = arguments[arg_idx].get();
            if (!isStringOrFixedString(arg))
                throw Exception(
-                    "Illegal type " + arg->getName() + " of argument " + std::to_string(arg_idx + 1) + " of function " + getName(),
-                    ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT);
+                    ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                    "Illegal type {} of argument {} of function {}",
+                    arg->getName(),
+                    arg_idx + 1,
+                    getName());
        }

        return std::make_shared<DataTypeString>();
@ -84,7 +82,7 @@ public:
        std::vector<const ColumnString::Chars *> data(arguments.size() - 1);
        std::vector<const ColumnString::Offsets *> offsets(arguments.size() - 1);
        std::vector<size_t> fixed_string_sizes(arguments.size() - 1);
-        std::vector<String> constant_strings(arguments.size() - 1);
+        std::vector<std::optional<String>> constant_strings(arguments.size() - 1);

        bool has_column_string = false;
        bool has_column_fixed_string = false;
--- a/src/Functions/formatString.h
+++ b/src/Functions/formatString.h
@ -4,8 +4,10 @@
 #include <base/types.h>
 #include <Common/Exception.h>
 #include <Common/StringUtils/StringUtils.h>
+#include <Common/format.h>
 #include <Common/memcpySmall.h>

+
 #include <algorithm>
 #include <optional>
 #include <string>
@ -15,15 +17,9 @@

 namespace DB
 {
-namespace ErrorCodes
-{
-    extern const int BAD_ARGUMENTS;
-}

 struct FormatImpl
 {
-    static constexpr size_t small_argument_threshold = 1024;
-    static constexpr size_t argument_threshold = std::numeric_limits<UInt32>::max();
    static constexpr size_t right_padding = 15;

    template <typename... Args>
@ -39,165 +35,10 @@ struct FormatImpl
            format<false, false>(std::forward<Args>(args)...);
    }

-    static void parseNumber(const String & description, UInt64 l, UInt64 r, UInt64 & res)
-    {
-        res = 0;
-        for (UInt64 pos = l; pos < r; ++pos)
-        {
-            if (!isNumericASCII(description[pos]))
-                throw Exception("Not a number in curly braces at position " + std::to_string(pos), ErrorCodes::BAD_ARGUMENTS);
-            res = res * 10 + description[pos] - '0';
-            if (res >= argument_threshold)
-                throw Exception(
-                    "Too big number for arguments, must be at most " + std::to_string(argument_threshold), ErrorCodes::BAD_ARGUMENTS);
-        }
-    }
-
-    static inline void init(
-        const String & pattern,
-        const std::vector<const ColumnString::Chars *> & data,
-        size_t argument_number,
-        const std::vector<String> & constant_strings,
-        UInt64 * index_positions_ptr,
-        std::vector<String> & substrings)
-    {
-        /// Is current position after open curly brace.
-        bool is_open_curly = false;
-        /// The position of last open token.
-        size_t last_open = -1;
-
-        /// Is formatting in a plain {} token.
-        std::optional<bool> is_plain_numbering;
-        UInt64 index_if_plain = 0;
-
-        /// Left position of adding substrings, just to the closed brace position or the start of the string.
-        /// Invariant --- the start of substring is in this position.
-        size_t start_pos = 0;
-
-        /// A flag to decide whether we should glue the constant strings.
-        bool glue_to_next = false;
-
-        /// Handling double braces (escaping).
-        auto double_brace_removal = [](String & str)
-        {
-            size_t i = 0;
-            bool should_delete = true;
-            str.erase(
-                std::remove_if(
-                    str.begin(),
-                    str.end(),
-                    [&i, &should_delete, &str](char)
-                    {
-                        bool is_double_brace = (str[i] == '{' && str[i + 1] == '{') || (str[i] == '}' && str[i + 1] == '}');
-                        ++i;
-                        if (is_double_brace && should_delete)
-                        {
-                            should_delete = false;
-                            return true;
-                        }
-                        should_delete = true;
-                        return false;
-                    }),
-                str.end());
-        };
-
-        for (size_t i = 0; i < pattern.size(); ++i)
-        {
-            if (pattern[i] == '{')
-            {
-                /// Escaping handling
-                /// It is safe to access because of null termination
-                if (pattern[i + 1] == '{')
-                {
-                    ++i;
-                    continue;
-                }
-
-                if (is_open_curly)
-                    throw Exception("Two open curly braces without close one at position " + std::to_string(i), ErrorCodes::BAD_ARGUMENTS);
-
-                String to_add = String(pattern.data() + start_pos, i - start_pos);
-                double_brace_removal(to_add);
-                if (!glue_to_next)
-                    substrings.emplace_back(to_add);
-                else
-                    substrings.back() += to_add;
-
-                glue_to_next = false;
-
-                is_open_curly = true;
-                last_open = i + 1;
-            }
-            else if (pattern[i] == '}')
-            {
-                if (pattern[i + 1] == '}')
-                {
-                    ++i;
-                    continue;
-                }
-
-                if (!is_open_curly)
-                    throw Exception("Closed curly brace without open one at position " + std::to_string(i), ErrorCodes::BAD_ARGUMENTS);
-
-                is_open_curly = false;
-
-                if (last_open == i)
-                {
-                    if (is_plain_numbering && !*is_plain_numbering)
-                        throw Exception(
-                            "Cannot switch from automatic field numbering to manual field specification", ErrorCodes::BAD_ARGUMENTS);
-                    is_plain_numbering = true;
-                    if (index_if_plain >= argument_number)
-                        throw Exception("Argument is too big for formatting", ErrorCodes::BAD_ARGUMENTS);
-                    *index_positions_ptr = index_if_plain++;
-                }
-                else
-                {
-                    if (is_plain_numbering && *is_plain_numbering)
-                        throw Exception(
-                            "Cannot switch from automatic field numbering to manual field specification", ErrorCodes::BAD_ARGUMENTS);
-                    is_plain_numbering = false;
-
-                    UInt64 arg;
-                    parseNumber(pattern, last_open, i, arg);
-
-                    if (arg >= argument_number)
-                        throw Exception(
-                            "Argument is too big for formatting. Note that indexing starts from zero", ErrorCodes::BAD_ARGUMENTS);
-
-                    *index_positions_ptr = arg;
-                }
-
-                /// Constant string.
-                if (!data[*index_positions_ptr])
-                {
-                    /// The next string should be glued to last `A {} C`.format('B') -> `A B C`.
-                    glue_to_next = true;
-                    substrings.back() += constant_strings[*index_positions_ptr];
-                }
-                else
-                    ++index_positions_ptr; /// Otherwise we commit arg number and proceed.
-
-                start_pos = i + 1;
-            }
-        }
-
-        if (is_open_curly)
-            throw Exception("Last open curly brace is not closed", ErrorCodes::BAD_ARGUMENTS);
-
-        String to_add = String(pattern.data() + start_pos, pattern.size() - start_pos);
-        double_brace_removal(to_add);
-
-        if (!glue_to_next)
-            substrings.emplace_back(to_add);
-        else
-            substrings.back() += to_add;
-    }
-
    /// data for ColumnString and ColumnFixed. Nullptr means no data, it is const string.
    /// offsets for ColumnString, nullptr is an indicator that there is a fixed string rather than ColumnString.
    /// fixed_string_N for savings N to fixed strings.
-    /// constant_strings for constant strings. If data[i] is nullptr, than it is constant string.
+    /// constant_strings for constant strings. If data[i] is nullptr, it is constant string.
    /// res_data is result_data, res_offsets is offset result.
    /// input_rows_count is the number of rows processed.
    /// Precondition: data.size() == offsets.size() == fixed_string_N.size() == constant_strings.size().
@ -207,29 +48,22 @@ struct FormatImpl
        const std::vector<const ColumnString::Chars *> & data,
        const std::vector<const ColumnString::Offsets *> & offsets,
        [[maybe_unused]] /* Because sometimes !has_column_fixed_string */ const std::vector<size_t> & fixed_string_N,
-        const std::vector<String> & constant_strings,
+        const std::vector<std::optional<String>> & constant_strings,
        ColumnString::Chars & res_data,
        ColumnString::Offsets & res_offsets,
        size_t input_rows_count)
    {
        const size_t argument_number = offsets.size();

-        UInt64 small_index_positions_buffer[small_argument_threshold];
-        /// The subsequent indexes of strings we should use. e.g `Hello world {1} {3} {1} {0}` this array will be filled with [1, 3, 1, 0, ... (garbage)] but without constant string indices.
-        UInt64 * index_positions = small_index_positions_buffer;
-
-        std::unique_ptr<UInt64[]> big_index_positions_buffer;
-        if (argument_number > small_argument_threshold)
-        {
-            big_index_positions_buffer.reset(new UInt64[argument_number]);
-            index_positions = big_index_positions_buffer.get();
-        }
+        /// The subsequent indexes of strings we should use. e.g `Hello world {1} {3} {1} {0}` this
+        /// array will be filled with [1, 3, 1, 0] but without constant string indices.
+        Format::IndexPositions index_positions;

        /// Vector of substrings of pattern that will be copied to the answer, not string view because of escaping and iterators invalidation.
        /// These are exactly what is between {} tokens, for `Hello {} world {}` we will have [`Hello `, ` world `, ``].
        std::vector<String> substrings;

-        init(pattern, data, argument_number, constant_strings, index_positions, substrings);
+        Format::init(pattern, argument_number, constant_strings, index_positions, substrings);

        UInt64 final_size = 0;

@ -271,7 +105,7 @@ struct FormatImpl
                for (size_t j = 1; j < substrings.size(); ++j)
                {
                    UInt64 arg = index_positions[j - 1];
-                    auto offset_ptr = offsets[arg];
+                    const auto * offset_ptr = offsets[arg];
                    UInt64 arg_offset = 0;
                    UInt64 size = 0;

--- a/src/Functions/makeDate.cpp
+++ b/src/Functions/makeDate.cpp
@ -0,0 +1,155 @@
+#include <Functions/IFunction.h>
+#include <Functions/FunctionFactory.h>
+#include <DataTypes/DataTypeDate.h>
+#include <DataTypes/DataTypeDate32.h>
+#include <DataTypes/DataTypesNumber.h>
+#include <Columns/ColumnsNumber.h>
+#include <Interpreters/castColumn.h>
+
+#include <Common/DateLUT.h>
+#include <Common/typeid_cast.h>
+
+#include <array>
+
+namespace DB
+{
+namespace ErrorCodes
+{
+    extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH;
+    extern const int ILLEGAL_TYPE_OF_ARGUMENT;
+}
+
+namespace
+{
+
+// A helper function to simplify comparisons of valid YYYY-MM-DD values for <,>,=
+inline constexpr Int64 YearMonthDayToSingleInt(Int64 year, Int64 month, Int64 day)
+{
+    return year * 512 + month * 32 + day;
+}
+
+// Common implementation for makeDate, makeDate32
+template <typename Traits>
+class FunctionMakeDate : public IFunction
+{
+private:
+    static constexpr std::array<const char*, 3> argument_names = {"year", "month", "day"};
+
+public:
+    static constexpr auto name = Traits::name;
+
+    static FunctionPtr create(ContextPtr) { return std::make_shared<FunctionMakeDate>(); }
+
+    String getName() const override { return name; }
+
+    bool isVariadic() const override { return false; }
+
+    size_t getNumberOfArguments() const override { return argument_names.size(); }
+
+    bool isInjective(const ColumnsWithTypeAndName &) const override
+    {
+        return false; // {year,month,day} that are out of supported range are converted into a default value
+    }
+
+    bool isSuitableForShortCircuitArgumentsExecution(const DataTypesWithConstInfo & /*arguments*/) const override { return true; }
+
+    bool useDefaultImplementationForNulls() const override { return true; }
+
+    bool useDefaultImplementationForConstants() const override { return true; }
+
+    DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
+    {
+        if (arguments.size() != argument_names.size())
+            throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH,
+                "Function {} requires 3 arguments, but {} given", getName(), arguments.size());
+
+        for (size_t i = 0; i < argument_names.size(); ++i)
+        {
+            DataTypePtr argument_type = arguments[i];
+            if (!isNumber(argument_type))
+                throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT,
+                    "Argument '{}' for function {} must be number", std::string(argument_names[i]), getName());
+        }
+
+        return std::make_shared<typename Traits::ReturnDataType>();
+    }
+
+    ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr &, size_t input_rows_count) const override
+    {
+        const DataTypePtr converted_argument_type = std::make_shared<DataTypeFloat32>();
+        Columns converted_arguments;
+        converted_arguments.reserve(arguments.size());
+        for (const auto & argument : arguments)
+        {
+            ColumnPtr argument_column = castColumn(argument, converted_argument_type);
+            argument_column = argument_column->convertToFullColumnIfConst();
+            converted_arguments.push_back(argument_column);
+        }
+
+        auto res_column = Traits::ReturnColumnType::create(input_rows_count);
+        auto & result_data = res_column->getData();
+
+        const auto & year_data = typeid_cast<const ColumnFloat32 &>(*converted_arguments[0]).getData();
+        const auto & month_data = typeid_cast<const ColumnFloat32 &>(*converted_arguments[1]).getData();
+        const auto & day_data = typeid_cast<const ColumnFloat32 &>(*converted_arguments[2]).getData();
+
+        const auto & date_lut = DateLUT::instance();
+
+        for (size_t i = 0; i < input_rows_count; ++i)
+        {
+            const auto year = year_data[i];
+            const auto month = month_data[i];
+            const auto day = day_data[i];
+
+            Int32 day_num = 0;
+
+            if (year >= Traits::MIN_YEAR &&
+                year <= Traits::MAX_YEAR &&
+                month >= 1 && month <= 12 &&
+                day >= 1 && day <= 31 &&
+                YearMonthDayToSingleInt(year, month, day) <= Traits::MAX_DATE)
+            {
+                day_num = date_lut.makeDayNum(year, month, day);
+            }
+
+            result_data[i] = day_num;
+        }
+
+        return res_column;
+    }
+};
+
+// makeDate(year, month, day)
+struct MakeDateTraits
+{
+    static constexpr auto name = "makeDate";
+    using ReturnDataType = DataTypeDate;
+    using ReturnColumnType = ColumnUInt16;
+
+    static constexpr auto MIN_YEAR = 1970;
+    static constexpr auto MAX_YEAR = 2149;
+    // This date has the maximum day number that fits in 16-bit uint
+    static constexpr auto MAX_DATE = YearMonthDayToSingleInt(MAX_YEAR, 6, 6);
+};
+
+// makeDate32(year, month, day)
+struct MakeDate32Traits
+{
+    static constexpr auto name = "makeDate32";
+    using ReturnDataType = DataTypeDate32;
+    using ReturnColumnType = ColumnInt32;
+
+    static constexpr auto MIN_YEAR = 1925;
+    static constexpr auto MAX_YEAR = 2283;
+    static constexpr auto MAX_DATE = YearMonthDayToSingleInt(MAX_YEAR, 11, 11);
+};
+
+}
+
+void registerFunctionsMakeDate(FunctionFactory & factory)
+{
+    factory.registerFunction<FunctionMakeDate<MakeDateTraits>>();
+    factory.registerFunction<FunctionMakeDate<MakeDate32Traits>>();
+}
+
+}
--- a/src/Functions/registerFunctions.cpp
+++ b/src/Functions/registerFunctions.cpp
@ -8,6 +8,7 @@ namespace DB
 void registerFunctionsArithmetic(FunctionFactory &);
 void registerFunctionsArray(FunctionFactory &);
 void registerFunctionsTuple(FunctionFactory &);
+void registerFunctionsMakeDate(FunctionFactory &);
 void registerFunctionsMap(FunctionFactory &);
 void registerFunctionsBitmap(FunctionFactory &);
 void registerFunctionsBinaryRepr(FunctionFactory &);
@ -73,6 +74,7 @@ void registerFunctions()
    registerFunctionsArithmetic(factory);
    registerFunctionsArray(factory);
    registerFunctionsTuple(factory);
+    registerFunctionsMakeDate(factory);
    registerFunctionsMap(factory);
    registerFunctionsBitmap(factory);
    registerFunctionsBinaryRepr(factory);
--- a/src/Functions/registerFunctionsMiscellaneous.cpp
+++ b/src/Functions/registerFunctionsMiscellaneous.cpp
@ -80,6 +80,7 @@ void registerFunctionInitialQueryID(FunctionFactory & factory);
 void registerFunctionServerUUID(FunctionFactory &);
 void registerFunctionZooKeeperSessionUptime(FunctionFactory &);
 void registerFunctionGetOSKernelVersion(FunctionFactory &);
+void registerFunctionFlattenTuple(FunctionFactory &);

 #if USE_ICU
 void registerFunctionConvertCharset(FunctionFactory &);
@ -166,6 +167,7 @@ void registerFunctionsMiscellaneous(FunctionFactory & factory)
    registerFunctionServerUUID(factory);
    registerFunctionZooKeeperSessionUptime(factory);
    registerFunctionGetOSKernelVersion(factory);
+    registerFunctionFlattenTuple(factory);

 #if USE_ICU
    registerFunctionConvertCharset(factory);
--- a/Show More
+++ b/Show More