init

docs/en/operations/settings/settings.md

@@ -3954,6 +3954,17 @@ Possible values:
 
 Default value: `''`.
 
+## preferred_optimize_projection_name {#preferred_optimize_projection_name}
+
+If it is set to a non-empty string, ClickHouse will try to apply specified projection in query.
+
+
+Possible values:
+
+- string: name of preferred projection
+
+Default value: `''`.
+
 ## alter_sync {#alter-sync}
 
 Allows to set up waiting for actions to be executed on replicas by [ALTER](../../sql-reference/statements/alter/index.md), [OPTIMIZE](../../sql-reference/statements/optimize.md) or [TRUNCATE](../../sql-reference/statements/truncate.md) queries.

src/Core/Settings.h

@@ -605,6 +605,7 @@ class IColumn;
     M(Bool, optimize_use_implicit_projections, true, "Automatically choose implicit projections to perform SELECT query", 0) \
     M(Bool, force_optimize_projection, false, "If projection optimization is enabled, SELECT queries need to use projection", 0) \
     M(String, force_optimize_projection_name, "", "If it is set to a non-empty string, check that this projection is used in the query at least once.", 0) \
+    M(String, preferred_optimize_projection_name, "", "If it is set to a non-empty string, ClickHouse tries to apply specified projection", 0) \
     M(Bool, async_socket_for_remote, true, "Asynchronously read from socket executing remote query", 0) \
     M(Bool, async_query_sending_for_remote, true, "Asynchronously create connections and send query to shards in remote query", 0) \
     M(Bool, insert_null_as_default, true, "Insert DEFAULT values instead of NULL in INSERT SELECT (UNION ALL)", 0) \

src/Processors/QueryPlan/Optimizations/QueryPlanOptimizationSettings.h

@@ -42,6 +42,7 @@ struct QueryPlanOptimizationSettings
     bool optimize_projection = false;
     bool force_use_projection = false;
     String force_projection_name;
+    String preferred_projection_name;
     bool optimize_use_implicit_projections = false;
 
     static QueryPlanOptimizationSettings fromSettings(const Settings & from);

src/Processors/QueryPlan/Optimizations/optimizeUseAggregateProjection.cpp

@@ -444,8 +444,33 @@ AggregateProjectionCandidates getAggregateProjectionCandidates(
 
     const auto & projections = metadata->projections;
     std::vector<const ProjectionDescription *> agg_projections;
-    for (const auto & projection : projections)
-        if (projection.type == ProjectionDescription::Type::Aggregate)
+
+    const auto & proj_name_from_settings = context->getSettings().preferred_optimize_projection_name.value;
+    bool is_projection_found = false;
+
+    // Here we iterate over the projections and check if we have the same projections as we specified in preferred_projection_name
+    if (!proj_name_from_settings.empty())
+    {
+        for (const auto & projection : projections)
+        {
+            if (projection.type == ProjectionDescription::Type::Aggregate)
+            {
+                size_t last_dot_pos = projection.name.find_last_of('.');
+                std::string projection_name = (last_dot_pos != std::string::npos) ? projection.name.substr(last_dot_pos + 1) : projection.name;
+                if (proj_name_from_settings == projection_name)
+                {
+                    agg_projections.push_back(&projection);
+                    is_projection_found = true;
+                    break;
+                }
+            }
+        }
+        if (!is_projection_found)
+            throw Exception(ErrorCodes::INCORRECT_DATA, "Projection {} is specified in setting force_optimize_projection_name but not used",
+                            proj_name_from_settings);
+    }
+    else
+        for (const auto & projection : projections)
             agg_projections.push_back(&projection);
 
     bool can_use_minmax_projection = allow_implicit_projections && metadata->minmax_count_projection

src/Processors/QueryPlan/Optimizations/optimizeUseNormalProjection.cpp

@@ -7,6 +7,8 @@
 #include <Processors/QueryPlan/ReadFromPreparedSource.h>
 #include <Processors/Sources/NullSource.h>
 #include <Common/logger_useful.h>
+#include <Storages/ProjectionsDescription.h>
+#include <Storages/SelectQueryInfo.h>
 #include <Storages/MergeTree/MergeTreeDataSelectExecutor.h>
 #include <stack>
 
@@ -131,6 +133,26 @@ bool optimizeUseNormalProjections(Stack & stack, QueryPlan::Nodes & nodes)
 
     std::shared_ptr<PartitionIdToMaxBlock> max_added_blocks = getMaxAddedBlocks(reading);
 
+    // Here we iterate over the projections and check if we have the same projections as we specified in preferred_projection_name
+    bool is_projection_found = false;
+    const auto & proj_name_from_settings = context->getSettings().preferred_optimize_projection_name.value;
+    if (!proj_name_from_settings.empty())
+    {
+        for (const auto * projection : normal_projections)
+        {
+            size_t last_dot_pos = projection->name.find_last_of('.');
+            std::string projection_name = (last_dot_pos != std::string::npos) ? projection->name.substr(last_dot_pos + 1) : projection->name;
+            if (projection_name == proj_name_from_settings)
+            {
+                is_projection_found = true;
+                break;
+            }
+        }
+        if (!is_projection_found)
+            throw Exception(ErrorCodes::INCORRECT_DATA, "Projection {} is specified in setting force_optimize_projection_name but not used",
+                            proj_name_from_settings);
+    }
+
     for (const auto * projection : normal_projections)
     {
         if (!hasAllRequiredColumns(projection, required_columns))
@@ -153,7 +175,12 @@ bool optimizeUseNormalProjections(Stack & stack, QueryPlan::Nodes & nodes)
         if (candidate.sum_marks >= ordinary_reading_marks)
             continue;
 
-        if (best_candidate == nullptr || candidate.sum_marks < best_candidate->sum_marks)
+        size_t last_dot_pos = projection->name.find_last_of('.');
+        std::string projection_name = (last_dot_pos != std::string::npos) ? projection->name.substr(last_dot_pos + 1) : projection->name;
+
+        if (!is_projection_found && (best_candidate == nullptr || candidate.sum_marks < best_candidate->sum_marks))
+            best_candidate = &candidate;
+        else if (is_projection_found && projection_name == proj_name_from_settings)
             best_candidate = &candidate;
     }
 

tests/queries/0_stateless/02907_preferred_optimize_projection_name.reference

@@ -0,0 +1,6 @@
+test
+projection_test_by_string
+Executing query with setting
+test
+projection_test_by_more
+0

tests/queries/0_stateless/02907_preferred_optimize_projection_name.sh

@@ -0,0 +1,85 @@
+#!/usr/bin/env bash
+# Tags: no-fasttest
+
+CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
+# shellcheck source=../shell_config.sh
+. "$CURDIR"/../shell_config.sh
+
+$CLICKHOUSE_CLIENT -q "
+CREATE TABLE test (
+    test_id UInt64,
+    test_name String,
+    test_count Nullable(Float64),
+    test_string String,
+    PROJECTION projection_test_by_string (
+        SELECT test_string,
+            sum(test_count)
+        GROUP BY test_id,
+            test_string,
+            test_name
+    ),
+    PROJECTION projection_test_by_more (
+        SELECT test_string,
+            test_name,
+            sum(test_count)
+        GROUP BY test_id,
+            test_string,
+            test_name
+    )
+) ENGINE = MergeTree
+ORDER BY test_string;"
+
+$CLICKHOUSE_CLIENT -q "
+INSERT INTO test
+SELECT number,
+    'test',
+    1.* (number / 2),
+    'test'
+FROM numbers(100, 500);"
+
+$CLICKHOUSE_CLIENT --query_id '02907_test' -q "
+SELECT test_string
+FROM test
+WHERE (test_id > 50)
+    AND (test_id < 150)
+GROUP BY test_string;"
+
+$CLICKHOUSE_CLIENT -q "
+SELECT projections
+FROM system.query_log
+WHERE query_id = '02907_test' AND arrayElement(projections, 1) LIKE '%projection_test_by_string'
+LIMIT 1;" | grep -o "projection_test_by_string" || true
+
+$CLICKHOUSE_CLIENT -q "
+SELECT projections
+FROM system.query_log 
+WHERE query_id = '02907_test' AND arrayElement(projections, 1) LIKE '%projection_test_by_more' 
+LIMIT 1;" | grep -o "projection_test_by_more" || true
+
+echo "Executing query with setting"
+
+$CLICKHOUSE_CLIENT --query_id '02907_test_1' --preferred_optimize_projection_name 'projection_test_by_more' -q "
+SELECT test_string
+FROM test
+WHERE (test_id > 50)
+    AND (test_id < 150)
+GROUP BY test_string;"
+
+$CLICKHOUSE_CLIENT -q "
+SELECT projections
+FROM system.query_log
+WHERE query_id = '02907_test_1' AND arrayElement(projections, 1) LIKE '%projection_test_by_more'
+LIMIT 1;" | grep -o "projection_test_by_more" || true
+
+$CLICKHOUSE_CLIENT -q "
+SELECT projections
+FROM system.query_log
+WHERE query_id = '02907_test_1' AND arrayElement(projections, 1) LIKE '%projection_test_by_string'
+LIMIT 1" | grep -o "projection_test_by_string" || true
+
+$CLICKHOUSE_CLIENT --query_id '02907_test_1' --preferred_optimize_projection_name 'non_existing_projection' -q "
+SELECT test_string
+FROM test
+WHERE (test_id > 50)
+    AND (test_id < 150)
+GROUP BY test_string;" 2>&1 | grep -c "BAD_ARGUMENTS" || true