Merge branch 'master' into ci-fuzzer-enable

This commit is contained in:
Yakov Olkhovskiy 2024-07-17 03:43:57 +00:00
commit f07a395bf1
48 changed files with 316 additions and 145 deletions

View File

@ -3,8 +3,9 @@
#include <base/defines.h>
#include <fstream>
#include <sstream>
#include <string>
namespace fs = std::filesystem;
bool cgroupsV2Enabled()
{
@ -13,11 +14,11 @@ bool cgroupsV2Enabled()
{
/// This file exists iff the host has cgroups v2 enabled.
auto controllers_file = default_cgroups_mount / "cgroup.controllers";
if (!std::filesystem::exists(controllers_file))
if (!fs::exists(controllers_file))
return false;
return true;
}
catch (const std::filesystem::filesystem_error &) /// all "underlying OS API errors", typically: permission denied
catch (const fs::filesystem_error &) /// all "underlying OS API errors", typically: permission denied
{
return false; /// not logging the exception as most callers fall back to cgroups v1
}
@ -33,8 +34,9 @@ bool cgroupsV2MemoryControllerEnabled()
/// According to https://docs.kernel.org/admin-guide/cgroup-v2.html, file "cgroup.controllers" defines which controllers are available
/// for the current + child cgroups. The set of available controllers can be restricted from level to level using file
/// "cgroups.subtree_control". It is therefore sufficient to check the bottom-most nested "cgroup.controllers" file.
std::string cgroup = cgroupV2OfProcess();
auto cgroup_dir = cgroup.empty() ? default_cgroups_mount : (default_cgroups_mount / cgroup);
fs::path cgroup_dir = cgroupV2PathOfProcess();
if (cgroup_dir.empty())
return false;
std::ifstream controllers_file(cgroup_dir / "cgroup.controllers");
if (!controllers_file.is_open())
return false;
@ -46,7 +48,7 @@ bool cgroupsV2MemoryControllerEnabled()
#endif
}
std::string cgroupV2OfProcess()
fs::path cgroupV2PathOfProcess()
{
#if defined(OS_LINUX)
chassert(cgroupsV2Enabled());
@ -54,17 +56,18 @@ std::string cgroupV2OfProcess()
/// A simpler way to get the membership is:
std::ifstream cgroup_name_file("/proc/self/cgroup");
if (!cgroup_name_file.is_open())
return "";
return {};
/// With cgroups v2, there will be a *single* line with prefix "0::/"
/// (see https://docs.kernel.org/admin-guide/cgroup-v2.html)
std::string cgroup;
std::getline(cgroup_name_file, cgroup);
static const std::string v2_prefix = "0::/";
if (!cgroup.starts_with(v2_prefix))
return "";
return {};
cgroup = cgroup.substr(v2_prefix.length());
return cgroup;
/// Note: The 'root' cgroup can have an empty cgroup name, this is valid
return default_cgroups_mount / cgroup;
#else
return "";
return {};
#endif
}

View File

@ -1,7 +1,6 @@
#pragma once
#include <filesystem>
#include <string>
#if defined(OS_LINUX)
/// I think it is possible to mount the cgroups hierarchy somewhere else (e.g. when in containers).
@ -16,7 +15,7 @@ bool cgroupsV2Enabled();
/// Assumes that cgroupsV2Enabled() is enabled.
bool cgroupsV2MemoryControllerEnabled();
/// Which cgroup does the process belong to?
/// Returns an empty string if the cgroup cannot be determined.
/// Detects which cgroup v2 the process belongs to and returns the filesystem path to the cgroup.
/// Returns an empty path the cgroup cannot be determined.
/// Assumes that cgroupsV2Enabled() is enabled.
std::string cgroupV2OfProcess();
std::filesystem::path cgroupV2PathOfProcess();

View File

@ -23,8 +23,9 @@ std::optional<uint64_t> getCgroupsV2MemoryLimit()
if (!cgroupsV2MemoryControllerEnabled())
return {};
std::string cgroup = cgroupV2OfProcess();
auto current_cgroup = cgroup.empty() ? default_cgroups_mount : (default_cgroups_mount / cgroup);
std::filesystem::path current_cgroup = cgroupV2PathOfProcess();
if (current_cgroup.empty())
return {};
/// Open the bottom-most nested memory limit setting file. If there is no such file at the current
/// level, try again at the parent level as memory settings are inherited.

2
contrib/libunwind vendored

@ -1 +1 @@
Subproject commit d6a01c46327e56fd86beb8aaa31591fcd9a6b7df
Subproject commit 8f28e64d15819d2d096badd598c7d85bebddb1f2

View File

@ -4,9 +4,6 @@ set(LIBUNWIND_CXX_SOURCES
"${LIBUNWIND_SOURCE_DIR}/src/libunwind.cpp"
"${LIBUNWIND_SOURCE_DIR}/src/Unwind-EHABI.cpp"
"${LIBUNWIND_SOURCE_DIR}/src/Unwind-seh.cpp")
if (APPLE)
set(LIBUNWIND_CXX_SOURCES ${LIBUNWIND_CXX_SOURCES} "${LIBUNWIND_SOURCE_DIR}/src/Unwind_AppleExtras.cpp")
endif ()
set(LIBUNWIND_C_SOURCES
"${LIBUNWIND_SOURCE_DIR}/src/UnwindLevel1.c"
@ -32,6 +29,7 @@ set_target_properties(unwind PROPERTIES FOLDER "contrib/libunwind-cmake")
target_include_directories(unwind SYSTEM BEFORE PUBLIC $<BUILD_INTERFACE:${LIBUNWIND_SOURCE_DIR}/include>)
target_compile_definitions(unwind PRIVATE -D_LIBUNWIND_NO_HEAP=1)
target_compile_definitions(unwind PRIVATE -D_LIBUNWIND_REMEMBER_STACK_ALLOC=1)
# NOTE: from this macros sizeof(unw_context_t)/sizeof(unw_cursor_t) is depends, so it should be set always
target_compile_definitions(unwind PUBLIC -D_LIBUNWIND_IS_NATIVE_ONLY)

View File

@ -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

View File

@ -11,7 +11,7 @@ This is for the case when you have Linux machine and want to use it to build `cl
The cross-build for RISC-V 64 is based on the [Build instructions](../development/build.md), follow them first.
## Install Clang-16
## Install Clang-18
Follow the instructions from https://apt.llvm.org/ for your Ubuntu or Debian setup or do
```

View File

@ -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) │
└────────────┴────────────────────────┴───────────────────────────┴────────────────────────┘
```

View File

@ -4,6 +4,9 @@ if (USE_CLANG_TIDY)
set (CMAKE_CXX_CLANG_TIDY "${CLANG_TIDY_PATH}")
endif ()
set(MAX_LINKER_MEMORY 3500)
include(../cmake/limit_jobs.cmake)
include(${ClickHouse_SOURCE_DIR}/cmake/split_debug_symbols.cmake)
# The `clickhouse` binary is a multi purpose tool that contains multiple execution modes (client, server, etc.),

View File

@ -12,6 +12,7 @@
#include <DataTypes/DataTypeLowCardinality.h>
#include <DataTypes/DataTypesNumber.h>
#include <DataTypes/DataTypeTuple.h>
namespace DB
{
@ -616,6 +617,7 @@ private:
bool is_any_nullable = false;
Tuple args;
args.reserve(equals_functions.size());
DataTypes tuple_element_types;
/// first we create tuple from RHS of equals functions
for (const auto & equals : equals_functions)
{
@ -628,16 +630,18 @@ private:
if (const auto * rhs_literal = equals_arguments[1]->as<ConstantNode>())
{
args.push_back(rhs_literal->getValue());
tuple_element_types.push_back(rhs_literal->getResultType());
}
else
{
const auto * lhs_literal = equals_arguments[0]->as<ConstantNode>();
assert(lhs_literal);
args.push_back(lhs_literal->getValue());
tuple_element_types.push_back(lhs_literal->getResultType());
}
}
auto rhs_node = std::make_shared<ConstantNode>(std::move(args));
auto rhs_node = std::make_shared<ConstantNode>(std::move(args), std::make_shared<DataTypeTuple>(std::move(tuple_element_types)));
auto in_function = std::make_shared<FunctionNode>("in");

View File

@ -3832,6 +3832,10 @@ ProjectionNames QueryAnalyzer::resolveExpressionNode(
node->convertToNullable();
break;
}
/// Check parent scopes until find current query scope.
if (scope_ptr->scope_node->getNodeType() == QueryTreeNodeType::QUERY)
break;
}
}

View File

@ -55,9 +55,9 @@ namespace
S3::PocoHTTPClientConfiguration client_configuration = S3::ClientFactory::instance().createClientConfiguration(
settings.auth_settings.region,
context->getRemoteHostFilter(),
static_cast<unsigned>(global_settings.s3_max_redirects),
static_cast<unsigned>(global_settings.s3_retry_attempts),
global_settings.enable_s3_requests_logging,
static_cast<unsigned>(local_settings.s3_max_redirects),
static_cast<unsigned>(local_settings.backup_restore_s3_retry_attempts),
local_settings.enable_s3_requests_logging,
/* for_disk_s3 = */ false,
request_settings.get_request_throttler,
request_settings.put_request_throttler,

View File

@ -80,6 +80,7 @@
#include <Common/config_version.h>
#include "config.h"
namespace fs = std::filesystem;
using namespace std::literals;
@ -2565,12 +2566,12 @@ void ClientBase::runInteractive()
word_break_characters,
highlight_callback);
#else
(void)word_break_characters;
LineReader lr(
history_file,
getClientConfiguration().has("multiline"),
query_extenders,
query_delimiters,
word_break_characters);
query_delimiters);
#endif
static const std::initializer_list<std::pair<String, String>> backslash_aliases =

View File

@ -46,8 +46,7 @@ public:
Patterns delimiters,
std::istream & input_stream_ = std::cin,
std::ostream & output_stream_ = std::cout,
int in_fd_ = STDIN_FILENO
);
int in_fd_ = STDIN_FILENO);
virtual ~LineReader() = default;

View File

@ -362,6 +362,9 @@ ReplxxLineReader::ReplxxLineReader(
rx.bind_key(Replxx::KEY::control('N'), [this](char32_t code) { return rx.invoke(Replxx::ACTION::HISTORY_NEXT, code); });
rx.bind_key(Replxx::KEY::control('P'), [this](char32_t code) { return rx.invoke(Replxx::ACTION::HISTORY_PREVIOUS, code); });
/// We don't want the default, "suspend" behavior, it confuses people.
rx.bind_key_internal(replxx::Replxx::KEY::control('Z'), "insert_character");
auto commit_action = [this](char32_t code)
{
/// If we allow multiline and there is already something in the input, start a newline.

View File

@ -49,6 +49,7 @@ void logAboutProgress(LoggerPtr log, size_t processed, size_t total, AtomicStopw
AsyncLoader::Pool::Pool(const AsyncLoader::PoolInitializer & init)
: name(init.name)
, priority(init.priority)
, max_threads(init.max_threads > 0 ? init.max_threads : getNumberOfPhysicalCPUCores())
, thread_pool(std::make_unique<ThreadPool>(
init.metric_threads,
init.metric_active_threads,
@ -56,17 +57,16 @@ AsyncLoader::Pool::Pool(const AsyncLoader::PoolInitializer & init)
/* max_threads = */ std::numeric_limits<size_t>::max(), // Unlimited number of threads, we do worker management ourselves
/* max_free_threads = */ 0, // We do not require free threads
/* queue_size = */0)) // Unlimited queue to avoid blocking during worker spawning
, max_threads(init.max_threads > 0 ? init.max_threads : getNumberOfPhysicalCPUCores())
{}
AsyncLoader::Pool::Pool(Pool&& o) noexcept
: name(o.name)
, priority(o.priority)
, thread_pool(std::move(o.thread_pool))
, ready_queue(std::move(o.ready_queue))
, max_threads(o.max_threads)
, workers(o.workers)
, suspended_workers(o.suspended_workers.load()) // All these constructors are needed because std::atomic is neither copy-constructible, nor move-constructible. We never move pools after init, so it is safe.
, thread_pool(std::move(o.thread_pool))
{}
void cancelOnDependencyFailure(const LoadJobPtr & self, const LoadJobPtr & dependency, std::exception_ptr & cancel)

View File

@ -365,11 +365,11 @@ private:
{
const String name;
const Priority priority;
std::unique_ptr<ThreadPool> thread_pool; // NOTE: we avoid using a `ThreadPool` queue to be able to move jobs between pools.
std::map<UInt64, LoadJobPtr> ready_queue; // FIFO queue of jobs to be executed in this pool. Map is used for faster erasing. Key is `ready_seqno`
size_t max_threads; // Max number of workers to be spawn
size_t workers = 0; // Number of currently executing workers
std::atomic<size_t> suspended_workers{0}; // Number of workers that are blocked by `wait()` call on a job executing in the same pool (for deadlock resolution)
std::unique_ptr<ThreadPool> thread_pool; // NOTE: we avoid using a `ThreadPool` queue to be able to move jobs between pools.
explicit Pool(const PoolInitializer & init);
Pool(Pool&& o) noexcept;

View File

@ -25,6 +25,7 @@
#endif
using namespace DB;
namespace fs = std::filesystem;
namespace DB
{
@ -69,7 +70,7 @@ uint64_t readMetricFromStatFile(ReadBufferFromFile & buf, const std::string & ke
struct CgroupsV1Reader : ICgroupsReader
{
explicit CgroupsV1Reader(const std::filesystem::path & stat_file_dir) : buf(stat_file_dir / "memory.stat") { }
explicit CgroupsV1Reader(const fs::path & stat_file_dir) : buf(stat_file_dir / "memory.stat") { }
uint64_t readMemoryUsage() override
{
@ -85,7 +86,7 @@ private:
struct CgroupsV2Reader : ICgroupsReader
{
explicit CgroupsV2Reader(const std::filesystem::path & stat_file_dir)
explicit CgroupsV2Reader(const fs::path & stat_file_dir)
: current_buf(stat_file_dir / "memory.current"), stat_buf(stat_file_dir / "memory.stat")
{
}
@ -129,8 +130,9 @@ std::optional<std::string> getCgroupsV2Path()
if (!cgroupsV2MemoryControllerEnabled())
return {};
String cgroup = cgroupV2OfProcess();
auto current_cgroup = cgroup.empty() ? default_cgroups_mount : (default_cgroups_mount / cgroup);
fs::path current_cgroup = cgroupV2PathOfProcess();
if (current_cgroup.empty())
return {};
/// Return the bottom-most nested current memory file. If there is no such file at the current
/// level, try again at the parent level as memory settings are inherited.
@ -138,7 +140,7 @@ std::optional<std::string> getCgroupsV2Path()
{
const auto current_path = current_cgroup / "memory.current";
const auto stat_path = current_cgroup / "memory.stat";
if (std::filesystem::exists(current_path) && std::filesystem::exists(stat_path))
if (fs::exists(current_path) && fs::exists(stat_path))
return {current_cgroup};
current_cgroup = current_cgroup.parent_path();
}
@ -148,7 +150,7 @@ std::optional<std::string> getCgroupsV2Path()
std::optional<std::string> getCgroupsV1Path()
{
auto path = default_cgroups_mount / "memory/memory.stat";
if (!std::filesystem::exists(path))
if (!fs::exists(path))
return {};
return {default_cgroups_mount / "memory"};
}

View File

@ -37,12 +37,12 @@ uint32_t getCGroupLimitedCPUCores(unsigned default_cpu_count)
/// cgroupsv2
if (cgroupsV2Enabled())
{
/// First, we identify the cgroup the process belongs
std::string cgroup = cgroupV2OfProcess();
if (cgroup.empty())
/// First, we identify the path of the cgroup the process belongs
std::filesystem::path cgroup_path = cgroupV2PathOfProcess();
if (cgroup_path.empty())
return default_cpu_count;
auto current_cgroup = cgroup.empty() ? default_cgroups_mount : (default_cgroups_mount / cgroup);
auto current_cgroup = cgroup_path;
// Looking for cpu.max in directories from the current cgroup to the top level
// It does not stop on the first time since the child could have a greater value than parent
@ -62,7 +62,7 @@ uint32_t getCGroupLimitedCPUCores(unsigned default_cpu_count)
}
current_cgroup = current_cgroup.parent_path();
}
current_cgroup = default_cgroups_mount / cgroup;
current_cgroup = cgroup_path;
// Looking for cpuset.cpus.effective in directories from the current cgroup to the top level
while (current_cgroup != default_cgroups_mount.parent_path())
{

View File

@ -80,7 +80,7 @@ inline ALWAYS_INLINE void * newImpl(std::size_t size, TAlign... align)
throw std::bad_alloc{};
}
inline ALWAYS_INLINE void * newNoExept(std::size_t size) noexcept
inline ALWAYS_INLINE void * newNoExcept(std::size_t size) noexcept
{
#if USE_GWP_ASAN
if (unlikely(GWPAsan::GuardedAlloc.shouldSample()))
@ -99,7 +99,7 @@ inline ALWAYS_INLINE void * newNoExept(std::size_t size) noexcept
return malloc(size);
}
inline ALWAYS_INLINE void * newNoExept(std::size_t size, std::align_val_t align) noexcept
inline ALWAYS_INLINE void * newNoExcept(std::size_t size, std::align_val_t align) noexcept
{
#if USE_GWP_ASAN
if (unlikely(GWPAsan::GuardedAlloc.shouldSample()))

View File

@ -87,7 +87,7 @@ void * operator new(std::size_t size, const std::nothrow_t &) noexcept
{
AllocationTrace trace;
std::size_t actual_size = Memory::trackMemory(size, trace);
void * ptr = Memory::newNoExept(size);
void * ptr = Memory::newNoExcept(size);
trace.onAlloc(ptr, actual_size);
return ptr;
}
@ -96,7 +96,7 @@ void * operator new[](std::size_t size, const std::nothrow_t &) noexcept
{
AllocationTrace trace;
std::size_t actual_size = Memory::trackMemory(size, trace);
void * ptr = Memory::newNoExept(size);
void * ptr = Memory::newNoExcept(size);
trace.onAlloc(ptr, actual_size);
return ptr;
}
@ -105,7 +105,7 @@ void * operator new(std::size_t size, std::align_val_t align, const std::nothrow
{
AllocationTrace trace;
std::size_t actual_size = Memory::trackMemory(size, trace, align);
void * ptr = Memory::newNoExept(size, align);
void * ptr = Memory::newNoExcept(size, align);
trace.onAlloc(ptr, actual_size);
return ptr;
}
@ -114,7 +114,7 @@ void * operator new[](std::size_t size, std::align_val_t align, const std::nothr
{
AllocationTrace trace;
std::size_t actual_size = Memory::trackMemory(size, trace, align);
void * ptr = Memory::newNoExept(size, align);
void * ptr = Memory::newNoExcept(size, align);
trace.onAlloc(ptr, actual_size);
return ptr;
}

View File

@ -502,6 +502,7 @@ class IColumn;
M(UInt64, backup_restore_keeper_value_max_size, 1048576, "Maximum size of data of a [Zoo]Keeper's node during backup", 0) \
M(UInt64, backup_restore_batch_size_for_keeper_multiread, 10000, "Maximum size of batch for multiread request to [Zoo]Keeper during backup or restore", 0) \
M(UInt64, backup_restore_batch_size_for_keeper_multi, 1000, "Maximum size of batch for multi request to [Zoo]Keeper during backup or restore", 0) \
M(UInt64, backup_restore_s3_retry_attempts, 1000, "Setting for Aws::Client::RetryStrategy, Aws::Client does retries itself, 0 means no retries. It takes place only for backup/restore.", 0) \
M(UInt64, max_backup_bandwidth, 0, "The maximum read speed in bytes per second for particular backup on server. Zero means unlimited.", 0) \
\
M(Bool, log_profile_events, true, "Log query performance statistics into the query_log, query_thread_log and query_views_log.", 0) \

View File

@ -77,6 +77,7 @@ static std::initializer_list<std::pair<ClickHouseVersion, SettingsChangesHistory
{"azure_sdk_retry_initial_backoff_ms", 10, 10, "Minimal backoff between retries in azure sdk"},
{"azure_sdk_retry_max_backoff_ms", 1000, 1000, "Maximal backoff between retries in azure sdk"},
{"ignore_on_cluster_for_replicated_named_collections_queries", false, false, "Ignore ON CLUSTER clause for replicated named collections management queries."},
{"backup_restore_s3_retry_attempts", 1000,1000, "Setting for Aws::Client::RetryStrategy, Aws::Client does retries itself, 0 means no retries. It takes place only for backup/restore."},
{"postgresql_connection_attempt_timeout", 2, 2, "Allow to control 'connect_timeout' parameter of PostgreSQL connection."},
{"postgresql_connection_pool_retries", 2, 2, "Allow to control the number of retries in PostgreSQL connection pool."}
}},

View File

@ -162,7 +162,7 @@ public:
class RetryStrategy : public Aws::Client::RetryStrategy
{
public:
explicit RetryStrategy(uint32_t maxRetries_ = 10, uint32_t scaleFactor_ = 25, uint32_t maxDelayMs_ = 90000);
explicit RetryStrategy(uint32_t maxRetries_ = 10, uint32_t scaleFactor_ = 25, uint32_t maxDelayMs_ = 5000);
/// NOLINTNEXTLINE(google-runtime-int)
bool ShouldRetry(const Aws::Client::AWSError<Aws::Client::CoreErrors>& error, long attemptedRetries) const override;

View File

@ -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())
{

View File

@ -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());

View File

@ -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,

View File

@ -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 +65,20 @@ 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;
else 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 +92,13 @@ 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;
}
}

View File

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

View File

@ -58,14 +58,14 @@ def decompress_fast(archive_path: Path, result_path: Optional[Path] = None) -> N
archive_path,
result_path,
)
program_part = "--use-compress-program='zstd --threads=0'"
program_part = "--use-compress-program='zstd --threads=0 -d'"
elif PIGZ.exists():
logging.info(
"pigz found, will compress and decompress faster ('%s' -> '%s')",
archive_path,
result_path,
)
program_part = "--use-compress-program='pigz'"
program_part = "--use-compress-program='pigz -d'"
else:
program_part = "-z"
logging.info(

View File

@ -3386,13 +3386,13 @@ def parse_args():
parser.add_argument(
"--replace-replicated-with-shared",
action="store_true",
default=os.environ.get("USE_META_IN_KEEPER_FOR_MERGE_TREE", False),
default=os.environ.get("REPLACE_RMT_WITH_SMT", False),
help="Replace ReplicatedMergeTree engine with SharedMergeTree",
)
parser.add_argument(
"--replace-non-replicated-with-shared",
action="store_true",
default=False,
default=os.environ.get("REPLACE_MT_WITH_SMT", False),
help="Replace ordinary MergeTree engine with SharedMergeTree",
)

View File

@ -2,6 +2,7 @@
<profiles>
<default>
<s3_retry_attempts>5</s3_retry_attempts>
<backup_restore_s3_retry_attempts>5</backup_restore_s3_retry_attempts>
</default>
</profiles>
<users>

View File

@ -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", "ANY LEFT", "SEMI RIGHT"]:
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...")

View File

@ -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

View File

@ -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"

View File

@ -1,102 +1,42 @@
flat
-- { echoOn }
INSERT INTO table_for_update_field_dictionary VALUES (1, 'First', now());
SELECT key, value FROM dict_flat ORDER BY key ASC;
1 First
INSERT INTO table_for_update_field_dictionary VALUES (2, 'Second', now());
SELECT sleepEachRow(1) FROM numbers(10) SETTINGS function_sleep_max_microseconds_per_block = 10000000 FORMAT Null;
SELECT key, value FROM dict_flat ORDER BY key ASC;
1 First
2 Second
INSERT INTO table_for_update_field_dictionary VALUES (2, 'SecondUpdated', now());
INSERT INTO table_for_update_field_dictionary VALUES (3, 'Third', now());
SELECT sleepEachRow(1) FROM numbers(20) SETTINGS function_sleep_max_microseconds_per_block = 20000000 FORMAT Null;
SELECT key, value FROM dict_flat ORDER BY key ASC;
1 First
2 SecondUpdated
3 Third
flat/custom
-- { echoOn }
INSERT INTO table_for_update_field_dictionary VALUES (1, 'First', now());
SELECT key, value FROM dict_flat_custom ORDER BY key ASC;
1 First
INSERT INTO table_for_update_field_dictionary VALUES (2, 'Second', now());
SELECT sleepEachRow(1) FROM numbers(10) SETTINGS function_sleep_max_microseconds_per_block = 10000000 FORMAT Null;
SELECT key, value FROM dict_flat_custom ORDER BY key ASC;
1 First
2 Second
INSERT INTO table_for_update_field_dictionary VALUES (2, 'SecondUpdated', now());
INSERT INTO table_for_update_field_dictionary VALUES (3, 'Third', now());
SELECT sleepEachRow(1) FROM numbers(20) SETTINGS function_sleep_max_microseconds_per_block = 20000000 FORMAT Null;
SELECT key, value FROM dict_flat_custom ORDER BY key ASC;
1 First
2 SecondUpdated
3 Third
hashed
-- { echoOn }
INSERT INTO table_for_update_field_dictionary VALUES (1, 'First', now());
SELECT key, value FROM dict_hashed ORDER BY key ASC;
1 First
INSERT INTO table_for_update_field_dictionary VALUES (2, 'Second', now());
SELECT sleepEachRow(1) FROM numbers(10) SETTINGS function_sleep_max_microseconds_per_block = 10000000 FORMAT Null;
SELECT key, value FROM dict_hashed ORDER BY key ASC;
1 First
2 Second
INSERT INTO table_for_update_field_dictionary VALUES (2, 'SecondUpdated', now());
INSERT INTO table_for_update_field_dictionary VALUES (3, 'Third', now());
SELECT sleepEachRow(1) FROM numbers(20) SETTINGS function_sleep_max_microseconds_per_block = 20000000 FORMAT Null;
SELECT key, value FROM dict_hashed ORDER BY key ASC;
1 First
2 SecondUpdated
3 Third
hashed/custom
-- { echoOn }
INSERT INTO table_for_update_field_dictionary VALUES (1, 'First', now());
SELECT key, value FROM dict_hashed_custom ORDER BY key ASC;
1 First
INSERT INTO table_for_update_field_dictionary VALUES (2, 'Second', now());
SELECT sleepEachRow(1) FROM numbers(10) SETTINGS function_sleep_max_microseconds_per_block = 10000000 FORMAT Null;
SELECT key, value FROM dict_hashed_custom ORDER BY key ASC;
1 First
2 Second
INSERT INTO table_for_update_field_dictionary VALUES (2, 'SecondUpdated', now());
INSERT INTO table_for_update_field_dictionary VALUES (3, 'Third', now());
SELECT sleepEachRow(1) FROM numbers(20) SETTINGS function_sleep_max_microseconds_per_block = 20000000 FORMAT Null;
SELECT key, value FROM dict_hashed_custom ORDER BY key ASC;
1 First
2 SecondUpdated
3 Third
complex_key_hashed
-- { echoOn }
INSERT INTO table_for_update_field_dictionary VALUES (1, 'First', now());
SELECT key, value FROM dict_complex_key_hashed ORDER BY key ASC;
1 First
INSERT INTO table_for_update_field_dictionary VALUES (2, 'Second', now());
SELECT sleepEachRow(1) FROM numbers(10) SETTINGS function_sleep_max_microseconds_per_block = 10000000 FORMAT Null;
SELECT key, value FROM dict_complex_key_hashed ORDER BY key ASC;
1 First
2 Second
INSERT INTO table_for_update_field_dictionary VALUES (2, 'SecondUpdated', now());
INSERT INTO table_for_update_field_dictionary VALUES (3, 'Third', now());
SELECT sleepEachRow(1) FROM numbers(20) SETTINGS function_sleep_max_microseconds_per_block = 20000000 FORMAT Null;
SELECT key, value FROM dict_complex_key_hashed ORDER BY key ASC;
1 First
2 SecondUpdated
3 Third
complex_key_hashed/custom
-- { echoOn }
INSERT INTO table_for_update_field_dictionary VALUES (1, 'First', now());
SELECT key, value FROM dict_complex_key_hashed_custom ORDER BY key ASC;
1 First
INSERT INTO table_for_update_field_dictionary VALUES (2, 'Second', now());
SELECT sleepEachRow(1) FROM numbers(10) SETTINGS function_sleep_max_microseconds_per_block = 10000000 FORMAT Null;
SELECT key, value FROM dict_complex_key_hashed_custom ORDER BY key ASC;
1 First
2 Second
INSERT INTO table_for_update_field_dictionary VALUES (2, 'SecondUpdated', now());
INSERT INTO table_for_update_field_dictionary VALUES (3, 'Third', now());
SELECT sleepEachRow(1) FROM numbers(20) SETTINGS function_sleep_max_microseconds_per_block = 20000000 FORMAT Null;
SELECT key, value FROM dict_complex_key_hashed_custom ORDER BY key ASC;
1 First
2 SecondUpdated
3 Third

View File

@ -35,7 +35,7 @@ for layout in "${layouts[@]}"; do
echo "$layout"
fi
$CLICKHOUSE_CLIENT -nm -q "
$CLICKHOUSE_CLIENT --multiquery "
TRUNCATE TABLE table_for_update_field_dictionary;
CREATE DICTIONARY $dictionary_name
@ -49,24 +49,31 @@ for layout in "${layouts[@]}"; do
LAYOUT($layout())
LIFETIME(1);
-- { echoOn }
INSERT INTO table_for_update_field_dictionary VALUES (1, 'First', now());
SELECT key, value FROM $dictionary_name ORDER BY key ASC;
INSERT INTO table_for_update_field_dictionary VALUES (1, 'First', now());"
INSERT INTO table_for_update_field_dictionary VALUES (2, 'Second', now());
SELECT sleepEachRow(1) FROM numbers(10) SETTINGS function_sleep_max_microseconds_per_block = 10000000 FORMAT Null;
while true
do
$CLICKHOUSE_CLIENT --query "SELECT key, value FROM $dictionary_name ORDER BY key ASC" | grep -A10 -B10 'First' && break;
sleep .1;
done
SELECT key, value FROM $dictionary_name ORDER BY key ASC;
$CLICKHOUSE_CLIENT --query "INSERT INTO table_for_update_field_dictionary VALUES (2, 'Second', now());"
INSERT INTO table_for_update_field_dictionary VALUES (2, 'SecondUpdated', now());
INSERT INTO table_for_update_field_dictionary VALUES (3, 'Third', now());
SELECT sleepEachRow(1) FROM numbers(20) SETTINGS function_sleep_max_microseconds_per_block = 20000000 FORMAT Null;
while true
do
$CLICKHOUSE_CLIENT --query "SELECT key, value FROM $dictionary_name ORDER BY key ASC" | grep -A10 -B10 'Second' && break;
sleep .1;
done
SELECT key, value FROM $dictionary_name ORDER BY key ASC;
-- { echoOff }
$CLICKHOUSE_CLIENT --query "INSERT INTO table_for_update_field_dictionary VALUES (2, 'SecondUpdated', now()), (3, 'Third', now())"
DROP DICTIONARY $dictionary_name;
"
while true
do
$CLICKHOUSE_CLIENT --query "SELECT key, value FROM $dictionary_name ORDER BY key ASC" | grep -A10 -B10 'SecondUpdated' && break;
sleep .1;
done
$CLICKHOUSE_CLIENT --query "DROP DICTIONARY $dictionary_name"
done
done

View File

@ -264,3 +264,44 @@ SETTINGS group_by_use_nulls = 1, max_bytes_before_external_sort=10;
9 \N 9
\N 0 20
\N 1 25
CREATE TABLE test
ENGINE = ReplacingMergeTree
PRIMARY KEY id
AS SELECT number AS id FROM numbers(100);
SELECT id
FROM test
GROUP BY id
WITH CUBE
HAVING id IN (
SELECT id
FROM test
)
FORMAT `NUll`
SETTINGS allow_experimental_analyzer = 1, group_by_use_nulls = true;
SELECT id
FROM test
FINAL
GROUP BY id
WITH CUBE
HAVING id IN (
SELECT DISTINCT id
FROM test
FINAL
)
FORMAT `NUll`
SETTINGS allow_experimental_analyzer = 1, group_by_use_nulls = true;
SELECT id
FROM test
FINAL
GROUP BY
GROUPING SETS ((id))
ORDER BY
id IN (
SELECT DISTINCT id
FROM test
FINAL
LIMIT 4
) ASC
LIMIT 256 BY id
FORMAT `NUll`
SETTINGS allow_experimental_analyzer = 1, group_by_use_nulls=true;

View File

@ -83,3 +83,48 @@ GROUP BY
)
ORDER BY 1, tuple(val)
SETTINGS group_by_use_nulls = 1, max_bytes_before_external_sort=10;
CREATE TABLE test
ENGINE = ReplacingMergeTree
PRIMARY KEY id
AS SELECT number AS id FROM numbers(100);
SELECT id
FROM test
GROUP BY id
WITH CUBE
HAVING id IN (
SELECT id
FROM test
)
FORMAT `NUll`
SETTINGS allow_experimental_analyzer = 1, group_by_use_nulls = true;
SELECT id
FROM test
FINAL
GROUP BY id
WITH CUBE
HAVING id IN (
SELECT DISTINCT id
FROM test
FINAL
)
FORMAT `NUll`
SETTINGS allow_experimental_analyzer = 1, group_by_use_nulls = true;
SELECT id
FROM test
FINAL
GROUP BY
GROUPING SETS ((id))
ORDER BY
id IN (
SELECT DISTINCT id
FROM test
FINAL
LIMIT 4
) ASC
LIMIT 256 BY id
FORMAT `NUll`
SETTINGS allow_experimental_analyzer = 1, group_by_use_nulls=true;

View File

@ -1,3 +0,0 @@
select * from remote('127.2', view(select sleep(3) from system.one)) settings receive_timeout=1, async_socket_for_remote=0, use_hedged_requests=1 format Null;
select * from remote('127.2', view(select sleep(3) from system.one)) settings receive_timeout=1, async_socket_for_remote=1, use_hedged_requests=0 format Null;
select * from remote('127.2', view(select sleep(3) from system.one)) settings receive_timeout=1, async_socket_for_remote=0, use_hedged_requests=0 format Null;

View File

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

View File

@ -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

View File

@ -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;

View File

@ -0,0 +1,2 @@
2020-01-01
2020-01-02

View File

@ -0,0 +1,12 @@
SET allow_experimental_analyzer=1;
CREATE TABLE foo (i Date) ENGINE MergeTree ORDER BY i;
INSERT INTO foo VALUES ('2020-01-01');
INSERT INTO foo VALUES ('2020-01-02');
SET optimize_min_equality_disjunction_chain_length = 3;
SELECT *
FROM foo
WHERE (foo.i = parseDateTimeBestEffort('2020-01-01'))
OR (foo.i = parseDateTimeBestEffort('2020-01-02'))
OR (foo.i = parseDateTimeBestEffort('2020-01-03'))
ORDER BY foo.i ASC

View File

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

View File

@ -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;