Merge branch 'master' into something-strange-with-the-test-about-refreshable-materialized-views

.github/actions/release/action.yml

@@ -62,8 +62,8 @@ runs:
       if: ${{ inputs.type == 'patch' }}
       shell: bash
       run: |
-        python3 ./tests/ci/create_release.py --set-progress-started --progress "update ChangeLog"
+        git checkout master
-        [ "$(git branch --show-current)" != "master" ] && echo "not on the master" && exit 1
+        python3 ./tests/ci/create_release.py --set-progress-started --progress "update changelog, docker version, security"
         echo "List versions"
         ./utils/list-versions/list-versions.sh > ./utils/list-versions/version_date.tsv
         echo "Update docker version"
@@ -96,17 +96,13 @@ runs:
           Update version_date.tsv and changelogs after ${{ env.RELEASE_TAG }}
           ### Changelog category (leave one):
           - Not for changelog (changelog entry is not required)
-    - name: Reset changes if Dry-run
+    - name: Complete previous steps and Restore git state
-      if: ${{ inputs.dry-run }}
+      if: ${{ inputs.type == 'patch' }}
       shell: bash
       run: |
-        git reset --hard HEAD
-    - name: Checkout back to GITHUB_REF
-      shell: bash
-      run: |
-        git checkout "$GITHUB_REF_NAME"
-        # set current progress to OK
         python3 ./tests/ci/create_release.py --set-progress-completed
+        git reset --hard HEAD
+        git checkout "$GITHUB_REF_NAME"
     - name: Create GH Release
       shell: bash
       if: ${{ inputs.type == 'patch' }}
@@ -146,24 +142,23 @@ runs:
       if: ${{ inputs.type == 'patch' }}
       shell: bash
       run: |
-        python3 ./tests/ci/create_release.py --set-progress-started --progress "docker server release"
         cd "./tests/ci"
+        python3 ./create_release.py --set-progress-started --progress "docker server release"
         export CHECK_NAME="Docker server image"
         python3 docker_server.py --release-type auto --version ${{ env.RELEASE_TAG }} --check-name "$CHECK_NAME" --sha ${{ env.COMMIT_SHA }} ${{ ! inputs.dry-run && '--push' || '' }}
-        python3 ./tests/ci/create_release.py --set-progress-completed
+        python3 ./create_release.py --set-progress-completed
     - name: Docker clickhouse/clickhouse-keeper building
       if: ${{ inputs.type == 'patch' }}
       shell: bash
       run: |
-        python3 ./tests/ci/create_release.py --set-progress-started --progress "docker keeper release"
         cd "./tests/ci"
+        python3 ./create_release.py --set-progress-started --progress "docker keeper release"
         export CHECK_NAME="Docker keeper image"
         python3 docker_server.py --release-type auto --version ${{ env.RELEASE_TAG }} --check-name "$CHECK_NAME" --sha ${{ env.COMMIT_SHA }} ${{ ! inputs.dry-run && '--push'  || '' }}
-        python3 ./tests/ci/create_release.py --set-progress-completed
+        python3 ./create_release.py --set-progress-completed
-    - name: Set Release progress completed
+    - name: Set current Release progress to Completed with OK
       shell: bash
       run: |
-        # If we here - set completed status, to post proper Slack OK or FAIL message in the next step
         python3 ./tests/ci/create_release.py --set-progress-started --progress "completed"
         python3 ./tests/ci/create_release.py --set-progress-completed
     - name: Post Slack Message

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

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

src/Analyzer/Passes/LogicalExpressionOptimizerPass.cpp

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

src/Common/Allocator.cpp

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

src/Common/CurrentMetrics.cpp

@@ -234,10 +234,10 @@
     M(PartsCommitted, "Deprecated. See PartsActive.") \
     M(PartsPreActive, "The part is in data_parts, but not used for SELECTs.") \
     M(PartsActive, "Active data part, used by current and upcoming SELECTs.") \
-    M(AttachedDatabase, "Active database, used by current and upcoming SELECTs.") \
+    M(AttachedDatabase, "Active databases.") \
-    M(AttachedTable, "Active table, used by current and upcoming SELECTs.") \
+    M(AttachedTable, "Active tables.") \
-    M(AttachedView, "Active view, used by current and upcoming SELECTs.") \
+    M(AttachedView, "Active views.") \
-    M(AttachedDictionary, "Active dictionary, used by current and upcoming SELECTs.") \
+    M(AttachedDictionary, "Active dictionaries.") \
     M(PartsOutdated, "Not active data part, but could be used by only current SELECTs, could be deleted after SELECTs finishes.") \
     M(PartsDeleting, "Not active data part with identity refcounter, it is deleting right now by a cleaner.") \
     M(PartsDeleteOnDestroy, "Part was moved to another disk and should be deleted in own destructor.") \

src/Common/getRandomASCIIString.cpp

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

src/Common/getRandomASCIIString.h

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

src/Databases/DatabaseLazy.cpp

@@ -195,7 +195,7 @@ void DatabaseLazy::attachTable(ContextPtr /* context_ */, const String & table_n
         snapshot_detached_tables.erase(table_name);
     }
 
-    CurrentMetrics::add(CurrentMetrics::AttachedTable, 1);
+    CurrentMetrics::add(CurrentMetrics::AttachedTable);
 }
 
 StoragePtr DatabaseLazy::detachTable(ContextPtr /* context */, const String & table_name)
@@ -221,7 +221,7 @@ StoragePtr DatabaseLazy::detachTable(ContextPtr /* context */, const String & ta
                 .metadata_path = getObjectMetadataPath(table_name),
                 .is_permanently = false});
 
-        CurrentMetrics::sub(CurrentMetrics::AttachedTable, 1);
+        CurrentMetrics::sub(CurrentMetrics::AttachedTable);
     }
     return res;
 }

src/Databases/DatabasesCommon.cpp

@@ -289,8 +289,8 @@ StoragePtr DatabaseWithOwnTablesBase::detachTableUnlocked(const String & table_n
     tables.erase(it);
     table_storage->is_detached = true;
 
-    if (table_storage->isSystemStorage() == false)
+    if (!table_storage->isSystemStorage() && database_name != DatabaseCatalog::SYSTEM_DATABASE)
-        CurrentMetrics::sub(getAttachedCounterForStorage(table_storage), 1);
+        CurrentMetrics::sub(getAttachedCounterForStorage(table_storage));
 
     auto table_id = table_storage->getStorageID();
     if (table_id.hasUUID())
@@ -334,8 +334,8 @@ void DatabaseWithOwnTablesBase::attachTableUnlocked(const String & table_name, c
     /// non-Atomic database the is_detached is set to true before RENAME.
     table->is_detached = false;
 
-    if (table->isSystemStorage() == false && table_id.database_name != DatabaseCatalog::SYSTEM_DATABASE)
+    if (!table->isSystemStorage() && table_id.database_name != DatabaseCatalog::SYSTEM_DATABASE)
-        CurrentMetrics::add(getAttachedCounterForStorage(table), 1);
+        CurrentMetrics::add(getAttachedCounterForStorage(table));
 }
 
 void DatabaseWithOwnTablesBase::shutdown()

src/Interpreters/InterpreterCreateQuery.cpp

@@ -16,7 +16,6 @@
 #include <Common/logger_useful.h>
 #include <Core/Settings.h>
 #include <Parsers/ASTSetQuery.h>
-#include <base/hex.h>
 
 #include <Core/Defines.h>
 #include <Core/SettingsEnums.h>
@@ -82,13 +81,13 @@
 #include <Interpreters/ApplyWithSubqueryVisitor.h>
 
 #include <TableFunctions/TableFunctionFactory.h>
-#include <DataTypes/DataTypeFixedString.h>
 
 #include <Functions/UserDefined/UserDefinedSQLFunctionFactory.h>
 #include <Functions/UserDefined/UserDefinedSQLFunctionVisitor.h>
 #include <Interpreters/ReplaceQueryParameterVisitor.h>
 #include <Parsers/QueryParameterVisitor.h>
 
+
 namespace CurrentMetrics
 {
     extern const Metric AttachedTable;
@@ -147,27 +146,27 @@ BlockIO InterpreterCreateQuery::createDatabase(ASTCreateQuery & create)
     }
 
     auto db_num_limit = getContext()->getGlobalContext()->getServerSettings().max_database_num_to_throw;
-    if (db_num_limit > 0)
+    if (db_num_limit > 0 && !internal)
     {
         size_t db_count = DatabaseCatalog::instance().getDatabases().size();
-        std::vector<String> system_databases = {
+        std::initializer_list<std::string_view> system_databases =
+        {
             DatabaseCatalog::TEMPORARY_DATABASE,
             DatabaseCatalog::SYSTEM_DATABASE,
             DatabaseCatalog::INFORMATION_SCHEMA,
             DatabaseCatalog::INFORMATION_SCHEMA_UPPERCASE,
-            DatabaseCatalog::DEFAULT_DATABASE
         };
 
         for (const auto & system_database : system_databases)
         {
-            if (db_count > 0 && DatabaseCatalog::instance().isDatabaseExist(system_database))
+            if (db_count > 0 && DatabaseCatalog::instance().isDatabaseExist(std::string(system_database)))
-                db_count--;
+                --db_count;
         }
 
         if (db_count >= db_num_limit)
             throw Exception(ErrorCodes::TOO_MANY_DATABASES,
-                            "Too many databases in the Clickhouse. "
+                            "Too many databases. "
-                            "The limit (setting 'max_database_num_to_throw') is set to {}, current number of databases is {}",
+                            "The limit (server configuration parameter `max_database_num_to_throw`) is set to {}, the current number of databases is {}",
                             db_num_limit, db_count);
     }
 
@@ -1601,13 +1600,13 @@ bool InterpreterCreateQuery::doCreateTable(ASTCreateQuery & create,
     }
 
     UInt64 table_num_limit = getContext()->getGlobalContext()->getServerSettings().max_table_num_to_throw;
-    if (table_num_limit > 0 && create.getDatabase() != DatabaseCatalog::SYSTEM_DATABASE)
+    if (table_num_limit > 0 && !internal)
     {
         UInt64 table_count = CurrentMetrics::get(CurrentMetrics::AttachedTable);
         if (table_count >= table_num_limit)
             throw Exception(ErrorCodes::TOO_MANY_TABLES,
-                            "Too many tables in the Clickhouse. "
+                            "Too many tables. "
-                            "The limit (setting 'max_table_num_to_throw') is set to {}, current number of tables is {}",
+                            "The limit (server configuration parameter `max_table_num_to_throw`) is set to {}, the current number of tables is {}",
                             table_num_limit, table_count);
     }
 

src/Interpreters/InterpreterSelectQuery.cpp

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

src/Parsers/ExpressionListParsers.cpp

@@ -2743,7 +2743,7 @@ Action ParserExpressionImpl::tryParseOperator(Layers & layers, IParser::Pos & po
     /// 'AND' can be both boolean function and part of the '... BETWEEN ... AND ...' operator
     if (op.function_name == "and" && layers.back()->between_counter)
     {
-        layers.back()->between_counter--;
+        --layers.back()->between_counter;
         op = finish_between_operator;
     }
 

src/Planner/PlannerJoinTree.cpp

@@ -77,7 +77,6 @@ namespace ErrorCodes
     extern const int INVALID_JOIN_ON_EXPRESSION;
     extern const int LOGICAL_ERROR;
     extern const int NOT_IMPLEMENTED;
-    extern const int SYNTAX_ERROR;
     extern const int ACCESS_DENIED;
     extern const int PARAMETER_OUT_OF_BOUND;
     extern const int TOO_MANY_COLUMNS;
@@ -1397,12 +1396,7 @@ JoinTreeQueryPlan buildQueryPlanForJoinNode(const QueryTreeNodePtr & join_table_
             {
                 if (!join_clause.hasASOF())
                     throw Exception(ErrorCodes::INVALID_JOIN_ON_EXPRESSION,
-                        "JOIN {} no inequality in ASOF JOIN ON section.",
+                        "JOIN {} no inequality in ASOF JOIN ON section",
-                        join_node.formatASTForErrorMessage());
-
-                if (table_join_clause.key_names_left.size() <= 1)
-                    throw Exception(ErrorCodes::SYNTAX_ERROR,
-                        "JOIN {} ASOF join needs at least one equi-join column",
                         join_node.formatASTForErrorMessage());
             }
 
@@ -1524,7 +1518,9 @@ JoinTreeQueryPlan buildQueryPlanForJoinNode(const QueryTreeNodePtr & join_table_
         {
             const auto & join_clause = table_join->getOnlyClause();
 
-            bool kind_allows_filtering = isInner(join_kind) || isLeft(join_kind) || isRight(join_kind);
+            bool join_type_allows_filtering = (join_strictness == JoinStrictness::All || join_strictness == JoinStrictness::Any)
+                                            && (isInner(join_kind) || isLeft(join_kind) || isRight(join_kind));
+
 
             auto has_non_const = [](const Block & block, const auto & keys)
             {
@@ -1544,7 +1540,7 @@ JoinTreeQueryPlan buildQueryPlanForJoinNode(const QueryTreeNodePtr & join_table_
             bool has_non_const_keys = has_non_const(left_plan.getCurrentDataStream().header, join_clause.key_names_left)
                 && has_non_const(right_plan.getCurrentDataStream().header, join_clause.key_names_right);
 
-            if (settings.max_rows_in_set_to_optimize_join > 0 && kind_allows_filtering && has_non_const_keys)
+            if (settings.max_rows_in_set_to_optimize_join > 0 && join_type_allows_filtering && has_non_const_keys)
             {
                 auto * left_set = add_create_set(left_plan, join_clause.key_names_left, JoinTableSide::Left);
                 auto * right_set = add_create_set(right_plan, join_clause.key_names_right, JoinTableSide::Right);

src/Processors/Transforms/MergeJoinTransform.cpp

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

src/Processors/Transforms/MergeJoinTransform.h

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

src/Processors/tests/gtest_full_sorting_join.cpp

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

tests/ci/artifactory.py

@@ -43,7 +43,6 @@ class R2MountPoint:
             self.bucket_name = self._PROD_BUCKET_NAME
 
         self.aux_mount_options = ""
-        self.async_mount = False
         if self.app == MountPointApp.S3FS:
             self.cache_dir = "/home/ubuntu/s3fs_cache"
             # self.aux_mount_options += "-o nomodtime " if self.NOMODTIME else "" not for s3fs
@@ -57,7 +56,6 @@ class R2MountPoint:
             self.mount_cmd = f"s3fs {self.bucket_name} {self.MOUNT_POINT} -o url={self.API_ENDPOINT} -o use_path_request_style -o umask=0000 -o nomultipart -o logfile={self.LOG_FILE} {self.aux_mount_options}"
         elif self.app == MountPointApp.RCLONE:
             # run rclone mount process asynchronously, otherwise subprocess.run(daemonized command) will not return
-            self.async_mount = True
             self.cache_dir = "/home/ubuntu/rclone_cache"
             self.aux_mount_options += "--no-modtime " if self.NOMODTIME else ""
             self.aux_mount_options += "-v " if self.DEBUG else ""  # -vv too verbose
@@ -85,10 +83,12 @@ class R2MountPoint:
         Shell.run(_UNMOUNT_CMD)
         Shell.run(_MKDIR_CMD)
         Shell.run(_MKDIR_FOR_CACHE)
-        # didn't manage to use simple run() and not block or fail
+        if self.app == MountPointApp.S3FS:
-        Shell.run_as_daemon(self.mount_cmd)
+            Shell.run(self.mount_cmd, check=True)
-        if self.async_mount:
+        else:
-            time.sleep(3)
+            # didn't manage to use simple run() and without blocking or failure
+            Shell.run_as_daemon(self.mount_cmd)
+        time.sleep(3)
         Shell.run(_TEST_MOUNT_CMD, check=True)
 
     @classmethod
@@ -107,6 +107,7 @@ class DebianArtifactory:
     _PROD_REPO_URL = "https://packages.clickhouse.com/deb"
 
     def __init__(self, release_info: ReleaseInfo, dry_run: bool):
+        self.release_info = release_info
         self.codename = release_info.codename
         self.version = release_info.version
         if dry_run:
@@ -154,9 +155,8 @@ class DebianArtifactory:
         print("Running test command:")
         print(f"  {cmd}")
         Shell.run(cmd, check=True)
-        release_info = ReleaseInfo.from_file()
+        self.release_info.debian_command = debian_command
-        release_info.debian_command = debian_command
+        self.release_info.dump()
-        release_info.dump()
 
 
 def _copy_if_not_exists(src: Path, dst: Path) -> Path:
@@ -177,6 +177,7 @@ class RpmArtifactory:
     _SIGN_KEY = "885E2BDCF96B0B45ABF058453E4AD4719DDE9A38"
 
     def __init__(self, release_info: ReleaseInfo, dry_run: bool):
+        self.release_info = release_info
         self.codename = release_info.codename
         self.version = release_info.version
         if dry_run:
@@ -230,9 +231,8 @@ class RpmArtifactory:
         print("Running test command:")
         print(f"  {cmd}")
         Shell.run(cmd, check=True)
-        release_info = ReleaseInfo.from_file()
+        self.release_info.rpm_command = rpm_command
-        release_info.rpm_command = rpm_command
+        self.release_info.dump()
-        release_info.dump()
 
 
 class TgzArtifactory:
@@ -240,6 +240,7 @@ class TgzArtifactory:
     _PROD_REPO_URL = "https://packages.clickhouse.com/tgz"
 
     def __init__(self, release_info: ReleaseInfo, dry_run: bool):
+        self.release_info = release_info
         self.codename = release_info.codename
         self.version = release_info.version
         if dry_run:
@@ -290,9 +291,8 @@ class TgzArtifactory:
             expected_checksum == actual_checksum
         ), f"[{actual_checksum} != {expected_checksum}]"
         Shell.run("rm /tmp/tmp.tgz*")
-        release_info = ReleaseInfo.from_file()
+        self.release_info.tgz_command = cmd
-        release_info.tgz_command = cmd
+        self.release_info.dump()
-        release_info.dump()
 
 
 def parse_args() -> argparse.Namespace:
@@ -340,9 +340,7 @@ def parse_args() -> argparse.Namespace:
 
 if __name__ == "__main__":
     args = parse_args()
-    assert args.dry_run
 
-    release_info = ReleaseInfo.from_file()
     """
     Use S3FS. RCLONE has some errors with r2 remote which I didn't figure out how to resolve:
            ERROR : IO error: NotImplemented: versionId not implemented
@@ -350,26 +348,38 @@ if __name__ == "__main__":
     """
     mp = R2MountPoint(MountPointApp.S3FS, dry_run=args.dry_run)
     if args.export_debian:
-        with ReleaseContextManager(release_progress=ReleaseProgress.EXPORT_DEB) as _:
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.EXPORT_DEB
+        ) as release_info:
             mp.init()
             DebianArtifactory(release_info, dry_run=args.dry_run).export_packages()
             mp.teardown()
     if args.export_rpm:
-        with ReleaseContextManager(release_progress=ReleaseProgress.EXPORT_RPM) as _:
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.EXPORT_RPM
+        ) as release_info:
             mp.init()
             RpmArtifactory(release_info, dry_run=args.dry_run).export_packages()
             mp.teardown()
     if args.export_tgz:
-        with ReleaseContextManager(release_progress=ReleaseProgress.EXPORT_TGZ) as _:
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.EXPORT_TGZ
+        ) as release_info:
             mp.init()
             TgzArtifactory(release_info, dry_run=args.dry_run).export_packages()
             mp.teardown()
     if args.test_debian:
-        with ReleaseContextManager(release_progress=ReleaseProgress.TEST_DEB) as _:
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.TEST_DEB
+        ) as release_info:
             DebianArtifactory(release_info, dry_run=args.dry_run).test_packages()
     if args.test_tgz:
-        with ReleaseContextManager(release_progress=ReleaseProgress.TEST_TGZ) as _:
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.TEST_TGZ
+        ) as release_info:
             TgzArtifactory(release_info, dry_run=args.dry_run).test_packages()
     if args.test_rpm:
-        with ReleaseContextManager(release_progress=ReleaseProgress.TEST_RPM) as _:
+        with ReleaseContextManager(
+            release_progress=ReleaseProgress.TEST_RPM
+        ) as release_info:
             RpmArtifactory(release_info, dry_run=args.dry_run).test_packages()

tests/ci/auto_release.py

@@ -191,7 +191,7 @@ def main():
                     title=f"Auto Release Status for {release_info.release_branch}",
                     body=release_info.to_dict(),
                 )
-    if args.post_auto_release_complete:
+    elif args.post_auto_release_complete:
         assert args.wf_status, "--wf-status Required with --post-auto-release-complete"
         if args.wf_status != SUCCESS:
             CIBuddy(dry_run=False).post_job_error(

tests/ci/ci_utils.py

@@ -182,10 +182,11 @@ class Shell:
             check=False,
         )
         if result.returncode == 0:
+            print(f"stdout: {result.stdout.strip()}")
             res = result.stdout
         else:
             print(
-                f"ERROR: stdout {result.stdout.strip()}, stderr {result.stderr.strip()}"
+                f"ERROR: stdout: {result.stdout.strip()}, stderr: {result.stderr.strip()}"
             )
             if check:
                 assert result.returncode == 0

tests/ci/create_release.py

@@ -43,6 +43,7 @@ class ReleaseProgress:
     TEST_TGZ = "test TGZ packages"
     TEST_RPM = "test RPM packages"
     TEST_DEB = "test DEB packages"
+    COMPLETED = "completed"
 
 
 class ReleaseProgressDescription:
@@ -108,6 +109,12 @@ class ReleaseInfo:
     release_progress: str = ""
     progress_description: str = ""
 
+    def is_patch(self):
+        return self.release_branch != "master"
+
+    def is_new_release_branch(self):
+        return self.release_branch == "master"
+
     @staticmethod
     def from_file() -> "ReleaseInfo":
         with open(RELEASE_INFO_FILE, "r", encoding="utf-8") as json_file:
@@ -126,12 +133,12 @@ class ReleaseInfo:
         release_tag = None
         previous_release_tag = None
         previous_release_sha = None
-        codename = None
+        codename = ""
         assert release_type in ("patch", "new")
         if release_type == "new":
             # check commit_ref is right and on a right branch
             Shell.run(
-                f"git merge-base --is-ancestor origin/{commit_ref} origin/master",
+                f"git merge-base --is-ancestor {commit_ref} origin/master",
                 check=True,
             )
             with checkout(commit_ref):
@@ -146,9 +153,6 @@ class ReleaseInfo:
                     git.latest_tag == expected_prev_tag
                 ), f"BUG: latest tag [{git.latest_tag}], expected [{expected_prev_tag}]"
                 release_tag = version.describe
-                codename = (
-                    VersionType.STABLE
-                )  # dummy value (artifactory won't be updated for new release)
                 previous_release_tag = expected_prev_tag
                 previous_release_sha = Shell.run_strict(
                     f"git rev-parse {previous_release_tag}"
@@ -205,7 +209,7 @@ class ReleaseInfo:
             and commit_sha
             and release_tag
             and version
-            and codename in ("lts", "stable")
+            and (codename in ("lts", "stable") or release_type == "new")
         )
 
         self.release_branch = release_branch
@@ -320,24 +324,27 @@ class ReleaseInfo:
                     Shell.run(
                         f"{GIT_PREFIX} checkout '{CMAKE_PATH}' '{CONTRIBUTORS_PATH}'"
                     )
-                self.version_bump_pr = GHActions.get_pr_url_by_branch(
+                    self.version_bump_pr = "dry-run"
-                    repo=GITHUB_REPOSITORY, branch=branch_upd_version_contributors
+                else:
-                )
+                    self.version_bump_pr = GHActions.get_pr_url_by_branch(
+                        repo=GITHUB_REPOSITORY, branch=branch_upd_version_contributors
+                    )
 
     def update_release_info(self, dry_run: bool) -> "ReleaseInfo":
-        branch = f"auto/{release_info.release_tag}"
+        if self.release_branch != "master":
-        if not dry_run:
+            branch = f"auto/{release_info.release_tag}"
-            url = GHActions.get_pr_url_by_branch(repo=GITHUB_REPOSITORY, branch=branch)
+            if not dry_run:
-        else:
+                url = GHActions.get_pr_url_by_branch(
-            url = "dry-run"
+                    repo=GITHUB_REPOSITORY, branch=branch
-
+                )
-        print(f"ChangeLog PR url [{url}]")
+            else:
-        self.changelog_pr = url
+                url = "dry-run"
-        print(f"Release url [{url}]")
+            print(f"ChangeLog PR url [{url}]")
-        self.release_url = (
+            self.changelog_pr = url
-            f"https://github.com/{GITHUB_REPOSITORY}/releases/tag/{self.release_tag}"
+            print(f"Release url [{url}]")
-        )
+            self.release_url = f"https://github.com/{GITHUB_REPOSITORY}/releases/tag/{self.release_tag}"
-        self.docker_command = f"docker run --rm clickhouse/clickhouse:{self.release_branch} clickhouse --version"
+            if self.release_progress == ReleaseProgress.COMPLETED:
+                self.docker_command = f"docker run --rm clickhouse/clickhouse:{self.version} clickhouse --version"
         self.dump()
         return self
 
@@ -712,13 +719,22 @@ if __name__ == "__main__":
     if args.post_status:
         release_info = ReleaseInfo.from_file()
         release_info.update_release_info(dry_run=args.dry_run)
-        if release_info.debian_command:
+        if release_info.is_new_release_branch():
+            title = "New release branch"
+        else:
+            title = "New release"
+        if (
+            release_info.progress_description == ReleaseProgressDescription.OK
+            and release_info.release_progress == ReleaseProgress.COMPLETED
+        ):
+            title = "Completed: " + title
             CIBuddy(dry_run=args.dry_run).post_done(
-                f"New release issued", dataclasses.asdict(release_info)
+                title, dataclasses.asdict(release_info)
             )
         else:
+            title = "Failed: " + title
             CIBuddy(dry_run=args.dry_run).post_critical(
-                f"Failed to issue new release", dataclasses.asdict(release_info)
+                title, dataclasses.asdict(release_info)
             )
 
     if args.set_progress_started:

tests/ci/worker/init_runner.sh

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

tests/integration/test_ssl_cert_authentication/test.py

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

tests/integration/test_table_db_num_limit/test.py

@@ -4,9 +4,7 @@ from helpers.cluster import ClickHouseCluster
 
 cluster = ClickHouseCluster(__file__)
 
-node1 = cluster.add_instance(
+node = cluster.add_instance("node", main_configs=["config/config.xml"])
-    "node1", main_configs=["config/config.xml"], with_zookeeper=True
-)
 
 
 @pytest.fixture(scope="module")
@@ -21,23 +19,28 @@ def started_cluster():
 
 
 def test_table_db_limit(started_cluster):
-    for i in range(10):
+    # By the way, default database already exists.
-        node1.query("create database db{}".format(i))
+    for i in range(9):
+        node.query("create database db{}".format(i))
 
     with pytest.raises(QueryRuntimeException) as exp_info:
-        node1.query("create database db_exp".format(i))
+        node.query("create database db_exp".format(i))
 
     assert "TOO_MANY_DATABASES" in str(exp_info)
 
     for i in range(10):
-        node1.query("create table t{} (a Int32) Engine = Log".format(i))
+        node.query("create table t{} (a Int32) Engine = Log".format(i))
+
+    # This checks that system tables are not accounted in the number of tables.
+    node.query("system flush logs")
 
-    node1.query("system flush logs")
     for i in range(10):
-        node1.query("drop table t{}".format(i))
+        node.query("drop table t{}".format(i))
+
     for i in range(10):
-        node1.query("create table t{} (a Int32) Engine = Log".format(i))
+        node.query("create table t{} (a Int32) Engine = Log".format(i))
 
     with pytest.raises(QueryRuntimeException) as exp_info:
-        node1.query("create table default.tx (a Int32) Engine = Log")
+        node.query("create table default.tx (a Int32) Engine = Log")
+
     assert "TOO_MANY_TABLES" in str(exp_info)

tests/queries/0_stateless/00927_asof_join_correct_bt.reference

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

tests/queries/0_stateless/00927_asof_join_correct_bt.sql

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

tests/queries/0_stateless/00927_asof_join_long.reference

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

tests/queries/0_stateless/00927_asof_join_long.sql

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

tests/queries/0_stateless/00927_asof_join_noninclusive.reference

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

tests/queries/0_stateless/00927_asof_join_noninclusive.sql

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

tests/queries/0_stateless/00927_asof_join_other_types.reference

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

tests/queries/0_stateless/00927_asof_join_other_types.sh

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

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

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

tests/queries/0_stateless/00927_asof_joins.reference

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

tests/queries/0_stateless/00927_asof_joins.sql

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

tests/queries/0_stateless/00976_asof_join_on.reference

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

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

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

tests/queries/0_stateless/01116_asof_join_dolbyzerr.reference

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

tests/queries/0_stateless/01116_asof_join_dolbyzerr.sql

@@ -16,3 +16,17 @@ GROUP BY
 ORDER BY
     visitorId ASC,
     orderId ASC;
+
+SELECT
+    visitorId,
+    orderId,
+    groupUniqArray(sessionId)
+FROM sessions
+ASOF INNER JOIN orders ON (sessions.visitorId = orders.visitorId) AND (sessions.date <= orders.date)
+GROUP BY
+    visitorId,
+    orderId
+ORDER BY
+    visitorId ASC,
+    orderId ASC
+SETTINGS join_algorithm = 'full_sorting_merge';

tests/queries/0_stateless/01194_http_query_id.sh

@@ -4,14 +4,22 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 # shellcheck source=../shell_config.sh
 . "$CURDIR"/../shell_config.sh
 
-url="http://${CLICKHOUSE_HOST}:${CLICKHOUSE_PORT_HTTP}/?session_id=test_01194"
+rnd="$CLICKHOUSE_DATABASE"
-rnd=$RANDOM
+url="${CLICKHOUSE_URL}&session_id=test_01194_${CLICKHOUSE_DATABASE}"
 
 ${CLICKHOUSE_CURL} -sS "$url&query=SELECT+'test_01194',$rnd,1" > /dev/null
 ${CLICKHOUSE_CURL} -sS "$url&query=SELECT+'test_01194',$rnd,2" > /dev/null
 ${CLICKHOUSE_CURL} -sS "$url" --data "SELECT 'test_01194',$rnd,3" > /dev/null
 ${CLICKHOUSE_CURL} -sS "$url" --data "SELECT 'test_01194',$rnd,4" > /dev/null
 
-${CLICKHOUSE_CURL} -sS "$url" --data "SYSTEM FLUSH LOGS"
+$CLICKHOUSE_CLIENT -q "SYSTEM FLUSH LOGS"
 
-${CLICKHOUSE_CURL} -sS "$url&query=SELECT+count(DISTINCT+query_id)+FROM+system.query_log+WHERE+current_database+LIKE+currentDatabase()+AND+query+LIKE+'SELECT+''test_01194'',$rnd%25'"
+$CLICKHOUSE_CLIENT -q "
+  SELECT
+    count(DISTINCT query_id)
+  FROM system.query_log
+  WHERE
+        current_database = currentDatabase()
+    AND event_date >= yesterday()
+    AND query LIKE 'SELECT ''test_01194'',$rnd%'
+    AND query_id != queryID()"

tests/queries/0_stateless/01338_long_select_and_alter.reference

@@ -1,3 +1,3 @@
-10
 5
-CREATE TABLE default.alter_mt\n(\n    `key` UInt64,\n    `value` UInt64\n)\nENGINE = MergeTree\nORDER BY key\nSETTINGS index_granularity = 8192
+5
+CREATE TABLE default.alter_mt\n(\n    `key` Int64,\n    `value` Int64\n)\nENGINE = MergeTree\nORDER BY key\nSETTINGS index_granularity = 8192

tests/queries/0_stateless/01338_long_select_and_alter.sh

@@ -7,16 +7,16 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 
 $CLICKHOUSE_CLIENT --query "DROP TABLE IF EXISTS alter_mt"
 
-$CLICKHOUSE_CLIENT --query "CREATE TABLE alter_mt (key UInt64, value String) ENGINE=MergeTree() ORDER BY key"
+$CLICKHOUSE_CLIENT --query "CREATE TABLE alter_mt (key Int64, value String) ENGINE=MergeTree() ORDER BY key"
 
-$CLICKHOUSE_CLIENT --query "INSERT INTO alter_mt SELECT number, toString(number) FROM numbers(5)"
+$CLICKHOUSE_CLIENT --query "INSERT INTO alter_mt SELECT number - 1 AS x, toString(x) FROM numbers(5)"
 
 $CLICKHOUSE_CLIENT --function_sleep_max_microseconds_per_block 10000000 --query "SELECT count(distinct concat(value, '_')) FROM alter_mt WHERE not sleepEachRow(2)" &
 
-# to be sure that select took all required locks
+# To be sure that select took all required locks for better test sensitivity, although it isn't guaranteed (then the test will also succeed).
 sleep 2
 
-$CLICKHOUSE_CLIENT --query "ALTER TABLE alter_mt MODIFY COLUMN value UInt64"
+$CLICKHOUSE_CLIENT --query "ALTER TABLE alter_mt MODIFY COLUMN value Int64"
 
 $CLICKHOUSE_CLIENT --query "SELECT sum(value) FROM alter_mt"
 

tests/queries/0_stateless/01338_long_select_and_alter_zookeeper.reference

@@ -1,3 +1,3 @@
-10
 5
-CREATE TABLE default.alter_mt\n(\n    `key` UInt64,\n    `value` UInt64\n)\nENGINE = ReplicatedMergeTree(\'/clickhouse/tables/01338_long_select_and_alter_zookeeper_default/alter_mt\', \'1\')\nORDER BY key\nSETTINGS index_granularity = 8192
+5
+CREATE TABLE default.alter_mt\n(\n    `key` Int64,\n    `value` Int64\n)\nENGINE = ReplicatedMergeTree(\'/clickhouse/tables/01338_long_select_and_alter_zookeeper_default/alter_mt\', \'1\')\nORDER BY key\nSETTINGS index_granularity = 8192

tests/queries/0_stateless/01338_long_select_and_alter_zookeeper.sh

@@ -7,16 +7,16 @@ CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 
 $CLICKHOUSE_CLIENT --query "DROP TABLE IF EXISTS alter_mt"
 
-$CLICKHOUSE_CLIENT --query "CREATE TABLE alter_mt (key UInt64, value String) ENGINE=ReplicatedMergeTree('/clickhouse/tables/$CLICKHOUSE_TEST_ZOOKEEPER_PREFIX/alter_mt', '1') ORDER BY key"
+$CLICKHOUSE_CLIENT --query "CREATE TABLE alter_mt (key Int64, value String) ENGINE=ReplicatedMergeTree('/clickhouse/tables/$CLICKHOUSE_TEST_ZOOKEEPER_PREFIX/alter_mt', '1') ORDER BY key"
 
-$CLICKHOUSE_CLIENT --query "INSERT INTO alter_mt SELECT number, toString(number) FROM numbers(5)"
+$CLICKHOUSE_CLIENT --query "INSERT INTO alter_mt SELECT number - 1 AS x, toString(x) FROM numbers(5)"
 
 $CLICKHOUSE_CLIENT --function_sleep_max_microseconds_per_block 10000000 --query "SELECT count(distinct concat(value, '_')) FROM alter_mt WHERE not sleepEachRow(2)" &
 
-# to be sure that select took all required locks
+# To be sure that select took all required locks for better test sensitivity, although it isn't guaranteed (then the test will also succeed).
 sleep 2
 
-$CLICKHOUSE_CLIENT --query "ALTER TABLE alter_mt MODIFY COLUMN value UInt64"
+$CLICKHOUSE_CLIENT --query "ALTER TABLE alter_mt MODIFY COLUMN value Int64"
 
 $CLICKHOUSE_CLIENT --query "SELECT sum(value) FROM alter_mt"
 

tests/queries/0_stateless/01592_long_window_functions1.sql

@@ -8,14 +8,14 @@ drop table if exists stack;
 set max_insert_threads = 4;
 
 create table stack(item_id Int64, brand_id Int64, rack_id Int64, dt DateTime, expiration_dt DateTime, quantity UInt64)
-Engine = MergeTree 
+Engine = MergeTree
-partition by toYYYYMM(dt) 
+partition by toYYYYMM(dt)
 order by (brand_id, toStartOfHour(dt)) SETTINGS index_granularity = 8192, index_granularity_bytes = '10Mi';
 
-insert into stack 
+insert into stack
-select number%99991, number%11, number%1111, toDateTime('2020-01-01 00:00:00')+number/100, 
+select number%99991, number%11, number%1111, toDateTime('2020-01-01 00:00:00')+number/100,
    toDateTime('2020-02-01 00:00:00')+number/10, intDiv(number,100)+1
-from numbers_mt(10000000);
+from numbers_mt(1000000);
 
 select '---- arrays ----';
 
@@ -32,8 +32,8 @@ select '---- window f ----';
 select cityHash64( toString( groupArray (tuple(*) ) )) from (
     select brand_id, rack_id,  quantity from
        ( select brand_id, rack_id, quantity, row_number() over (partition by brand_id, rack_id order by quantity) rn
-         from stack ) as t0 
+         from stack ) as t0
-    where rn <= 2  
+    where rn <= 2
     order by brand_id, rack_id, quantity
 ) t;
 

tests/queries/0_stateless/02240_asof_join_biginteger.reference

@@ -2,3 +2,7 @@
 0	340282366920938463463374607431768211457
 0	18446744073709551617
 0	340282366920938463463374607431768211457
+0	18446744073709551617
+0	340282366920938463463374607431768211457
+0	18446744073709551617
+0	340282366920938463463374607431768211457

tests/queries/0_stateless/02240_asof_join_biginteger.sql

@@ -3,3 +3,11 @@ select * from (select 0 as k, toInt256('340282366920938463463374607431768211457'
 
 select * from (select 0 as k, toUInt128('18446744073709551617') as v) t1 asof join (select 0 as k, toUInt128('18446744073709551616') as v) t2 using(k, v);
 select * from (select 0 as k, toUInt256('340282366920938463463374607431768211457') as v) t1 asof join (select 0 as k, toUInt256('340282366920938463463374607431768211456') as v) t2 using(k, v);
+
+SET join_algorithm = 'full_sorting_merge';
+
+select * from (select 0 as k, toInt128('18446744073709551617') as v) t1 asof join (select 0 as k, toInt128('18446744073709551616') as v) t2 using(k, v);
+select * from (select 0 as k, toInt256('340282366920938463463374607431768211457') as v) t1 asof join (select 0 as k, toInt256('340282366920938463463374607431768211456') as v) t2 using(k, v);
+
+select * from (select 0 as k, toUInt128('18446744073709551617') as v) t1 asof join (select 0 as k, toUInt128('18446744073709551616') as v) t2 using(k, v);
+select * from (select 0 as k, toUInt256('340282366920938463463374607431768211457') as v) t1 asof join (select 0 as k, toUInt256('340282366920938463463374607431768211456') as v) t2 using(k, v);

tests/queries/0_stateless/02241_join_rocksdb_bs.reference

@@ -10,59 +10,3 @@
 1
 1
 1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1

tests/queries/0_stateless/02241_join_rocksdb_bs.sql.j2

@@ -1,4 +1,4 @@
--- Tags: use-rocksdb, long, no-s3-storage, no-random-settings, no-random-merge-tree-settings
+-- Tags: use-rocksdb, long, no-s3-storage
 
 SET join_algorithm = 'direct';
 
@@ -13,27 +13,21 @@ INSERT INTO rdb_{{ table_size }}
     SELECT (sipHash64(number) % {{ table_size }}) as key, ('val' || toString(key)) AS value
     FROM numbers_mt({{ table_size }});
 
-{% for block_size in [10, 11, 128, 129, 65505, 65506, 70000] -%}
-
-{% if block_size * 5000 > table_size -%}
-
-SET max_block_size = {{ block_size }};
-
 {% for right_size in [table_size // 2, table_size + table_size // 4 + 1] -%}
 
 SELECT count() == (SELECT count() FROM rdb_{{ table_size }} WHERE key < {{ right_size }})
 FROM (SELECT number as k FROM numbers_mt({{ right_size }})) as t1
 INNER JOIN rdb_{{ table_size }} as rdb
-ON rdb.key == t1.k;
+ON rdb.key == t1.k
+{% if table_size < 100 %}SETTINGS max_block_size = 1{% endif -%}
+;
 
 SELECT count() == {{ right_size }} and countIf(value != '') == (SELECT count() FROM rdb_{{ table_size }} WHERE key < {{ right_size }})
 FROM (SELECT number as k FROM numbers_mt({{ right_size }})) as t1
 LEFT JOIN rdb_{{ table_size }} as rdb
-ON rdb.key == t1.k;
+ON rdb.key == t1.k
-
+{% if table_size < 100 %}SETTINGS max_block_size = 1{% endif -%}
-{% endfor -%}
+;
-
-{% endif -%}
 
 {% endfor -%}
 {% endfor -%}

tests/queries/0_stateless/02276_full_sort_join_unsupported.sql

@@ -19,8 +19,6 @@ SELECT * FROM t1 ANTI JOIN t2 ON t1.key = t2.key; -- { serverError NOT_IMPLEMENT
 
 SELECT * FROM t1 SEMI JOIN t2 ON t1.key = t2.key; -- { serverError NOT_IMPLEMENTED }
 
-SELECT * FROM t1 ASOF JOIN t2 ON t1.key = t2.key AND t1.val > t2.val; -- { serverError NOT_IMPLEMENTED }
-
 SELECT * FROM t1 ANY JOIN t2 ON t1.key = t2.key SETTINGS any_join_distinct_right_table_keys = 1; -- { serverError NOT_IMPLEMENTED }
 
 SELECT * FROM t1 JOIN t2 USING (key) SETTINGS join_use_nulls = 1; -- { serverError NOT_IMPLEMENTED }

tests/queries/0_stateless/02724_function_in_left_table_clause_asof_join.reference

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

tests/queries/0_stateless/02724_function_in_left_table_clause_asof_join.sql

@@ -6,3 +6,15 @@ ASOF LEFT JOIN (
   select 1 as session_id, 4 as id
 ) as visitors
 ON visitors.session_id <= sessions.id AND arrayFirst(a -> a, arrayMap((a) -> a, sessions.arr)) = visitors.id
+;
+
+select count(*)
+from (
+  select 1 as id, [1, 2, 3] as arr
+) as sessions
+ASOF LEFT JOIN (
+  select 1 as session_id, 4 as id
+) as visitors
+ON visitors.session_id <= sessions.id AND arrayFirst(a -> a, arrayMap((a) -> a, sessions.arr)) = visitors.id
+SETTINGS join_algorithm = 'full_sorting_merge'
+;

tests/queries/0_stateless/02911_join_on_nullsafe_optimization.reference

@@ -35,6 +35,10 @@ SELECT * FROM t1 JOIN t2 ON (t1.x == t2.x AND ((t2.x IS NOT NULL) AND (t1.x IS N
 2	2	2	2
 3	3	3	33
 \N	\N	\N	\N
+SELECT * FROM t1 JOIN t2 ON (t1.x == t2.x AND ((t2.x IS NOT NULL) AND (t1.x IS NOT NULL)) ) OR ( t2.x <> t1.x AND (t2.x IS NULL) AND (t1.x IS NULL) ) ORDER BY t1.x NULLS LAST;
+2	2	2	2
+3	3	3	33
+\N	\N	\N	\N
 -- aliases defined in the join condition are valid
 -- FIXME(@vdimir) broken query formatting for the following queries:
 -- SELECT *, e, e2 FROM t1 FULL JOIN t2 ON ( ( ((t1.x == t2.x) AS e) AND ((t2.x IS NOT NULL) AND (t1.x IS NOT NULL)) ) OR ( (t2.x IS NULL) AND (t1.x IS NULL) ) AS e2 ) ORDER BY t1.x NULLS LAST, t2.x NULLS LAST;

tests/queries/0_stateless/02911_join_on_nullsafe_optimization.sql

@@ -34,6 +34,7 @@ SELECT * FROM t1 JOIN t2 ON (t1.x <=> t2.x OR (t1.x IS NULL AND t2.x IS NULL)) A
 SELECT x = y OR (x IS NULL AND y IS NULL) FROM t1 ORDER BY x NULLS LAST;
 
 SELECT * FROM t1 JOIN t2 ON (t1.x == t2.x AND ((t2.x IS NOT NULL) AND (t1.x IS NOT NULL)) ) OR ( (t2.x IS NULL) AND (t1.x IS NULL) ) ORDER BY t1.x NULLS LAST;
+SELECT * FROM t1 JOIN t2 ON (t1.x == t2.x AND ((t2.x IS NOT NULL) AND (t1.x IS NOT NULL)) ) OR ( t2.x <> t1.x AND (t2.x IS NULL) AND (t1.x IS NULL) ) ORDER BY t1.x NULLS LAST;
 
 -- aliases defined in the join condition are valid
 -- FIXME(@vdimir) broken query formatting for the following queries:

tests/queries/0_stateless/03002_part_log_rmt_fetch_merge_error.reference

@@ -1,10 +0,0 @@
-before
-rmt_master	NewPart	0	1
-rmt_master	MergeParts	0	1
-rmt_slave	MergeParts	1	0
-rmt_slave	DownloadPart	0	1
-after
-rmt_master	NewPart	0	1
-rmt_master	MergeParts	0	1
-rmt_slave	MergeParts	1	0
-rmt_slave	DownloadPart	0	2

tests/queries/0_stateless/03002_part_log_rmt_fetch_merge_error.sql

@@ -1,35 +0,0 @@
--- Tags: no-replicated-database, no-parallel, no-shared-merge-tree
--- SMT: The merge process is completely different from RMT
-
-drop table if exists rmt_master;
-drop table if exists rmt_slave;
-
-create table rmt_master (key Int) engine=ReplicatedMergeTree('/clickhouse/{database}', 'master') order by key settings always_fetch_merged_part=0;
--- always_fetch_merged_part=1, consider this table as a "slave"
-create table rmt_slave (key Int) engine=ReplicatedMergeTree('/clickhouse/{database}', 'slave') order by key settings always_fetch_merged_part=1;
-
-insert into rmt_master values (1);
-
-system sync replica rmt_master;
-system sync replica rmt_slave;
-system stop replicated sends rmt_master;
-optimize table rmt_master final settings alter_sync=1, optimize_throw_if_noop=1;
-
-select sleep(3) format Null;
-
-system flush logs;
-select 'before';
-select table, event_type, error>0, countIf(error=0) from system.part_log where database = currentDatabase() group by 1, 2, 3 order by 1, 2, 3;
-
-system start replicated sends rmt_master;
--- sleep few seconds to try rmt_slave to fetch the part and reflect this error
--- in system.part_log
-select sleep(3) format Null;
-system sync replica rmt_slave;
-
-system flush logs;
-select 'after';
-select table, event_type, error>0, countIf(error=0) from system.part_log where database = currentDatabase() group by 1, 2, 3 order by 1, 2, 3;
-
-drop table rmt_master;
-drop table rmt_slave;

tests/queries/0_stateless/03002_part_log_rmt_fetch_mutate_error.reference

@@ -1,10 +0,0 @@
-before
-rmt_master	NewPart	0	1
-rmt_master	MutatePart	0	1
-rmt_slave	DownloadPart	0	1
-rmt_slave	MutatePart	1	0
-after
-rmt_master	NewPart	0	1
-rmt_master	MutatePart	0	1
-rmt_slave	DownloadPart	0	2
-rmt_slave	MutatePart	1	0

tests/queries/0_stateless/03002_part_log_rmt_fetch_mutate_error.sql

@@ -1,41 +0,0 @@
--- Tags: no-replicated-database, no-parallel, no-shared-merge-tree
--- SMT: The merge process is completely different from RMT
-
-drop table if exists rmt_master;
-drop table if exists rmt_slave;
-
-create table rmt_master (key Int) engine=ReplicatedMergeTree('/clickhouse/{database}', 'master') order by tuple() settings always_fetch_merged_part=0, old_parts_lifetime=600;
--- prefer_fetch_merged_part_*_threshold=0, consider this table as a "slave"
-create table rmt_slave (key Int) engine=ReplicatedMergeTree('/clickhouse/{database}', 'slave') order by tuple() settings prefer_fetch_merged_part_time_threshold=0, prefer_fetch_merged_part_size_threshold=0, old_parts_lifetime=600;
-
-insert into rmt_master values (1);
-
-system sync replica rmt_master;
-system sync replica rmt_slave;
-system stop replicated sends rmt_master;
-system stop pulling replication log rmt_slave;
-alter table rmt_master update key=key+100 where 1 settings alter_sync=1;
-
--- first we need to make the rmt_master execute mutation so that it will have
--- the part, and rmt_slave will consider it instead of performing mutation on
--- it's own, otherwise prefer_fetch_merged_part_*_threshold will be simply ignored
-select sleep(3) format Null;
-system start pulling replication log rmt_slave;
--- and sleep few more seconds to try rmt_slave to fetch the part and reflect
--- this error in system.part_log
-select sleep(3) format Null;
-
-system flush logs;
-select 'before';
-select table, event_type, error>0, countIf(error=0) from system.part_log where database = currentDatabase() group by 1, 2, 3 order by 1, 2, 3;
-
-system start replicated sends rmt_master;
-select sleep(3) format Null;
-system sync replica rmt_slave;
-
-system flush logs;
-select 'after';
-select table, event_type, error>0, countIf(error=0) from system.part_log where database = currentDatabase() group by 1, 2, 3 order by 1, 2, 3;
-
-drop table rmt_master;
-drop table rmt_slave;

tests/queries/0_stateless/03143_asof_join_ddb_long.reference

@@ -0,0 +1,2 @@
+49999983751397	10000032
+49999983751397	10000032

tests/queries/0_stateless/03143_asof_join_ddb_long.sql

@@ -0,0 +1,50 @@
+-- Tags: long
+
+DROP TABLE IF EXISTS build;
+DROP TABLE IF EXISTS skewed_probe;
+
+SET session_timezone = 'UTC';
+
+CREATE TABLE build ENGINE = MergeTree ORDER BY (key, begin)
+AS
+    SELECT
+        toDateTime('1990-03-21 13:00:00') + INTERVAL number MINUTE AS begin,
+        number % 4 AS key,
+        number AS value
+    FROM numbers(0, 10000000);
+
+CREATE TABLE skewed_probe ENGINE = MergeTree ORDER BY (key, begin)
+AS
+    SELECT
+        toDateTime('1990-04-21 13:00:01') + INTERVAL number MINUTE AS begin,
+        0 AS key
+    FROM numbers(0, 5)
+    UNION ALL
+    SELECT
+        toDateTime('1990-05-21 13:00:01') + INTERVAL number MINUTE AS begin,
+        1 AS key
+    FROM numbers(0, 10)
+    UNION ALL
+    SELECT
+        toDateTime('1990-06-21 13:00:01') + INTERVAL number MINUTE AS begin,
+        2 AS key
+    FROM numbers(0, 20)
+    UNION ALL
+    SELECT
+        toDateTime('1990-03-21 13:00:01') + INTERVAL number MINUTE AS begin,
+        3 AS key
+    FROM numbers(0, 10000000);
+
+
+SELECT SUM(value), COUNT(*)
+FROM skewed_probe
+ASOF JOIN build
+USING (key, begin)
+;
+
+SELECT SUM(value), COUNT(*)
+FROM skewed_probe
+ASOF JOIN build
+USING (key, begin)
+SETTINGS join_algorithm = 'full_sorting_merge'
+;

tests/queries/0_stateless/03144_asof_join_ddb_doubles.reference

@@ -0,0 +1,58 @@
+1	0
+2	0
+3	1
+4	1
+5	1
+6	2
+7	2
+8	3
+9	3
+0	0
+1	0
+2	0
+3	1
+4	1
+5	1
+6	2
+7	2
+8	3
+9	3
+1	1	0
+1	2	0
+1	3	1
+1	4	1
+1	5	1
+1	6	2
+1	7	2
+1	8	3
+1	9	3
+2	0	10
+2	1	10
+2	2	10
+2	3	10
+2	4	10
+2	5	10
+2	6	10
+2	7	20
+2	8	20
+2	9	20
+1	0	0
+1	1	0
+1	2	0
+1	3	1
+1	4	1
+1	5	1
+1	6	2
+1	7	2
+1	8	3
+1	9	3
+2	0	10
+2	1	10
+2	2	10
+2	3	10
+2	4	10
+2	5	10
+2	6	10
+2	7	20
+2	8	20
+2	9	20

tests/queries/0_stateless/03144_asof_join_ddb_doubles.sql

@@ -0,0 +1,65 @@
+SET join_algorithm = 'full_sorting_merge';
+SET allow_experimental_analyzer = 1;
+
+DROP TABLE IF EXISTS events0;
+
+CREATE TABLE events0 (
+    begin Float64,
+    value Int32
+) ENGINE = MergeTree ORDER BY begin;
+
+INSERT INTO events0 VALUES (1.0, 0), (3.0, 1), (6.0, 2), (8.0, 3);
+
+SELECT p.ts, e.value
+FROM
+    (SELECT number :: Float64 AS ts FROM numbers(10)) p
+ASOF  JOIN events0 e
+ON p.ts >= e.begin
+ORDER BY p.ts ASC;
+
+SELECT p.ts, e.value
+FROM
+    (SELECT number :: Float64 AS ts FROM numbers(10)) p
+ASOF LEFT JOIN events0 e
+ON p.ts >= e.begin
+ORDER BY p.ts ASC
+-- SETTINGS join_use_nulls = 1
+;
+
+DROP TABLE IF EXISTS events0;
+
+DROP TABLE IF EXISTS events;
+DROP TABLE IF EXISTS probes;
+
+CREATE TABLE events (
+    key Int32,
+    begin Float64,
+    value Int32
+) ENGINE = MergeTree ORDER BY (key, begin);
+
+INSERT INTO events VALUES (1, 1.0, 0), (1, 3.0, 1), (1, 6.0, 2), (1, 8.0, 3), (2, 0.0, 10), (2, 7.0, 20), (2, 11.0, 30);
+
+CREATE TABLE probes (
+    key Int32,
+    ts Float64
+) ENGINE = MergeTree ORDER BY (key, ts) AS
+SELECT
+    key.number,
+    ts.number
+FROM
+    numbers(1, 2) as key,
+    numbers(10) as ts
+SETTINGS join_algorithm = 'hash';
+
+SELECT p.key, p.ts, e.value
+FROM probes p
+ASOF JOIN events e
+ON p.key = e.key AND p.ts >= e.begin
+ORDER BY p.key, p.ts ASC;
+
+SELECT p.key, p.ts, e.value
+FROM probes p
+ASOF LEFT JOIN events e
+ON p.key = e.key AND p.ts >= e.begin
+ORDER BY p.key, p.ts ASC NULLS FIRST;
+

tests/queries/0_stateless/03145_asof_join_ddb_inequalities.reference

@@ -0,0 +1,73 @@
+-
+2023-03-21 12:00:00	1970-01-01 00:00:00	-1
+2023-03-21 13:00:00	1970-01-01 00:00:00	-1
+2023-03-21 14:00:00	2023-03-21 13:00:00	0
+2023-03-21 15:00:00	2023-03-21 14:00:00	1
+2023-03-21 16:00:00	2023-03-21 15:00:00	2
+2023-03-21 17:00:00	2023-03-21 16:00:00	3
+2023-03-21 18:00:00	2023-03-21 16:00:00	3
+2023-03-21 19:00:00	2023-03-21 16:00:00	3
+2023-03-21 20:00:00	2023-03-21 16:00:00	3
+2023-03-21 21:00:00	2023-03-21 16:00:00	3
+2027-10-18 11:03:27	2023-03-21 16:00:00	3
+-
+2023-03-21 12:00:00	1970-01-01 00:00:00	-1
+2023-03-21 13:00:00	1970-01-01 00:00:00	-1
+2023-03-21 14:00:00	2023-03-21 13:00:00	0
+2023-03-21 15:00:00	2023-03-21 14:00:00	1
+2023-03-21 16:00:00	2023-03-21 15:00:00	2
+2023-03-21 17:00:00	2023-03-21 16:00:00	3
+2023-03-21 18:00:00	2023-03-21 16:00:00	3
+2023-03-21 19:00:00	2023-03-21 16:00:00	3
+2023-03-21 20:00:00	2023-03-21 16:00:00	3
+2023-03-21 21:00:00	2023-03-21 16:00:00	3
+2027-10-18 11:03:27	2023-03-21 16:00:00	3
+\N	\N	\N
+2023-03-21 12:00:00	2023-03-21 13:00:00	0
+2023-03-21 13:00:00	2023-03-21 13:00:00	0
+2023-03-21 14:00:00	2023-03-21 14:00:00	1
+2023-03-21 15:00:00	2023-03-21 15:00:00	2
+2023-03-21 16:00:00	2023-03-21 16:00:00	3
+2023-03-21 17:00:00	2027-10-18 11:03:27	9
+2023-03-21 18:00:00	2027-10-18 11:03:27	9
+2023-03-21 19:00:00	2027-10-18 11:03:27	9
+2023-03-21 20:00:00	2027-10-18 11:03:27	9
+2023-03-21 21:00:00	2027-10-18 11:03:27	9
+2027-10-18 11:03:27	2027-10-18 11:03:27	9
+-
+2023-03-21 12:00:00	2023-03-21 13:00:00	0
+2023-03-21 13:00:00	2023-03-21 13:00:00	0
+2023-03-21 14:00:00	2023-03-21 14:00:00	1
+2023-03-21 15:00:00	2023-03-21 15:00:00	2
+2023-03-21 16:00:00	2023-03-21 16:00:00	3
+2023-03-21 17:00:00	2027-10-18 11:03:27	9
+2023-03-21 18:00:00	2027-10-18 11:03:27	9
+2023-03-21 19:00:00	2027-10-18 11:03:27	9
+2023-03-21 20:00:00	2027-10-18 11:03:27	9
+2023-03-21 21:00:00	2027-10-18 11:03:27	9
+2027-10-18 11:03:27	2027-10-18 11:03:27	9
+\N	\N	\N
+-
+2023-03-21 12:00:00	2023-03-21 13:00:00	0
+2023-03-21 13:00:00	2023-03-21 14:00:00	1
+2023-03-21 14:00:00	2023-03-21 15:00:00	2
+2023-03-21 15:00:00	2023-03-21 16:00:00	3
+2023-03-21 16:00:00	2027-10-18 11:03:27	9
+2023-03-21 17:00:00	2027-10-18 11:03:27	9
+2023-03-21 18:00:00	2027-10-18 11:03:27	9
+2023-03-21 19:00:00	2027-10-18 11:03:27	9
+2023-03-21 20:00:00	2027-10-18 11:03:27	9
+2023-03-21 21:00:00	2027-10-18 11:03:27	9
+-
+2023-03-21 12:00:00	2023-03-21 13:00:00	0
+2023-03-21 13:00:00	2023-03-21 14:00:00	1
+2023-03-21 14:00:00	2023-03-21 15:00:00	2
+2023-03-21 15:00:00	2023-03-21 16:00:00	3
+2023-03-21 16:00:00	2027-10-18 11:03:27	9
+2023-03-21 17:00:00	2027-10-18 11:03:27	9
+2023-03-21 18:00:00	2027-10-18 11:03:27	9
+2023-03-21 19:00:00	2027-10-18 11:03:27	9
+2023-03-21 20:00:00	2027-10-18 11:03:27	9
+2023-03-21 21:00:00	2027-10-18 11:03:27	9
+2027-10-18 11:03:27	\N	\N
+\N	\N	\N

tests/queries/0_stateless/03145_asof_join_ddb_inequalities.sql

@@ -0,0 +1,66 @@
+DROP TABLE IF EXISTS events0;
+DROP TABLE IF EXISTS probe0;
+
+SET allow_experimental_analyzer = 1;
+SET join_algorithm = 'full_sorting_merge';
+
+CREATE TABLE events0 (
+    begin Nullable(DateTime('UTC')),
+    value Int32
+) ENGINE = MergeTree ORDER BY tuple();
+
+INSERT INTO events0 SELECT toDateTime('2023-03-21 13:00:00', 'UTC') + INTERVAL number HOUR, number FROM numbers(4);
+INSERT INTO events0 VALUES (NULL, -10),('0000-01-01 00:00:00', -1), ('9999-12-31 23:59:59', 9);
+
+CREATE TABLE probe0 (
+    begin Nullable(DateTime('UTC'))
+) ENGINE = MergeTree ORDER BY tuple();
+
+INSERT INTO probe0 SELECT toDateTime('2023-03-21 12:00:00', 'UTC') + INTERVAl number HOUR FROM numbers(10);
+INSERT INTO probe0 VALUES (NULL),('9999-12-31 23:59:59');
+
+SET join_use_nulls = 1;
+
+SELECT '-';
+SELECT p.begin, e.begin, e.value
+FROM probe0 p
+ASOF JOIN events0 e
+ON p.begin > e.begin
+ORDER BY p.begin ASC;
+
+SELECT '-';
+SELECT p.begin, e.begin, e.value
+FROM probe0 p
+ASOF LEFT JOIN events0 e
+ON p.begin > e.begin
+ORDER BY p.begin ASC;
+
+SELECT p.begin, e.begin, e.value
+FROM probe0 p
+ASOF JOIN events0 e
+ON p.begin <= e.begin
+ORDER BY p.begin ASC;
+
+SELECT '-';
+SELECT p.begin, e.begin, e.value
+FROM probe0 p
+ASOF LEFT JOIN events0 e
+ON p.begin <= e.begin
+ORDER BY p.begin ASC;
+
+SELECT '-';
+SELECT p.begin, e.begin, e.value
+FROM probe0 p
+ASOF JOIN events0 e
+ON p.begin < e.begin
+ORDER BY p.begin ASC;
+
+SELECT '-';
+SELECT p.begin, e.begin, e.value
+FROM probe0 p
+ASOF LEFT JOIN events0 e
+ON p.begin < e.begin
+ORDER BY p.begin ASC;
+
+DROP TABLE IF EXISTS events0;
+DROP TABLE IF EXISTS probe0;

tests/queries/0_stateless/03146_asof_join_ddb_merge_long.reference

@@ -0,0 +1,2 @@
+26790	1488
+26790	1488

tests/queries/0_stateless/03146_asof_join_ddb_merge_long.sql.j2

@@ -0,0 +1,39 @@
+-- Tags: long
+
+SET allow_experimental_analyzer=1;
+
+SET session_timezone = 'UTC';
+
+{% for join_algorithm in ['default', 'full_sorting_merge'] -%}
+
+SET join_algorithm = '{{ join_algorithm }}';
+
+-- TODO: enable once USING and `join_use_nulls` is supported by `full_sorting_merge`
+-- SET join_use_nulls = 1;
+
+WITH build AS (
+    SELECT
+        tk.number AS k,
+        toDateTime('2021-01-01 00:00:00') + INTERVAL i.number SECONDS AS t,
+        i.number % 37 AS v
+    FROM numbers(3000000) AS i
+    CROSS JOIN numbers(2) AS tk
+    SETTINGS join_algorithm = 'hash', join_use_nulls = 0
+),
+probe AS (
+    SELECT
+        tk.number AS k,
+        toDateTime('2021-01-01 00:00:30') + INTERVAL tt.number HOUR AS t
+    FROM numbers(2) AS tk
+    CROSS JOIN numbers(toUInt32((toDateTime('2021-02-01 00:00:30') - toDateTime('2021-01-01 00:00:30')) / 3600)) AS tt
+    SETTINGS join_algorithm = 'hash', join_use_nulls = 0
+)
+SELECT
+    SUM(v) AS v,
+    COUNT(*) AS n
+FROM probe
+ASOF LEFT JOIN build
+USING (k, t)
+;
+
+{% endfor -%}

tests/queries/0_stateless/03147_asof_join_ddb_missing.reference

@@ -0,0 +1,10 @@
+108
+108	27
+513
+1218
+3528
+14553
+121275
+1495503
+12462525
+1249625025

tests/queries/0_stateless/03147_asof_join_ddb_missing.sql

@@ -0,0 +1,186 @@
+SET allow_experimental_analyzer=1;
+
+SET session_timezone = 'UTC';
+SET joined_subquery_requires_alias = 0;
+SET allow_experimental_analyzer = 1;
+SET join_algorithm = 'full_sorting_merge';
+
+-- # 10 dates, 5 keys
+WITH build AS (
+    SELECT
+        k,
+        toDateTime('2001-01-01 00:00:00') + INTERVAL number MINUTE AS t,
+        number AS v
+    FROM numbers(10), (SELECT number AS k FROM numbers(5))
+    SETTINGS join_algorithm = 'default'
+),
+probe AS (
+    SELECT
+        k * 2 AS k,
+        t - INTERVAL 30 SECOND AS t
+    FROM build
+)
+SELECT SUM(v)
+FROM probe ASOF JOIN build USING (k, t);
+
+-- # Coverage: Missing right side bin
+WITH build AS (
+    SELECT
+        k * 2 AS k,
+        toDateTime('2001-01-01 00:00:00') + INTERVAL number MINUTE AS t,
+        number AS v
+    FROM numbers(10), (SELECT number AS k FROM numbers(5))
+    SETTINGS join_algorithm = 'default'
+),
+probe AS (
+    SELECT
+        intDiv(k, 2) AS k,
+        t - INTERVAL 30 SECOND AS t
+    FROM build
+)
+SELECT SUM(v), COUNT(*)
+FROM probe ASOF JOIN build USING (k, t);
+
+-- # 20 dates, 5 keys
+WITH build AS (
+    SELECT
+        k,
+        toDateTime('2001-01-01 00:00:00') + INTERVAL number MINUTE AS t,
+        number AS v
+    FROM numbers(20), (SELECT number AS k FROM numbers(5))
+    SETTINGS join_algorithm = 'default'
+),
+probe AS (
+    SELECT
+        k * 2 AS k,
+        t - INTERVAL 30 SECOND AS t
+    FROM build
+)
+SELECT SUM(v)
+FROM probe ASOF JOIN build USING (k, t);
+
+-- # 30 dates, 5 keys
+WITH build AS (
+    SELECT
+        k,
+        toDateTime('2001-01-01 00:00:00') + INTERVAL number MINUTE AS t,
+        number AS v
+    FROM numbers(30), (SELECT number AS k FROM numbers(5))
+    SETTINGS join_algorithm = 'default'
+),
+probe AS (
+    SELECT
+        k * 2 AS k,
+        t - INTERVAL 30 SECOND AS t
+    FROM build
+)
+SELECT SUM(v)
+FROM probe ASOF JOIN build USING (k, t);
+
+-- # 50 dates, 5 keys
+WITH build AS (
+    SELECT
+        k,
+        toDateTime('2001-01-01 00:00:00') + INTERVAL number MINUTE AS t,
+        number AS v
+    FROM numbers(50), (SELECT number AS k FROM numbers(5))
+    SETTINGS join_algorithm = 'default'
+),
+probe AS (
+    SELECT
+        k * 2 AS k,
+        t - INTERVAL 30 SECOND AS t
+    FROM build
+)
+SELECT SUM(v)
+FROM probe ASOF JOIN build USING (k, t);
+
+-- # 100 dates, 5 keys
+WITH build AS (
+    SELECT
+        k,
+        toDateTime('2001-01-01 00:00:00') + INTERVAL number MINUTE AS t,
+        number AS v
+    FROM numbers(100), (SELECT number AS k FROM numbers(5))
+    SETTINGS join_algorithm = 'default'
+),
+probe AS (
+    SELECT
+        k * 2 AS k,
+        t - INTERVAL 30 SECOND AS t
+    FROM build
+)
+SELECT SUM(v)
+FROM probe ASOF JOIN build USING (k, t);
+
+-- # 100 dates, 50 keys
+WITH build AS (
+    SELECT
+        k,
+        toDateTime('2001-01-01 00:00:00') + INTERVAL number MINUTE AS t,
+        number AS v
+    FROM numbers(100), (SELECT number AS k FROM numbers(50))
+    SETTINGS join_algorithm = 'default'
+),
+probe AS (
+    SELECT
+        k * 2 AS k,
+        t - INTERVAL 30 SECOND AS t
+    FROM build
+)
+SELECT SUM(v)
+FROM probe ASOF JOIN build USING (k, t);
+
+-- # 1000 dates, 5 keys
+WITH build AS (
+    SELECT
+        k,
+        toDateTime('2001-01-01 00:00:00') + INTERVAL number MINUTE AS t,
+        number AS v
+    FROM numbers(1000), (SELECT number AS k FROM numbers(5))
+    SETTINGS join_algorithm = 'default'
+),
+probe AS (
+    SELECT
+        k * 2 AS k,
+        t - INTERVAL 30 SECOND AS t
+    FROM build
+)
+SELECT SUM(v)
+FROM probe ASOF JOIN build USING (k, t);
+
+-- # 1000 dates, 50 keys
+WITH build AS (
+    SELECT
+        k,
+        toDateTime('2001-01-01 00:00:00') + INTERVAL number MINUTE AS t,
+        number AS v
+    FROM numbers(1000), (SELECT number AS k FROM numbers(50))
+    SETTINGS join_algorithm = 'default'
+),
+probe AS (
+    SELECT
+        k * 2 AS k,
+        t - INTERVAL 30 SECOND AS t
+    FROM build
+)
+SELECT SUM(v)
+FROM probe ASOF JOIN build USING (k, t);
+
+-- # 10000 dates, 50 keys
+WITH build AS (
+    SELECT
+        k,
+        toDateTime('2001-01-01 00:00:00') + INTERVAL number MINUTE AS t,
+        number AS v
+    FROM numbers(10000), (SELECT number AS k FROM numbers(50))
+    SETTINGS join_algorithm = 'default'
+),
+probe AS (
+    SELECT
+        k * 2 AS k,
+        t - INTERVAL 30 SECOND AS t
+    FROM build
+)
+SELECT SUM(v)
+FROM probe ASOF JOIN build USING (k, t);

tests/queries/0_stateless/03148_asof_join_ddb_subquery.reference

@@ -0,0 +1,4 @@
+1	1
+3	1
+6	1
+8	1

tests/queries/0_stateless/03148_asof_join_ddb_subquery.sql

@@ -0,0 +1,29 @@
+DROP TABLE IF EXISTS events;
+CREATE TABLE events (begin Float64, value Int32) ENGINE = MergeTree() ORDER BY begin;
+
+INSERT INTO events VALUES (1, 0), (3, 1), (6, 2), (8, 3);
+
+SET allow_experimental_analyzer = 1;
+SET join_algorithm = 'full_sorting_merge';
+SET joined_subquery_requires_alias = 0;
+
+SELECT
+    begin,
+    value IN (
+        SELECT e1.value
+        FROM (
+            SELECT *
+            FROM events e1
+            WHERE e1.value = events.value
+        ) AS e1
+        ASOF JOIN (
+            SELECT number :: Float64 AS begin
+            FROM numbers(10)
+            WHERE number >= 1 AND number < 10
+        )
+        USING (begin)
+    )
+FROM events
+ORDER BY begin ASC;
+
+DROP TABLE IF EXISTS events;

tests/queries/0_stateless/03149_asof_join_ddb_timestamps.reference

@@ -0,0 +1,56 @@
+2023-03-21 13:00:00	0
+2023-03-21 14:00:00	1
+2023-03-21 15:00:00	2
+2023-03-21 16:00:00	3
+2023-03-21 17:00:00	3
+2023-03-21 18:00:00	3
+2023-03-21 19:00:00	3
+2023-03-21 20:00:00	3
+2023-03-21 21:00:00	3
+2106-02-07 06:28:15	9
+2023-03-21 13:00:00	0
+2023-03-21 14:00:00	1
+2023-03-21 15:00:00	2
+2023-03-21 16:00:00	3
+2023-03-21 17:00:00	3
+2023-03-21 18:00:00	3
+2023-03-21 19:00:00	3
+2023-03-21 20:00:00	3
+2023-03-21 21:00:00	3
+2106-02-07 06:28:15	9
+2023-03-21 12:00:00	\N
+2023-03-21 13:00:00	0
+2023-03-21 14:00:00	1
+2023-03-21 15:00:00	2
+2023-03-21 16:00:00	3
+2023-03-21 17:00:00	3
+2023-03-21 18:00:00	3
+2023-03-21 19:00:00	3
+2023-03-21 20:00:00	3
+2023-03-21 21:00:00	3
+2106-02-07 06:28:15	9
+\N	\N
+2023-03-21 12:00:00	0
+2023-03-21 13:00:00	0
+2023-03-21 14:00:00	1
+2023-03-21 15:00:00	2
+2023-03-21 16:00:00	3
+2023-03-21 17:00:00	3
+2023-03-21 18:00:00	3
+2023-03-21 19:00:00	3
+2023-03-21 20:00:00	3
+2023-03-21 21:00:00	3
+2106-02-07 06:28:15	9
+\N	0
+2023-03-21 12:00:00	\N
+2023-03-21 13:00:00	\N
+2023-03-21 14:00:00	\N
+2023-03-21 15:00:00	\N
+2023-03-21 16:00:00	\N
+2023-03-21 17:00:00	\N
+2023-03-21 18:00:00	\N
+2023-03-21 19:00:00	\N
+2023-03-21 20:00:00	\N
+2023-03-21 21:00:00	\N
+2106-02-07 06:28:15	\N
+\N	\N

tests/queries/0_stateless/03149_asof_join_ddb_timestamps.sql

@@ -0,0 +1,95 @@
+DROP TABLE IF EXISTS events0;
+DROP TABLE IF EXISTS probe0;
+
+SET session_timezone = 'UTC';
+SET allow_experimental_analyzer = 1;
+SET join_algorithm = 'full_sorting_merge';
+SET join_use_nulls = 1;
+
+CREATE TABLE events0
+ENGINE = MergeTree()
+ORDER BY COALESCE(begin, toDateTime('9999-12-31 23:59:59'))
+AS
+SELECT
+    toNullable(toDateTime('2023-03-21 13:00:00') + INTERVAL number HOUR) AS begin,
+    number AS value
+FROM numbers(4);
+
+INSERT INTO events0 VALUES (NULL, -1), (toDateTime('9999-12-31 23:59:59'), 9);
+
+CREATE TABLE probe0
+ENGINE = MergeTree()
+ORDER BY COALESCE(begin, toDateTime('9999-12-31 23:59:59'))
+AS
+SELECT
+    toNullable(toDateTime('2023-03-21 12:00:00') + INTERVAL number HOUR) AS begin
+FROM numbers(10);
+
+INSERT INTO probe0 VALUES (NULL), (toDateTime('9999-12-31 23:59:59'));
+
+SELECT
+    p.begin,
+    e.value
+FROM
+    probe0 p
+    ASOF JOIN events0 e ON p.begin >= e.begin
+ORDER BY p.begin ASC;
+
+SELECT
+    p.begin,
+    e.value
+FROM
+    probe0 p
+    ASOF JOIN events0 e USING (begin)
+ORDER BY p.begin ASC
+SETTINGS join_use_nulls = 0
+;
+
+SELECT
+    p.begin,
+    e.value
+FROM
+    probe0 p
+    ASOF LEFT JOIN events0 e ON p.begin >= e.begin
+ORDER BY p.begin ASC;
+
+SELECT
+    p.begin,
+    e.value
+FROM
+    probe0 p
+    ASOF LEFT JOIN events0 e USING (begin)
+ORDER BY p.begin ASC
+SETTINGS join_use_nulls = 0
+;
+
+SELECT
+    p.begin,
+    e.value
+FROM
+    probe0 p
+    ASOF RIGHT JOIN events0 e ON p.begin >= e.begin
+ORDER BY e.begin ASC; -- { serverError NOT_IMPLEMENTED}
+
+SELECT
+    p.begin,
+    e.value
+FROM
+    probe0 p
+    ASOF RIGHT JOIN events0 e USING (begin)
+ORDER BY e.begin ASC; -- { serverError NOT_IMPLEMENTED}
+
+
+SELECT
+    p.begin,
+    e.value
+FROM
+    probe0 p
+    ASOF LEFT JOIN (
+        SELECT * FROM events0 WHERE log(value + 5) > 10
+    ) e ON p.begin + INTERVAL 2 HOUR >= e.begin + INTERVAL 1 HOUR
+ORDER BY p.begin ASC;
+
+
+DROP TABLE IF EXISTS events0;
+DROP TABLE IF EXISTS probe0;