merge output blocks if need in hash join

dbms/src/Core/Block.cpp

@@ -514,6 +514,36 @@ static ReturnType checkBlockStructure(const Block & lhs, const Block & rhs, cons
     return ReturnType(true);
 }
 
+Block mergeBlocks(Blocks && blocks)
+{
+    assert(!blocks.empty());
+    auto & first_block = blocks[0];
+    size_t result_rows = 0;
+    for (const auto & block : blocks)
+    {
+        result_rows += block.rows();
+    }
+
+    MutableColumns dst_columns(first_block.columns());
+
+    for (size_t i = 0; i < first_block.columns(); i++)
+    {
+        dst_columns[i] = (*std::move(first_block.getByPosition(i).column)).mutate();
+        dst_columns[i]->reserve(result_rows);
+    }
+
+    for (size_t i = 1; i < blocks.size(); ++i)
+    {
+        if (likely(blocks[i].rows()) > 0)
+        {
+            for (size_t column = 0; column < blocks[i].columns(); ++column)
+            {
+                dst_columns[column]->insertRangeFrom(*blocks[i].getByPosition(column).column, 0, blocks[i].rows());
+            }
+        }
+    }
+    return first_block.cloneWithColumns(std::move(dst_columns));
+}
 
 bool blocksHaveEqualStructure(const Block & lhs, const Block & rhs)
 {

dbms/src/Core/Block.h

@@ -149,6 +149,7 @@ class Block
       */
     void updateHash(SipHash & hash) const;
 
+
 private:
     void eraseImpl(size_t position);
     void initializeIndexByName();
@@ -157,6 +158,7 @@ class Block
 using Blocks = std::vector<Block>;
 using BlocksList = std::list<Block>;
 
+Block mergeBlocks(Blocks && blocks);
 
 /// Compare number of columns, data types, column types, column names, and values of constant columns.
 bool blocksHaveEqualStructure(const Block & lhs, const Block & rhs);

dbms/src/DataStreams/HashJoinProbeBlockInputStream.cpp

@@ -13,7 +13,6 @@
 // limitations under the License.
 
 #include <DataStreams/HashJoinProbeBlockInputStream.h>
-#include <Interpreters/ExpressionActions.h>
 
 namespace DB
 {
@@ -25,6 +24,7 @@ HashJoinProbeBlockInputStream::HashJoinProbeBlockInputStream(
     : log(Logger::get(req_id))
     , join(join_)
     , probe_process_info(max_block_size)
+    , squashing_transform(max_block_size)
 {
     children.push_back(input);
 
@@ -67,17 +67,34 @@ Block HashJoinProbeBlockInputStream::getHeader() const
 
 Block HashJoinProbeBlockInputStream::readImpl()
 {
-    if (probe_process_info.all_rows_joined_finish)
+    // if join finished, return {} directly.
+    if (squashing_transform.isJoinFinished())
+    {
+        return Block{};
+    }
+
+    while (squashing_transform.needAppendBlock())
+    {
+        Block result_block = getOutputBlock(probe_process_info);
+        squashing_transform.appendBlock(result_block);
+    }
+    return squashing_transform.getFinalOutputBlock();
+}
+
+Block HashJoinProbeBlockInputStream::getOutputBlock(ProbeProcessInfo & probe_process_info_) const
+{
+    if (probe_process_info_.all_rows_joined_finish)
     {
         Block block = children.back()->read();
         if (!block)
+        {
             return block;
+        }
         join->checkTypes(block);
-        probe_process_info.resetBlock(std::move(block));
+        probe_process_info_.resetBlock(std::move(block));
     }
 
-    return join->joinBlock(probe_process_info);
+    return join->joinBlock(probe_process_info_);
 }
 
-
 } // namespace DB

dbms/src/DataStreams/HashJoinProbeBlockInputStream.h

@@ -15,12 +15,12 @@
 #pragma once
 
 #include <DataStreams/IProfilingBlockInputStream.h>
+#include <DataStreams/SquashingHashJoinBlockTransform.h>
 #include <Interpreters/Join.h>
 
 namespace DB
 {
 
-
 /** Executes a certain expression over the block.
   * Basically the same as ExpressionBlockInputStream,
   * but requires that there must be a join probe action in the Expression.
@@ -44,6 +44,7 @@ class HashJoinProbeBlockInputStream : public IProfilingBlockInputStream
     String getName() const override { return name; }
     Block getTotals() override;
     Block getHeader() const override;
+    Block getOutputBlock(ProbeProcessInfo & probe_process_info_) const;
 
 protected:
     Block readImpl() override;
@@ -52,6 +53,9 @@ class HashJoinProbeBlockInputStream : public IProfilingBlockInputStream
     const LoggerPtr log;
     JoinPtr join;
     ProbeProcessInfo probe_process_info;
+    Blocks result_blocks;
+    Block over_limit_block;
+    SquashingHashJoinBlockTransform squashing_transform;
 };
 
 } // namespace DB

dbms/src/DataStreams/SquashingHashJoinBlockTransform.cpp

@@ -0,0 +1,98 @@
+// Copyright 2022 PingCAP, Ltd.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include <DataStreams/SquashingHashJoinBlockTransform.h>
+
+namespace DB
+{
+
+SquashingHashJoinBlockTransform::SquashingHashJoinBlockTransform(UInt64 max_block_size_)
+    : output_rows(0)
+    , max_block_size(max_block_size_)
+    , join_finished(false)
+    , over_limit(false)
+{}
+
+void SquashingHashJoinBlockTransform::handleOverLimitBlock()
+{
+    // if over_limit_block is not null, we need to push it into blocks first.
+    if (over_limit_block.has_value())
+    {
+        output_rows += over_limit_block->rows();
+        blocks.push_back(std::move(over_limit_block.value()));
+        over_limit_block = {};
+    }
+}
+
+void SquashingHashJoinBlockTransform::appendBlock(Block block)
+{
+    handleOverLimitBlock();
+    if (!block)
+    {
+        // if append block is {}, mark join finished.
+        join_finished = true;
+        return;
+    }
+    size_t current_rows = block.rows();
+
+    over_limit = output_rows && output_rows + current_rows > max_block_size;
+    if (!over_limit)
+    {
+        blocks.push_back(block);
+        output_rows += current_rows;
+    }
+    else
+    {
+        // if output_rows + current_rows > max block size, put the current result block into over_limit_block and handle it in next read.
+        over_limit_block = block;
+    }
+}
+
+Block SquashingHashJoinBlockTransform::getFinalOutputBlock()
+{
+    Block final_block;
+    if (blocks.empty())
+    {
+        final_block = {};
+    }
+    else if (blocks.size() == 1)
+    {
+        final_block = blocks[0];
+    }
+    else
+    {
+        final_block = mergeBlocks(std::move(blocks));
+    }
+    reset();
+    return final_block;
+}
+
+void SquashingHashJoinBlockTransform::reset()
+{
+    blocks.clear();
+    output_rows = 0;
+    over_limit = false;
+}
+
+bool SquashingHashJoinBlockTransform::isJoinFinished() const
+{
+    return join_finished;
+}
+
+bool SquashingHashJoinBlockTransform::needAppendBlock() const
+{
+    return !over_limit && !join_finished;
+}
+
+} // namespace DB

dbms/src/DataStreams/SquashingHashJoinBlockTransform.h

@@ -0,0 +1,46 @@
+// Copyright 2022 PingCAP, Ltd.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include <Core/Block.h>
+
+namespace DB
+{
+
+class SquashingHashJoinBlockTransform
+{
+public:
+    SquashingHashJoinBlockTransform(UInt64 max_block_size_);
+
+    void handleOverLimitBlock();
+    void appendBlock(Block block);
+    Block getFinalOutputBlock();
+    void reset();
+    bool isJoinFinished() const;
+    bool needAppendBlock() const;
+
+
+#ifndef DBMS_PUBLIC_GTEST
+private:
+#endif
+    Blocks blocks;
+    std::optional<Block> over_limit_block;
+    size_t output_rows;
+    UInt64 max_block_size;
+    bool join_finished;
+    bool over_limit;
+};
+
+} // namespace DB

dbms/src/Flash/tests/gtest_join_executor.cpp

@@ -700,7 +700,7 @@ CATCH
 TEST_F(JoinExecutorTestRunner, SplitJoinResult)
 try
 {
-    context.addMockTable("split_test", "t1", {{"a", TiDB::TP::TypeLong}}, {toVec<Int32>("a", {1, 1, 1, 1, 1, 1, 1, 1, 1, 1})});
+    context.addMockTable("split_test", "t1", {{"a", TiDB::TP::TypeLong}, {"b", TiDB::TP::TypeLong}}, {toVec<Int32>("a", {1, 1, 1, 1, 1, 1, 1, 1, 1, 1}), toVec<Int32>("b", {1, 1, 3, 3, 1, 1, 3, 3, 1, 3})});
     context.addMockTable("split_test", "t2", {{"a", TiDB::TP::TypeLong}}, {toVec<Int32>("a", {1, 1, 1, 1, 1})});
 
     auto request = context
@@ -720,6 +720,25 @@ try
             ASSERT_EQ(expect[i][j], blocks[j].rows());
         }
     }
+
+    // with other condition
+    const auto cond = gt(col("b"), lit(Field(static_cast<Int64>(2))));
+    request = context
+                  .scan("split_test", "t1")
+                  .join(context.scan("split_test", "t2"), tipb::JoinType::TypeInnerJoin, {col("a")}, {}, {}, {cond}, {})
+
+                  .build(context);
+    expect = {{5, 5, 5, 5, 5}, {5, 5, 5, 5, 5}, {5, 5, 5, 5, 5}, {25}, {25}, {25}, {25}, {25}};
+    for (size_t i = 0; i < block_sizes.size(); ++i)
+    {
+        context.context.setSetting("max_block_size", Field(static_cast<UInt64>(block_sizes[i])));
+        auto blocks = getExecuteStreamsReturnBlocks(request);
+        ASSERT_EQ(expect[i].size(), blocks.size());
+        for (size_t j = 0; j < blocks.size(); ++j)
+        {
+            ASSERT_EQ(expect[i][j], blocks[j].rows());
+        }
+    }
 }
 CATCH