Merge branch 'master' of github.com:ClickHouse/ClickHouse into fix-wrong-diagnistic-in-local

docker/test/stateless/Dockerfile

@@ -69,8 +69,8 @@ ENV MAX_RUN_TIME=0
 
 # Unrelated to vars in setup_minio.sh, but should be the same there
 # to have the same binaries for local running scenario
-ARG MINIO_SERVER_VERSION=2022-01-03T18-22-58Z
+ARG MINIO_SERVER_VERSION=2024-08-03T04-33-23Z
-ARG MINIO_CLIENT_VERSION=2022-01-05T23-52-51Z
+ARG MINIO_CLIENT_VERSION=2024-07-31T15-58-33Z
 ARG TARGETARCH
 
 # Download Minio-related binaries

docker/test/stateless/setup_minio.sh

@@ -59,8 +59,8 @@ find_os() {
 download_minio() {
   local os
   local arch
-  local minio_server_version=${MINIO_SERVER_VERSION:-2022-09-07T22-25-02Z}
+  local minio_server_version=${MINIO_SERVER_VERSION:-2024-08-03T04-33-23Z}
-  local minio_client_version=${MINIO_CLIENT_VERSION:-2022-08-28T20-08-11Z}
+  local minio_client_version=${MINIO_CLIENT_VERSION:-2024-07-31T15-58-33Z}
 
   os=$(find_os)
   arch=$(find_arch)
@@ -82,10 +82,10 @@ setup_minio() {
   local test_type=$1
   ./mc alias set clickminio http://localhost:11111 clickhouse clickhouse
   ./mc admin user add clickminio test testtest
-  ./mc admin policy set clickminio readwrite user=test
+  ./mc admin policy attach clickminio readwrite --user=test
   ./mc mb --ignore-existing clickminio/test
   if [ "$test_type" = "stateless" ]; then
-    ./mc policy set public clickminio/test
+    ./mc anonymous set public clickminio/test
   fi
 }
 
@@ -148,4 +148,4 @@ main() {
   setup_aws_credentials
 }
 
-main "$@"
+main "$@"

src/Backups/WithRetries.cpp

@@ -68,13 +68,19 @@ const WithRetries::KeeperSettings & WithRetries::getKeeperSettings() const
 
 WithRetries::FaultyKeeper WithRetries::getFaultyZooKeeper() const
 {
-    /// We need to create new instance of ZooKeeperWithFaultInjection each time a copy a pointer to ZooKeeper client there
+    zkutil::ZooKeeperPtr current_zookeeper;
+    {
+        std::lock_guard lock(zookeeper_mutex);
+        current_zookeeper = zookeeper;
+    }
+
+    /// We need to create new instance of ZooKeeperWithFaultInjection each time and copy a pointer to ZooKeeper client there
     /// The reason is that ZooKeeperWithFaultInjection may reset the underlying pointer and there could be a race condition
     /// when the same object is used from multiple threads.
     auto faulty_zookeeper = ZooKeeperWithFaultInjection::createInstance(
         settings.keeper_fault_injection_probability,
         settings.keeper_fault_injection_seed,
-        zookeeper,
+        current_zookeeper,
         log->name(),
         log);
 

src/Client/ClientApplicationBase.cpp

@@ -200,8 +200,6 @@ void ClientApplicationBase::init(int argc, char ** argv)
         ("pager", po::value<std::string>(), "Pipe all output into this command (less or similar)")
         ("max_memory_usage_in_client", po::value<std::string>(), "Set memory limit in client/local server")
 
-        ("fuzzer-args", po::value<std::string>(), "Command line arguments for the LLVM's libFuzzer driver. Only relevant if the application is compiled with libFuzzer.")
-
         ("client_logs_file", po::value<std::string>(), "Path to a file for writing client logs. Currently we only have fatal logs (when the client crashes)")
     ;
 

src/Common/ShellCommand.cpp

@@ -237,7 +237,14 @@ std::unique_ptr<ShellCommand> ShellCommand::executeImpl(
         res->write_fds.emplace(fd, fds.fds_rw[1]);
     }
 
-    LOG_TRACE(getLogger(), "Started shell command '{}' with pid {}", filename, pid);
+    LOG_TRACE(
+        getLogger(),
+        "Started shell command '{}' with pid {} and file descriptors: out {}, err {}",
+        filename,
+        pid,
+        res->out.getFD(),
+        res->err.getFD());
+
     return res;
 }
 

src/Core/Settings.h

@@ -605,7 +605,7 @@ class IColumn;
     M(Bool, optimize_if_chain_to_multiif, false, "Replace if(cond1, then1, if(cond2, ...)) chains to multiIf. Currently it's not beneficial for numeric types.", 0) \
     M(Bool, optimize_multiif_to_if, true, "Replace 'multiIf' with only one condition to 'if'.", 0) \
     M(Bool, optimize_if_transform_strings_to_enum, false, "Replaces string-type arguments in If and Transform to enum. Disabled by default cause it could make inconsistent change in distributed query that would lead to its fail.", 0) \
-    M(Bool, optimize_functions_to_subcolumns, false, "Transform functions to subcolumns, if possible, to reduce amount of read data. E.g. 'length(arr)' -> 'arr.size0', 'col IS NULL' -> 'col.null' ", 0) \
+    M(Bool, optimize_functions_to_subcolumns, true, "Transform functions to subcolumns, if possible, to reduce amount of read data. E.g. 'length(arr)' -> 'arr.size0', 'col IS NULL' -> 'col.null' ", 0) \
     M(Bool, optimize_using_constraints, false, "Use constraints for query optimization", 0)                                                                                                                                           \
     M(Bool, optimize_substitute_columns, false, "Use constraints for column substitution", 0)                                                                                                                                         \
     M(Bool, optimize_append_index, false, "Use constraints in order to append index condition (indexHint)", 0) \

src/Core/SettingsChangesHistory.cpp

@@ -84,6 +84,7 @@ static std::initializer_list<std::pair<ClickHouseVersion, SettingsChangesHistory
             {"allow_archive_path_syntax", true, true, "Added new setting to allow disabling archive path syntax."},
             {"allow_experimental_time_series_table", false, false, "Added new setting to allow the TimeSeries table engine"},
             {"enable_analyzer", 1, 1, "Added an alias to a setting `allow_experimental_analyzer`."},
+            {"optimize_functions_to_subcolumns", false, true, "Enabled settings by default"},
         }
     },
     {"24.7",

src/Disks/ObjectStorages/S3/S3ObjectStorage.cpp

@@ -305,7 +305,8 @@ void S3ObjectStorage::listObjects(const std::string & path, RelativePathsWithMet
 
     S3::ListObjectsV2Request request;
     request.SetBucket(uri.bucket);
-    request.SetPrefix(path);
+    if (path != "/")
+        request.SetPrefix(path);
     if (max_keys)
         request.SetMaxKeys(static_cast<int>(max_keys));
     else

src/Functions/CastOverloadResolver.cpp

@@ -3,6 +3,7 @@
 #include <Functions/FunctionHelpers.h>
 #include <DataTypes/DataTypeFactory.h>
 #include <DataTypes/DataTypeNullable.h>
+#include <DataTypes/DataTypeString.h>
 #include <Columns/ColumnString.h>
 #include <Core/Settings.h>
 #include <Interpreters/parseColumnsListForTableFunction.h>
@@ -35,7 +36,7 @@ FunctionBasePtr createFunctionBaseCast(
 class CastOverloadResolverImpl : public IFunctionOverloadResolver
 {
 public:
-    const char * getNameImpl() const
+    static const char * getNameImpl(CastType cast_type, bool internal)
     {
         if (cast_type == CastType::accurate)
             return "accurateCast";
@@ -49,7 +50,7 @@ public:
 
     String getName() const override
     {
-        return getNameImpl();
+        return getNameImpl(cast_type, internal);
     }
 
     size_t getNumberOfArguments() const override { return 2; }
@@ -79,10 +80,22 @@ public:
         }
     }
 
+    static FunctionBasePtr createInternalCast(ColumnWithTypeAndName from, DataTypePtr to, CastType cast_type, std::optional<CastDiagnostic> diagnostic)
+    {
+        if (cast_type == CastType::accurateOrNull && !isVariant(to))
+            to = makeNullable(to);
+
+        ColumnsWithTypeAndName arguments;
+        arguments.emplace_back(std::move(from));
+        arguments.emplace_back().type = std::make_unique<DataTypeString>();
+
+        return createFunctionBaseCast(nullptr, getNameImpl(cast_type, true), arguments, to, diagnostic, cast_type);
+    }
+
 protected:
     FunctionBasePtr buildImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & return_type) const override
     {
-        return createFunctionBaseCast(context, getNameImpl(), arguments, return_type, diagnostic, cast_type);
+        return createFunctionBaseCast(context, getNameImpl(cast_type, internal), arguments, return_type, diagnostic, cast_type);
     }
 
     DataTypePtr getReturnTypeImpl(const ColumnsWithTypeAndName & arguments) const override
@@ -130,9 +143,9 @@ private:
 };
 
 
-FunctionOverloadResolverPtr createInternalCastOverloadResolver(CastType type, std::optional<CastDiagnostic> diagnostic)
+FunctionBasePtr createInternalCast(ColumnWithTypeAndName from, DataTypePtr to, CastType cast_type, std::optional<CastDiagnostic> diagnostic)
 {
-    return CastOverloadResolverImpl::create(ContextPtr{}, type, true, diagnostic);
+    return CastOverloadResolverImpl::createInternalCast(std::move(from), std::move(to), cast_type, std::move(diagnostic));
 }
 
 REGISTER_FUNCTION(CastOverloadResolvers)

src/Functions/CastOverloadResolver.h

@@ -3,6 +3,7 @@
 #include <memory>
 #include <optional>
 #include <Interpreters/Context_fwd.h>
+#include <Core/ColumnWithTypeAndName.h>
 
 
 namespace DB
@@ -11,6 +12,9 @@ namespace DB
 class IFunctionOverloadResolver;
 using FunctionOverloadResolverPtr = std::shared_ptr<IFunctionOverloadResolver>;
 
+class IFunctionBase;
+using FunctionBasePtr = std::shared_ptr<const IFunctionBase>;
+
 enum class CastType : uint8_t
 {
     nonAccurate,
@@ -24,6 +28,6 @@ struct CastDiagnostic
     std::string column_to;
 };
 
-FunctionOverloadResolverPtr createInternalCastOverloadResolver(CastType type, std::optional<CastDiagnostic> diagnostic);
+FunctionBasePtr createInternalCast(ColumnWithTypeAndName from, DataTypePtr to, CastType cast_type, std::optional<CastDiagnostic> diagnostic);
 
 }

src/Functions/toBool.cpp

@@ -54,8 +54,7 @@ namespace
                 }
             };
 
-            FunctionOverloadResolverPtr func_builder_cast = createInternalCastOverloadResolver(CastType::nonAccurate, {});
+            auto func_cast = createInternalCast(arguments[0], result_type, CastType::nonAccurate, {});
-            auto func_cast = func_builder_cast->build(cast_args);
             return func_cast->execute(cast_args, result_type, arguments[0].column->size());
         }
     };

src/Interpreters/ActionsDAG.cpp

@@ -301,11 +301,11 @@ const ActionsDAG::Node & ActionsDAG::addCast(const Node & node_to_cast, const Da
     column.column = DataTypeString().createColumnConst(0, cast_type_constant_value);
     column.type = std::make_shared<DataTypeString>();
 
-    const auto * cast_type_constant_node = &addColumn(std::move(column));
+    const auto * cast_type_constant_node = &addColumn(column);
     ActionsDAG::NodeRawConstPtrs children = {&node_to_cast, cast_type_constant_node};
-    FunctionOverloadResolverPtr func_builder_cast = createInternalCastOverloadResolver(CastType::nonAccurate, {});
+    auto func_base_cast = createInternalCast(ColumnWithTypeAndName{node_to_cast.result_type, node_to_cast.result_name}, cast_type, CastType::nonAccurate, {});
 
-    return addFunction(func_builder_cast, std::move(children), result_name);
+    return addFunction(func_base_cast, std::move(children), result_name);
 }
 
 const ActionsDAG::Node & ActionsDAG::addFunctionImpl(
@@ -1547,11 +1547,11 @@ ActionsDAG ActionsDAG::makeConvertingActions(
             const auto * left_arg = dst_node;
 
             CastDiagnostic diagnostic = {dst_node->result_name, res_elem.name};
-            FunctionOverloadResolverPtr func_builder_cast
+            ColumnWithTypeAndName left_column{nullptr, dst_node->result_type, {}};
-                = createInternalCastOverloadResolver(CastType::nonAccurate, std::move(diagnostic));
+            auto func_base_cast = createInternalCast(std::move(left_column), res_elem.type, CastType::nonAccurate, std::move(diagnostic));
 
             NodeRawConstPtrs children = { left_arg, right_arg };
-            dst_node = &actions_dag.addFunction(func_builder_cast, std::move(children), {});
+            dst_node = &actions_dag.addFunction(func_base_cast, std::move(children), {});
         }
 
         if (dst_node->column && isColumnConst(*dst_node->column) && !(res_elem.column && isColumnConst(*res_elem.column)))

src/Interpreters/castColumn.cpp

@@ -26,11 +26,9 @@ static ColumnPtr castColumn(CastType cast_type, const ColumnWithTypeAndName & ar
             ""
         }
     };
-    auto get_cast_func = [cast_type, &arguments]
+    auto get_cast_func = [from = arg, to = type, cast_type]
     {
-
+        return createInternalCast(from, to, cast_type, {});
-        FunctionOverloadResolverPtr func_builder_cast = createInternalCastOverloadResolver(cast_type, {});
-        return func_builder_cast->build(arguments);
     };
 
     FunctionBasePtr func_cast = cache ? cache->getOrSet(cast_type, from_name, to_name, std::move(get_cast_func)) : get_cast_func();

src/Processors/QueryPlan/Optimizations/optimizeUseAggregateProjection.cpp

@@ -255,20 +255,13 @@ static void appendAggregateFunctions(
 
         const auto * node = input;
 
-        if (node->result_name != aggregate.column_name)
+        if (!DataTypeAggregateFunction::strictEquals(type, node->result_type))
-        {
+            /// Cast to aggregate types specified in query if it's not
-            if (DataTypeAggregateFunction::strictEquals(type, node->result_type))
+            /// strictly the same as the one specified in projection. This
-            {
+            /// is required to generate correct results during finalization.
-                node = &proj_dag.addAlias(*node, aggregate.column_name);
+            node = &proj_dag.addCast(*node, type, aggregate.column_name);
-            }
+        else if (node->result_name != aggregate.column_name)
-            else
+            node = &proj_dag.addAlias(*node, aggregate.column_name);
-            {
-                /// Cast to aggregate types specified in query if it's not
-                /// strictly the same as the one specified in projection. This
-                /// is required to generate correct results during finalization.
-                node = &proj_dag.addCast(*node, type, aggregate.column_name);
-            }
-        }
 
         proj_dag_outputs.push_back(node);
     }

src/Processors/Sources/ShellCommandSource.cpp

@@ -8,13 +8,15 @@
 #include <IO/WriteHelpers.h>
 #include <IO/ReadHelpers.h>
 
-#include <QueryPipeline/Pipe.h>
-#include <Processors/ISimpleTransform.h>
-#include <Processors/Formats/IOutputFormat.h>
-#include <Processors/Executors/CompletedPipelineExecutor.h>
 #include <Interpreters/Context.h>
+#include <Processors/Executors/CompletedPipelineExecutor.h>
+#include <Processors/Formats/IOutputFormat.h>
+#include <Processors/ISimpleTransform.h>
+#include <QueryPipeline/Pipe.h>
+
 #include <boost/circular_buffer.hpp>
 
+#include <ranges>
 
 namespace DB
 {
@@ -68,11 +70,17 @@ static void makeFdBlocking(int fd)
 
 static int pollWithTimeout(pollfd * pfds, size_t num, size_t timeout_milliseconds)
 {
+    auto logger = getLogger("TimeoutReadBufferFromFileDescriptor");
+    auto describe_fd = [](const auto & pollfd) { return fmt::format("(fd={}, flags={})", pollfd.fd, fcntl(pollfd.fd, F_GETFL)); };
+
     int res;
 
     while (true)
     {
         Stopwatch watch;
+
+        LOG_TEST(logger, "Polling descriptors: {}", fmt::join(std::span(pfds, pfds + num) | std::views::transform(describe_fd), ", "));
+
         res = poll(pfds, static_cast<nfds_t>(num), static_cast<int>(timeout_milliseconds));
 
         if (res < 0)
@@ -82,7 +90,10 @@ static int pollWithTimeout(pollfd * pfds, size_t num, size_t timeout_millisecond
 
             const auto elapsed = watch.elapsedMilliseconds();
             if (timeout_milliseconds <= elapsed)
+            {
+                LOG_TEST(logger, "Timeout exceeded: elapsed={}, timeout={}", elapsed, timeout_milliseconds);
                 break;
+            }
             timeout_milliseconds -= elapsed;
         }
         else
@@ -91,6 +102,12 @@ static int pollWithTimeout(pollfd * pfds, size_t num, size_t timeout_millisecond
         }
     }
 
+    LOG_TEST(
+        logger,
+        "Poll for descriptors: {} returned {}",
+        fmt::join(std::span(pfds, pfds + num) | std::views::transform(describe_fd), ", "),
+        res);
+
     return res;
 }
 
@@ -200,12 +217,6 @@ public:
         return true;
     }
 
-    void reset() const
-    {
-        makeFdBlocking(stdout_fd);
-        makeFdBlocking(stderr_fd);
-    }
-
     ~TimeoutReadBufferFromFileDescriptor() override
     {
         tryMakeFdBlocking(stdout_fd);

src/Processors/Transforms/WindowTransform.cpp

@@ -2337,22 +2337,9 @@ struct WindowFunctionLagLeadInFrame final : public WindowFunction
                 argument_types[2]->getName());
         }
 
-        const auto from_name = argument_types[2]->getName();
+        auto get_cast_func = [from = argument_types[2], to = argument_types[0]]
-        const auto to_name = argument_types[0]->getName();
-        ColumnsWithTypeAndName arguments
         {
-            { argument_types[2], "" },
+            return createInternalCast({from, {}}, to, CastType::accurate, {});
-            {
-                DataTypeString().createColumnConst(0, to_name),
-                std::make_shared<DataTypeString>(),
-                ""
-            }
-        };
-
-        auto get_cast_func = [&arguments]
-        {
-            FunctionOverloadResolverPtr func_builder_cast = createInternalCastOverloadResolver(CastType::accurate, {});
-            return func_builder_cast->build(arguments);
         };
 
         func_cast = get_cast_func();

src/Storages/MergeTree/IMergeTreeDataPart.cpp

@@ -749,8 +749,16 @@ void IMergeTreeDataPart::loadColumnsChecksumsIndexes(bool require_columns_checks
                 /// Probably there is something wrong with files of this part.
                 /// So it can be helpful to add to the error message some information about those files.
                 String files_in_part;
+
                 for (auto it = getDataPartStorage().iterate(); it->isValid(); it->next())
-                    files_in_part += fmt::format("{}{} ({} bytes)", (files_in_part.empty() ? "" : ", "), it->name(), getDataPartStorage().getFileSize(it->name()));
+                {
+                    std::string file_info;
+                    if (!getDataPartStorage().isDirectory(it->name()))
+                        file_info = fmt::format(" ({} bytes)", getDataPartStorage().getFileSize(it->name()));
+
+                    files_in_part += fmt::format("{}{}{}", (files_in_part.empty() ? "" : ", "), it->name(), file_info);
+
+                }
                 if (!files_in_part.empty())
                     e->addMessage("Part contains files: {}", files_in_part);
                 if (isEmpty())
@@ -2141,7 +2149,27 @@ void IMergeTreeDataPart::checkConsistencyBase() const
             }
         }
 
-        checksums.checkSizes(getDataPartStorage());
+        const auto & data_part_storage = getDataPartStorage();
+        for (const auto & [filename, checksum] : checksums.files)
+        {
+            try
+            {
+                checksum.checkSize(data_part_storage, filename);
+            }
+            catch (const Exception & ex)
+            {
+                /// For projection parts check will mark them broken in loadProjections
+                if (!parent_part && filename.ends_with(".proj"))
+                {
+                    std::string projection_name = fs::path(filename).stem();
+                    LOG_INFO(storage.log, "Projection {} doesn't exist on start for part {}, marking it as broken", projection_name, name);
+                    if (hasProjection(projection_name))
+                        markProjectionPartAsBroken(projection_name, ex.message(), ex.code());
+                }
+                else
+                    throw;
+            }
+        }
     }
     else
     {

src/Storages/MergeTree/KeyCondition.cpp

@@ -1956,11 +1956,8 @@ bool KeyCondition::extractAtomFromTree(const RPNBuilderTreeNode & node, RPNEleme
                         auto common_type_maybe_nullable = (key_expr_type_is_nullable && !common_type->isNullable())
                             ? DataTypePtr(std::make_shared<DataTypeNullable>(common_type))
                             : common_type;
-                        ColumnsWithTypeAndName arguments{
+
-                            {nullptr, key_expr_type, ""},
+                        auto func_cast = createInternalCast({key_expr_type, {}}, common_type_maybe_nullable, CastType::nonAccurate, {});
-                            {DataTypeString().createColumnConst(1, common_type_maybe_nullable->getName()), common_type_maybe_nullable, ""}};
-                        FunctionOverloadResolverPtr func_builder_cast = createInternalCastOverloadResolver(CastType::nonAccurate, {});
-                        auto func_cast = func_builder_cast->build(arguments);
 
                         /// If we know the given range only contains one value, then we treat all functions as positive monotonic.
                         if (!single_point && !func_cast->hasInformationAboutMonotonicity())

src/Storages/MergeTree/MergeTreeData.cpp

@@ -1146,7 +1146,7 @@ std::optional<UInt64> MergeTreeData::totalRowsByPartitionPredicateImpl(
     auto metadata_snapshot = getInMemoryMetadataPtr();
     auto virtual_columns_block = getBlockWithVirtualsForFilter(metadata_snapshot, {parts[0]});
 
-    auto filter_dag = VirtualColumnUtils::splitFilterDagForAllowedInputs(filter_actions_dag.getOutputs().at(0), nullptr);
+    auto filter_dag = VirtualColumnUtils::splitFilterDagForAllowedInputs(filter_actions_dag.getOutputs().at(0), nullptr, /*allow_partial_result=*/ false);
     if (!filter_dag)
         return {};
 
@@ -6932,7 +6932,8 @@ Block MergeTreeData::getMinMaxCountProjectionBlock(
         const auto * predicate = filter_dag->getOutputs().at(0);
 
         // Generate valid expressions for filtering
-        VirtualColumnUtils::filterBlockWithPredicate(predicate, virtual_columns_block, query_context);
+        VirtualColumnUtils::filterBlockWithPredicate(
+            predicate, virtual_columns_block, query_context, /*allow_filtering_with_partial_predicate =*/true);
 
         rows = virtual_columns_block.rows();
         part_name_column = virtual_columns_block.getByName("_part").column;

src/Storages/MergeTree/MergeTreeDataPartChecksum.cpp

@@ -100,12 +100,6 @@ void MergeTreeDataPartChecksums::checkEqual(const MergeTreeDataPartChecksums & r
     }
 }
 
-void MergeTreeDataPartChecksums::checkSizes(const IDataPartStorage & storage) const
-{
-    for (const auto & [name, checksum] : files)
-        checksum.checkSize(storage, name);
-}
-
 UInt64 MergeTreeDataPartChecksums::getTotalSizeOnDisk() const
 {
     UInt64 res = 0;

src/Storages/MergeTree/MergeTreeDataPartChecksum.h

@@ -65,9 +65,6 @@ struct MergeTreeDataPartChecksums
 
     static bool isBadChecksumsErrorCode(int code);
 
-    /// Checks that the directory contains all the needed files of the correct size. Does not check the checksum.
-    void checkSizes(const IDataPartStorage & storage) const;
-
     /// Returns false if the checksum is too old.
     bool read(ReadBuffer & in);
     /// Assume that header with version (the first line) is read

src/Storages/MergeTree/MergeTreeSplitPrewhereIntoReadSteps.cpp

@@ -152,23 +152,15 @@ const ActionsDAG::Node & addFunction(
 const ActionsDAG::Node & addCast(
         const ActionsDAGPtr & dag,
         const ActionsDAG::Node & node_to_cast,
-        const String & type_name,
+        const DataTypePtr & to_type,
         OriginalToNewNodeMap & node_remap)
 {
-    if (node_to_cast.result_type->getName() == type_name)
+    if (!node_to_cast.result_type->equals(*to_type))
         return node_to_cast;
 
-    Field cast_type_constant_value(type_name);
+    const auto & new_node = dag->addCast(node_to_cast, to_type, {});
-
+    node_remap[new_node.result_name] = {dag.get(), &new_node};
-    ColumnWithTypeAndName column;
+    return new_node;
-    column.column = DataTypeString().createColumnConst(0, cast_type_constant_value);
-    column.type = std::make_shared<DataTypeString>();
-
-    const auto * cast_type_constant_node = &dag->addColumn(std::move(column));
-    ActionsDAG::NodeRawConstPtrs children = {&node_to_cast, cast_type_constant_node};
-    FunctionOverloadResolverPtr func_builder_cast = createInternalCastOverloadResolver(CastType::nonAccurate, {});
-
-    return addFunction(dag, func_builder_cast, std::move(children), node_remap);
 }
 
 /// Normalizes the filter node by adding AND with a constant true.
@@ -332,7 +324,7 @@ bool tryBuildPrewhereSteps(PrewhereInfoPtr prewhere_info, const ExpressionAction
             /// Build AND(last_step_result_node, true)
             const auto & and_node = addAndTrue(last_step_dag, *last_step_result_node_info.node, node_remap);
             /// Build CAST(and_node, type of PREWHERE column)
-            const auto & cast_node = addCast(last_step_dag, and_node, output->result_type->getName(), node_remap);
+            const auto & cast_node = addCast(last_step_dag, and_node, output->result_type, node_remap);
             /// Add alias for the result with the name of the PREWHERE column
             const auto & prewhere_result_node = last_step_dag->addAlias(cast_node, output->result_name);
             last_step_dag->addOrReplaceInOutputs(prewhere_result_node);

src/Storages/VirtualColumnUtils.cpp

@@ -1,51 +1,46 @@
-#include <algorithm>
+#include <Storages/VirtualColumnUtils.h>
+
 #include <memory>
 #include <stack>
+#include <Columns/ColumnConst.h>
+#include <Columns/ColumnSet.h>
+#include <Columns/ColumnsCommon.h>
+#include <Columns/ColumnsNumber.h>
+#include <Columns/FilterDescription.h>
 #include <Core/NamesAndTypes.h>
 #include <Core/TypeId.h>
-
+#include <DataTypes/DataTypeDateTime.h>
+#include <DataTypes/DataTypeLowCardinality.h>
+#include <DataTypes/DataTypeString.h>
+#include <DataTypes/DataTypesNumber.h>
+#include <Functions/FunctionHelpers.h>
+#include <Functions/FunctionsLogical.h>
+#include <Functions/IFunction.h>
+#include <Functions/IFunctionAdaptors.h>
+#include <Functions/indexHint.h>
+#include <IO/WriteHelpers.h>
+#include <Interpreters/ActionsDAG.h>
+#include <Interpreters/ActionsVisitor.h>
 #include <Interpreters/Context.h>
-#include <Interpreters/TreeRewriter.h>
-#include <Interpreters/ExpressionAnalyzer.h>
 #include <Interpreters/ExpressionActions.h>
+#include <Interpreters/ExpressionAnalyzer.h>
 #include <Interpreters/IdentifierSemantic.h>
+#include <Interpreters/TreeRewriter.h>
 #include <Interpreters/misc.h>
-
-#include <Parsers/ASTIdentifier.h>
 #include <Parsers/ASTExpressionList.h>
-#include <Parsers/ASTLiteral.h>
 #include <Parsers/ASTFunction.h>
+#include <Parsers/ASTIdentifier.h>
+#include <Parsers/ASTLiteral.h>
 #include <Parsers/ASTSelectQuery.h>
 #include <Parsers/ASTSubquery.h>
-
+#include <Parsers/makeASTForLogicalFunction.h>
-#include <Columns/ColumnConst.h>
+#include <Processors/Executors/CompletedPipelineExecutor.h>
-#include <Columns/ColumnsNumber.h>
-#include <Columns/ColumnsCommon.h>
-#include <Columns/FilterDescription.h>
-
-#include <DataTypes/DataTypesNumber.h>
-#include <DataTypes/DataTypeString.h>
-#include <DataTypes/DataTypeLowCardinality.h>
-#include <DataTypes/DataTypeDateTime.h>
-
-#include <Processors/QueryPlan/QueryPlan.h>
 #include <Processors/QueryPlan/BuildQueryPipelineSettings.h>
 #include <Processors/QueryPlan/Optimizations/QueryPlanOptimizationSettings.h>
+#include <Processors/QueryPlan/QueryPlan.h>
 #include <Processors/Sinks/EmptySink.h>
-#include <Processors/Executors/CompletedPipelineExecutor.h>
 #include <QueryPipeline/QueryPipelineBuilder.h>
-
-#include <Storages/VirtualColumnUtils.h>
-#include <IO/WriteHelpers.h>
 #include <Common/typeid_cast.h>
-#include "Functions/FunctionsLogical.h"
-#include "Functions/IFunction.h"
-#include "Functions/IFunctionAdaptors.h"
-#include "Functions/indexHint.h"
-#include <Parsers/makeASTForLogicalFunction.h>
-#include <Columns/ColumnSet.h>
-#include <Functions/FunctionHelpers.h>
-#include <Interpreters/ActionsVisitor.h>
 
 
 namespace DB
@@ -281,9 +276,7 @@ bool isDeterministicInScopeOfQuery(const ActionsDAG::Node * node)
 }
 
 static const ActionsDAG::Node * splitFilterNodeForAllowedInputs(
-    const ActionsDAG::Node * node,
+    const ActionsDAG::Node * node, const Block * allowed_inputs, ActionsDAG::Nodes & additional_nodes, bool allow_partial_result)
-    const Block * allowed_inputs,
-    ActionsDAG::Nodes & additional_nodes)
 {
     if (node->type == ActionsDAG::ActionType::FUNCTION)
     {
@@ -292,8 +285,15 @@ static const ActionsDAG::Node * splitFilterNodeForAllowedInputs(
             auto & node_copy = additional_nodes.emplace_back(*node);
             node_copy.children.clear();
             for (const auto * child : node->children)
-                if (const auto * child_copy = splitFilterNodeForAllowedInputs(child, allowed_inputs, additional_nodes))
+                if (const auto * child_copy
+                    = splitFilterNodeForAllowedInputs(child, allowed_inputs, additional_nodes, allow_partial_result))
                     node_copy.children.push_back(child_copy);
+                /// Expression like (now_allowed AND allowed) is not allowed if allow_partial_result = true. This is important for
+                /// trivial count optimization, otherwise we can get incorrect results. For example, if the query is
+                /// SELECT count() FROM table WHERE _partition_id = '0' AND rowNumberInBlock() = 1, we cannot apply
+                /// trivial count.
+                else if (!allow_partial_result)
+                    return nullptr;
 
             if (node_copy.children.empty())
                 return nullptr;
@@ -301,7 +301,7 @@ static const ActionsDAG::Node * splitFilterNodeForAllowedInputs(
             if (node_copy.children.size() == 1)
             {
                 const ActionsDAG::Node * res = node_copy.children.front();
-                /// Expression like (not_allowed AND 256) can't be resuced to (and(256)) because AND requires
+                /// Expression like (not_allowed AND 256) can't be reduced to (and(256)) because AND requires
                 /// at least two arguments; also it can't be reduced to (256) because result type is different.
                 if (!res->result_type->equals(*node->result_type))
                 {
@@ -319,7 +319,7 @@ static const ActionsDAG::Node * splitFilterNodeForAllowedInputs(
         {
             auto & node_copy = additional_nodes.emplace_back(*node);
             for (auto & child : node_copy.children)
-                if (child = splitFilterNodeForAllowedInputs(child, allowed_inputs, additional_nodes); !child)
+                if (child = splitFilterNodeForAllowedInputs(child, allowed_inputs, additional_nodes, allow_partial_result); !child)
                     return nullptr;
 
             return &node_copy;
@@ -333,7 +333,8 @@ static const ActionsDAG::Node * splitFilterNodeForAllowedInputs(
                     auto index_hint_dag = index_hint->getActions().clone();
                     ActionsDAG::NodeRawConstPtrs atoms;
                     for (const auto & output : index_hint_dag.getOutputs())
-                        if (const auto * child_copy = splitFilterNodeForAllowedInputs(output, allowed_inputs, additional_nodes))
+                        if (const auto * child_copy
+                            = splitFilterNodeForAllowedInputs(output, allowed_inputs, additional_nodes, allow_partial_result))
                             atoms.push_back(child_copy);
 
                     if (!atoms.empty())
@@ -367,22 +368,24 @@ static const ActionsDAG::Node * splitFilterNodeForAllowedInputs(
     return node;
 }
 
-std::optional<ActionsDAG> splitFilterDagForAllowedInputs(const ActionsDAG::Node * predicate, const Block * allowed_inputs)
+std::optional<ActionsDAG>
+splitFilterDagForAllowedInputs(const ActionsDAG::Node * predicate, const Block * allowed_inputs, bool allow_partial_result)
 {
     if (!predicate)
         return {};
 
     ActionsDAG::Nodes additional_nodes;
-    const auto * res = splitFilterNodeForAllowedInputs(predicate, allowed_inputs, additional_nodes);
+    const auto * res = splitFilterNodeForAllowedInputs(predicate, allowed_inputs, additional_nodes, allow_partial_result);
     if (!res)
         return {};
 
     return ActionsDAG::cloneSubDAG({res}, true);
 }
 
-void filterBlockWithPredicate(const ActionsDAG::Node * predicate, Block & block, ContextPtr context)
+void filterBlockWithPredicate(
+    const ActionsDAG::Node * predicate, Block & block, ContextPtr context, bool allow_filtering_with_partial_predicate)
 {
-    auto dag = splitFilterDagForAllowedInputs(predicate, &block);
+    auto dag = splitFilterDagForAllowedInputs(predicate, &block, /*allow_partial_result=*/allow_filtering_with_partial_predicate);
     if (dag)
         filterBlockWithExpression(buildFilterExpression(std::move(*dag), context), block);
 }

src/Storages/VirtualColumnUtils.h

@@ -26,9 +26,13 @@ namespace VirtualColumnUtils
 ///
 /// Otherwise calling filter*() outside applyFilters() will throw "Not-ready Set is passed"
 /// if there are subqueries.
+///
+/// Similar to filterBlockWithExpression(buildFilterExpression(splitFilterDagForAllowedInputs(...)))./// Similar to filterBlockWithQuery, but uses ActionsDAG as a predicate.
+/// Basically it is filterBlockWithDAG(splitFilterDagForAllowedInputs).
+/// If allow_filtering_with_partial_predicate is true, then the filtering will be done even if some part of the predicate
+/// cannot be evaluated using the columns from the block.
+void filterBlockWithPredicate(const ActionsDAG::Node * predicate, Block & block, ContextPtr context, bool allow_filtering_with_partial_predicate = true);
 
-/// Similar to filterBlockWithExpression(buildFilterExpression(splitFilterDagForAllowedInputs(...))).
-void filterBlockWithPredicate(const ActionsDAG::Node * predicate, Block & block, ContextPtr context);
 
 /// Just filters block. Block should contain all the required columns.
 ExpressionActionsPtr buildFilterExpression(ActionsDAG dag, ContextPtr context);
@@ -41,7 +45,15 @@ void buildSetsForDAG(const ActionsDAG & dag, const ContextPtr & context);
 bool isDeterministicInScopeOfQuery(const ActionsDAG::Node * node);
 
 /// Extract a part of predicate that can be evaluated using only columns from input_names.
-std::optional<ActionsDAG> splitFilterDagForAllowedInputs(const ActionsDAG::Node * predicate, const Block * allowed_inputs);
+/// When allow_partial_result is false, then the result will be empty if any part of if cannot be evaluated deterministically
+/// on the given inputs.
+/// allow_partial_result must be false when we are going to use the result to filter parts in
+/// MergeTreeData::totalRowsByPartitionPredicateImp. For example, if the query is
+/// `SELECT count() FROM table  WHERE _partition_id = '0' AND rowNumberInBlock() = 1`
+/// The predicate will be `_partition_id = '0' AND rowNumberInBlock() = 1`, and `rowNumberInBlock()` is
+/// non-deterministic. If we still extract the part `_partition_id = '0'` for filtering parts, then trivial
+/// count optimization will be mistakenly applied to the query.
+std::optional<ActionsDAG> splitFilterDagForAllowedInputs(const ActionsDAG::Node * predicate, const Block * allowed_inputs, bool allow_partial_result = true);
 
 /// Extract from the input stream a set of `name` column values
 template <typename T>

tests/ci/ci.py

@@ -985,6 +985,7 @@ def _run_test(job_name: str, run_command: str) -> int:
     else:
         print("Use run command from the workflow")
     env["CHECK_NAME"] = job_name
+    env["MAX_RUN_TIME"] = str(timeout or 0)
     print(f"Going to start run command [{run_command}]")
     stopwatch = Stopwatch()
     job_log = Path(TEMP_PATH) / "job_log.txt"

tests/ci/functional_test_check.py

@@ -114,6 +114,9 @@ def get_run_command(
     if flaky_check:
         envs.append("-e NUM_TRIES=50")
         envs.append("-e MAX_RUN_TIME=2800")
+    else:
+        max_run_time = os.getenv("MAX_RUN_TIME", "0")
+        envs.append(f"-e MAX_RUN_TIME={max_run_time}")
 
     envs += [f"-e {e}" for e in additional_envs]
 

tests/ci/report.py

@@ -738,7 +738,7 @@ def create_test_html_report(
     if test_results:
         rows_part = []
         num_fails = 0
-        has_test_time = False
+        has_test_time = any(tr.time is not None for tr in test_results)
         has_log_urls = False
 
         # Display entires with logs at the top (they correspond to failed tests)
@@ -770,12 +770,12 @@ def create_test_html_report(
             row.append(f'<td {fail_id}style="{style}">{test_result.status}</td>')
             colspan += 1
 
-            row.append("<td>")
+            if has_test_time:
-            if test_result.time is not None:
+                if test_result.time is not None:
-                has_test_time = True
+                    row.append(f"<td>{test_result.time}</td>")
-                row.append(str(test_result.time))
+                else:
-            row.append("</td>")
+                    row.append("<td></td>")
-            colspan += 1
+                colspan += 1
 
             if test_result.log_urls is not None:
                 has_log_urls = True

tests/integration/test_broken_projections/config.d/dont_start_broken.xml

@@ -0,0 +1,6 @@
+<?xml version="1.0" encoding="utf-8"?>
+<clickhouse>
+  <merge_tree>
+    <max_suspicious_broken_parts_bytes>0</max_suspicious_broken_parts_bytes>
+  </merge_tree>
+</clickhouse>

tests/integration/test_broken_projections/test.py

@@ -4,6 +4,7 @@ import logging
 import string
 import random
 from helpers.cluster import ClickHouseCluster
+from multiprocessing.dummy import Pool
 
 cluster = ClickHouseCluster(__file__)
 
@@ -18,6 +19,12 @@ def cluster():
             stay_alive=True,
             with_zookeeper=True,
         )
+        cluster.add_instance(
+            "node_restart",
+            main_configs=["config.d/dont_start_broken.xml"],
+            stay_alive=True,
+            with_zookeeper=True,
+        )
 
         logging.info("Starting cluster...")
         cluster.start()
@@ -632,6 +639,49 @@ def test_broken_on_start(cluster):
     check(node, table_name, 0)
 
 
+def test_disappeared_projection_on_start(cluster):
+    node = cluster.instances["node_restart"]
+
+    table_name = "test_disapperead_projection"
+    create_table(node, table_name, 1)
+
+    node.query(f"SYSTEM STOP MERGES {table_name}")
+
+    insert(node, table_name, 0, 5)
+    insert(node, table_name, 5, 5)
+    insert(node, table_name, 10, 5)
+    insert(node, table_name, 15, 5)
+
+    assert ["all_0_0_0", "all_1_1_0", "all_2_2_0", "all_3_3_0"] == get_parts(
+        node, table_name
+    )
+
+    def drop_projection():
+        node.query(
+            f"ALTER TABLE {table_name} DROP PROJECTION proj2",
+            settings={"mutations_sync": "0"},
+        )
+
+    p = Pool(2)
+    p.apply_async(drop_projection)
+
+    for i in range(30):
+        create_query = node.query(f"SHOW CREATE TABLE {table_name}")
+        if "proj2" not in create_query:
+            break
+        time.sleep(0.5)
+
+    assert "proj2" not in create_query
+
+    # Remove 'proj2' for part all_2_2_0
+    break_projection(node, table_name, "proj2", "all_2_2_0", "part")
+
+    node.restart_clickhouse()
+
+    # proj2 is not broken, it doesn't exist, but ok
+    check(node, table_name, 0, expect_broken_part="proj2", do_check_command=0)
+
+
 def test_mutation_with_broken_projection(cluster):
     node = cluster.instances["node"]
 

tests/integration/test_drop_is_lock_free/test.py

@@ -176,7 +176,7 @@ def test_query_is_permanent(transaction, permanent, exclusive_table):
 
     select_handler = node.get_query_request(
         f"""
-            SELECT sleepEachRow(3) FROM {exclusive_table} SETTINGS function_sleep_max_microseconds_per_block = 0;
+            SELECT sleepEachRow(3) FROM {exclusive_table} SETTINGS function_sleep_max_microseconds_per_block = 0, max_threads=1;
         """,
         query_id=query_id,
     )

tests/integration/test_storage_s3_queue/test.py

@@ -771,7 +771,11 @@ def test_multiple_tables_streaming_sync_distributed(started_cluster, mode):
             table_name,
             mode,
             files_path,
-            additional_settings={"keeper_path": keeper_path, "s3queue_buckets": 2},
+            additional_settings={
+                "keeper_path": keeper_path,
+                "s3queue_buckets": 2,
+                **({"s3queue_processing_threads_num": 1} if mode == "ordered" else {}),
+            },
         )
 
     for instance in [node, node_2]:
@@ -806,6 +810,10 @@ def test_multiple_tables_streaming_sync_distributed(started_cluster, mode):
         list(map(int, l.split())) for l in run_query(node_2, get_query).splitlines()
     ]
 
+    logging.debug(
+        f"res1 size: {len(res1)}, res2 size: {len(res2)}, total_rows: {total_rows}"
+    )
+
     assert len(res1) + len(res2) == total_rows
 
     # Checking that all engines have made progress

tests/performance/optimize_functions_to_subcolumns.xml

@@ -0,0 +1,26 @@
+<test>
+    <settings>
+        <max_insert_threads>4</max_insert_threads>
+    </settings>
+
+    <create_query>
+        CREATE TABLE t_subcolumns (a Array(UInt64), s Nullable(String), m Map(String, UInt64)) ENGINE = MergeTree ORDER BY tuple()
+    </create_query>
+
+    <fill_query>
+        INSERT INTO t_subcolumns SELECT range(number % 20), toString(number), mapFromArrays(range(number % 20), range(number % 20)) FROM numbers_mt(50000000)
+    </fill_query>
+
+    <fill_query>
+        OPTIMIZE TABLE t_subcolumns FINAL
+    </fill_query>
+
+    <query>SELECT count() FROM t_subcolumns WHERE NOT ignore(length(a))</query>
+    <query>SELECT count() FROM t_subcolumns WHERE notEmpty(a)</query>
+    <query>SELECT count() FROM t_subcolumns WHERE NOT ignore(length(m))</query>
+    <query>SELECT count() FROM t_subcolumns WHERE notEmpty(m)</query>
+    <query>SELECT count() FROM t_subcolumns WHERE isNotNull(s)</query>
+    <query>SELECT count(s) FROM t_subcolumns</query>
+
+    <drop_query>DROP TABLE t_subcolumns</drop_query>
+</test>

tests/queries/0_stateless/01246_buffer_flush.sh

@@ -27,7 +27,7 @@ function wait_until()
 function get_buffer_delay()
 {
     local buffer_insert_id=$1 && shift
-    query "SYSTEM FLUSH LOGS"
+    $CLICKHOUSE_CLIENT -q "SYSTEM FLUSH LOGS"
     query "
         WITH
             (SELECT event_time_microseconds FROM system.query_log WHERE current_database = '$CLICKHOUSE_DATABASE' AND type = 'QueryStart' AND query_id = '$buffer_insert_id') AS begin_,

tests/queries/0_stateless/01600_parts_states_metrics_long.sh

@@ -1,10 +1,16 @@
 #!/usr/bin/env bash
-# Tags: long
+# Tags: long, no-parallel
 
 CURDIR=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
 # shellcheck source=../shell_config.sh
 . "$CURDIR"/../shell_config.sh
 
+function query()
+{
+    # NOTE: database_atomic_wait_for_drop_and_detach_synchronously needed only for local env, CI has it ON
+    ${CLICKHOUSE_CURL} -sS "${CLICKHOUSE_URL}&database_atomic_wait_for_drop_and_detach_synchronously=1" -d "$*"
+}
+
 # NOTE: database = $CLICKHOUSE_DATABASE is unwanted
 verify_sql="SELECT
     (SELECT sumIf(value, metric = 'PartsActive'), sumIf(value, metric = 'PartsOutdated') FROM system.metrics)
@@ -18,13 +24,13 @@ verify()
 {
     for i in {1..5000}
     do
-        result=$( $CLICKHOUSE_CLIENT --query="$verify_sql" )
+        result=$( query "$verify_sql" )
         [ "$result" = "1" ] && echo "$result" && break
         sleep 0.1
 
         if [[ $i -eq 5000 ]]
         then
-            $CLICKHOUSE_CLIENT "
+            query "
               SELECT sumIf(value, metric = 'PartsActive'), sumIf(value, metric = 'PartsOutdated') FROM system.metrics;
               SELECT sum(active), sum(NOT active) FROM system.parts;
               SELECT sum(active), sum(NOT active) FROM system.projection_parts;
@@ -34,17 +40,17 @@ verify()
     done
 }
 
-$CLICKHOUSE_CLIENT --database_atomic_wait_for_drop_and_detach_synchronously=1 --query="DROP TABLE IF EXISTS test_table"
+query "DROP TABLE IF EXISTS test_table"
-$CLICKHOUSE_CLIENT --query="CREATE TABLE test_table (data Date) ENGINE = MergeTree PARTITION BY toYear(data) ORDER BY data;"
+query "CREATE TABLE test_table (data Date) ENGINE = MergeTree PARTITION BY toYear(data) ORDER BY data;"
 
-$CLICKHOUSE_CLIENT --query="INSERT INTO test_table VALUES ('1992-01-01')"
+query "INSERT INTO test_table VALUES ('1992-01-01')"
 verify
 
-$CLICKHOUSE_CLIENT --query="INSERT INTO test_table VALUES ('1992-01-02')"
+query "INSERT INTO test_table VALUES ('1992-01-02')"
 verify
 
-$CLICKHOUSE_CLIENT --query="OPTIMIZE TABLE test_table FINAL"
+query "OPTIMIZE TABLE test_table FINAL"
 verify
 
-$CLICKHOUSE_CLIENT --database_atomic_wait_for_drop_and_detach_synchronously=1 --query="DROP TABLE test_table"
+query "DROP TABLE test_table"
 verify

tests/queries/0_stateless/01656_test_query_log_factories_info.reference

@@ -17,7 +17,7 @@ used_functions
 ['repeat']
 
 arraySort(used_data_type_families)
-['Array','Int32','Nullable','String']
+['Int32','Nullable','String']
 
 used_database_engines
 ['Atomic']

tests/queries/0_stateless/01710_aggregate_projection_with_normalized_states.reference

@@ -1,2 +1,3 @@
 3
 950	990	500	2000
+[950]	[999]

tests/queries/0_stateless/01710_aggregate_projection_with_normalized_states.sql

@@ -29,4 +29,6 @@ FROM cluster('test_cluster_two_shards', currentDatabase(), r)
 WHERE a = 'x'
 settings prefer_localhost_replica=0;
 
+SELECT quantilesTimingMerge(0.95)(q), quantilesTimingMerge(toInt64(1))(q) FROM remote('127.0.0.{1,2}', currentDatabase(), r);
+
 DROP TABLE r;

tests/queries/0_stateless/03203_count_with_non_deterministic_function.reference

@@ -0,0 +1,2 @@
+1
+1

tests/queries/0_stateless/03203_count_with_non_deterministic_function.sql

@@ -0,0 +1,4 @@
+CREATE TABLE t (p UInt8, x UInt64) Engine = MergeTree PARTITION BY p ORDER BY x;
+INSERT INTO t SELECT 0, number FROM numbers(10) SETTINGS max_block_size = 100;
+SELECT count() FROM t WHERE p = 0 AND rowNumberInAllBlocks() = 1 SETTINGS allow_experimental_analyzer = 0;
+SELECT count() FROM t WHERE p = 0 AND rowNumberInAllBlocks() = 1 SETTINGS allow_experimental_analyzer = 1;

tests/queries/0_stateless/03217_filtering_in_storage_merge.reference

@@ -0,0 +1,6 @@
+Expression ((Project names + Projection))
+  Aggregating
+    Expression (Before GROUP BY)
+      ReadFromMerge
+        Filter (( + ( + )))
+          ReadFromMergeTree (default.test_03217_merge_replica_1)

tests/queries/0_stateless/03217_filtering_in_storage_merge.sql

@@ -0,0 +1,16 @@
+CREATE TABLE test_03217_merge_replica_1(x UInt32)
+    ENGINE ReplicatedMergeTree('/clickhouse/tables/{database}/test_03217_merge_replica', 'r1')
+    ORDER BY x;
+CREATE TABLE test_03217_merge_replica_2(x UInt32)
+    ENGINE ReplicatedMergeTree('/clickhouse/tables/{database}/test_03217_merge_replica', 'r2')
+    ORDER BY x;
+
+
+CREATE TABLE test_03217_all_replicas (x UInt32)
+    ENGINE = Merge(currentDatabase(), 'test_03217_merge_replica_*');
+
+INSERT INTO test_03217_merge_replica_1 SELECT number AS x FROM numbers(10);
+SYSTEM SYNC REPLICA test_03217_merge_replica_2;
+
+-- If the filter on _table is not applied, then the plan will show both replicas
+EXPLAIN SELECT _table, count() FROM test_03217_all_replicas WHERE  _table = 'test_03217_merge_replica_1' AND x >= 0 GROUP BY _table SETTINGS allow_experimental_analyzer=1;

tests/queries/0_stateless/03217_filtering_in_system_tables.reference

@@ -0,0 +1,6 @@
+information_schema	tables
+both	default	test_03217_system_tables_replica_1	r1
+both	default	test_03217_system_tables_replica_2	r2
+default	test_03217_system_tables_replica_1	r1
+1
+1

tests/queries/0_stateless/03217_filtering_in_system_tables.sql

@@ -0,0 +1,30 @@
+-- If filtering is not done correctly on databases, then this query report to read 3 rows, which are: `system.tables`, `information_schema.tables` and `INFORMATION_SCHEMA.tables`
+SELECT database, table FROM system.tables WHERE database = 'information_schema' AND table = 'tables';
+
+CREATE TABLE test_03217_system_tables_replica_1(x UInt32)
+    ENGINE ReplicatedMergeTree('/clickhouse/tables/{database}/test_03217_system_tables_replica', 'r1')
+    ORDER BY x;
+CREATE TABLE test_03217_system_tables_replica_2(x UInt32)
+    ENGINE ReplicatedMergeTree('/clickhouse/tables/{database}/test_03217_system_tables_replica', 'r2')
+    ORDER BY x;
+
+-- Make sure we can read both replicas
+-- The replica name might be altered because of `_functional_tests_helper_database_replicated_replace_args_macros`,
+-- thus we need to use `left`
+SELECT 'both', database, table, left(replica_name, 2) FROM system.replicas WHERE database = currentDatabase();
+-- If filtering is not done correctly on database-table column, then this query report to read 2 rows, which are the above tables
+SELECT database, table, left(replica_name, 2) FROM system.replicas WHERE database = currentDatabase() AND table = 'test_03217_system_tables_replica_1' AND replica_name LIKE 'r1%';
+SYSTEM FLUSH LOGS;
+-- argMax is necessary to make the test repeatable
+
+-- StorageSystemTables
+SELECT argMax(read_rows, event_time_microseconds) FROM system.query_log WHERE 1
+    AND current_database = currentDatabase()
+    AND query LIKE '%SELECT database, table FROM system.tables WHERE database = \'information_schema\' AND table = \'tables\';'
+    AND type = 'QueryFinish';
+
+-- StorageSystemReplicas
+SELECT argMax(read_rows, event_time_microseconds) FROM system.query_log WHERE 1
+    AND current_database = currentDatabase()
+    AND query LIKE '%SELECT database, table, left(replica_name, 2) FROM system.replicas WHERE database = currentDatabase() AND table = \'test_03217_system_tables_replica_1\' AND replica_name LIKE \'r1\%\';'
+    AND type = 'QueryFinish';

tests/queries/0_stateless/03218_materialize_msan.reference

@@ -0,0 +1 @@
+[(NULL,'11\01111111\011111','1111')]		-2147483648	\N

tests/queries/0_stateless/03218_materialize_msan.sql

@@ -0,0 +1,23 @@
+SET enable_analyzer = 1;
+
+SELECT
+    materialize([(NULL, '11\01111111\011111', '1111')]) AS t,
+    (t[1048576]).2,
+    materialize(-2147483648),
+    (t[-2147483648]).1
+GROUP BY
+    materialize([(NULL, '1')]),
+    '',
+    (materialize((t[1023]).2), (materialize(''), (t[2147483647]).1, materialize(9223372036854775807)), (materialize(''), materialize(NULL, 2147483647, t[65535], 256)), materialize(NULL))
+; -- { serverError NUMBER_OF_ARGUMENTS_DOESNT_MATCH}
+
+SELECT
+    materialize([(NULL, '11\01111111\011111', '1111')]) AS t,
+    (t[1048576]).2,
+    materialize(-2147483648),
+    (t[-2147483648]).1
+GROUP BY
+    materialize([(NULL, '1')]),
+    '',
+    (materialize((t[1023]).2), (materialize(''), (t[2147483647]).1, materialize(9223372036854775807)), (materialize(''), materialize(NULL), materialize(2147483647), materialize(t[65535]), materialize(256)), materialize(NULL))
+;