Merge branch 'master' into fix-weak-hash-for-sparce

docker/test/stateless/run.sh

@@ -12,7 +12,7 @@ MAX_RUN_TIME=$((MAX_RUN_TIME == 0 ? 7200 : MAX_RUN_TIME))
 USE_DATABASE_REPLICATED=${USE_DATABASE_REPLICATED:=0}
 USE_SHARED_CATALOG=${USE_SHARED_CATALOG:=0}
 
-RUN_SEQUENTIAL_TESTS_IN_PARALLEL=1
+RUN_SEQUENTIAL_TESTS_IN_PARALLEL=0
 
 if [[ "$USE_DATABASE_REPLICATED" -eq 1 ]] || [[ "$USE_SHARED_CATALOG" -eq 1 ]]; then
   RUN_SEQUENTIAL_TESTS_IN_PARALLEL=0

docs/en/sql-reference/data-types/array.md

@@ -96,3 +96,22 @@ Result:
 │         1 │ [2]       │ [[4,1]]   │
 └───────────┴───────────┴───────────┘
 ```
+
+## Reading nested subcolumns from Array
+
+If nested type `T` inside `Array` has subcolumns (for example, if it's a [named tuple](./tuple.md)), you can read its subcolumns from an `Array(T)` type with the same subcolumn names. The type of a subcolumn will be `Array` of the type of original subcolumn.
+
+**Example**
+
+```sql
+CREATE TABLE t_arr (arr Array(Tuple(field1 UInt32, field2 String))) ENGINE = MergeTree ORDER BY tuple();
+INSERT INTO t_arr VALUES ([(1, 'Hello'), (2, 'World')]), ([(3, 'This'), (4, 'is'), (5, 'subcolumn')]);
+SELECT arr.field1, toTypeName(arr.field1), arr.field2, toTypeName(arr.field2) from t_arr;
+```
+
+```test
+┌─arr.field1─┬─toTypeName(arr.field1)─┬─arr.field2────────────────┬─toTypeName(arr.field2)─┐
+│ [1,2]      │ Array(UInt32)          │ ['Hello','World']         │ Array(String)          │
+│ [3,4,5]    │ Array(UInt32)          │ ['This','is','subcolumn'] │ Array(String)          │
+└────────────┴────────────────────────┴───────────────────────────┴────────────────────────┘
+```

src/Interpreters/HashJoin/HashJoin.cpp

@@ -125,7 +125,7 @@ HashJoin::HashJoin(std::shared_ptr<TableJoin> table_join_, const Block & right_s
     if (isCrossOrComma(kind))
     {
         data->type = Type::CROSS;
-        sample_block_with_columns_to_add = right_sample_block;
+        sample_block_with_columns_to_add = materializeBlock(right_sample_block);
     }
     else if (table_join->getClauses().empty())
     {

src/Storages/RocksDB/EmbeddedRocksDBBulkSink.cpp

@@ -226,10 +226,11 @@ void EmbeddedRocksDBBulkSink::consume(Chunk chunk_)
     if (chunks_to_write.empty())
         return;
 
+    size_t num_chunks = chunks_to_write.size();
     auto [serialized_key_column, serialized_value_column]
         = storage.ttl > 0 ? serializeChunks<true>(std::move(chunks_to_write)) : serializeChunks<false>(std::move(chunks_to_write));
     auto sst_file_path = getTemporarySSTFilePath();
-    LOG_DEBUG(getLogger("EmbeddedRocksDBBulkSink"), "Writing {} rows to SST file {}", serialized_key_column->size(), sst_file_path);
+    LOG_DEBUG(getLogger("EmbeddedRocksDBBulkSink"), "Writing {} rows from {} chunks to SST file {}", serialized_key_column->size(), num_chunks, sst_file_path);
     if (auto status = buildSSTFile(sst_file_path, *serialized_key_column, *serialized_value_column); !status.ok())
         throw Exception(ErrorCodes::ROCKSDB_ERROR, "RocksDB write error: {}", status.ToString());
 

src/Storages/transformQueryForExternalDatabase.cpp

@@ -420,7 +420,7 @@ String transformQueryForExternalDatabase(
             throw Exception(ErrorCodes::UNSUPPORTED_METHOD, "No column names for query '{}' to external table '{}.{}'",
                             query_info.query_tree->formatASTForErrorMessage(), database, table);
 
-        auto clone_query = getASTForExternalDatabaseFromQueryTree(query_info.query_tree);
+        auto clone_query = getASTForExternalDatabaseFromQueryTree(query_info.query_tree, query_info.table_expression);
 
         return transformQueryForExternalDatabaseImpl(
             clone_query,

src/Storages/transformQueryForExternalDatabaseAnalyzer.cpp

@@ -3,6 +3,7 @@
 #include <Storages/transformQueryForExternalDatabaseAnalyzer.h>
 
 #include <Parsers/ASTSelectWithUnionQuery.h>
+#include <Parsers/ASTSelectQuery.h>
 #include <Analyzer/InDepthQueryTreeVisitor.h>
 
 #include <Columns/ColumnConst.h>
@@ -10,7 +11,7 @@
 #include <Analyzer/QueryNode.h>
 #include <Analyzer/ConstantNode.h>
 #include <Analyzer/ConstantValue.h>
-
+#include <Analyzer/JoinNode.h>
 
 #include <DataTypes/DataTypesNumber.h>
 
@@ -20,6 +21,7 @@ namespace DB
 namespace ErrorCodes
 {
     extern const int UNSUPPORTED_METHOD;
+    extern const int LOGICAL_ERROR;
 }
 
 namespace
@@ -55,7 +57,7 @@ public:
 
 }
 
-ASTPtr getASTForExternalDatabaseFromQueryTree(const QueryTreeNodePtr & query_tree)
+ASTPtr getASTForExternalDatabaseFromQueryTree(const QueryTreeNodePtr & query_tree, const QueryTreeNodePtr & table_expression)
 {
     auto new_tree = query_tree->clone();
 
@@ -63,6 +65,21 @@ ASTPtr getASTForExternalDatabaseFromQueryTree(const QueryTreeNodePtr & query_tre
     visitor.visit(new_tree);
     const auto * query_node = new_tree->as<QueryNode>();
 
+    const auto & join_tree = query_node->getJoinTree();
+    bool allow_where = true;
+    if (const auto * join_node = join_tree->as<JoinNode>())
+    {
+        if (join_node->getStrictness() != JoinStrictness::All)
+            allow_where = false;
+
+        if (join_node->getKind() == JoinKind::Left)
+            allow_where = join_node->getLeftTableExpression()->isEqual(*table_expression);
+        else if (join_node->getKind() == JoinKind::Right)
+            allow_where = join_node->getRightTableExpression()->isEqual(*table_expression);
+        else
+            allow_where = (join_node->getKind() == JoinKind::Inner);
+    }
+
     auto query_node_ast = query_node->toAST({ .add_cast_for_constants = false, .fully_qualified_identifiers = false });
     const IAST * ast = query_node_ast.get();
 
@@ -76,7 +93,13 @@ ASTPtr getASTForExternalDatabaseFromQueryTree(const QueryTreeNodePtr & query_tre
     if (union_ast->list_of_selects->children.size() != 1)
         throw Exception(ErrorCodes::UNSUPPORTED_METHOD, "QueryNode AST is not a single ASTSelectQuery, got {}", union_ast->list_of_selects->children.size());
 
-    return union_ast->list_of_selects->children.at(0);
+    ASTPtr select_query = union_ast->list_of_selects->children.at(0);
+    auto * select_query_typed = select_query->as<ASTSelectQuery>();
+    if (!select_query_typed)
+        throw Exception(ErrorCodes::LOGICAL_ERROR, "Expected ASTSelectQuery, got {}", select_query ? select_query->formatForErrorMessage() : "nullptr");
+    if (!allow_where)
+        select_query_typed->setExpression(ASTSelectQuery::Expression::WHERE, nullptr);
+    return select_query;
 }
 
 }

src/Storages/transformQueryForExternalDatabaseAnalyzer.h

@@ -6,6 +6,6 @@
 namespace DB
 {
 
-ASTPtr getASTForExternalDatabaseFromQueryTree(const QueryTreeNodePtr & query_tree);
+ASTPtr getASTForExternalDatabaseFromQueryTree(const QueryTreeNodePtr & query_tree, const QueryTreeNodePtr & table_expression);
 
 }

tests/integration/test_storage_postgresql/test.py

@@ -834,6 +834,60 @@ def test_literal_escaping(started_cluster):
     cursor.execute(f"DROP TABLE escaping")
 
 
+def test_filter_pushdown(started_cluster):
+    cursor = started_cluster.postgres_conn.cursor()
+    cursor.execute("CREATE SCHEMA test_filter_pushdown")
+    cursor.execute(
+        "CREATE TABLE test_filter_pushdown.test_table (id integer, value integer)"
+    )
+    cursor.execute(
+        "INSERT INTO test_filter_pushdown.test_table VALUES (1, 10), (1, 110), (2, 0), (3, 33), (4, 0)"
+    )
+
+    node1.query(
+        """
+        CREATE TABLE test_filter_pushdown_pg_table (id UInt32, value UInt32)
+        ENGINE PostgreSQL('postgres1:5432', 'postgres', 'test_table', 'postgres', 'mysecretpassword', 'test_filter_pushdown');
+    """
+    )
+
+    node1.query(
+        """
+        CREATE TABLE test_filter_pushdown_local_table (id UInt32, value UInt32) ENGINE Memory AS SELECT * FROM test_filter_pushdown_pg_table
+    """
+    )
+
+    node1.query(
+        "CREATE TABLE ch_table (id UInt32, pg_id UInt32) ENGINE MergeTree ORDER BY id"
+    )
+    node1.query("INSERT INTO ch_table VALUES (1, 1), (2, 2), (3, 1), (4, 2), (5, 999)")
+
+    def compare_results(query, **kwargs):
+        result1 = node1.query(
+            query.format(pg_table="test_filter_pushdown_pg_table", **kwargs)
+        )
+        result2 = node1.query(
+            query.format(pg_table="test_filter_pushdown_local_table", **kwargs)
+        )
+        assert result1 == result2
+
+    for kind in ["INNER", "LEFT", "RIGHT", "FULL"]:
+        for value in [0, 10]:
+            compare_results(
+                "SELECT * FROM ch_table {kind} JOIN {pg_table} as p ON ch_table.pg_id = p.id WHERE value = {value} ORDER BY ALL",
+                kind=kind,
+                value=value,
+            )
+
+            compare_results(
+                "SELECT * FROM {pg_table} as p {kind} JOIN ch_table ON ch_table.pg_id = p.id WHERE value = {value} ORDER BY ALL",
+                kind=kind,
+                value=value,
+            )
+
+    cursor.execute("DROP SCHEMA test_filter_pushdown CASCADE")
+
+
 if __name__ == "__main__":
     cluster.start()
     input("Cluster created, press any key to destroy...")

tests/queries/0_stateless/01396_inactive_replica_cleanup_nodes_zookeeper.sh

@@ -30,7 +30,7 @@ $CLICKHOUSE_CLIENT --max_block_size 1 --min_insert_block_size_rows 1 --min_inser
 
 for _ in {1..60}; do
     $CLICKHOUSE_CLIENT --query "SYSTEM FLUSH LOGS"
-    [[ $($CLICKHOUSE_CLIENT --query "SELECT sum(toUInt32(extract(message, 'Removed (\d+) old log entries'))) FROM system.text_log WHERE event_date >= yesterday() AND logger_name LIKE '%' || '$CLICKHOUSE_DATABASE' || '%r1%(ReplicatedMergeTreeCleanupThread)%' AND message LIKE '%Removed % old log entries%'") -gt $((SCALE - 100)) ]] && break;
+    [[ $($CLICKHOUSE_CLIENT --query "SELECT sum(toUInt32(extract(message, 'Removed (\d+) old log entries'))) FROM system.text_log WHERE event_date >= yesterday() AND logger_name LIKE '%' || '$CLICKHOUSE_DATABASE' || '%r1%(ReplicatedMergeTreeCleanupThread)%' AND message LIKE '%Removed % old log entries%'") -gt $((SCALE - 10)) ]] && break;
     sleep 1
 done
 

tests/queries/0_stateless/02443_detach_attach_partition.sh

@@ -73,7 +73,7 @@ kill -TERM $PID_1 && kill -TERM $PID_2 && kill -TERM $PID_3 && kill -TERM $PID_4
 wait
 
 $CLICKHOUSE_CLIENT -q "SELECT '$CLICKHOUSE_DATABASE', 'threads finished'"
-wait_for_queries_to_finish
+wait_for_queries_to_finish 60
 
 $CLICKHOUSE_CLIENT -q "SYSTEM SYNC REPLICA alter_table0"
 $CLICKHOUSE_CLIENT -q "SYSTEM SYNC REPLICA alter_table1"

tests/queries/0_stateless/02956_rocksdb_bulk_sink.reference

@@ -3,7 +3,7 @@
 1000
 1
 1000
-2
+1
 1000000
 1000
 0	999001

tests/queries/0_stateless/02956_rocksdb_bulk_sink.sh

@@ -29,7 +29,7 @@ ${CLICKHOUSE_CLIENT} --query "SELECT count() FROM rocksdb_worm;"
 ${CLICKHOUSE_CLIENT} --query "TRUNCATE TABLE rocksdb_worm;"
 # Must set both max_threads and max_insert_threads to 2 to make sure there is only two sinks
 ${CLICKHOUSE_CLIENT} --query "INSERT INTO rocksdb_worm SELECT number, number+1 FROM numbers_mt(1000000) SETTINGS max_threads = 2, max_insert_threads = 2, max_block_size = 10000, min_insert_block_size_rows = 0, min_insert_block_size_bytes = 0, insert_deduplication_token = '', optimize_trivial_insert_select = 1;"
-${CLICKHOUSE_CLIENT} --query "SELECT sum(value) FROM system.rocksdb WHERE database = currentDatabase() AND table = 'rocksdb_worm' AND name = 'no.file.opens';" # should be 2 because default bulk sink size is ~1M rows / SST file
+${CLICKHOUSE_CLIENT} --query "SELECT sum(value) IN (1, 2) FROM system.rocksdb WHERE database = currentDatabase() AND table = 'rocksdb_worm' AND name = 'no.file.opens';" # should be not more than 2 because default bulk sink size is ~1M rows / SST file.
 ${CLICKHOUSE_CLIENT} --query "SELECT count() FROM rocksdb_worm;"
 
 # Testing insert with duplicated keys

tests/queries/0_stateless/03001_consider_lwd_when_merge.sql

@@ -7,12 +7,14 @@ SETTINGS max_bytes_to_merge_at_max_space_in_pool = 80000, exclude_deleted_rows_f
 INSERT INTO lwd_merge SELECT number FROM numbers(10000);
 INSERT INTO lwd_merge SELECT number FROM numbers(10000, 10000);
 
-OPTIMIZE TABLE lwd_merge;
+SET optimize_throw_if_noop = 1;
+
+OPTIMIZE TABLE lwd_merge; -- { serverError CANNOT_ASSIGN_OPTIMIZE }
 SELECT count() FROM system.parts WHERE database = currentDatabase() AND table = 'lwd_merge' AND active = 1;
 
 DELETE FROM lwd_merge WHERE id % 10 > 0;
 
-OPTIMIZE TABLE lwd_merge;
+OPTIMIZE TABLE lwd_merge; -- { serverError CANNOT_ASSIGN_OPTIMIZE }
 SELECT count() FROM system.parts WHERE database = currentDatabase() AND table = 'lwd_merge' AND active = 1;
 
 ALTER TABLE lwd_merge MODIFY SETTING exclude_deleted_rows_for_part_size_in_merge = 1;

tests/queries/0_stateless/03205_column_type_check.reference

@@ -0,0 +1,2 @@
+1	nan	1048575	2
+1	1	1	1	1

tests/queries/0_stateless/03205_column_type_check.sql

@@ -0,0 +1,7 @@
+SELECT * FROM (SELECT toUInt256(1)) AS t, (SELECT greatCircleAngle(toLowCardinality(toNullable(toUInt256(1048575))), 257, -9223372036854775808, 1048576), 1048575, materialize(2)) AS u;
+
+
+SET join_algorithm='hash';
+SET allow_experimental_join_condition=1;
+SELECT * FROM ( SELECT 1 AS a, toLowCardinality(1), 1) AS t1 CROSS  JOIN (SELECT toLowCardinality(1 AS a), 1 AS b) AS t2;
+