diff --git a/contrib/azure b/contrib/azure index 92c94d7f37a..ea3e19a7be0 160000 --- a/contrib/azure +++ b/contrib/azure @@ -1 +1 @@ -Subproject commit 92c94d7f37a43cc8fc4d466884a95f610c0593bf +Subproject commit ea3e19a7be08519134c643177d56c7484dfec884 diff --git a/contrib/pocketfft b/contrib/pocketfft index 9efd4da52cf..f4c1aa8aa9c 160000 --- a/contrib/pocketfft +++ b/contrib/pocketfft @@ -1 +1 @@ -Subproject commit 9efd4da52cf8d28d14531d14e43ad9d913807546 +Subproject commit f4c1aa8aa9ce79ad39e80f2c9c41b92ead90fda3 diff --git a/contrib/rocksdb b/contrib/rocksdb index 078fa563869..be366233921 160000 --- a/contrib/rocksdb +++ b/contrib/rocksdb @@ -1 +1 @@ -Subproject commit 078fa5638690004e1f744076d1bdcc4e93767304 +Subproject commit be366233921293bd07a84dc4ea6991858665f202 diff --git a/contrib/rocksdb-cmake/CMakeLists.txt b/contrib/rocksdb-cmake/CMakeLists.txt index 943e1d8acbd..3a14407166c 100644 --- a/contrib/rocksdb-cmake/CMakeLists.txt +++ b/contrib/rocksdb-cmake/CMakeLists.txt @@ -5,20 +5,13 @@ if (NOT ENABLE_ROCKSDB) return() endif() -## this file is extracted from `contrib/rocksdb/CMakeLists.txt` -set(ROCKSDB_SOURCE_DIR "${ClickHouse_SOURCE_DIR}/contrib/rocksdb") -list(APPEND CMAKE_MODULE_PATH "${ROCKSDB_SOURCE_DIR}/cmake/modules/") - -set(PORTABLE ON) -## always disable jemalloc for rocksdb by default -## because it introduces non-standard jemalloc APIs +# Always disable jemalloc for rocksdb by default because it introduces non-standard jemalloc APIs option(WITH_JEMALLOC "build with JeMalloc" OFF) -set(USE_SNAPPY OFF) -if (TARGET ch_contrib::snappy) - set(USE_SNAPPY ON) -endif() -option(WITH_SNAPPY "build with SNAPPY" ${USE_SNAPPY}) -## lz4, zlib, zstd is enabled in ClickHouse by default + +option(WITH_LIBURING "build with liburing" OFF) # TODO could try to enable this conditionally, depending on ClickHouse's ENABLE_LIBURING + +# ClickHouse cannot be compiled without snappy, lz4, zlib, zstd +option(WITH_SNAPPY "build with SNAPPY" ON) option(WITH_LZ4 "build with lz4" ON) option(WITH_ZLIB "build with zlib" ON) option(WITH_ZSTD "build with zstd" ON) @@ -26,78 +19,46 @@ option(WITH_ZSTD "build with zstd" ON) # third-party/folly is only validated to work on Linux and Windows for now. # So only turn it on there by default. if(CMAKE_SYSTEM_NAME MATCHES "Linux|Windows") - if(MSVC AND MSVC_VERSION LESS 1910) - # Folly does not compile with MSVC older than VS2017 - option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" OFF) - else() - option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" ON) - endif() + option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" ON) else() option(WITH_FOLLY_DISTRIBUTED_MUTEX "build with folly::DistributedMutex" OFF) endif() -if( NOT DEFINED CMAKE_CXX_STANDARD ) - set(CMAKE_CXX_STANDARD 11) +if(WITH_SNAPPY) + add_definitions(-DSNAPPY) + list(APPEND THIRDPARTY_LIBS ch_contrib::snappy) endif() -if(MSVC) - option(WITH_XPRESS "build with windows built in compression" OFF) - include("${ROCKSDB_SOURCE_DIR}/thirdparty.inc") -else() - if(CMAKE_SYSTEM_NAME MATCHES "FreeBSD" AND NOT CMAKE_SYSTEM_NAME MATCHES "kFreeBSD") - # FreeBSD has jemalloc as default malloc - # but it does not have all the jemalloc files in include/... - set(WITH_JEMALLOC ON) - else() - if(WITH_JEMALLOC AND TARGET ch_contrib::jemalloc) - add_definitions(-DROCKSDB_JEMALLOC -DJEMALLOC_NO_DEMANGLE) - list(APPEND THIRDPARTY_LIBS ch_contrib::jemalloc) - endif() - endif() - - if(WITH_SNAPPY) - add_definitions(-DSNAPPY) - list(APPEND THIRDPARTY_LIBS ch_contrib::snappy) - endif() - - if(WITH_ZLIB) - add_definitions(-DZLIB) - list(APPEND THIRDPARTY_LIBS ch_contrib::zlib) - endif() - - if(WITH_LZ4) - add_definitions(-DLZ4) - list(APPEND THIRDPARTY_LIBS ch_contrib::lz4) - endif() - - if(WITH_ZSTD) - add_definitions(-DZSTD) - list(APPEND THIRDPARTY_LIBS ch_contrib::zstd) - endif() +if(WITH_ZLIB) + add_definitions(-DZLIB) + list(APPEND THIRDPARTY_LIBS ch_contrib::zlib) endif() -if(CMAKE_SYSTEM_PROCESSOR MATCHES "^(powerpc|ppc)64") - if(POWER9) - set(HAS_POWER9 1) - set(HAS_ALTIVEC 1) - else() - set(HAS_POWER8 1) - set(HAS_ALTIVEC 1) - endif(POWER9) -endif(CMAKE_SYSTEM_PROCESSOR MATCHES "^(powerpc|ppc)64") +if(WITH_LZ4) + add_definitions(-DLZ4) + list(APPEND THIRDPARTY_LIBS ch_contrib::lz4) +endif() -if(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64|AARCH64|arm64|ARM64") - set(HAS_ARMV8_CRC 1) - set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=armv8-a+crc+crypto -Wno-unused-function") - set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=armv8-a+crc+crypto -Wno-unused-function") -endif(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64|AARCH64|arm64|ARM64") +if(WITH_ZSTD) + add_definitions(-DZSTD) + list(APPEND THIRDPARTY_LIBS ch_contrib::zstd) +endif() +option(PORTABLE "build a portable binary" ON) -if(ENABLE_AVX2 AND ENABLE_PCLMULQDQ) +if(ENABLE_SSE42 AND ENABLE_PCLMULQDQ) add_definitions(-DHAVE_SSE42) add_definitions(-DHAVE_PCLMUL) endif() +if(CMAKE_SYSTEM_PROCESSOR MATCHES "arm64|aarch64|AARCH64") + set (HAS_ARMV8_CRC 1) + # the original build descriptions set specific flags for ARM. These flags are already subsumed by ClickHouse's general + # ARM flags, see cmake/cpu_features.cmake + # set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=armv8-a+crc+crypto -Wno-unused-function") + # set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -march=armv8-a+crc+crypto -Wno-unused-function") +endif() + set (HAVE_THREAD_LOCAL 1) if(HAVE_THREAD_LOCAL) add_definitions(-DROCKSDB_SUPPORT_THREAD_LOCAL) @@ -107,8 +68,6 @@ if(CMAKE_SYSTEM_NAME MATCHES "Darwin") add_definitions(-DOS_MACOSX) elseif(CMAKE_SYSTEM_NAME MATCHES "Linux") add_definitions(-DOS_LINUX) -elseif(CMAKE_SYSTEM_NAME MATCHES "SunOS") - add_definitions(-DOS_SOLARIS) elseif(CMAKE_SYSTEM_NAME MATCHES "FreeBSD") add_definitions(-DOS_FREEBSD) elseif(CMAKE_SYSTEM_NAME MATCHES "Android") @@ -123,12 +82,10 @@ endif() if (OS_LINUX) add_definitions(-DROCKSDB_SCHED_GETCPU_PRESENT) - add_definitions(-DROCKSDB_AUXV_SYSAUXV_PRESENT) add_definitions(-DROCKSDB_AUXV_GETAUXVAL_PRESENT) -elseif (OS_FREEBSD) - add_definitions(-DROCKSDB_AUXV_SYSAUXV_PRESENT) endif() +set(ROCKSDB_SOURCE_DIR "${ClickHouse_SOURCE_DIR}/contrib/rocksdb") include_directories(${ROCKSDB_SOURCE_DIR}) include_directories("${ROCKSDB_SOURCE_DIR}/include") @@ -136,11 +93,11 @@ if(WITH_FOLLY_DISTRIBUTED_MUTEX) include_directories("${ROCKSDB_SOURCE_DIR}/third-party/folly") endif() -# Main library source code - set(SOURCES ${ROCKSDB_SOURCE_DIR}/cache/cache.cc ${ROCKSDB_SOURCE_DIR}/cache/cache_entry_roles.cc + ${ROCKSDB_SOURCE_DIR}/cache/cache_key.cc + ${ROCKSDB_SOURCE_DIR}/cache/cache_reservation_manager.cc ${ROCKSDB_SOURCE_DIR}/cache/clock_cache.cc ${ROCKSDB_SOURCE_DIR}/cache/lru_cache.cc ${ROCKSDB_SOURCE_DIR}/cache/sharded_cache.cc @@ -156,6 +113,7 @@ set(SOURCES ${ROCKSDB_SOURCE_DIR}/db/blob/blob_log_format.cc ${ROCKSDB_SOURCE_DIR}/db/blob/blob_log_sequential_reader.cc ${ROCKSDB_SOURCE_DIR}/db/blob/blob_log_writer.cc + ${ROCKSDB_SOURCE_DIR}/db/blob/prefetch_buffer_collection.cc ${ROCKSDB_SOURCE_DIR}/db/builder.cc ${ROCKSDB_SOURCE_DIR}/db/c.cc ${ROCKSDB_SOURCE_DIR}/db/column_family.cc @@ -229,6 +187,7 @@ set(SOURCES ${ROCKSDB_SOURCE_DIR}/env/file_system_tracer.cc ${ROCKSDB_SOURCE_DIR}/env/fs_remap.cc ${ROCKSDB_SOURCE_DIR}/env/mock_env.cc + ${ROCKSDB_SOURCE_DIR}/env/unique_id_gen.cc ${ROCKSDB_SOURCE_DIR}/file/delete_scheduler.cc ${ROCKSDB_SOURCE_DIR}/file/file_prefetch_buffer.cc ${ROCKSDB_SOURCE_DIR}/file/file_util.cc @@ -247,6 +206,7 @@ set(SOURCES ${ROCKSDB_SOURCE_DIR}/memory/concurrent_arena.cc ${ROCKSDB_SOURCE_DIR}/memory/jemalloc_nodump_allocator.cc ${ROCKSDB_SOURCE_DIR}/memory/memkind_kmem_allocator.cc + ${ROCKSDB_SOURCE_DIR}/memory/memory_allocator.cc ${ROCKSDB_SOURCE_DIR}/memtable/alloc_tracker.cc ${ROCKSDB_SOURCE_DIR}/memtable/hash_linklist_rep.cc ${ROCKSDB_SOURCE_DIR}/memtable/hash_skiplist_rep.cc @@ -322,6 +282,7 @@ set(SOURCES ${ROCKSDB_SOURCE_DIR}/table/table_factory.cc ${ROCKSDB_SOURCE_DIR}/table/table_properties.cc ${ROCKSDB_SOURCE_DIR}/table/two_level_iterator.cc + ${ROCKSDB_SOURCE_DIR}/table/unique_id.cc ${ROCKSDB_SOURCE_DIR}/test_util/sync_point.cc ${ROCKSDB_SOURCE_DIR}/test_util/sync_point_impl.cc ${ROCKSDB_SOURCE_DIR}/test_util/testutil.cc @@ -333,9 +294,12 @@ set(SOURCES ${ROCKSDB_SOURCE_DIR}/tools/ldb_tool.cc ${ROCKSDB_SOURCE_DIR}/tools/sst_dump_tool.cc ${ROCKSDB_SOURCE_DIR}/tools/trace_analyzer_tool.cc - ${ROCKSDB_SOURCE_DIR}/trace_replay/trace_replay.cc ${ROCKSDB_SOURCE_DIR}/trace_replay/block_cache_tracer.cc ${ROCKSDB_SOURCE_DIR}/trace_replay/io_tracer.cc + ${ROCKSDB_SOURCE_DIR}/trace_replay/trace_record_handler.cc + ${ROCKSDB_SOURCE_DIR}/trace_replay/trace_record_result.cc + ${ROCKSDB_SOURCE_DIR}/trace_replay/trace_record.cc + ${ROCKSDB_SOURCE_DIR}/trace_replay/trace_replay.cc ${ROCKSDB_SOURCE_DIR}/util/coding.cc ${ROCKSDB_SOURCE_DIR}/util/compaction_job_stats_impl.cc ${ROCKSDB_SOURCE_DIR}/util/comparator.cc @@ -347,6 +311,7 @@ set(SOURCES ${ROCKSDB_SOURCE_DIR}/util/murmurhash.cc ${ROCKSDB_SOURCE_DIR}/util/random.cc ${ROCKSDB_SOURCE_DIR}/util/rate_limiter.cc + ${ROCKSDB_SOURCE_DIR}/util/regex.cc ${ROCKSDB_SOURCE_DIR}/util/ribbon_config.cc ${ROCKSDB_SOURCE_DIR}/util/slice.cc ${ROCKSDB_SOURCE_DIR}/util/file_checksum_helper.cc @@ -362,18 +327,23 @@ set(SOURCES ${ROCKSDB_SOURCE_DIR}/utilities/blob_db/blob_db_impl_filesnapshot.cc ${ROCKSDB_SOURCE_DIR}/utilities/blob_db/blob_dump_tool.cc ${ROCKSDB_SOURCE_DIR}/utilities/blob_db/blob_file.cc + ${ROCKSDB_SOURCE_DIR}/utilities/cache_dump_load.cc + ${ROCKSDB_SOURCE_DIR}/utilities/cache_dump_load_impl.cc ${ROCKSDB_SOURCE_DIR}/utilities/cassandra/cassandra_compaction_filter.cc ${ROCKSDB_SOURCE_DIR}/utilities/cassandra/format.cc ${ROCKSDB_SOURCE_DIR}/utilities/cassandra/merge_operator.cc ${ROCKSDB_SOURCE_DIR}/utilities/checkpoint/checkpoint_impl.cc + ${ROCKSDB_SOURCE_DIR}/utilities/compaction_filters.cc ${ROCKSDB_SOURCE_DIR}/utilities/compaction_filters/remove_emptyvalue_compactionfilter.cc ${ROCKSDB_SOURCE_DIR}/utilities/debug.cc ${ROCKSDB_SOURCE_DIR}/utilities/env_mirror.cc ${ROCKSDB_SOURCE_DIR}/utilities/env_timed.cc ${ROCKSDB_SOURCE_DIR}/utilities/fault_injection_env.cc ${ROCKSDB_SOURCE_DIR}/utilities/fault_injection_fs.cc + ${ROCKSDB_SOURCE_DIR}/utilities/fault_injection_secondary_cache.cc ${ROCKSDB_SOURCE_DIR}/utilities/leveldb_options/leveldb_options.cc ${ROCKSDB_SOURCE_DIR}/utilities/memory/memory_util.cc + ${ROCKSDB_SOURCE_DIR}/utilities/merge_operators.cc ${ROCKSDB_SOURCE_DIR}/utilities/merge_operators/bytesxor.cc ${ROCKSDB_SOURCE_DIR}/utilities/merge_operators/max.cc ${ROCKSDB_SOURCE_DIR}/utilities/merge_operators/put.cc @@ -393,6 +363,7 @@ set(SOURCES ${ROCKSDB_SOURCE_DIR}/utilities/simulator_cache/sim_cache.cc ${ROCKSDB_SOURCE_DIR}/utilities/table_properties_collectors/compact_on_deletion_collector.cc ${ROCKSDB_SOURCE_DIR}/utilities/trace/file_trace_reader_writer.cc + ${ROCKSDB_SOURCE_DIR}/utilities/trace/replayer_impl.cc ${ROCKSDB_SOURCE_DIR}/utilities/transactions/lock/lock_manager.cc ${ROCKSDB_SOURCE_DIR}/utilities/transactions/lock/point/point_lock_tracker.cc ${ROCKSDB_SOURCE_DIR}/utilities/transactions/lock/point/point_lock_manager.cc @@ -411,6 +382,7 @@ set(SOURCES ${ROCKSDB_SOURCE_DIR}/utilities/transactions/write_unprepared_txn.cc ${ROCKSDB_SOURCE_DIR}/utilities/transactions/write_unprepared_txn_db.cc ${ROCKSDB_SOURCE_DIR}/utilities/ttl/db_ttl_impl.cc + ${ROCKSDB_SOURCE_DIR}/utilities/wal_filter.cc ${ROCKSDB_SOURCE_DIR}/utilities/write_batch_with_index/write_batch_with_index.cc ${ROCKSDB_SOURCE_DIR}/utilities/write_batch_with_index/write_batch_with_index_internal.cc ${ROCKSDB_SOURCE_DIR}/utilities/transactions/lock/range/range_tree/lib/locktree/concurrent_tree.cc @@ -425,7 +397,7 @@ set(SOURCES ${ROCKSDB_SOURCE_DIR}/utilities/transactions/lock/range/range_tree/lib/standalone_port.cc ${ROCKSDB_SOURCE_DIR}/utilities/transactions/lock/range/range_tree/lib/util/dbt.cc ${ROCKSDB_SOURCE_DIR}/utilities/transactions/lock/range/range_tree/lib/util/memarena.cc - rocksdb_build_version.cc) + build_version.cc) # generated by hand if(ENABLE_SSE42 AND ENABLE_PCLMULQDQ) set_source_files_properties( @@ -462,5 +434,6 @@ endif() add_library(_rocksdb ${SOURCES}) add_library(ch_contrib::rocksdb ALIAS _rocksdb) target_link_libraries(_rocksdb PRIVATE ${THIRDPARTY_LIBS} ${SYSTEM_LIBS}) + # SYSTEM is required to overcome some issues target_include_directories(_rocksdb SYSTEM BEFORE INTERFACE "${ROCKSDB_SOURCE_DIR}/include") diff --git a/contrib/rocksdb-cmake/rocksdb_build_version.cc b/contrib/rocksdb-cmake/build_version.cc similarity index 100% rename from contrib/rocksdb-cmake/rocksdb_build_version.cc rename to contrib/rocksdb-cmake/build_version.cc diff --git a/docker/test/stateless/run.sh b/docker/test/stateless/run.sh index bc34c45a3e4..5f2cb95de75 100755 --- a/docker/test/stateless/run.sh +++ b/docker/test/stateless/run.sh @@ -309,7 +309,7 @@ function run_tests() try_run_with_retry 10 clickhouse-client -q "insert into system.zookeeper (name, path, value) values ('auxiliary_zookeeper2', '/test/chroot/', '')" set +e - clickhouse-test --testname --shard --zookeeper --check-zookeeper-session --hung-check --print-time \ + timeout -k 60m -s TERM --preserve-status 140m clickhouse-test --testname --shard --zookeeper --check-zookeeper-session --hung-check --print-time \ --no-drop-if-fail --test-runs "$NUM_TRIES" "${ADDITIONAL_OPTIONS[@]}" 2>&1 \ | ts '%Y-%m-%d %H:%M:%S' \ | tee -a test_output/test_result.txt diff --git a/docs/en/operations/utilities/clickhouse-local.md b/docs/en/operations/utilities/clickhouse-local.md index f19643a3fa5..c20e4fc3b09 100644 --- a/docs/en/operations/utilities/clickhouse-local.md +++ b/docs/en/operations/utilities/clickhouse-local.md @@ -16,7 +16,7 @@ sidebar_label: clickhouse-local While `clickhouse-local` is a great tool for development and testing purposes, and for processing files, it is not suitable for serving end users or applications. In these scenarios, it is recommended to use the open-source [ClickHouse](https://clickhouse.com/docs/en/install). ClickHouse is a powerful OLAP database that is designed to handle large-scale analytical workloads. It provides fast and efficient processing of complex queries on large datasets, making it ideal for use in production environments where high-performance is critical. Additionally, ClickHouse offers a wide range of features such as replication, sharding, and high availability, which are essential for scaling up to handle large datasets and serving applications. If you need to handle larger datasets or serve end users or applications, we recommend using open-source ClickHouse instead of `clickhouse-local`. -Please read the docs below that show example use cases for `clickhouse-local`, such as [querying local CSVs](#query-data-in-a-csv-file-using-sql) or [reading a parquet file in S3](#query-data-in-a-parquet-file-in-aws-s3). +Please read the docs below that show example use cases for `clickhouse-local`, such as [querying local file](#query_data_in_file) or [reading a parquet file in S3](#query-data-in-a-parquet-file-in-aws-s3). ## Download clickhouse-local diff --git a/docs/en/sql-reference/aggregate-functions/index.md b/docs/en/sql-reference/aggregate-functions/index.md index 96bf0c5d93b..5056ef2c7aa 100644 --- a/docs/en/sql-reference/aggregate-functions/index.md +++ b/docs/en/sql-reference/aggregate-functions/index.md @@ -18,7 +18,7 @@ ClickHouse also supports: During aggregation, all `NULL` arguments are skipped. If the aggregation has several arguments it will ignore any row in which one or more of them are NULL. -There is an exception to this rule, which are the functions [`first_value`](../../sql-reference/aggregate-functions/reference/first_value.md), [`last_value`](../../sql-reference/aggregate-functions/reference/last_value.md) and their aliases when followed by the modifier `RESPECT NULLS`: `FIRST_VALUE(b) RESPECT NULLS`. +There is an exception to this rule, which are the functions [`first_value`](../../sql-reference/aggregate-functions/reference/first_value.md), [`last_value`](../../sql-reference/aggregate-functions/reference/last_value.md) and their aliases (`any` and `anyLast` respectively) when followed by the modifier `RESPECT NULLS`. For example, `FIRST_VALUE(b) RESPECT NULLS`. **Examples:** diff --git a/docs/en/sql-reference/aggregate-functions/reference/any.md b/docs/en/sql-reference/aggregate-functions/reference/any.md index cdff7dde4a9..972263585f2 100644 --- a/docs/en/sql-reference/aggregate-functions/reference/any.md +++ b/docs/en/sql-reference/aggregate-functions/reference/any.md @@ -5,12 +5,12 @@ sidebar_position: 102 # any -Selects the first encountered value of a column. +Selects the first encountered value of a column, ignoring any `NULL` values. **Syntax** ```sql -any(column) +any(column) [RESPECT NULLS] ``` Aliases: `any_value`, [`first_value`](../reference/first_value.md). @@ -20,7 +20,9 @@ Aliases: `any_value`, [`first_value`](../reference/first_value.md). **Returned value** -By default, it ignores NULL values and returns the first NOT NULL value found in the column. Like [`first_value`](../../../sql-reference/aggregate-functions/reference/first_value.md) it supports `RESPECT NULLS`, in which case it will select the first value passed, independently on whether it's NULL or not. +:::note +Supports the `RESPECT NULLS` modifier after the function name. Using this modifier will ensure the function selects the first value passed, regardless of whether it is `NULL` or not. +::: :::note The return type of the function is the same as the input, except for LowCardinality which is discarded. This means that given no rows as input it will return the default value of that type (0 for integers, or Null for a Nullable() column). You might use the `-OrNull` [combinator](../../../sql-reference/aggregate-functions/combinators.md) ) to modify this behaviour. diff --git a/docs/en/sql-reference/aggregate-functions/reference/any_respect_nulls.md b/docs/en/sql-reference/aggregate-functions/reference/any_respect_nulls.md deleted file mode 100644 index 99104a9b8c7..00000000000 --- a/docs/en/sql-reference/aggregate-functions/reference/any_respect_nulls.md +++ /dev/null @@ -1,44 +0,0 @@ ---- -slug: /en/sql-reference/aggregate-functions/reference/any_respect_nulls -sidebar_position: 103 ---- - -# any_respect_nulls - -Selects the first encountered value of a column, irregardless of whether it is a `NULL` value or not. - -Alias: `any_value_respect_nulls`, `first_value_repect_nulls`. - -**Syntax** - -```sql -any_respect_nulls(column) -``` - -**Parameters** -- `column`: The column name. - -**Returned value** - -- The last value encountered, irregardless of whether it is a `NULL` value or not. - -**Example** - -Query: - -```sql -CREATE TABLE any_nulls (city Nullable(String)) ENGINE=Log; - -INSERT INTO any_nulls (city) VALUES (NULL), ('Amsterdam'), ('New York'), ('Tokyo'), ('Valencia'), (NULL); - -SELECT any(city), any_respect_nulls(city) FROM any_nulls; -``` - -```response -┌─any(city)─┬─any_respect_nulls(city)─┐ -│ Amsterdam │ ᴺᵁᴸᴸ │ -└───────────┴─────────────────────────┘ -``` - -**See Also** -- [any](../reference/any.md) diff --git a/docs/en/sql-reference/aggregate-functions/reference/anylast.md b/docs/en/sql-reference/aggregate-functions/reference/anylast.md index e43bc07fbdc..202d2e9fb10 100644 --- a/docs/en/sql-reference/aggregate-functions/reference/anylast.md +++ b/docs/en/sql-reference/aggregate-functions/reference/anylast.md @@ -5,17 +5,21 @@ sidebar_position: 105 # anyLast -Selects the last value encountered. The result is just as indeterminate as for the [any](../../../sql-reference/aggregate-functions/reference/any.md) function. +Selects the last value encountered, ignoring any `NULL` values by default. The result is just as indeterminate as for the [any](../../../sql-reference/aggregate-functions/reference/any.md) function. **Syntax** ```sql -anyLast(column) +anyLast(column) [RESPECT NULLS] ``` **Parameters** - `column`: The column name. +:::note +Supports the `RESPECT NULLS` modifier after the function name. Using this modifier will ensure the function selects the first value passed, regardless of whether it is `NULL` or not. +::: + **Returned value** - The last value encountered. diff --git a/docs/en/sql-reference/aggregate-functions/reference/anylast_respect_nulls.md b/docs/en/sql-reference/aggregate-functions/reference/anylast_respect_nulls.md deleted file mode 100644 index 8f093cfdb61..00000000000 --- a/docs/en/sql-reference/aggregate-functions/reference/anylast_respect_nulls.md +++ /dev/null @@ -1,39 +0,0 @@ ---- -slug: /en/sql-reference/aggregate-functions/reference/anylast_respect_nulls -sidebar_position: 106 ---- - -# anyLast_respect_nulls - -Selects the last value encountered, irregardless of whether it is `NULL` or not. - -**Syntax** - -```sql -anyLast_respect_nulls(column) -``` - -**Parameters** -- `column`: The column name. - -**Returned value** - -- The last value encountered, irregardless of whether it is `NULL` or not. - -**Example** - -Query: - -```sql -CREATE TABLE any_last_nulls (city Nullable(String)) ENGINE=Log; - -INSERT INTO any_last_nulls (city) VALUES ('Amsterdam'),(NULL),('New York'),('Tokyo'),('Valencia'),(NULL); - -SELECT anyLast(city), anyLast_respect_nulls(city) FROM any_last_nulls; -``` - -```response -┌─anyLast(city)─┬─anyLast_respect_nulls(city)─┐ -│ Valencia │ ᴺᵁᴸᴸ │ -└───────────────┴─────────────────────────────┘ -``` \ No newline at end of file diff --git a/docs/en/sql-reference/aggregate-functions/reference/index.md b/docs/en/sql-reference/aggregate-functions/reference/index.md index b0e5582bd87..2dce0afe2e1 100644 --- a/docs/en/sql-reference/aggregate-functions/reference/index.md +++ b/docs/en/sql-reference/aggregate-functions/reference/index.md @@ -45,10 +45,9 @@ ClickHouse-specific aggregate functions: - [aggThrow](../reference/aggthrow.md) - [analysisOfVariance](../reference/analysis_of_variance.md) -- [any](../reference/any_respect_nulls.md) +- [any](../reference/any.md) - [anyHeavy](../reference/anyheavy.md) - [anyLast](../reference/anylast.md) -- [anyLast](../reference/anylast_respect_nulls.md) - [boundingRatio](../reference/boundrat.md) - [first_value](../reference/first_value.md) - [last_value](../reference/last_value.md) diff --git a/docs/en/sql-reference/data-types/dynamic.md b/docs/en/sql-reference/data-types/dynamic.md index f45feb9fe5e..8be81471377 100644 --- a/docs/en/sql-reference/data-types/dynamic.md +++ b/docs/en/sql-reference/data-types/dynamic.md @@ -1,6 +1,6 @@ --- slug: /en/sql-reference/data-types/dynamic -sidebar_position: 56 +sidebar_position: 62 sidebar_label: Dynamic --- @@ -494,13 +494,43 @@ SELECT count(), dynamicType(d), _part FROM test GROUP BY _part, dynamicType(d) O As we can see, ClickHouse kept the most frequent types `UInt64` and `Array(UInt64)` and casted all other types to `String`. +## JSONExtract functions with Dynamic + +All `JSONExtract*` functions support `Dynamic` type: + +```sql +SELECT JSONExtract('{"a" : [1, 2, 3]}', 'a', 'Dynamic') AS dynamic, dynamicType(dynamic) AS dynamic_type; +``` + +```text +┌─dynamic─┬─dynamic_type───────────┐ +│ [1,2,3] │ Array(Nullable(Int64)) │ +└─────────┴────────────────────────┘ +``` + +```sql +SELECT JSONExtract('{"obj" : {"a" : 42, "b" : "Hello", "c" : [1,2,3]}}', 'obj', 'Map(String, Variant(UInt32, String, Array(UInt32)))') AS map_of_dynamics, mapApply((k, v) -> (k, variantType(v)), map_of_dynamics) AS map_of_dynamic_types``` + +```text +┌─map_of_dynamics──────────────────┬─map_of_dynamic_types────────────────────────────┐ +│ {'a':42,'b':'Hello','c':[1,2,3]} │ {'a':'UInt32','b':'String','c':'Array(UInt32)'} │ +└──────────────────────────────────┴─────────────────────────────────────────────────┘ +``` + +```sql +SELECT JSONExtractKeysAndValues('{"a" : 42, "b" : "Hello", "c" : [1,2,3]}', 'Variant(UInt32, String, Array(UInt32))') AS dynamics, arrayMap(x -> (x.1, variantType(x.2)), dynamics) AS dynamic_types``` +``` + +```text +┌─dynamics───────────────────────────────┬─dynamic_types─────────────────────────────────────────┐ +│ [('a',42),('b','Hello'),('c',[1,2,3])] │ [('a','UInt32'),('b','String'),('c','Array(UInt32)')] │ +└────────────────────────────────────────┴───────────────────────────────────────────────────────┘ +``` ### Binary output format -In [RowBinary](../../interfaces/formats.md#rowbinary-rowbinary) format values of `Dynamic` type are serialized in the following format: +In RowBinary format values of `Dynamic` type are serialized in the following format: ```text ``` - -See the [data types binary encoding specification](../../sql-reference/data-types/data-types-binary-encoding.md) diff --git a/docs/en/sql-reference/functions/array-functions.md b/docs/en/sql-reference/functions/array-functions.md index d87ca4a0fe7..1b52440903d 100644 --- a/docs/en/sql-reference/functions/array-functions.md +++ b/docs/en/sql-reference/functions/array-functions.md @@ -3080,4 +3080,4 @@ Result: ## Distance functions -All supported functions are described in [distance functions documentation](../../sql-reference/functions/distance-functions.md). +All supported functions are described in [distance functions documentation](../../sql-reference/functions/distance-functions.md). \ No newline at end of file diff --git a/docs/en/sql-reference/functions/hash-functions.md b/docs/en/sql-reference/functions/hash-functions.md index e431ed75465..7c977e7d6dc 100644 --- a/docs/en/sql-reference/functions/hash-functions.md +++ b/docs/en/sql-reference/functions/hash-functions.md @@ -314,10 +314,71 @@ SELECT groupBitXor(cityHash64(*)) FROM table Calculates a 32-bit hash code from any type of integer. This is a relatively fast non-cryptographic hash function of average quality for numbers. +**Syntax** + +```sql +intHash32(int) +``` + +**Arguments** + +- `int` — Integer to hash. [(U)Int*](../data-types/int-uint.md). + +**Returned value** + +- 32-bit hash code. [UInt32](../data-types/int-uint.md). + +**Example** + +Query: + +```sql +SELECT intHash32(42); +``` + +Result: + +```response +┌─intHash32(42)─┐ +│ 1228623923 │ +└───────────────┘ +``` + ## intHash64 Calculates a 64-bit hash code from any type of integer. -It works faster than intHash32. Average quality. +This is a relatively fast non-cryptographic hash function of average quality for numbers. +It works faster than [intHash32](#inthash32). + +**Syntax** + +```sql +intHash64(int) +``` + +**Arguments** + +- `int` — Integer to hash. [(U)Int*](../data-types/int-uint.md). + +**Returned value** + +- 64-bit hash code. [UInt64](../data-types/int-uint.md). + +**Example** + +Query: + +```sql +SELECT intHash64(42); +``` + +Result: + +```response +┌────────intHash64(42)─┐ +│ 11490350930367293593 │ +└──────────────────────┘ +``` ## SHA1, SHA224, SHA256, SHA512, SHA512_256 diff --git a/docs/en/sql-reference/window-functions/dense_rank.md b/docs/en/sql-reference/window-functions/dense_rank.md new file mode 100644 index 00000000000..d6445b68c55 --- /dev/null +++ b/docs/en/sql-reference/window-functions/dense_rank.md @@ -0,0 +1,73 @@ +--- +slug: /en/sql-reference/window-functions/dense_rank +sidebar_label: dense_rank +sidebar_position: 7 +--- + +# dense_rank + +Ranks the current row within its partition without gaps. In other words, if the value of any new row encountered is equal to the value of one of the previous rows then it will receive the next successive rank without any gaps in ranking. + +The [rank](./rank.md) function provides the same behaviour, but with gaps in ranking. + +**Syntax** + +```sql +dense_rank (column_name) + OVER ([[PARTITION BY grouping_column] [ORDER BY sorting_column] + [ROWS or RANGE expression_to_bound_rows_withing_the_group]] | [window_name]) +FROM table_name +WINDOW window_name as ([[PARTITION BY grouping_column] [ORDER BY sorting_column]) +``` + +For more detail on window function syntax see: [Window Functions - Syntax](./index.md/#syntax). + +**Returned value** + +- A number for the current row within its partition, without gaps in ranking. [UInt64](../data-types/int-uint.md). + +**Example** + +The following example is based on the example provided in the video instructional [Ranking window functions in ClickHouse](https://youtu.be/Yku9mmBYm_4?si=XIMu1jpYucCQEoXA). + +Query: + +```sql +CREATE TABLE salaries +( + `team` String, + `player` String, + `salary` UInt32, + `position` String +) +Engine = Memory; + +INSERT INTO salaries FORMAT Values + ('Port Elizabeth Barbarians', 'Gary Chen', 195000, 'F'), + ('New Coreystad Archdukes', 'Charles Juarez', 190000, 'F'), + ('Port Elizabeth Barbarians', 'Michael Stanley', 150000, 'D'), + ('New Coreystad Archdukes', 'Scott Harrison', 150000, 'D'), + ('Port Elizabeth Barbarians', 'Robert George', 195000, 'M'), + ('South Hampton Seagulls', 'Douglas Benson', 150000, 'M'), + ('South Hampton Seagulls', 'James Henderson', 140000, 'M'); +``` + +```sql +SELECT player, salary, + dense_rank() OVER (ORDER BY salary DESC) AS dense_rank +FROM salaries; +``` + +Result: + +```response + ┌─player──────────┬─salary─┬─dense_rank─┐ +1. │ Gary Chen │ 195000 │ 1 │ +2. │ Robert George │ 195000 │ 1 │ +3. │ Charles Juarez │ 190000 │ 2 │ +4. │ Michael Stanley │ 150000 │ 3 │ +5. │ Douglas Benson │ 150000 │ 3 │ +6. │ Scott Harrison │ 150000 │ 3 │ +7. │ James Henderson │ 140000 │ 4 │ + └─────────────────┴────────┴────────────┘ +``` \ No newline at end of file diff --git a/docs/en/sql-reference/window-functions/first_value.md b/docs/en/sql-reference/window-functions/first_value.md new file mode 100644 index 00000000000..30c3b1f99dc --- /dev/null +++ b/docs/en/sql-reference/window-functions/first_value.md @@ -0,0 +1,79 @@ +--- +slug: /en/sql-reference/window-functions/first_value +sidebar_label: first_value +sidebar_position: 3 +--- + +# first_value + +Returns the first value evaluated within its ordered frame. By default, NULL arguments are skipped, however the `RESPECT NULLS` modifier can be used to override this behaviour. + +**Syntax** + +```sql +first_value (column_name) [[RESPECT NULLS] | [IGNORE NULLS]] + OVER ([[PARTITION BY grouping_column] [ORDER BY sorting_column] + [ROWS or RANGE expression_to_bound_rows_withing_the_group]] | [window_name]) +FROM table_name +WINDOW window_name as ([[PARTITION BY grouping_column] [ORDER BY sorting_column]) +``` + +Alias: `any`. + +:::note +Using the optional modifier `RESPECT NULLS` after `first_value(column_name)` will ensure that `NULL` arguments are not skipped. +See [NULL processing](../aggregate-functions/index.md/#null-processing) for more information. +::: + +For more detail on window function syntax see: [Window Functions - Syntax](./index.md/#syntax). + +**Returned value** + +- The first value evaluated within its ordered frame. + +**Example** + +In this example the `first_value` function is used to find the highest paid footballer from a fictional dataset of salaries of Premier League football players. + +Query: + +```sql +DROP TABLE IF EXISTS salaries; +CREATE TABLE salaries +( + `team` String, + `player` String, + `salary` UInt32, + `position` String +) +Engine = Memory; + +INSERT INTO salaries FORMAT Values + ('Port Elizabeth Barbarians', 'Gary Chen', 196000, 'F'), + ('New Coreystad Archdukes', 'Charles Juarez', 190000, 'F'), + ('Port Elizabeth Barbarians', 'Michael Stanley', 100000, 'D'), + ('New Coreystad Archdukes', 'Scott Harrison', 180000, 'D'), + ('Port Elizabeth Barbarians', 'Robert George', 195000, 'M'), + ('South Hampton Seagulls', 'Douglas Benson', 150000, 'M'), + ('South Hampton Seagulls', 'James Henderson', 140000, 'M'); +``` + +```sql +SELECT player, salary, + first_value(player) OVER (ORDER BY salary DESC) AS highest_paid_player +FROM salaries; +``` + +Result: + +```response + ┌─player──────────┬─salary─┬─highest_paid_player─┐ +1. │ Gary Chen │ 196000 │ Gary Chen │ +2. │ Robert George │ 195000 │ Gary Chen │ +3. │ Charles Juarez │ 190000 │ Gary Chen │ +4. │ Scott Harrison │ 180000 │ Gary Chen │ +5. │ Douglas Benson │ 150000 │ Gary Chen │ +6. │ James Henderson │ 140000 │ Gary Chen │ +7. │ Michael Stanley │ 100000 │ Gary Chen │ + └─────────────────┴────────┴─────────────────────┘ +``` \ No newline at end of file diff --git a/docs/en/sql-reference/window-functions/index.md b/docs/en/sql-reference/window-functions/index.md index 16225d4b0e2..0c3e2ea1cb6 100644 --- a/docs/en/sql-reference/window-functions/index.md +++ b/docs/en/sql-reference/window-functions/index.md @@ -1,10 +1,11 @@ --- slug: /en/sql-reference/window-functions/ -sidebar_position: 62 sidebar_label: Window Functions -title: Window Functions +sidebar_position: 1 --- +# Window Functions + Windows functions let you perform calculations across a set of rows that are related to the current row. Some of the calculations that you can do are similar to those that can be done with an aggregate function, but a window function doesn't cause rows to be grouped into a single output - the individual rows are still returned. @@ -12,8 +13,8 @@ Some of the calculations that you can do are similar to those that can be done w ClickHouse supports the standard grammar for defining windows and window functions. The table below indicates whether a feature is currently supported. -| Feature | Supported? | -|------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Feature | Supported? | +|--------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| | ad hoc window specification (`count(*) over (partition by id order by time desc)`) | ✅ | | expressions involving window functions, e.g. `(count(*) over ()) / 2)` | ✅ | | `WINDOW` clause (`select ... from table window w as (partition by id)`) | ✅ | @@ -75,14 +76,14 @@ WINDOW window_name as ([[PARTITION BY grouping_column] [ORDER BY sorting_column] These functions can be used only as a window function. -- `row_number()` - Number the current row within its partition starting from 1. -- `first_value(x)` - Return the first non-NULL value evaluated within its ordered frame. -- `last_value(x)` - Return the last non-NULL value evaluated within its ordered frame. -- `nth_value(x, offset)` - Return the first non-NULL value evaluated against the nth row (offset) in its ordered frame. -- `rank()` - Rank the current row within its partition with gaps. -- `dense_rank()` - Rank the current row within its partition without gaps. -- `lagInFrame(x[, offset[, default]])` - Return a value evaluated at the row that is at a specified physical offset row before the current row within the ordered frame. The offset parameter, if not specified, defaults to 1, meaning it will fetch the value from the next row. If the calculated row exceeds the boundaries of the window frame, the specified default value is returned. -- `leadInFrame(x[, offset[, default]])` - Return a value evaluated at the row that is offset rows after the current row within the ordered frame. If offset is not provided, it defaults to 1. If the offset leads to a position outside the window frame, the specified default value is used. +- [`row_number()`](./row_number.md) - Number the current row within its partition starting from 1. +- [`first_value(x)`](./first_value.md) - Return the first value evaluated within its ordered frame. +- [`last_value(x)`](./last_value.md) - Return the last value evaluated within its ordered frame. +- [`nth_value(x, offset)`](./nth_value.md) - Return the first non-NULL value evaluated against the nth row (offset) in its ordered frame. +- [`rank()`](./rank.md) - Rank the current row within its partition with gaps. +- [`dense_rank()`](./dense_rank.md) - Rank the current row within its partition without gaps. +- [`lagInFrame(x)`](./lagInFrame.md) - Return a value evaluated at the row that is at a specified physical offset row before the current row within the ordered frame. +- [`leadInFrame(x)`](./leadInFrame.md) - Return a value evaluated at the row that is offset rows after the current row within the ordered frame. ## Examples diff --git a/docs/en/sql-reference/window-functions/lagInFrame.md b/docs/en/sql-reference/window-functions/lagInFrame.md new file mode 100644 index 00000000000..049e095c10f --- /dev/null +++ b/docs/en/sql-reference/window-functions/lagInFrame.md @@ -0,0 +1,79 @@ +--- +slug: /en/sql-reference/window-functions/lagInFrame +sidebar_label: lagInFrame +sidebar_position: 8 +--- + +# lagInFrame + +Returns a value evaluated at the row that is at a specified physical offset row before the current row within the ordered frame. + +**Syntax** + +```sql +lagInFrame(x[, offset[, default]]) + OVER ([[PARTITION BY grouping_column] [ORDER BY sorting_column] + [ROWS or RANGE expression_to_bound_rows_withing_the_group]] | [window_name]) +FROM table_name +WINDOW window_name as ([[PARTITION BY grouping_column] [ORDER BY sorting_column]) +``` + +For more detail on window function syntax see: [Window Functions - Syntax](./index.md/#syntax). + +**Parameters** +- `x` — Column name. +- `offset` — Offset to apply. [(U)Int*](../data-types/int-uint.md). (Optional - `1` by default). +- `default` — Value to return if calculated row exceeds the boundaries of the window frame. (Optional - `null` by default). + +**Returned value** + +- Value evaluated at the row that is at a specified physical offset before the current row within the ordered frame. + +**Example** + +This example looks at historical data for a specific stock and uses the `lagInFrame` function to calculate a day-to-day delta and percentage change in the closing price of the stock. + +Query: + +```sql +CREATE TABLE stock_prices +( + `date` Date, + `open` Float32, -- opening price + `high` Float32, -- daily high + `low` Float32, -- daily low + `close` Float32, -- closing price + `volume` UInt32 -- trade volume +) +Engine = Memory; + +INSERT INTO stock_prices FORMAT Values + ('2024-06-03', 113.62, 115.00, 112.00, 115.00, 438392000), + ('2024-06-04', 115.72, 116.60, 114.04, 116.44, 403324000), + ('2024-06-05', 118.37, 122.45, 117.47, 122.44, 528402000), + ('2024-06-06', 124.05, 125.59, 118.32, 121.00, 664696000), + ('2024-06-07', 119.77, 121.69, 118.02, 120.89, 412386000); +``` + +```sql +SELECT + date, + close, + lagInFrame(close, 1, close) OVER (ORDER BY date ASC) AS previous_day_close, + COALESCE(ROUND(close - previous_day_close, 2)) AS delta, + COALESCE(ROUND((delta / previous_day_close) * 100, 2)) AS percent_change +FROM stock_prices +ORDER BY date DESC; +``` + +Result: + +```response + ┌───────date─┬──close─┬─previous_day_close─┬─delta─┬─percent_change─┐ +1. │ 2024-06-07 │ 120.89 │ 121 │ -0.11 │ -0.09 │ +2. │ 2024-06-06 │ 121 │ 122.44 │ -1.44 │ -1.18 │ +3. │ 2024-06-05 │ 122.44 │ 116.44 │ 6 │ 5.15 │ +4. │ 2024-06-04 │ 116.44 │ 115 │ 1.44 │ 1.25 │ +5. │ 2024-06-03 │ 115 │ 115 │ 0 │ 0 │ + └────────────┴────────┴────────────────────┴───────┴────────────────┘ +``` \ No newline at end of file diff --git a/docs/en/sql-reference/window-functions/last_value.md b/docs/en/sql-reference/window-functions/last_value.md new file mode 100644 index 00000000000..dd7f5fa078a --- /dev/null +++ b/docs/en/sql-reference/window-functions/last_value.md @@ -0,0 +1,79 @@ +--- +slug: /en/sql-reference/window-functions/last_value +sidebar_label: last_value +sidebar_position: 4 +--- + +# last_value + +Returns the last value evaluated within its ordered frame. By default, NULL arguments are skipped, however the `RESPECT NULLS` modifier can be used to override this behaviour. + +**Syntax** + +```sql +last_value (column_name) [[RESPECT NULLS] | [IGNORE NULLS]] + OVER ([[PARTITION BY grouping_column] [ORDER BY sorting_column] + [ROWS or RANGE expression_to_bound_rows_withing_the_group]] | [window_name]) +FROM table_name +WINDOW window_name as ([[PARTITION BY grouping_column] [ORDER BY sorting_column]) +``` + +Alias: `anyLast`. + +:::note +Using the optional modifier `RESPECT NULLS` after `first_value(column_name)` will ensure that `NULL` arguments are not skipped. +See [NULL processing](../aggregate-functions/index.md/#null-processing) for more information. +::: + +For more detail on window function syntax see: [Window Functions - Syntax](./index.md/#syntax). + +**Returned value** + +- The last value evaluated within its ordered frame. + +**Example** + +In this example the `last_value` function is used to find the highest paid footballer from a fictional dataset of salaries of Premier League football players. + +Query: + +```sql +DROP TABLE IF EXISTS salaries; +CREATE TABLE salaries +( + `team` String, + `player` String, + `salary` UInt32, + `position` String +) +Engine = Memory; + +INSERT INTO salaries FORMAT Values + ('Port Elizabeth Barbarians', 'Gary Chen', 196000, 'F'), + ('New Coreystad Archdukes', 'Charles Juarez', 190000, 'F'), + ('Port Elizabeth Barbarians', 'Michael Stanley', 100000, 'D'), + ('New Coreystad Archdukes', 'Scott Harrison', 180000, 'D'), + ('Port Elizabeth Barbarians', 'Robert George', 195000, 'M'), + ('South Hampton Seagulls', 'Douglas Benson', 150000, 'M'), + ('South Hampton Seagulls', 'James Henderson', 140000, 'M'); +``` + +```sql +SELECT player, salary, + last_value(player) OVER (ORDER BY salary DESC RANGE BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING) AS lowest_paid_player +FROM salaries; +``` + +Result: + +```response + ┌─player──────────┬─salary─┬─lowest_paid_player─┐ +1. │ Gary Chen │ 196000 │ Michael Stanley │ +2. │ Robert George │ 195000 │ Michael Stanley │ +3. │ Charles Juarez │ 190000 │ Michael Stanley │ +4. │ Scott Harrison │ 180000 │ Michael Stanley │ +5. │ Douglas Benson │ 150000 │ Michael Stanley │ +6. │ James Henderson │ 140000 │ Michael Stanley │ +7. │ Michael Stanley │ 100000 │ Michael Stanley │ + └─────────────────┴────────┴────────────────────┘ +``` \ No newline at end of file diff --git a/docs/en/sql-reference/window-functions/leadInFrame.md b/docs/en/sql-reference/window-functions/leadInFrame.md new file mode 100644 index 00000000000..fc1b92cc266 --- /dev/null +++ b/docs/en/sql-reference/window-functions/leadInFrame.md @@ -0,0 +1,60 @@ +--- +slug: /en/sql-reference/window-functions/leadInFrame +sidebar_label: leadInFrame +sidebar_position: 9 +--- + +# leadInFrame + +Returns a value evaluated at the row that is offset rows after the current row within the ordered frame. + +**Syntax** + +```sql +leadInFrame(x[, offset[, default]]) + OVER ([[PARTITION BY grouping_column] [ORDER BY sorting_column] + [ROWS or RANGE expression_to_bound_rows_withing_the_group]] | [window_name]) +FROM table_name +WINDOW window_name as ([[PARTITION BY grouping_column] [ORDER BY sorting_column]) +``` + +For more detail on window function syntax see: [Window Functions - Syntax](./index.md/#syntax). + +**Parameters** +- `x` — Column name. +- `offset` — Offset to apply. [(U)Int*](../data-types/int-uint.md). (Optional - `1` by default). +- `default` — Value to return if calculated row exceeds the boundaries of the window frame. (Optional - `null` by default). + +**Returned value** + +- value evaluated at the row that is offset rows after the current row within the ordered frame. + +**Example** + +This example looks at [historical data](https://www.kaggle.com/datasets/sazidthe1/nobel-prize-data) for Nobel Prize winners and uses the `leadInFrame` function to return a list of successive winners in the physics category. + +Query: + +```sql +CREATE OR REPLACE VIEW nobel_prize_laureates AS FROM file('nobel_laureates_data.csv') SELECT *; +``` + +```sql +FROM nobel_prize_laureates SELECT fullName, leadInFrame(year, 1, year) OVER (PARTITION BY category ORDER BY year) AS year, category, motivation WHERE category == 'physics' ORDER BY year DESC LIMIT 9; +``` + +Result: + +```response + ┌─fullName─────────┬─year─┬─category─┬─motivation─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┐ +1. │ Pierre Agostini │ 2023 │ physics │ for experimental methods that generate attosecond pulses of light for the study of electron dynamics in matter │ +2. │ Ferenc Krausz │ 2023 │ physics │ for experimental methods that generate attosecond pulses of light for the study of electron dynamics in matter │ +3. │ Anne L Huillier │ 2023 │ physics │ for experimental methods that generate attosecond pulses of light for the study of electron dynamics in matter │ +4. │ Alain Aspect │ 2022 │ physics │ for experiments with entangled photons establishing the violation of Bell inequalities and pioneering quantum information science │ +5. │ Anton Zeilinger │ 2022 │ physics │ for experiments with entangled photons establishing the violation of Bell inequalities and pioneering quantum information science │ +6. │ John Clauser │ 2022 │ physics │ for experiments with entangled photons establishing the violation of Bell inequalities and pioneering quantum information science │ +7. │ Syukuro Manabe │ 2021 │ physics │ for the physical modelling of Earths climate quantifying variability and reliably predicting global warming │ +8. │ Klaus Hasselmann │ 2021 │ physics │ for the physical modelling of Earths climate quantifying variability and reliably predicting global warming │ +9. │ Giorgio Parisi │ 2021 │ physics │ for the discovery of the interplay of disorder and fluctuations in physical systems from atomic to planetary scales │ + └──────────────────┴──────┴──────────┴────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┘ +``` \ No newline at end of file diff --git a/docs/en/sql-reference/window-functions/nth_value.md b/docs/en/sql-reference/window-functions/nth_value.md new file mode 100644 index 00000000000..aa5baf651a8 --- /dev/null +++ b/docs/en/sql-reference/window-functions/nth_value.md @@ -0,0 +1,75 @@ +--- +slug: /en/sql-reference/window-functions/nth_value +sidebar_label: nth_value +sidebar_position: 5 +--- + +# nth_value + +Returns the first non-NULL value evaluated against the nth row (offset) in its ordered frame. + +**Syntax** + +```sql +nth_value (x, offset) + OVER ([[PARTITION BY grouping_column] [ORDER BY sorting_column] + [ROWS or RANGE expression_to_bound_rows_withing_the_group]] | [window_name]) +FROM table_name +WINDOW window_name as ([[PARTITION BY grouping_column] [ORDER BY sorting_column]) +``` + +For more detail on window function syntax see: [Window Functions - Syntax](./index.md/#syntax). + +**Parameters** + +- `x` — Column name. +- `offset` — nth row to evaluate current row against. + +**Returned value** + +- The first non-NULL value evaluated against the nth row (offset) in its ordered frame. + +**Example** + +In this example the `nth-value` function is used to find the third-highest salary from a fictional dataset of salaries of Premier League football players. + +Query: + +```sql +DROP TABLE IF EXISTS salaries; +CREATE TABLE salaries +( + `team` String, + `player` String, + `salary` UInt32, + `position` String +) +Engine = Memory; + +INSERT INTO salaries FORMAT Values + ('Port Elizabeth Barbarians', 'Gary Chen', 195000, 'F'), + ('New Coreystad Archdukes', 'Charles Juarez', 190000, 'F'), + ('Port Elizabeth Barbarians', 'Michael Stanley', 100000, 'D'), + ('New Coreystad Archdukes', 'Scott Harrison', 180000, 'D'), + ('Port Elizabeth Barbarians', 'Robert George', 195000, 'M'), + ('South Hampton Seagulls', 'Douglas Benson', 150000, 'M'), + ('South Hampton Seagulls', 'James Henderson', 140000, 'M'); +``` + +```sql +SELECT player, salary, nth_value(player,3) OVER(ORDER BY salary DESC) AS third_highest_salary FROM salaries; +``` + +Result: + +```response + ┌─player──────────┬─salary─┬─third_highest_salary─┐ +1. │ Gary Chen │ 195000 │ │ +2. │ Robert George │ 195000 │ │ +3. │ Charles Juarez │ 190000 │ Charles Juarez │ +4. │ Scott Harrison │ 180000 │ Charles Juarez │ +5. │ Douglas Benson │ 150000 │ Charles Juarez │ +6. │ James Henderson │ 140000 │ Charles Juarez │ +7. │ Michael Stanley │ 100000 │ Charles Juarez │ + └─────────────────┴────────┴──────────────────────┘ +``` \ No newline at end of file diff --git a/docs/en/sql-reference/window-functions/rank.md b/docs/en/sql-reference/window-functions/rank.md new file mode 100644 index 00000000000..dff5e154151 --- /dev/null +++ b/docs/en/sql-reference/window-functions/rank.md @@ -0,0 +1,74 @@ +--- +slug: /en/sql-reference/window-functions/rank +sidebar_label: rank +sidebar_position: 6 +--- + +# rank + +Ranks the current row within its partition with gaps. In other words, if the value of any row it encounters is equal to the value of a previous row then it will receive the same rank as that previous row. +The rank of the next row is then equal to the rank of the previous row plus a gap equal to the number of times the previous rank was given. + +The [dense_rank](./dense_rank.md) function provides the same behaviour but without gaps in ranking. + +**Syntax** + +```sql +rank (column_name) + OVER ([[PARTITION BY grouping_column] [ORDER BY sorting_column] + [ROWS or RANGE expression_to_bound_rows_withing_the_group]] | [window_name]) +FROM table_name +WINDOW window_name as ([[PARTITION BY grouping_column] [ORDER BY sorting_column]) +``` + +For more detail on window function syntax see: [Window Functions - Syntax](./index.md/#syntax). + +**Returned value** + +- A number for the current row within its partition, including gaps. [UInt64](../data-types/int-uint.md). + +**Example** + +The following example is based on the example provided in the video instructional [Ranking window functions in ClickHouse](https://youtu.be/Yku9mmBYm_4?si=XIMu1jpYucCQEoXA). + +Query: + +```sql +CREATE TABLE salaries +( + `team` String, + `player` String, + `salary` UInt32, + `position` String +) +Engine = Memory; + +INSERT INTO salaries FORMAT Values + ('Port Elizabeth Barbarians', 'Gary Chen', 195000, 'F'), + ('New Coreystad Archdukes', 'Charles Juarez', 190000, 'F'), + ('Port Elizabeth Barbarians', 'Michael Stanley', 150000, 'D'), + ('New Coreystad Archdukes', 'Scott Harrison', 150000, 'D'), + ('Port Elizabeth Barbarians', 'Robert George', 195000, 'M'), + ('South Hampton Seagulls', 'Douglas Benson', 150000, 'M'), + ('South Hampton Seagulls', 'James Henderson', 140000, 'M'); +``` + +```sql +SELECT player, salary, + rank() OVER (ORDER BY salary DESC) AS rank +FROM salaries; +``` + +Result: + +```response + ┌─player──────────┬─salary─┬─rank─┐ +1. │ Gary Chen │ 195000 │ 1 │ +2. │ Robert George │ 195000 │ 1 │ +3. │ Charles Juarez │ 190000 │ 3 │ +4. │ Douglas Benson │ 150000 │ 4 │ +5. │ Michael Stanley │ 150000 │ 4 │ +6. │ Scott Harrison │ 150000 │ 4 │ +7. │ James Henderson │ 140000 │ 7 │ + └─────────────────┴────────┴──────┘ +``` \ No newline at end of file diff --git a/docs/en/sql-reference/window-functions/row_number.md b/docs/en/sql-reference/window-functions/row_number.md new file mode 100644 index 00000000000..f1c331f89a3 --- /dev/null +++ b/docs/en/sql-reference/window-functions/row_number.md @@ -0,0 +1,67 @@ +--- +slug: /en/sql-reference/window-functions/row_number +sidebar_label: row_number +sidebar_position: 2 +--- + +# row_number + +Numbers the current row within its partition starting from 1. + +**Syntax** + +```sql +row_number (column_name) + OVER ([[PARTITION BY grouping_column] [ORDER BY sorting_column] + [ROWS or RANGE expression_to_bound_rows_withing_the_group]] | [window_name]) +FROM table_name +WINDOW window_name as ([[PARTITION BY grouping_column] [ORDER BY sorting_column]) +``` + +For more detail on window function syntax see: [Window Functions - Syntax](./index.md/#syntax). + +**Returned value** + +- A number for the current row within its partition. [UInt64](../data-types/int-uint.md). + +**Example** + +The following example is based on the example provided in the video instructional [Ranking window functions in ClickHouse](https://youtu.be/Yku9mmBYm_4?si=XIMu1jpYucCQEoXA). + +Query: + +```sql +CREATE TABLE salaries +( + `team` String, + `player` String, + `salary` UInt32, + `position` String +) +Engine = Memory; + +INSERT INTO salaries FORMAT Values + ('Port Elizabeth Barbarians', 'Gary Chen', 195000, 'F'), + ('New Coreystad Archdukes', 'Charles Juarez', 190000, 'F'), + ('Port Elizabeth Barbarians', 'Michael Stanley', 150000, 'D'), + ('New Coreystad Archdukes', 'Scott Harrison', 150000, 'D'), + ('Port Elizabeth Barbarians', 'Robert George', 195000, 'M'); +``` + +```sql +SELECT player, salary, + row_number() OVER (ORDER BY salary DESC) AS row_number +FROM salaries; +``` + +Result: + +```response + ┌─player──────────┬─salary─┬─row_number─┐ +1. │ Gary Chen │ 195000 │ 1 │ +2. │ Robert George │ 195000 │ 2 │ +3. │ Charles Juarez │ 190000 │ 3 │ +4. │ Scott Harrison │ 150000 │ 4 │ +5. │ Michael Stanley │ 150000 │ 5 │ + └─────────────────┴────────┴────────────┘ +``` \ No newline at end of file diff --git a/pyproject.toml b/pyproject.toml index 279d077a695..90f089afa41 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -35,10 +35,9 @@ disable = ''' broad-except, bare-except, no-else-return, - global-statement + global-statement, ''' [tool.pylint.SIMILARITIES] # due to SQL min-similarity-lines=1000 - diff --git a/src/Common/Exception.cpp b/src/Common/Exception.cpp index 1f4b0aea8f2..111280074dd 100644 --- a/src/Common/Exception.cpp +++ b/src/Common/Exception.cpp @@ -38,10 +38,19 @@ namespace ErrorCodes extern const int CANNOT_MREMAP; } +void abortOnFailedAssertion(const String & description, void * const * trace, size_t trace_offset, size_t trace_size) +{ + auto & logger = Poco::Logger::root(); + LOG_FATAL(&logger, "Logical error: '{}'.", description); + if (trace) + LOG_FATAL(&logger, "Stack trace (when copying this message, always include the lines below):\n\n{}", StackTrace::toString(trace, trace_offset, trace_size)); + abort(); +} + void abortOnFailedAssertion(const String & description) { - LOG_FATAL(&Poco::Logger::root(), "Logical error: '{}'.", description); - abort(); + StackTrace st; + abortOnFailedAssertion(description, st.getFramePointers().data(), st.getOffset(), st.getSize()); } bool terminate_on_any_exception = false; @@ -58,7 +67,7 @@ void handle_error_code(const std::string & msg, int code, bool remote, const Exc #ifdef ABORT_ON_LOGICAL_ERROR if (code == ErrorCodes::LOGICAL_ERROR) { - abortOnFailedAssertion(msg); + abortOnFailedAssertion(msg, trace.data(), 0, trace.size()); } #endif diff --git a/src/Common/Exception.h b/src/Common/Exception.h index 87ef7101cdc..a4774a89f6a 100644 --- a/src/Common/Exception.h +++ b/src/Common/Exception.h @@ -25,8 +25,6 @@ namespace DB class AtomicLogger; -[[noreturn]] void abortOnFailedAssertion(const String & description); - /// This flag can be set for testing purposes - to check that no exceptions are thrown. extern bool terminate_on_any_exception; @@ -167,6 +165,8 @@ protected: mutable std::vector capture_thread_frame_pointers; }; +[[noreturn]] void abortOnFailedAssertion(const String & description, void * const * trace, size_t trace_offset, size_t trace_size); +[[noreturn]] void abortOnFailedAssertion(const String & description); std::string getExceptionStackTraceString(const std::exception & e); std::string getExceptionStackTraceString(std::exception_ptr e); diff --git a/src/Common/JSONParsers/SimdJSONParser.h b/src/Common/JSONParsers/SimdJSONParser.h index 827d142266a..db679b14f52 100644 --- a/src/Common/JSONParsers/SimdJSONParser.h +++ b/src/Common/JSONParsers/SimdJSONParser.h @@ -14,6 +14,7 @@ namespace DB { + namespace ErrorCodes { extern const int CANNOT_ALLOCATE_MEMORY; diff --git a/src/Common/NamedCollections/NamedCollectionsFactory.cpp b/src/Common/NamedCollections/NamedCollectionsFactory.cpp index 14105a8651d..2faea1957ba 100644 --- a/src/Common/NamedCollections/NamedCollectionsFactory.cpp +++ b/src/Common/NamedCollections/NamedCollectionsFactory.cpp @@ -235,7 +235,7 @@ bool NamedCollectionFactory::loadIfNot(std::lock_guard & lock) loadFromConfig(context->getConfigRef(), lock); loadFromSQL(lock); - if (metadata_storage->supportsPeriodicUpdate()) + if (metadata_storage->isReplicated()) { update_task = context->getSchedulePool().createTask("NamedCollectionsMetadataStorage", [this]{ updateFunc(); }); update_task->activate(); @@ -357,6 +357,13 @@ void NamedCollectionFactory::reloadFromSQL() add(std::move(collections), lock); } +bool NamedCollectionFactory::usesReplicatedStorage() +{ + std::lock_guard lock(mutex); + loadIfNot(lock); + return metadata_storage->isReplicated(); +} + void NamedCollectionFactory::updateFunc() { LOG_TRACE(log, "Named collections background updating thread started"); diff --git a/src/Common/NamedCollections/NamedCollectionsFactory.h b/src/Common/NamedCollections/NamedCollectionsFactory.h index 6ee5940e686..a0721ad8a50 100644 --- a/src/Common/NamedCollections/NamedCollectionsFactory.h +++ b/src/Common/NamedCollections/NamedCollectionsFactory.h @@ -34,6 +34,8 @@ public: void updateFromSQL(const ASTAlterNamedCollectionQuery & query); + bool usesReplicatedStorage(); + void loadIfNot(); void shutdown(); diff --git a/src/Common/NamedCollections/NamedCollectionsMetadataStorage.cpp b/src/Common/NamedCollections/NamedCollectionsMetadataStorage.cpp index 32fdb25abd3..b3671350f92 100644 --- a/src/Common/NamedCollections/NamedCollectionsMetadataStorage.cpp +++ b/src/Common/NamedCollections/NamedCollectionsMetadataStorage.cpp @@ -67,7 +67,7 @@ public: virtual bool removeIfExists(const std::string & path) = 0; - virtual bool supportsPeriodicUpdate() const = 0; + virtual bool isReplicated() const = 0; virtual bool waitUpdate(size_t /* timeout */) { return false; } }; @@ -89,7 +89,7 @@ public: ~LocalStorage() override = default; - bool supportsPeriodicUpdate() const override { return false; } + bool isReplicated() const override { return false; } std::vector list() const override { @@ -221,7 +221,7 @@ public: ~ZooKeeperStorage() override = default; - bool supportsPeriodicUpdate() const override { return true; } + bool isReplicated() const override { return true; } /// Return true if children changed. bool waitUpdate(size_t timeout) override @@ -465,14 +465,14 @@ void NamedCollectionsMetadataStorage::writeCreateQuery(const ASTCreateNamedColle storage->write(getFileName(query.collection_name), serializeAST(*normalized_query), replace); } -bool NamedCollectionsMetadataStorage::supportsPeriodicUpdate() const +bool NamedCollectionsMetadataStorage::isReplicated() const { - return storage->supportsPeriodicUpdate(); + return storage->isReplicated(); } bool NamedCollectionsMetadataStorage::waitUpdate() { - if (!storage->supportsPeriodicUpdate()) + if (!storage->isReplicated()) throw Exception(ErrorCodes::LOGICAL_ERROR, "Periodic updates are not supported"); const auto & config = Context::getGlobalContextInstance()->getConfigRef(); diff --git a/src/Common/NamedCollections/NamedCollectionsMetadataStorage.h b/src/Common/NamedCollections/NamedCollectionsMetadataStorage.h index 3c089fe2fa2..c3468fbc468 100644 --- a/src/Common/NamedCollections/NamedCollectionsMetadataStorage.h +++ b/src/Common/NamedCollections/NamedCollectionsMetadataStorage.h @@ -30,7 +30,7 @@ public: /// Return true if update was made bool waitUpdate(); - bool supportsPeriodicUpdate() const; + bool isReplicated() const; private: class INamedCollectionsStorage; diff --git a/src/Common/StackTrace.cpp b/src/Common/StackTrace.cpp index 239e957bdfe..34f6f0b7535 100644 --- a/src/Common/StackTrace.cpp +++ b/src/Common/StackTrace.cpp @@ -545,7 +545,7 @@ std::string StackTrace::toString() const return toStringCached(frame_pointers, offset, size); } -std::string StackTrace::toString(void ** frame_pointers_raw, size_t offset, size_t size) +std::string StackTrace::toString(void * const * frame_pointers_raw, size_t offset, size_t size) { __msan_unpoison(frame_pointers_raw, size * sizeof(*frame_pointers_raw)); diff --git a/src/Common/StackTrace.h b/src/Common/StackTrace.h index 4ce9a9281f3..2078828f3d7 100644 --- a/src/Common/StackTrace.h +++ b/src/Common/StackTrace.h @@ -59,7 +59,7 @@ public: const FramePointers & getFramePointers() const { return frame_pointers; } std::string toString() const; - static std::string toString(void ** frame_pointers, size_t offset, size_t size); + static std::string toString(void * const * frame_pointers, size_t offset, size_t size); static void dropCache(); /// @param fatal - if true, will process inline frames (slower) diff --git a/src/Common/assert_cast.h b/src/Common/assert_cast.h index 0b73ba1cc12..f9d0bf0e595 100644 --- a/src/Common/assert_cast.h +++ b/src/Common/assert_cast.h @@ -25,7 +25,7 @@ namespace DB template inline To assert_cast(From && from) { -#ifndef NDEBUG +#ifdef ABORT_ON_LOGICAL_ERROR try { if constexpr (std::is_pointer_v) diff --git a/src/Common/mysqlxx/Pool.cpp b/src/Common/mysqlxx/Pool.cpp index cc5b18214c8..546e9e91dc7 100644 --- a/src/Common/mysqlxx/Pool.cpp +++ b/src/Common/mysqlxx/Pool.cpp @@ -228,7 +228,6 @@ Pool::Entry Pool::tryGet() for (auto connection_it = connections.cbegin(); connection_it != connections.cend();) { Connection * connection_ptr = *connection_it; - /// Fixme: There is a race condition here b/c we do not synchronize with Pool::Entry's copy-assignment operator if (connection_ptr->ref_count == 0) { { diff --git a/src/Common/mysqlxx/mysqlxx/Pool.h b/src/Common/mysqlxx/mysqlxx/Pool.h index 6e509d8bdd6..f1ef81e28dd 100644 --- a/src/Common/mysqlxx/mysqlxx/Pool.h +++ b/src/Common/mysqlxx/mysqlxx/Pool.h @@ -64,17 +64,6 @@ public: decrementRefCount(); } - Entry & operator= (const Entry & src) /// NOLINT - { - pool = src.pool; - if (data) - decrementRefCount(); - data = src.data; - if (data) - incrementRefCount(); - return * this; - } - bool isNull() const { return data == nullptr; diff --git a/src/Common/mysqlxx/tests/mysqlxx_pool_test.cpp b/src/Common/mysqlxx/tests/mysqlxx_pool_test.cpp index 61d6a117285..121767edc84 100644 --- a/src/Common/mysqlxx/tests/mysqlxx_pool_test.cpp +++ b/src/Common/mysqlxx/tests/mysqlxx_pool_test.cpp @@ -13,13 +13,11 @@ mysqlxx::Pool::Entry getWithFailover(mysqlxx::Pool & connections_pool) constexpr size_t max_tries = 3; - mysqlxx::Pool::Entry worker_connection; - for (size_t try_no = 1; try_no <= max_tries; ++try_no) { try { - worker_connection = connections_pool.tryGet(); + mysqlxx::Pool::Entry worker_connection = connections_pool.tryGet(); if (!worker_connection.isNull()) { diff --git a/src/Core/Settings.h b/src/Core/Settings.h index 580c7253b03..2391c02752c 100644 --- a/src/Core/Settings.h +++ b/src/Core/Settings.h @@ -346,6 +346,7 @@ class IColumn; \ M(Bool, ignore_on_cluster_for_replicated_udf_queries, false, "Ignore ON CLUSTER clause for replicated UDF management queries.", 0) \ M(Bool, ignore_on_cluster_for_replicated_access_entities_queries, false, "Ignore ON CLUSTER clause for replicated access entities management queries.", 0) \ + M(Bool, ignore_on_cluster_for_replicated_named_collections_queries, false, "Ignore ON CLUSTER clause for replicated named collections management queries.", 0) \ /** Settings for testing hedged requests */ \ M(Milliseconds, sleep_in_send_tables_status_ms, 0, "Time to sleep in sending tables status response in TCPHandler", 0) \ M(Milliseconds, sleep_in_send_data_ms, 0, "Time to sleep in sending data in TCPHandler", 0) \ diff --git a/src/Core/SettingsChangesHistory.cpp b/src/Core/SettingsChangesHistory.cpp index f9cbe77b90b..f26c9a701f0 100644 --- a/src/Core/SettingsChangesHistory.cpp +++ b/src/Core/SettingsChangesHistory.cpp @@ -77,6 +77,7 @@ static std::initializer_list #include #include +#include #include #include #include @@ -211,6 +212,13 @@ void DDLLoadingDependencyVisitor::extractTableNameFromArgument(const ASTFunction qualified_name.database = table_identifier->getDatabaseName(); qualified_name.table = table_identifier->shortName(); } + else if (arg->as()) + { + /// Allow IN subquery. + /// Do not add tables from the subquery into dependencies, + /// because CREATE will succeed anyway. + return; + } else { assert(false); diff --git a/src/Databases/DatabaseAtomic.cpp b/src/Databases/DatabaseAtomic.cpp index 0ed995c6180..f58b90f1134 100644 --- a/src/Databases/DatabaseAtomic.cpp +++ b/src/Databases/DatabaseAtomic.cpp @@ -107,12 +107,24 @@ void DatabaseAtomic::attachTable(ContextPtr /* context_ */, const String & name, StoragePtr DatabaseAtomic::detachTable(ContextPtr /* context */, const String & name) { + // it is important to call the destructors of not_in_use without + // locked mutex to avoid potential deadlock. DetachedTables not_in_use; - std::lock_guard lock(mutex); - auto table = DatabaseOrdinary::detachTableUnlocked(name); - table_name_to_path.erase(name); - detached_tables.emplace(table->getStorageID().uuid, table); - not_in_use = cleanupDetachedTables(); + StoragePtr table; + { + std::lock_guard lock(mutex); + table = DatabaseOrdinary::detachTableUnlocked(name); + table_name_to_path.erase(name); + detached_tables.emplace(table->getStorageID().uuid, table); + not_in_use = cleanupDetachedTables(); + } + + if (!not_in_use.empty()) + { + not_in_use.clear(); + LOG_DEBUG(log, "Finished removing not used detached tables"); + } + return table; } diff --git a/src/Databases/MySQL/MaterializedMySQLSyncThread.cpp b/src/Databases/MySQL/MaterializedMySQLSyncThread.cpp index cb9fc76b1a8..2c342755337 100644 --- a/src/Databases/MySQL/MaterializedMySQLSyncThread.cpp +++ b/src/Databases/MySQL/MaterializedMySQLSyncThread.cpp @@ -29,6 +29,7 @@ #include #include #include +#include #include #include #include @@ -532,13 +533,17 @@ static inline void dumpDataForTables( bool MaterializedMySQLSyncThread::prepareSynchronized(MaterializeMetadata & metadata) { bool opened_transaction = false; - mysqlxx::PoolWithFailover::Entry connection; while (!isCancelled()) { try { - connection = pool.tryGet(); + mysqlxx::PoolWithFailover::Entry connection = pool.tryGet(); + SCOPE_EXIT({ + if (opened_transaction) + connection->query("ROLLBACK").execute(); + }); + if (connection.isNull()) { if (settings->max_wait_time_when_mysql_unavailable < 0) @@ -602,9 +607,6 @@ bool MaterializedMySQLSyncThread::prepareSynchronized(MaterializeMetadata & meta { tryLogCurrentException(log); - if (opened_transaction) - connection->query("ROLLBACK").execute(); - if (settings->max_wait_time_when_mysql_unavailable < 0) throw; diff --git a/src/Formats/JSONExtractTree.cpp b/src/Formats/JSONExtractTree.cpp new file mode 100644 index 00000000000..242d2dc9f80 --- /dev/null +++ b/src/Formats/JSONExtractTree.cpp @@ -0,0 +1,1660 @@ +#include +#include + +#include +#if USE_SIMDJSON +#include +#endif +#if USE_RAPIDJSON +#include +#endif + +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + + +#include +#include +#include +#include + +namespace DB +{ + +namespace ErrorCodes +{ + extern const int ILLEGAL_TYPE_OF_ARGUMENT; +} + +template +void jsonElementToString(const typename JSONParser::Element & element, WriteBuffer & buf, const FormatSettings & format_settings) +{ + if (element.isInt64()) + { + writeIntText(element.getInt64(), buf); + return; + } + if (element.isUInt64()) + { + writeIntText(element.getUInt64(), buf); + return; + } + if (element.isDouble()) + { + writeFloatText(element.getDouble(), buf); + return; + } + if (element.isBool()) + { + if (element.getBool()) + writeCString("true", buf); + else + writeCString("false", buf); + return; + } + if (element.isString()) + { + writeJSONString(element.getString(), buf, format_settings); + return; + } + if (element.isArray()) + { + writeChar('[', buf); + bool need_comma = false; + for (auto value : element.getArray()) + { + if (std::exchange(need_comma, true)) + writeChar(',', buf); + jsonElementToString(value, buf, format_settings); + } + writeChar(']', buf); + return; + } + if (element.isObject()) + { + writeChar('{', buf); + bool need_comma = false; + for (auto [key, value] : element.getObject()) + { + if (std::exchange(need_comma, true)) + writeChar(',', buf); + writeJSONString(key, buf, format_settings); + writeChar(':', buf); + jsonElementToString(value, buf, format_settings); + } + writeChar('}', buf); + return; + } + if (element.isNull()) + { + writeCString("null", buf); + return; + } +} + +template +bool tryGetNumericValueFromJSONElement( + NumberType & value, const typename JSONParser::Element & element, bool convert_bool_to_integer, String & error) +{ + switch (element.type()) + { + case ElementType::DOUBLE: + if constexpr (std::is_floating_point_v) + { + /// We permit inaccurate conversion of double to float. + /// Example: double 0.1 from JSON is not representable in float. + /// But it will be more convenient for user to perform conversion. + value = static_cast(element.getDouble()); + } + else if (!accurate::convertNumeric(element.getDouble(), value)) + { + error = fmt::format("cannot convert double value {} to {}", element.getDouble(), TypeName); + return false; + } + break; + case ElementType::UINT64: + if (!accurate::convertNumeric(element.getUInt64(), value)) + { + error = fmt::format("cannot convert UInt64 value {} to {}", element.getUInt64(), TypeName); + return false; + } + break; + case ElementType::INT64: + if (!accurate::convertNumeric(element.getInt64(), value)) + { + error = fmt::format("cannot convert Int64 value {} to {}", element.getInt64(), TypeName); + return false; + } + break; + case ElementType::BOOL: + if constexpr (is_integer) + { + if (convert_bool_to_integer) + { + value = static_cast(element.getBool()); + break; + } + } + error = fmt::format("cannot convert bool value to {}", TypeName); + return false; + case ElementType::STRING: { + auto rb = ReadBufferFromMemory{element.getString()}; + if constexpr (std::is_floating_point_v) + { + if (!tryReadFloatText(value, rb) || !rb.eof()) + { + error = fmt::format("cannot parse {} value here: {}", TypeName, element.getString()); + return false; + } + } + else + { + if (tryReadIntText(value, rb) && rb.eof()) + break; + + /// Try to parse float and convert it to integer. + Float64 tmp_float; + rb.position() = rb.buffer().begin(); + if (!tryReadFloatText(tmp_float, rb) || !rb.eof()) + { + error = fmt::format("cannot parse {} value here: {}", TypeName, element.getString()); + return false; + } + + if (!accurate::convertNumeric(tmp_float, value)) + { + error = fmt::format("cannot parse {} value here: {}", TypeName, element.getString()); + return false; + } + } + break; + } + default: + return false; + } + + return true; +} + +namespace +{ + +template +String jsonElementToString(const typename JSONParser::Element & element, const FormatSettings & format_settings) +{ + WriteBufferFromOwnString buf; + jsonElementToString(element, buf, format_settings); + return buf.str(); +} + +template +class NumericNode : public JSONExtractTreeNode +{ +public: + explicit NumericNode(bool is_bool_type_ = false) : is_bool_type(is_bool_type_) { } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings & insert_settings, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull()) + { + if (format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + error = fmt::format("cannot parse {} value from null", TypeName); + return false; + } + + NumberType value; + if (!tryGetNumericValueFromJSONElement(value, element, insert_settings.convert_bool_to_integer || is_bool_type, error)) + { + if (error.empty()) + error = fmt::format("cannot read {} value from JSON element: {}", TypeName, jsonElementToString(element, format_settings)); + return false; + } + + if (is_bool_type) + value = static_cast(value); + + auto & col_vec = assert_cast &>(column); + col_vec.insertValue(value); + return true; + } + +protected: + bool is_bool_type; +}; + +template +class LowCardinalityNumericNode : public NumericNode +{ +public: + explicit LowCardinalityNumericNode(bool is_nullable_, bool is_bool_type_ = false) + : NumericNode(is_bool_type_), is_nullable(is_nullable_) + { + } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings & insert_settings, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull()) + { + if (is_nullable || format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + error = fmt::format("cannot parse {} value from null", TypeName); + return false; + } + + NumberType value; + if (!tryGetNumericValueFromJSONElement(value, element, insert_settings.convert_bool_to_integer || this->is_bool_type, error)) + { + if (error.empty()) + error = fmt::format("cannot read {} value from JSON element: {}", TypeName, jsonElementToString(element, format_settings)); + return false; + } + + if (this->is_bool_type) + value = static_cast(value); + + auto & col_lc = assert_cast(column); + col_lc.insertData(reinterpret_cast(&value), sizeof(value)); + return true; + } + +private: + bool is_nullable; +}; + +template +class StringNode : public JSONExtractTreeNode +{ +public: + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull()) + { + if (format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + error = "cannot parse String value from null"; + return false; + } + + if (!element.isString()) + { + auto & col_str = assert_cast(column); + auto & chars = col_str.getChars(); + WriteBufferFromVector buf(chars, AppendModeTag()); + jsonElementToString(element, buf, format_settings); + buf.finalize(); + chars.push_back(0); + col_str.getOffsets().push_back(chars.size()); + } + else + { + auto value = element.getString(); + auto & col_str = assert_cast(column); + col_str.insertData(value.data(), value.size()); + } + return true; + } +}; + +template +class LowCardinalityStringNode : public JSONExtractTreeNode +{ +public: + explicit LowCardinalityStringNode(bool is_nullable_) : is_nullable(is_nullable_) { } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull()) + { + if (is_nullable || format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + error = "cannot parse String value from null"; + return false; + } + + if (!element.isString()) + { + auto value = jsonElementToString(element, format_settings); + assert_cast(column).insertData(value.data(), value.size()); + } + else + { + auto value = element.getString(); + assert_cast(column).insertData(value.data(), value.size()); + } + + return true; + } + +private: + bool is_nullable; +}; + +template +class FixedStringNode : public JSONExtractTreeNode +{ +public: + explicit FixedStringNode(size_t fixed_length_) : fixed_length(fixed_length_) { } + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull()) + { + if (format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + error = "cannot parse FixedString value from null"; + return false; + } + + if (!element.isString()) + return checkValueSizeAndInsert(column, jsonElementToString(element, format_settings), error); + return checkValueSizeAndInsert(column, element.getString(), error); + } + +private: + template + bool checkValueSizeAndInsert(IColumn & column, const T & value, String & error) const + { + if (value.size() > fixed_length) + { + error = fmt::format("too large string for FixedString({}): {}", fixed_length, value); + return false; + } + assert_cast(column).insertData(value.data(), value.size()); + return true; + } + + size_t fixed_length; +}; + +template +class LowCardinalityFixedStringNode : public JSONExtractTreeNode +{ +public: + explicit LowCardinalityFixedStringNode(bool is_nullable_, size_t fixed_length_) : is_nullable(is_nullable_), fixed_length(fixed_length_) + { + } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull()) + { + if (is_nullable || format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + error = "cannot parse FixedString value from null"; + return false; + } + + if (!element.isString()) + return checkValueSizeAndInsert(column, jsonElementToString(element, format_settings), error); + return checkValueSizeAndInsert(column, element.getString(), error); + } + +private: + template + bool checkValueSizeAndInsert(IColumn & column, const T & value, String & error) const + { + if (value.size() > fixed_length) + { + error = fmt::format("too large string for FixedString({}): {}", fixed_length, value); + return false; + } + + // For the non low cardinality case of FixedString, the padding is done in the FixedString Column implementation. + // In order to avoid having to pass the data to a FixedString Column and read it back (which would slow down the execution) + // the data is padded here and written directly to the Low Cardinality Column + if (value.size() == fixed_length) + { + assert_cast(column).insertData(value.data(), value.size()); + } + else + { + String padded_value(value); + padded_value.resize(fixed_length, '\0'); + assert_cast(column).insertData(padded_value.data(), padded_value.size()); + } + return true; + } + + bool is_nullable; + size_t fixed_length; +}; + +template +class UUIDNode : public JSONExtractTreeNode +{ +public: + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull() && format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + if (!element.isString()) + { + error = fmt::format("cannot read UUID value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + + auto data = element.getString(); + UUID uuid; + if (!tryParse(uuid, data)) + { + error = fmt::format("cannot parse UUID value here: {}", data); + return false; + } + + assert_cast(column).insert(uuid); + return true; + } + + + static bool tryParse(UUID & uuid, std::string_view data) + { + ReadBufferFromMemory buf(data.data(), data.size()); + return tryReadUUIDText(uuid, buf) && buf.eof(); + } +}; + +template +class LowCardinalityUUIDNode : public JSONExtractTreeNode +{ +public: + explicit LowCardinalityUUIDNode(bool is_nullable_) : is_nullable(is_nullable_) { } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull() && (is_nullable || format_settings.null_as_default)) + { + column.insertDefault(); + return true; + } + + if (!element.isString()) + { + error = fmt::format("cannot read UUID value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + + auto data = element.getString(); + ReadBufferFromMemory buf(data.data(), data.size()); + UUID uuid; + if (!tryReadUUIDText(uuid, buf) || !buf.eof()) + { + error = fmt::format("cannot parse UUID value here: {}", data); + return false; + } + assert_cast(column).insertData(reinterpret_cast(&uuid), sizeof(uuid)); + return true; + } + +private: + bool is_nullable; +}; + +template +class DateNode : public JSONExtractTreeNode +{ +public: + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull() && format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + if (!element.isString()) + { + error = fmt::format("cannot read Date value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + + auto data = element.getString(); + ReadBufferFromMemory buf(data.data(), data.size()); + DateType date; + if (!tryReadDateText(date, buf) || !buf.eof()) + { + error = fmt::format("cannot parse Date value here: {}", data); + return false; + } + + assert_cast &>(column).insertValue(date); + return true; + } +}; + +template +class DateTimeNode : public JSONExtractTreeNode, public TimezoneMixin +{ +public: + explicit DateTimeNode(const DataTypeDateTime & datetime_type) : TimezoneMixin(datetime_type) { } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull() && format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + time_t value; + if (element.isString()) + { + if (!tryParse(value, element.getString(), format_settings.date_time_input_format)) + { + error = fmt::format("cannot parse DateTime value here: {}", element.getString()); + return false; + } + } + else if (element.isUInt64()) + { + value = element.getUInt64(); + } + else + { + error = fmt::format("cannot read DateTime value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + + assert_cast(column).insert(value); + return true; + } + + bool tryParse(time_t & value, std::string_view data, FormatSettings::DateTimeInputFormat date_time_input_format) const + { + ReadBufferFromMemory buf(data.data(), data.size()); + switch (date_time_input_format) + { + case FormatSettings::DateTimeInputFormat::Basic: + if (tryReadDateTimeText(value, buf, time_zone) && buf.eof()) + return true; + break; + case FormatSettings::DateTimeInputFormat::BestEffort: + if (tryParseDateTimeBestEffort(value, buf, time_zone, utc_time_zone) && buf.eof()) + return true; + break; + case FormatSettings::DateTimeInputFormat::BestEffortUS: + if (tryParseDateTimeBestEffortUS(value, buf, time_zone, utc_time_zone) && buf.eof()) + return true; + break; + } + + return false; + } +}; + +template +class DecimalNode : public JSONExtractTreeNode +{ +public: + explicit DecimalNode(const DataTypePtr & type) : scale(assert_cast &>(*type).getScale()) { } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + DecimalType value{}; + + switch (element.type()) + { + case ElementType::DOUBLE: + value = convertToDecimal, DataTypeDecimal>(element.getDouble(), scale); + break; + case ElementType::UINT64: + value = convertToDecimal, DataTypeDecimal>(element.getUInt64(), scale); + break; + case ElementType::INT64: + value = convertToDecimal, DataTypeDecimal>(element.getInt64(), scale); + break; + case ElementType::STRING: { + auto rb = ReadBufferFromMemory{element.getString()}; + if (!SerializationDecimal::tryReadText(value, rb, DecimalUtils::max_precision, scale)) + { + error = fmt::format("cannot parse Decimal value here: {}", element.getString()); + return false; + } + break; + } + case ElementType::NULL_VALUE: { + if (!format_settings.null_as_default) + { + error = "cannot convert null to Decimal value"; + return false; + } + break; + } + default: + { + error = fmt::format("cannot read Decimal value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + } + + assert_cast &>(column).insertValue(value); + return true; + } + +private: + UInt32 scale; +}; + + +template +class DateTime64Node : public JSONExtractTreeNode, public TimezoneMixin +{ +public: + explicit DateTime64Node(const DataTypeDateTime64 & datetime64_type) : TimezoneMixin(datetime64_type), scale(datetime64_type.getScale()) + { + } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull() && format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + DateTime64 value; + if (element.isString()) + { + if (!tryParse(value, element.getString(), format_settings.date_time_input_format)) + { + error = fmt::format("cannot parse DateTime64 value here: {}", element.getString()); + return false; + } + } + else + { + switch (element.type()) + { + case ElementType::DOUBLE: + value = convertToDecimal, DataTypeDecimal>(element.getDouble(), scale); + break; + case ElementType::UINT64: + value = convertToDecimal, DataTypeDecimal>(element.getUInt64(), scale); + break; + case ElementType::INT64: + value = convertToDecimal, DataTypeDecimal>(element.getInt64(), scale); + break; + default: + error = fmt::format("cannot read DateTime64 value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + } + + assert_cast(column).insert(value); + return true; + } + + bool tryParse(DateTime64 & value, std::string_view data, FormatSettings::DateTimeInputFormat date_time_input_format) const + { + ReadBufferFromMemory buf(data.data(), data.size()); + switch (date_time_input_format) + { + case FormatSettings::DateTimeInputFormat::Basic: + if (tryReadDateTime64Text(value, scale, buf, time_zone) && buf.eof()) + return true; + break; + case FormatSettings::DateTimeInputFormat::BestEffort: + if (tryParseDateTime64BestEffort(value, scale, buf, time_zone, utc_time_zone) && buf.eof()) + return true; + break; + case FormatSettings::DateTimeInputFormat::BestEffortUS: + if (tryParseDateTime64BestEffortUS(value, scale, buf, time_zone, utc_time_zone) && buf.eof()) + return true; + break; + } + + return false; + } + +private: + UInt32 scale; +}; + +template +class EnumNode : public JSONExtractTreeNode +{ +public: + explicit EnumNode(const std::vector> & name_value_pairs_) : name_value_pairs(name_value_pairs_) + { + for (const auto & name_value_pair : name_value_pairs) + { + name_to_value_map.emplace(name_value_pair.first, name_value_pair.second); + only_values.emplace(name_value_pair.second); + } + } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull()) + { + if (format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + error = "cannot convert null to Enum value"; + return false; + } + + auto & col_vec = assert_cast &>(column); + + if (element.isInt64()) + { + Type value; + if (!accurate::convertNumeric(element.getInt64(), value) || !only_values.contains(value)) + { + error = fmt::format("cannot convert value {} to enum: there is no such value in enum", element.getInt64()); + return false; + } + col_vec.insertValue(value); + return true; + } + + if (element.isUInt64()) + { + Type value; + if (!accurate::convertNumeric(element.getUInt64(), value) || !only_values.contains(value)) + { + error = fmt::format("cannot convert value {} to enum: there is no such value in enum", element.getUInt64()); + return false; + } + col_vec.insertValue(value); + return true; + } + + if (element.isString()) + { + auto value = name_to_value_map.find(element.getString()); + if (value == name_to_value_map.end()) + { + error = fmt::format("cannot convert value {} to enum: there is no such value in enum", element.getString()); + return false; + } + col_vec.insertValue(value->second); + return true; + } + + error = fmt::format("cannot read Enum value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + +private: + std::vector> name_value_pairs; + std::unordered_map name_to_value_map; + std::unordered_set only_values; +}; + +template +class IPv4Node : public JSONExtractTreeNode +{ +public: + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull() && format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + if (!element.isString()) + { + error = fmt::format("cannot read IPv4 value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + + auto data = element.getString(); + IPv4 value; + if (!tryParse(value, data)) + { + error = fmt::format("cannot parse IPv4 value here: {}", data); + return false; + } + + assert_cast(column).insert(value); + return true; + } + + static bool tryParse(IPv4 & value, std::string_view data) + { + ReadBufferFromMemory buf(data.data(), data.size()); + return tryReadIPv4Text(value, buf) && buf.eof(); + } +}; + +template +class IPv6Node : public JSONExtractTreeNode +{ +public: + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings &, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull() && format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + if (!element.isString()) + { + error = fmt::format("cannot read IPv6 value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + + auto data = element.getString(); + IPv6 value; + if (!tryParse(value, data)) + { + error = fmt::format("cannot parse IPv6 value here: {}", data); + return false; + } + + assert_cast(column).insert(value); + return true; + } + + + static bool tryParse(IPv6 & value, std::string_view data) + { + ReadBufferFromMemory buf(data.data(), data.size()); + return tryReadIPv6Text(value, buf) && buf.eof(); + } +}; + +template +class NullableNode : public JSONExtractTreeNode +{ +public: + explicit NullableNode(std::unique_ptr> nested_) : nested(std::move(nested_)) { } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings & insert_settings, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull()) + { + column.insertDefault(); + return true; + } + + auto & col_null = assert_cast(column); + if (!nested->insertResultToColumn(col_null.getNestedColumn(), element, insert_settings, format_settings, error)) + return false; + col_null.getNullMapColumn().insertValue(0); + return true; + } + +private: + std::unique_ptr> nested; +}; + +template +class LowCardinalityNode : public JSONExtractTreeNode +{ +public: + explicit LowCardinalityNode(bool is_nullable_, std::unique_ptr> nested_) + : is_nullable(is_nullable_), nested(std::move(nested_)) + { + } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings & insert_settings, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull() && (is_nullable || format_settings.null_as_default)) + { + column.insertDefault(); + return true; + } + + auto & col_lc = assert_cast(column); + auto tmp_nested = col_lc.getDictionary().getNestedColumn()->cloneEmpty(); + if (!nested->insertResultToColumn(*tmp_nested, element, insert_settings, format_settings, error)) + return false; + + col_lc.insertFromFullColumn(*tmp_nested, 0); + return true; + } + +private: + bool is_nullable; + std::unique_ptr> nested; +}; + +template +class ArrayNode : public JSONExtractTreeNode +{ +public: + explicit ArrayNode(std::unique_ptr> nested_) : nested(std::move(nested_)) { } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings & insert_settings, + const FormatSettings & format_settings, + String & error) const override + { + if (element.isNull() && format_settings.null_as_default) + { + column.insertDefault(); + return true; + } + + if (!element.isArray()) + { + error = fmt::format("cannot read Array value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + + auto array = element.getArray(); + + auto & col_arr = assert_cast(column); + auto & data = col_arr.getData(); + size_t old_size = data.size(); + bool were_valid_elements = false; + + for (auto value : array) + { + if (nested->insertResultToColumn(data, value, insert_settings, format_settings, error)) + { + were_valid_elements = true; + } + else if (insert_settings.insert_default_on_invalid_elements_in_complex_types) + { + data.insertDefault(); + } + else + { + data.popBack(data.size() - old_size); + return false; + } + } + + if (!were_valid_elements) + { + data.popBack(data.size() - old_size); + return false; + } + + col_arr.getOffsets().push_back(data.size()); + return true; + } + +private: + std::unique_ptr> nested; +}; + +template +class TupleNode : public JSONExtractTreeNode +{ +public: + TupleNode(std::vector>> nested_, const std::vector & explicit_names_) + : nested(std::move(nested_)), explicit_names(explicit_names_) + { + for (size_t i = 0; i != explicit_names.size(); ++i) + name_to_index_map.emplace(explicit_names[i], i); + } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings & insert_settings, + const FormatSettings & format_settings, + String & error) const override + { + auto & tuple = assert_cast(column); + size_t old_size = column.size(); + bool were_valid_elements = false; + + auto set_size = [&](size_t size) + { + for (size_t i = 0; i != tuple.tupleSize(); ++i) + { + auto & col = tuple.getColumn(i); + if (col.size() != size) + { + if (col.size() > size) + col.popBack(col.size() - size); + else + while (col.size() < size) + col.insertDefault(); + } + } + }; + + if (element.isArray()) + { + auto array = element.getArray(); + auto it = array.begin(); + + for (size_t index = 0; (index != nested.size()) && (it != array.end()); ++index) + { + if (nested[index]->insertResultToColumn(tuple.getColumn(index), *it++, insert_settings, format_settings, error)) + { + were_valid_elements = true; + } + else if (insert_settings.insert_default_on_invalid_elements_in_complex_types) + { + tuple.getColumn(index).insertDefault(); + } + else + { + set_size(old_size); + error += fmt::format("(during reading tuple {} element)", index); + return false; + } + } + + set_size(old_size + static_cast(were_valid_elements)); + return were_valid_elements; + } + + if (element.isObject()) + { + auto object = element.getObject(); + if (name_to_index_map.empty()) + { + auto it = object.begin(); + for (size_t index = 0; (index != nested.size()) && (it != object.end()); ++index) + { + if (nested[index]->insertResultToColumn(tuple.getColumn(index), (*it++).second, insert_settings, format_settings, error)) + { + were_valid_elements = true; + } + else if (insert_settings.insert_default_on_invalid_elements_in_complex_types) + { + tuple.getColumn(index).insertDefault(); + } + else + { + set_size(old_size); + error += fmt::format("(during reading tuple {} element)", index); + return false; + } + } + } + else + { + for (const auto & [key, value] : object) + { + auto index = name_to_index_map.find(key); + if (index != name_to_index_map.end()) + { + if (nested[index->second]->insertResultToColumn(tuple.getColumn(index->second), value, insert_settings, format_settings, error)) + { + were_valid_elements = true; + } + else if (!insert_settings.insert_default_on_invalid_elements_in_complex_types) + { + set_size(old_size); + error += fmt::format("(during reading tuple element \"{}\")", key); + return false; + } + } + } + } + + set_size(old_size + static_cast(were_valid_elements)); + return were_valid_elements; + } + + error = fmt::format("cannot read Tuple value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + +private: + std::vector>> nested; + std::vector explicit_names; + std::unordered_map name_to_index_map; +}; + +template +class MapNode : public JSONExtractTreeNode +{ +public: + explicit MapNode(std::unique_ptr> value_) : value(std::move(value_)) { } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings & insert_settings, + const FormatSettings & format_settings, + String & error) const override + { + if (!element.isObject()) + { + error = fmt::format("cannot read Map value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + + auto & map_col = assert_cast(column); + auto & offsets = map_col.getNestedColumn().getOffsets(); + auto & tuple_col = map_col.getNestedData(); + auto & key_col = tuple_col.getColumn(0); + auto & value_col = tuple_col.getColumn(1); + size_t old_size = tuple_col.size(); + + auto object = element.getObject(); + auto it = object.begin(); + for (; it != object.end(); ++it) + { + auto pair = *it; + + /// Insert key + key_col.insertData(pair.first.data(), pair.first.size()); + + /// Insert value + if (!value->insertResultToColumn(value_col, pair.second, insert_settings, format_settings, error)) + { + if (insert_settings.insert_default_on_invalid_elements_in_complex_types) + { + value_col.insertDefault(); + } + else + { + key_col.popBack(key_col.size() - offsets.back()); + value_col.popBack(value_col.size() - offsets.back()); + error += fmt::format("(during reading value of key \"{}\")", pair.first); + return false; + } + } + } + + offsets.push_back(old_size + object.size()); + return true; + } + +private: + std::unique_ptr> value; +}; + +template +class VariantNode : public JSONExtractTreeNode +{ +public: + VariantNode(std::vector>> variant_nodes_, std::vector order_) + : variant_nodes(std::move(variant_nodes_)), order(std::move(order_)) + { + } + + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings & insert_settings, + const FormatSettings & format_settings, + String & error) const override + { + auto & column_variant = assert_cast(column); + for (size_t i : order) + { + auto & variant = column_variant.getVariantByGlobalDiscriminator(i); + if (variant_nodes[i]->insertResultToColumn(variant, element, insert_settings, format_settings, error)) + { + column_variant.getLocalDiscriminators().push_back(column_variant.localDiscriminatorByGlobal(i)); + column_variant.getOffsets().push_back(variant.size() - 1); + return true; + } + } + + error = fmt::format("cannot read Map value from JSON element: {}", jsonElementToString(element, format_settings)); + return false; + } + +private: + std::vector>> variant_nodes; + /// Order in which we should try variants nodes. + /// For example, String should be always the last one. + std::vector order; +}; + + +template +class DynamicNode : public JSONExtractTreeNode +{ +public: + bool insertResultToColumn( + IColumn & column, + const typename JSONParser::Element & element, + const JSONExtractInsertSettings & insert_settings, + const FormatSettings & format_settings, + String & error) const override + { + auto & column_dynamic = assert_cast(column); + /// First, check if element is NULL. + if (element.isNull()) + { + column_dynamic.insertDefault(); + return true; + } + + auto & variant_column = column_dynamic.getVariantColumn(); + auto variant_info = column_dynamic.getVariantInfo(); + /// Second, infer ClickHouse type for this element and add it as a new variant. + auto element_type = elementToDataType(element, format_settings); + if (column_dynamic.addNewVariant(element_type)) + { + auto node = buildJSONExtractTree(element_type, "Dynamic inference"); + auto global_discriminator = variant_info.variant_name_to_discriminator[element_type->getName()]; + auto & variant = variant_column.getVariantByGlobalDiscriminator(global_discriminator); + if (!node->insertResultToColumn(variant, element, insert_settings, format_settings, error)) + return false; + variant_column.getLocalDiscriminators().push_back(variant_column.localDiscriminatorByGlobal(global_discriminator)); + variant_column.getOffsets().push_back(variant.size() - 1); + return true; + } + + /// We couldn't add new variant. Try to insert element into current variants. + auto variant_node = buildJSONExtractTree(variant_info.variant_type, "Dynamic inference"); + if (variant_node->insertResultToColumn(variant_column, element, insert_settings, format_settings, error)) + return true; + + /// We couldn't insert element into any existing variant, add String variant and read value as String. + column_dynamic.addStringVariant(); + auto string_global_discriminator = variant_info.variant_name_to_discriminator["String"]; + auto & string_column = variant_column.getVariantByGlobalDiscriminator(string_global_discriminator); + if (!getStringNode()->insertResultToColumn(string_column, element, insert_settings, format_settings, error)) + return false; + variant_column.getLocalDiscriminators().push_back(variant_column.localDiscriminatorByGlobal(string_global_discriminator)); + variant_column.getOffsets().push_back(string_column.size() - 1); + return true; + } + + static const std::unique_ptr> & getStringNode() + { + static const std::unique_ptr> string_node + = buildJSONExtractTree(std::make_shared(), "Dynamic inference"); + return string_node; + } + + static DataTypePtr elementToDataType(const typename JSONParser::Element & element, const FormatSettings & format_settings) + { + JSONInferenceInfo json_inference_info; + auto type = elementToDataTypeImpl(element, format_settings, json_inference_info); + transformFinalInferredJSONTypeIfNeeded(type, format_settings, &json_inference_info); + return type; + } + +private: + static DataTypePtr elementToDataTypeImpl(const typename JSONParser::Element & element, const FormatSettings & format_settings, JSONInferenceInfo & json_inference_info) + { + switch (element.type()) + { + case ElementType::NULL_VALUE: + return makeNullable(std::make_shared()); + case ElementType::BOOL: + return DataTypeFactory::instance().get("Bool"); + case ElementType::INT64: + { + auto type = std::make_shared(); + if (element.getInt64() < 0) + json_inference_info.negative_integers.insert(type.get()); + return type; + } + case ElementType::UINT64: + return std::make_shared(); + case ElementType::DOUBLE: + return std::make_shared(); + case ElementType::STRING: + { + auto data = element.getString(); + + if (auto type = tryInferDateOrDateTimeFromString(data, format_settings)) + return type; + + if (format_settings.json.try_infer_numbers_from_strings) + { + if (auto type = tryInferJSONNumberFromString(data, format_settings, &json_inference_info)) + { + json_inference_info.numbers_parsed_from_json_strings.insert(type.get()); + return type; + } + } + + return std::make_shared(); + } + case ElementType::ARRAY: + { + auto array = element.getArray(); + DataTypes types; + types.reserve(array.size()); + for (auto value : array) + types.push_back(makeNullableSafe(elementToDataTypeImpl(value, format_settings, json_inference_info))); + + if (types.empty()) + return std::make_shared(makeNullable(std::make_shared())); + + if (checkIfTypesAreEqual(types)) + return std::make_shared(types.back()); + + /// For JSON if we have not complete types, we should not try to transform them + /// and return it as a Tuple. + /// For example, if we have types [Nullable(Float64), Nullable(Nothing), Nullable(Float64)] + /// it can be Array(Nullable(Float64)) or Tuple(Nullable(Float64), , Nullable(Float64)) and + /// we can't determine which one it is right now. But we will be able to do it later + /// when we will have the final top level type. + /// For example, we can have JSON element [[42.42, null, 43.43], [44.44, "Some string", 45.45]] and we should + /// determine the type for this element as Tuple(Nullable(Float64), Nullable(String), Nullable(Float64)). + for (const auto & type : types) + { + if (!checkIfTypeIsComplete(type)) + return std::make_shared(types); + } + + auto types_copy = types; + transformInferredJSONTypesIfNeeded(types_copy, format_settings, &json_inference_info); + + if (checkIfTypesAreEqual(types_copy)) + return std::make_shared(types_copy.back()); + + return std::make_shared(types); + } + case ElementType::OBJECT: { + /// TODO: Use new JSON type here when it's ready. + return std::make_shared(std::make_shared(), makeNullable(std::make_shared())); + } + } + } +}; + +} + +template +std::unique_ptr> buildJSONExtractTree(const DataTypePtr & type, const char * source_for_exception_message) +{ + switch (type->getTypeId()) + { + case TypeIndex::UInt8: + return std::make_unique>(isBool(type)); + case TypeIndex::UInt16: + return std::make_unique>(); + case TypeIndex::UInt32: + return std::make_unique>(); + case TypeIndex::UInt64: + return std::make_unique>(); + case TypeIndex::UInt128: + return std::make_unique>(); + case TypeIndex::UInt256: + return std::make_unique>(); + case TypeIndex::Int8: + return std::make_unique>(); + case TypeIndex::Int16: + return std::make_unique>(); + case TypeIndex::Int32: + return std::make_unique>(); + case TypeIndex::Int64: + return std::make_unique>(); + case TypeIndex::Int128: + return std::make_unique>(); + case TypeIndex::Int256: + return std::make_unique>(); + case TypeIndex::Float32: + return std::make_unique>(); + case TypeIndex::Float64: + return std::make_unique>(); + case TypeIndex::String: + return std::make_unique>(); + case TypeIndex::FixedString: + return std::make_unique>(assert_cast(*type).getN()); + case TypeIndex::UUID: + return std::make_unique>(); + case TypeIndex::IPv4: + return std::make_unique>(); + case TypeIndex::IPv6: + return std::make_unique>(); + case TypeIndex::Date:; + return std::make_unique>(); + case TypeIndex::Date32: + return std::make_unique>(); + case TypeIndex::DateTime: + return std::make_unique>(assert_cast(*type)); + case TypeIndex::DateTime64: + return std::make_unique>(assert_cast(*type)); + case TypeIndex::Decimal32: + return std::make_unique>(type); + case TypeIndex::Decimal64: + return std::make_unique>(type); + case TypeIndex::Decimal128: + return std::make_unique>(type); + case TypeIndex::Decimal256: + return std::make_unique>(type); + case TypeIndex::Enum8: + return std::make_unique>(assert_cast(*type).getValues()); + case TypeIndex::Enum16: + return std::make_unique>(assert_cast(*type).getValues()); + case TypeIndex::LowCardinality: + { + /// To optimize inserting into LowCardinality we have special nodes for LowCardinality of numeric and string types. + const auto & lc_type = assert_cast(*type); + auto dictionary_type = removeNullable(lc_type.getDictionaryType()); + bool is_nullable = lc_type.isLowCardinalityNullable(); + + switch (dictionary_type->getTypeId()) + { + case TypeIndex::UInt8: + return std::make_unique>(is_nullable, isBool(type)); + case TypeIndex::UInt16: + return std::make_unique>(is_nullable); + case TypeIndex::UInt32: + return std::make_unique>(is_nullable); + case TypeIndex::UInt64: + return std::make_unique>(is_nullable); + case TypeIndex::Int8: + return std::make_unique>(is_nullable); + case TypeIndex::Int16: + return std::make_unique>(is_nullable); + case TypeIndex::Int32: + return std::make_unique>(is_nullable); + case TypeIndex::Int64: + return std::make_unique>(is_nullable); + case TypeIndex::Float32: + return std::make_unique>(is_nullable); + case TypeIndex::Float64: + return std::make_unique>(is_nullable); + case TypeIndex::String: + return std::make_unique>(is_nullable); + case TypeIndex::FixedString: + return std::make_unique>(is_nullable, assert_cast(*dictionary_type).getN()); + case TypeIndex::UUID: + return std::make_unique>(is_nullable); + default: + return std::make_unique>(is_nullable, buildJSONExtractTree(dictionary_type, source_for_exception_message)); + } + } + case TypeIndex::Nullable: + return std::make_unique>(buildJSONExtractTree(assert_cast(*type).getNestedType(), source_for_exception_message)); + case TypeIndex::Array: + return std::make_unique>(buildJSONExtractTree(assert_cast(*type).getNestedType(), source_for_exception_message)); + case TypeIndex::Tuple: + { + const auto & tuple = assert_cast(*type); + const auto & tuple_elements = tuple.getElements(); + std::vector>> elements; + elements.reserve(tuple_elements.size()); + for (const auto & tuple_element : tuple_elements) + elements.emplace_back(buildJSONExtractTree(tuple_element, source_for_exception_message)); + return std::make_unique>(std::move(elements), tuple.haveExplicitNames() ? tuple.getElementNames() : Strings{}); + } + case TypeIndex::Map: + { + const auto & map_type = assert_cast(*type); + const auto & key_type = map_type.getKeyType(); + if (!isString(removeLowCardinality(key_type))) + throw Exception( + ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, + "{} doesn't support the return type schema: {} with key type not String", + source_for_exception_message, + type->getName()); + + const auto & value_type = map_type.getValueType(); + return std::make_unique>(buildJSONExtractTree(value_type, source_for_exception_message)); + } + case TypeIndex::Variant: + { + const auto & variant_type = assert_cast(*type); + const auto & variants = variant_type.getVariants(); + std::vector>> variant_nodes; + variant_nodes.reserve(variants.size()); + for (const auto & variant : variants) + variant_nodes.push_back(buildJSONExtractTree(variant, source_for_exception_message)); + return std::make_unique>(std::move(variant_nodes), SerializationVariant::getVariantsDeserializeTextOrder(variants)); + } + case TypeIndex::Dynamic: + return std::make_unique>(); + default: + throw Exception( + ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, + "{} doesn't support the return type schema: {}", + source_for_exception_message, + type->getName()); + } +} + +#if USE_SIMDJSON +template void jsonElementToString(const SimdJSONParser::Element & element, WriteBuffer & buf, const FormatSettings & format_settings); +template std::unique_ptr> buildJSONExtractTree(const DataTypePtr & type, const char * source_for_exception_message); +#endif + +#if USE_RAPIDJSON +template void jsonElementToString(const RapidJSONParser::Element & element, WriteBuffer & buf, const FormatSettings & format_settings); +template std::unique_ptr> buildJSONExtractTree(const DataTypePtr & type, const char * source_for_exception_message); +template bool tryGetNumericValueFromJSONElement(Float64 & value, const RapidJSONParser::Element & element, bool convert_bool_to_integer, String & error); +#else +template void jsonElementToString(const DummyJSONParser::Element & element, WriteBuffer & buf, const FormatSettings & format_settings); +template std::unique_ptr> buildJSONExtractTree(const DataTypePtr & type, const char * source_for_exception_message); +#endif + +} diff --git a/src/Formats/JSONExtractTree.h b/src/Formats/JSONExtractTree.h new file mode 100644 index 00000000000..b5e82506548 --- /dev/null +++ b/src/Formats/JSONExtractTree.h @@ -0,0 +1,41 @@ +#pragma once +#include +#include +#include + + +namespace DB +{ + +struct JSONExtractInsertSettings +{ + /// If false, JSON boolean values won't be inserted into columns with integer types + /// It's used in JSONExtractInt64/JSONExtractUInt64/... functions. + bool convert_bool_to_integer = true; + /// If true, when complex type like Array/Map has both valid and invalid elements, + /// the default value will be inserted on invalid elements. + /// For example, if we have [1, "hello", 2] and type Array(UInt32), + /// we will insert [1, 0, 2] in the column. Used in all JSONExtract functions. + bool insert_default_on_invalid_elements_in_complex_types = false; +}; + +template +class JSONExtractTreeNode +{ +public: + JSONExtractTreeNode() = default; + virtual ~JSONExtractTreeNode() = default; + virtual bool insertResultToColumn(IColumn &, const typename JSONParser::Element &, const JSONExtractInsertSettings & insert_setting, const FormatSettings & format_settings, String & error) const = 0; +}; + +/// Build a tree for insertion JSON element into a column with provided data type. +template +std::unique_ptr> buildJSONExtractTree(const DataTypePtr & type, const char * source_for_exception_message); + +template +void jsonElementToString(const typename JSONParser::Element & element, WriteBuffer & buf, const FormatSettings & format_settings); + +template +bool tryGetNumericValueFromJSONElement(NumberType & value, const typename JSONParser::Element & element, bool convert_bool_to_integer, String & error); + +} diff --git a/src/Formats/SchemaInferenceUtils.cpp b/src/Formats/SchemaInferenceUtils.cpp index 31faea2e13e..3c374ada9e6 100644 --- a/src/Formats/SchemaInferenceUtils.cpp +++ b/src/Formats/SchemaInferenceUtils.cpp @@ -225,19 +225,6 @@ namespace Paths paths; }; - bool checkIfTypesAreEqual(const DataTypes & types) - { - if (types.empty()) - return true; - - for (size_t i = 1; i < types.size(); ++i) - { - if (!types[0]->equals(*types[i])) - return false; - } - return true; - } - void updateTypeIndexes(DataTypes & data_types, TypeIndexesSet & type_indexes) { type_indexes.clear(); @@ -272,24 +259,31 @@ namespace type_indexes.erase(TypeIndex::Nothing); } - /// If we have both Int64 and UInt64, convert all Int64 to UInt64, + /// If we have both Int64 and UInt64, convert all not-negative Int64 to UInt64, /// because UInt64 is inferred only in case of Int64 overflow. - void transformIntegers(DataTypes & data_types, TypeIndexesSet & type_indexes) + void transformIntegers(DataTypes & data_types, TypeIndexesSet & type_indexes, JSONInferenceInfo * json_info) { if (!type_indexes.contains(TypeIndex::Int64) || !type_indexes.contains(TypeIndex::UInt64)) return; + bool have_negative_integers = false; for (auto & type : data_types) { if (WhichDataType(type).isInt64()) - type = std::make_shared(); + { + bool is_negative = json_info && json_info->negative_integers.contains(type.get()); + have_negative_integers |= is_negative; + if (!is_negative) + type = std::make_shared(); + } } - type_indexes.erase(TypeIndex::Int64); + if (!have_negative_integers) + type_indexes.erase(TypeIndex::Int64); } /// If we have both Int64 and Float64 types, convert all Int64 to Float64. - void transformIntegersAndFloatsToFloats(DataTypes & data_types, TypeIndexesSet & type_indexes) + void transformIntegersAndFloatsToFloats(DataTypes & data_types, TypeIndexesSet & type_indexes, JSONInferenceInfo * json_info) { bool have_floats = type_indexes.contains(TypeIndex::Float64); bool have_integers = type_indexes.contains(TypeIndex::Int64) || type_indexes.contains(TypeIndex::UInt64); @@ -300,7 +294,12 @@ namespace { WhichDataType which(type); if (which.isInt64() || which.isUInt64()) - type = std::make_shared(); + { + auto new_type = std::make_shared(); + if (json_info && json_info->numbers_parsed_from_json_strings.erase(type.get())) + json_info->numbers_parsed_from_json_strings.insert(new_type.get()); + type = new_type; + } } type_indexes.erase(TypeIndex::Int64); @@ -635,9 +634,9 @@ namespace if (settings.try_infer_integers) { /// Transform Int64 to UInt64 if needed. - transformIntegers(data_types, type_indexes); + transformIntegers(data_types, type_indexes, json_info); /// Transform integers to floats if needed. - transformIntegersAndFloatsToFloats(data_types, type_indexes); + transformIntegersAndFloatsToFloats(data_types, type_indexes, json_info); } /// Transform Date to DateTime or both to String if needed. @@ -887,7 +886,7 @@ namespace } template - DataTypePtr tryInferNumber(ReadBuffer & buf, const FormatSettings & settings) + DataTypePtr tryInferNumber(ReadBuffer & buf, const FormatSettings & settings, JSONInferenceInfo * json_info) { if (buf.eof()) return nullptr; @@ -911,7 +910,12 @@ namespace Int64 tmp_int; buf.position() = number_start; if (tryReadIntText(tmp_int, buf)) - return std::make_shared(); + { + auto type = std::make_shared(); + if (json_info && tmp_int < 0) + json_info->negative_integers.insert(type.get()); + return type; + } /// In case of Int64 overflow we can try to infer UInt64. UInt64 tmp_uint; @@ -934,7 +938,12 @@ namespace Int64 tmp_int; if (tryReadIntText(tmp_int, peekable_buf)) - return std::make_shared(); + { + auto type = std::make_shared(); + if (json_info && tmp_int < 0) + json_info->negative_integers.insert(type.get()); + return type; + } peekable_buf.rollbackToCheckpoint(/* drop= */ true); /// In case of Int64 overflow we can try to infer UInt64. @@ -952,7 +961,7 @@ namespace } template - DataTypePtr tryInferNumberFromStringImpl(std::string_view field, const FormatSettings & settings) + DataTypePtr tryInferNumberFromStringImpl(std::string_view field, const FormatSettings & settings, JSONInferenceInfo * json_inference_info = nullptr) { ReadBufferFromString buf(field); @@ -960,7 +969,12 @@ namespace { Int64 tmp_int; if (tryReadIntText(tmp_int, buf) && buf.eof()) - return std::make_shared(); + { + auto type = std::make_shared(); + if (json_inference_info && tmp_int < 0) + json_inference_info->negative_integers.insert(type.get()); + return type; + } /// We can safely get back to the start of buffer, because we read from a string and we didn't reach eof. buf.position() = buf.buffer().begin(); @@ -1011,7 +1025,7 @@ namespace { if (settings.json.try_infer_numbers_from_strings) { - if (auto number_type = tryInferNumberFromStringImpl(field, settings)) + if (auto number_type = tryInferNumberFromStringImpl(field, settings, json_info)) { json_info->numbers_parsed_from_json_strings.insert(number_type.get()); return number_type; @@ -1254,10 +1268,23 @@ namespace } /// Number - return tryInferNumber(buf, settings); + return tryInferNumber(buf, settings, json_info); } } +bool checkIfTypesAreEqual(const DataTypes & types) +{ + if (types.empty()) + return true; + + for (size_t i = 1; i < types.size(); ++i) + { + if (!types[0]->equals(*types[i])) + return false; + } + return true; +} + void transformInferredTypesIfNeeded(DataTypePtr & first, DataTypePtr & second, const FormatSettings & settings) { DataTypes types = {first, second}; @@ -1275,6 +1302,11 @@ void transformInferredJSONTypesIfNeeded( second = std::move(types[1]); } +void transformInferredJSONTypesIfNeeded(DataTypes & types, const FormatSettings & settings, JSONInferenceInfo * json_info) +{ + transformInferredTypesIfNeededImpl(types, settings, json_info); +} + void transformInferredJSONTypesFromDifferentFilesIfNeeded(DataTypePtr & first, DataTypePtr & second, const FormatSettings & settings) { JSONInferenceInfo json_info; @@ -1396,6 +1428,12 @@ DataTypePtr tryInferNumberFromString(std::string_view field, const FormatSetting return tryInferNumberFromStringImpl(field, settings); } +DataTypePtr tryInferJSONNumberFromString(std::string_view field, const FormatSettings & settings, JSONInferenceInfo * json_info) +{ + return tryInferNumberFromStringImpl(field, settings, json_info); + +} + DataTypePtr tryInferDateOrDateTimeFromString(std::string_view field, const FormatSettings & settings) { if (settings.try_infer_dates && tryInferDate(field)) diff --git a/src/Formats/SchemaInferenceUtils.h b/src/Formats/SchemaInferenceUtils.h index bcf3d194825..06c14c0797a 100644 --- a/src/Formats/SchemaInferenceUtils.h +++ b/src/Formats/SchemaInferenceUtils.h @@ -2,6 +2,7 @@ #include #include +#include #include @@ -18,6 +19,11 @@ struct JSONInferenceInfo /// We store numbers that were parsed from strings. /// It's used in types transformation to change such numbers back to string if needed. std::unordered_set numbers_parsed_from_json_strings; + /// Store integer types that were inferred from negative numbers. + /// It's used to determine common type for Int64 and UInt64 + /// TODO: check it not only in JSON formats. + std::unordered_set negative_integers; + /// Indicates if currently we are inferring type for Map/Object key. bool is_object_key = false; /// When we transform types for the same column from different files @@ -48,6 +54,7 @@ DataTypePtr tryInferDateOrDateTimeFromString(std::string_view field, const Forma /// Try to parse a number value from a string. By default, it tries to parse Float64, /// but if setting try_infer_integers is enabled, it also tries to parse Int64. DataTypePtr tryInferNumberFromString(std::string_view field, const FormatSettings & settings); +DataTypePtr tryInferJSONNumberFromString(std::string_view field, const FormatSettings & settings, JSONInferenceInfo * json_info); /// It takes two types inferred for the same column and tries to transform them to a common type if possible. /// It's also used when we try to infer some not ordinary types from another types. @@ -77,6 +84,7 @@ void transformInferredTypesIfNeeded(DataTypePtr & first, DataTypePtr & second, c /// Example 2: /// We merge DataTypeJSONPaths types to a single DataTypeJSONPaths type with union of all JSON paths. void transformInferredJSONTypesIfNeeded(DataTypePtr & first, DataTypePtr & second, const FormatSettings & settings, JSONInferenceInfo * json_info); +void transformInferredJSONTypesIfNeeded(DataTypes & types, const FormatSettings & settings, JSONInferenceInfo * json_info); /// Make final transform for types inferred in JSON format. It does 3 types of transformation: /// 1) Checks if type is unnamed Tuple(...), tries to transform nested types to find a common type for them and if all nested types @@ -107,4 +115,6 @@ NamesAndTypesList getNamesAndRecursivelyNullableTypes(const Block & header); /// Check if type contains Nothing, like Array(Tuple(Nullable(Nothing), String)) bool checkIfTypeIsComplete(const DataTypePtr & type); +bool checkIfTypesAreEqual(const DataTypes & types); + } diff --git a/src/Functions/FunctionsJSON.cpp b/src/Functions/FunctionsJSON.cpp index fbd987577e9..db1602b1939 100644 --- a/src/Functions/FunctionsJSON.cpp +++ b/src/Functions/FunctionsJSON.cpp @@ -1,10 +1,1063 @@ -#include +#include +#include + +#include + +#include +#include + +#include +#include + +#include + +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include #include +#include +#include +#include +#include + +#include + +#include "config.h" namespace DB { +namespace ErrorCodes +{ +extern const int ILLEGAL_TYPE_OF_ARGUMENT; +extern const int ILLEGAL_COLUMN; +extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH; +} + +template +concept HasIndexOperator = requires (T t) +{ + t[0]; +}; + +/// Functions to parse JSONs and extract values from it. +/// The first argument of all these functions gets a JSON, +/// after that there are any number of arguments specifying path to a desired part from the JSON's root. +/// For example, +/// select JSONExtractInt('{"a": "hello", "b": [-100, 200.0, 300]}', 'b', 1) = -100 + +class FunctionJSONHelpers +{ +public: + template typename Impl, class JSONParser> + class Executor + { + public: + static ColumnPtr run(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count, const FormatSettings & format_settings) + { + MutableColumnPtr to{result_type->createColumn()}; + to->reserve(input_rows_count); + + if (arguments.empty()) + throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, "Function {} requires at least one argument", String(Name::name)); + + const auto & first_column = arguments[0]; + if (!isString(first_column.type)) + throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, + "The first argument of function {} should be a string containing JSON, illegal type: " + "{}", String(Name::name), first_column.type->getName()); + + const ColumnPtr & arg_json = first_column.column; + const auto * col_json_const = typeid_cast(arg_json.get()); + const auto * col_json_string + = typeid_cast(col_json_const ? col_json_const->getDataColumnPtr().get() : arg_json.get()); + + if (!col_json_string) + throw Exception(ErrorCodes::ILLEGAL_COLUMN, "Illegal column {}", arg_json->getName()); + + const ColumnString::Chars & chars = col_json_string->getChars(); + const ColumnString::Offsets & offsets = col_json_string->getOffsets(); + + size_t num_index_arguments = Impl::getNumberOfIndexArguments(arguments); + std::vector moves = prepareMoves(Name::name, arguments, 1, num_index_arguments); + + /// Preallocate memory in parser if necessary. + JSONParser parser; + if constexpr (has_member_function_reserve::value) + { + size_t max_size = calculateMaxSize(offsets); + if (max_size) + parser.reserve(max_size); + } + + Impl impl; + + /// prepare() does Impl-specific preparation before handling each row. + if constexpr (has_member_function_prepare::*)(const char *, const ColumnsWithTypeAndName &, const DataTypePtr &)>::value) + impl.prepare(Name::name, arguments, result_type); + + using Element = typename JSONParser::Element; + + Element document; + bool document_ok = false; + if (col_json_const) + { + std::string_view json{reinterpret_cast(chars.data()), offsets[0] - 1}; + document_ok = parser.parse(json, document); + } + + String error; + for (const auto i : collections::range(0, input_rows_count)) + { + if (!col_json_const) + { + std::string_view json{reinterpret_cast(&chars[offsets[i - 1]]), offsets[i] - offsets[i - 1] - 1}; + document_ok = parser.parse(json, document); + } + + bool added_to_column = false; + if (document_ok) + { + /// Perform moves. + Element element; + std::string_view last_key; + bool moves_ok = performMoves(arguments, i, document, moves, element, last_key); + + if (moves_ok) + added_to_column = impl.insertResultToColumn(*to, element, last_key, format_settings, error); + } + + /// We add default value (=null or zero) if something goes wrong, we don't throw exceptions in these JSON functions. + if (!added_to_column) + to->insertDefault(); + } + return to; + } + }; + +private: + BOOST_TTI_HAS_MEMBER_FUNCTION(reserve) + BOOST_TTI_HAS_MEMBER_FUNCTION(prepare) + + /// Represents a move of a JSON iterator described by a single argument passed to a JSON function. + /// For example, the call JSONExtractInt('{"a": "hello", "b": [-100, 200.0, 300]}', 'b', 1) + /// contains two moves: {MoveType::ConstKey, "b"} and {MoveType::ConstIndex, 1}. + /// Keys and indices can be nonconst, in this case they are calculated for each row. + enum class MoveType : uint8_t + { + Key, + Index, + ConstKey, + ConstIndex, + }; + + struct Move + { + explicit Move(MoveType type_, size_t index_ = 0) : type(type_), index(index_) {} + Move(MoveType type_, const String & key_) : type(type_), key(key_) {} + MoveType type; + size_t index = 0; + String key; + }; + + static std::vector prepareMoves( + const char * function_name, + const ColumnsWithTypeAndName & columns, + size_t first_index_argument, + size_t num_index_arguments) + { + std::vector moves; + moves.reserve(num_index_arguments); + for (const auto i : collections::range(first_index_argument, first_index_argument + num_index_arguments)) + { + const auto & column = columns[i]; + if (!isString(column.type) && !isNativeInteger(column.type)) + throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, + "The argument {} of function {} should be a string specifying key " + "or an integer specifying index, illegal type: {}", + std::to_string(i + 1), String(function_name), column.type->getName()); + + if (column.column && isColumnConst(*column.column)) + { + const auto & column_const = assert_cast(*column.column); + if (isString(column.type)) + moves.emplace_back(MoveType::ConstKey, column_const.getValue()); + else + moves.emplace_back(MoveType::ConstIndex, column_const.getInt(0)); + } + else + { + if (isString(column.type)) + moves.emplace_back(MoveType::Key, ""); + else + moves.emplace_back(MoveType::Index, 0); + } + } + return moves; + } + + + /// Performs moves of types MoveType::Index and MoveType::ConstIndex. + template + static bool performMoves(const ColumnsWithTypeAndName & arguments, size_t row, + const typename JSONParser::Element & document, const std::vector & moves, + typename JSONParser::Element & element, std::string_view & last_key) + { + typename JSONParser::Element res_element = document; + std::string_view key; + + for (size_t j = 0; j != moves.size(); ++j) + { + switch (moves[j].type) + { + case MoveType::ConstIndex: + { + if (!moveToElementByIndex(res_element, static_cast(moves[j].index), key)) + return false; + break; + } + case MoveType::ConstKey: + { + key = moves[j].key; + if (!moveToElementByKey(res_element, key)) + return false; + break; + } + case MoveType::Index: + { + Int64 index = (*arguments[j + 1].column)[row].get(); + if (!moveToElementByIndex(res_element, static_cast(index), key)) + return false; + break; + } + case MoveType::Key: + { + key = arguments[j + 1].column->getDataAt(row).toView(); + if (!moveToElementByKey(res_element, key)) + return false; + break; + } + } + } + + element = res_element; + last_key = key; + return true; + } + + template + static bool moveToElementByIndex(typename JSONParser::Element & element, int index, std::string_view & out_key) + { + if (element.isArray()) + { + auto array = element.getArray(); + if (index >= 0) + --index; + else + index += array.size(); + + if (static_cast(index) >= array.size()) + return false; + element = array[index]; + out_key = {}; + return true; + } + + if constexpr (HasIndexOperator) + { + if (element.isObject()) + { + auto object = element.getObject(); + if (index >= 0) + --index; + else + index += object.size(); + + if (static_cast(index) >= object.size()) + return false; + std::tie(out_key, element) = object[index]; + return true; + } + } + + return {}; + } + + /// Performs moves of types MoveType::Key and MoveType::ConstKey. + template + static bool moveToElementByKey(typename JSONParser::Element & element, std::string_view key) + { + if (!element.isObject()) + return false; + auto object = element.getObject(); + return object.find(key, element); + } + + static size_t calculateMaxSize(const ColumnString::Offsets & offsets) + { + size_t max_size = 0; + for (const auto i : collections::range(0, offsets.size())) + { + size_t size = offsets[i] - offsets[i - 1]; + max_size = std::max(max_size, size); + } + if (max_size) + --max_size; + return max_size; + } + +}; + +template +class JSONExtractImpl; + +template +class JSONExtractKeysAndValuesImpl; + +/** +* Functions JSONExtract and JSONExtractKeysAndValues force the return type - it is specified in the last argument. +* For example - `SELECT JSONExtract(materialize('{"a": 131231, "b": 1234}'), 'b', 'LowCardinality(FixedString(4))')` +* But by default ClickHouse decides on its own whether the return type will be LowCardinality based on the types of +* input arguments. +* And for these specific functions we cannot rely on this mechanism, so these functions have their own implementation - +* just convert all of the LowCardinality input columns to full ones, execute and wrap the resulting column in LowCardinality +* if needed. +*/ +template typename Impl> +constexpr bool functionForcesTheReturnType() +{ + return std::is_same_v, JSONExtractImpl> || std::is_same_v, JSONExtractKeysAndValuesImpl>; +} + +template typename Impl> +class ExecutableFunctionJSON : public IExecutableFunction +{ + +public: + explicit ExecutableFunctionJSON(const NullPresence & null_presence_, bool allow_simdjson_, const DataTypePtr & json_return_type_, const FormatSettings & format_settings_) + : null_presence(null_presence_), allow_simdjson(allow_simdjson_), json_return_type(json_return_type_), format_settings(format_settings_) + { + /// Don't escape forward slashes during converting JSON elements to raw string. + format_settings.json.escape_forward_slashes = false; + /// Don't insert default values on null during traversing the JSON element. + /// We allow to insert null only to Nullable columns in JSONExtract functions. + format_settings.null_as_default = false; + } + + String getName() const override { return Name::name; } + bool useDefaultImplementationForNulls() const override { return false; } + bool useDefaultImplementationForConstants() const override { return true; } + bool useDefaultImplementationForLowCardinalityColumns() const override + { + return !functionForcesTheReturnType(); + } + + ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override + { + if (null_presence.has_null_constant) + return result_type->createColumnConstWithDefaultValue(input_rows_count); + + if constexpr (functionForcesTheReturnType()) + { + ColumnsWithTypeAndName columns_without_low_cardinality = arguments; + + for (auto & column : columns_without_low_cardinality) + { + column.column = recursiveRemoveLowCardinality(column.column); + column.type = recursiveRemoveLowCardinality(column.type); + } + + ColumnsWithTypeAndName temporary_columns = null_presence.has_nullable ? createBlockWithNestedColumns(columns_without_low_cardinality) : columns_without_low_cardinality; + ColumnPtr temporary_result = chooseAndRunJSONParser(temporary_columns, json_return_type, input_rows_count); + + if (null_presence.has_nullable) + temporary_result = wrapInNullable(temporary_result, columns_without_low_cardinality, result_type, input_rows_count); + + if (result_type->lowCardinality()) + temporary_result = recursiveLowCardinalityTypeConversion(temporary_result, json_return_type, result_type); + + return temporary_result; + } + else + { + ColumnsWithTypeAndName temporary_columns = null_presence.has_nullable ? createBlockWithNestedColumns(arguments) : arguments; + ColumnPtr temporary_result = chooseAndRunJSONParser(temporary_columns, json_return_type, input_rows_count); + + if (null_presence.has_nullable) + temporary_result = wrapInNullable(temporary_result, arguments, result_type, input_rows_count); + + if (result_type->lowCardinality()) + temporary_result = recursiveLowCardinalityTypeConversion(temporary_result, json_return_type, result_type); + + return temporary_result; + } + } + +private: + + ColumnPtr + chooseAndRunJSONParser(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const + { +#if USE_SIMDJSON + if (allow_simdjson) + return FunctionJSONHelpers::Executor::run(arguments, result_type, input_rows_count, format_settings); +#endif + +#if USE_RAPIDJSON + return FunctionJSONHelpers::Executor::run(arguments, result_type, input_rows_count, format_settings); +#else + return FunctionJSONHelpers::Executor::run(arguments, result_type, input_rows_count, format_settings); +#endif + } + + NullPresence null_presence; + bool allow_simdjson; + DataTypePtr json_return_type; + FormatSettings format_settings; +}; + + +template typename Impl> +class FunctionBaseFunctionJSON : public IFunctionBase +{ +public: + explicit FunctionBaseFunctionJSON( + const NullPresence & null_presence_, + bool allow_simdjson_, + DataTypes argument_types_, + DataTypePtr return_type_, + DataTypePtr json_return_type_, + const FormatSettings & format_settings_) + : null_presence(null_presence_) + , allow_simdjson(allow_simdjson_) + , argument_types(std::move(argument_types_)) + , return_type(std::move(return_type_)) + , json_return_type(std::move(json_return_type_)) + , format_settings(format_settings_) + { + } + + String getName() const override { return Name::name; } + + const DataTypes & getArgumentTypes() const override + { + return argument_types; + } + + const DataTypePtr & getResultType() const override + { + return return_type; + } + + bool isSuitableForShortCircuitArgumentsExecution(const DataTypesWithConstInfo & /*arguments*/) const override { return true; } + + ExecutableFunctionPtr prepare(const ColumnsWithTypeAndName &) const override + { + return std::make_unique>(null_presence, allow_simdjson, json_return_type, format_settings); + } + +private: + NullPresence null_presence; + bool allow_simdjson; + DataTypes argument_types; + DataTypePtr return_type; + DataTypePtr json_return_type; + FormatSettings format_settings; +}; + +/// We use IFunctionOverloadResolver instead of IFunction to handle non-default NULL processing. +/// Both NULL and JSON NULL should generate NULL value. If any argument is NULL, return NULL. +template typename Impl> +class JSONOverloadResolver : public IFunctionOverloadResolver, WithContext +{ +public: + static constexpr auto name = Name::name; + + String getName() const override { return name; } + + static FunctionOverloadResolverPtr create(ContextPtr context_) + { + return std::make_unique(context_); + } + + explicit JSONOverloadResolver(ContextPtr context_) : WithContext(context_) {} + + bool isVariadic() const override { return true; } + size_t getNumberOfArguments() const override { return 0; } + bool useDefaultImplementationForNulls() const override { return false; } + bool useDefaultImplementationForLowCardinalityColumns() const override + { + return !functionForcesTheReturnType(); + } + + FunctionBasePtr build(const ColumnsWithTypeAndName & arguments) const override + { + bool has_nothing_argument = false; + for (const auto & arg : arguments) + has_nothing_argument |= isNothing(arg.type); + + DataTypePtr json_return_type = Impl::getReturnType(Name::name, createBlockWithNestedColumns(arguments)); + NullPresence null_presence = getNullPresense(arguments); + DataTypePtr return_type; + if (has_nothing_argument) + return_type = std::make_shared(); + else if (null_presence.has_null_constant) + return_type = makeNullable(std::make_shared()); + else if (null_presence.has_nullable) + return_type = makeNullable(json_return_type); + else + return_type = json_return_type; + + DataTypes argument_types; + argument_types.reserve(arguments.size()); + for (const auto & argument : arguments) + argument_types.emplace_back(argument.type); + return std::make_unique>( + null_presence, getContext()->getSettingsRef().allow_simdjson, argument_types, return_type, json_return_type, getFormatSettings(getContext())); + } +}; + +struct NameJSONHas { static constexpr auto name{"JSONHas"}; }; +struct NameIsValidJSON { static constexpr auto name{"isValidJSON"}; }; +struct NameJSONLength { static constexpr auto name{"JSONLength"}; }; +struct NameJSONKey { static constexpr auto name{"JSONKey"}; }; +struct NameJSONType { static constexpr auto name{"JSONType"}; }; +struct NameJSONExtractInt { static constexpr auto name{"JSONExtractInt"}; }; +struct NameJSONExtractUInt { static constexpr auto name{"JSONExtractUInt"}; }; +struct NameJSONExtractFloat { static constexpr auto name{"JSONExtractFloat"}; }; +struct NameJSONExtractBool { static constexpr auto name{"JSONExtractBool"}; }; +struct NameJSONExtractString { static constexpr auto name{"JSONExtractString"}; }; +struct NameJSONExtract { static constexpr auto name{"JSONExtract"}; }; +struct NameJSONExtractKeysAndValues { static constexpr auto name{"JSONExtractKeysAndValues"}; }; +struct NameJSONExtractRaw { static constexpr auto name{"JSONExtractRaw"}; }; +struct NameJSONExtractArrayRaw { static constexpr auto name{"JSONExtractArrayRaw"}; }; +struct NameJSONExtractKeysAndValuesRaw { static constexpr auto name{"JSONExtractKeysAndValuesRaw"}; }; +struct NameJSONExtractKeys { static constexpr auto name{"JSONExtractKeys"}; }; + + +template +class JSONHasImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) { return std::make_shared(); } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + static bool insertResultToColumn(IColumn & dest, const Element &, std::string_view, const FormatSettings &, String &) + { + ColumnVector & col_vec = assert_cast &>(dest); + col_vec.insertValue(1); + return true; + } +}; + + +template +class IsValidJSONImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char * function_name, const ColumnsWithTypeAndName & arguments) + { + if (arguments.size() != 1) + { + /// IsValidJSON() shouldn't get parameters other than JSON. + throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, "Function {} needs exactly one argument", + String(function_name)); + } + return std::make_shared(); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName &) { return 0; } + + static bool insertResultToColumn(IColumn & dest, const Element &, std::string_view, const FormatSettings &, String &) + { + /// This function is called only if JSON is valid. + /// If JSON isn't valid then `FunctionJSON::Executor::run()` adds default value (=zero) to `dest` without calling this function. + ColumnVector & col_vec = assert_cast &>(dest); + col_vec.insertValue(1); + return true; + } +}; + + +template +class JSONLengthImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) + { + return std::make_shared(); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings &, String &) + { + size_t size; + if (element.isArray()) + size = element.getArray().size(); + else if (element.isObject()) + size = element.getObject().size(); + else + return false; + + ColumnVector & col_vec = assert_cast &>(dest); + col_vec.insertValue(size); + return true; + } +}; + + +template +class JSONKeyImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) + { + return std::make_shared(); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + static bool insertResultToColumn(IColumn & dest, const Element &, std::string_view last_key, const FormatSettings &, String &) + { + if (last_key.empty()) + return false; + ColumnString & col_str = assert_cast(dest); + col_str.insertData(last_key.data(), last_key.size()); + return true; + } +}; + + +template +class JSONTypeImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) + { + static const std::vector> values = { + {"Array", '['}, + {"Object", '{'}, + {"String", '"'}, + {"Int64", 'i'}, + {"UInt64", 'u'}, + {"Double", 'd'}, + {"Bool", 'b'}, + {"Null", 0}, /// the default value for the column. + }; + return std::make_shared>(values); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings &, String &) + { + UInt8 type; + switch (element.type()) + { + case ElementType::INT64: + type = 'i'; + break; + case ElementType::UINT64: + type = 'u'; + break; + case ElementType::DOUBLE: + type = 'd'; + break; + case ElementType::STRING: + type = '"'; + break; + case ElementType::ARRAY: + type = '['; + break; + case ElementType::OBJECT: + type = '{'; + break; + case ElementType::BOOL: + type = 'b'; + break; + case ElementType::NULL_VALUE: + type = 0; + break; + } + + ColumnVector & col_vec = assert_cast &>(dest); + col_vec.insertValue(type); + return true; + } +}; + + +template +class JSONExtractNumericImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) + { + return std::make_shared>(); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + static const std::unique_ptr> & getInsertNode() + { + static const std::unique_ptr> node = buildJSONExtractTree(std::make_shared>()); + } + + static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings &, String & error) + { + NumberType value; + + tryGetNumericValueFromJSONElement(value, element, convert_bool_to_integer, error); + auto & col_vec = assert_cast &>(dest); + col_vec.insertValue(value); + return true; + } +}; + + +template +using JSONExtractInt64Impl = JSONExtractNumericImpl; +template +using JSONExtractUInt64Impl = JSONExtractNumericImpl; +template +using JSONExtractFloat64Impl = JSONExtractNumericImpl; + + +template +class JSONExtractBoolImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) + { + return std::make_shared(); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings &, String &) + { + bool value; + switch (element.type()) + { + case ElementType::BOOL: + value = element.getBool(); + break; + case ElementType::INT64: + value = element.getInt64() != 0; + break; + case ElementType::UINT64: + value = element.getUInt64() != 0; + break; + default: + return false; + } + + auto & col_vec = assert_cast &>(dest); + col_vec.insertValue(static_cast(value)); + return true; + } +}; + +template +class JSONExtractRawImpl; + +template +class JSONExtractStringImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) + { + return std::make_shared(); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings & format_settings, String & error) + { + if (element.isNull()) + return false; + + if (!element.isString()) + return JSONExtractRawImpl::insertResultToColumn(dest, element, {}, format_settings, error); + + auto str = element.getString(); + ColumnString & col_str = assert_cast(dest); + col_str.insertData(str.data(), str.size()); + return true; + } +}; + +template +class JSONExtractImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char * function_name, const ColumnsWithTypeAndName & arguments) + { + if (arguments.size() < 2) + throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, "Function {} requires at least two arguments", String(function_name)); + + const auto & col = arguments.back(); + const auto * col_type_const = typeid_cast(col.column.get()); + if (!col_type_const || !isString(col.type)) + throw Exception(ErrorCodes::ILLEGAL_COLUMN, + "The last argument of function {} should " + "be a constant string specifying the return data type, illegal value: {}", + String(function_name), col.name); + + return DataTypeFactory::instance().get(col_type_const->getValue()); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 2; } + + void prepare(const char * function_name, const ColumnsWithTypeAndName &, const DataTypePtr & result_type) + { + extract_tree = buildJSONExtractTree(result_type, function_name); + insert_settings.insert_default_on_invalid_elements_in_complex_types = true; + } + + bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings & format_settings, String & error) + { + return extract_tree->insertResultToColumn(dest, element, insert_settings, format_settings, error); + } + +protected: + std::unique_ptr> extract_tree; + JSONExtractInsertSettings insert_settings; +}; + + +template +class JSONExtractKeysAndValuesImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char * function_name, const ColumnsWithTypeAndName & arguments) + { + if (arguments.size() < 2) + throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, "Function {} requires at least two arguments", String(function_name)); + + const auto & col = arguments.back(); + const auto * col_type_const = typeid_cast(col.column.get()); + if (!col_type_const || !isString(col.type)) + throw Exception(ErrorCodes::ILLEGAL_COLUMN, + "The last argument of function {} should " + "be a constant string specifying the values' data type, illegal value: {}", + String(function_name), col.name); + + DataTypePtr key_type = std::make_unique(); + DataTypePtr value_type = DataTypeFactory::instance().get(col_type_const->getValue()); + DataTypePtr tuple_type = std::make_unique(DataTypes{key_type, value_type}); + return std::make_unique(tuple_type); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 2; } + + void prepare(const char * function_name, const ColumnsWithTypeAndName &, const DataTypePtr & result_type) + { + const auto tuple_type = typeid_cast(result_type.get())->getNestedType(); + const auto value_type = typeid_cast(tuple_type.get())->getElements()[1]; + extract_tree = buildJSONExtractTree(value_type, function_name); + insert_settings.insert_default_on_invalid_elements_in_complex_types = true; + } + + bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings & format_settings, String & error) + { + if (!element.isObject()) + return false; + + auto object = element.getObject(); + + auto & col_arr = assert_cast(dest); + auto & col_tuple = assert_cast(col_arr.getData()); + size_t old_size = col_tuple.size(); + auto & col_key = assert_cast(col_tuple.getColumn(0)); + auto & col_value = col_tuple.getColumn(1); + + for (const auto & [key, value] : object) + { + if (extract_tree->insertResultToColumn(col_value, value, insert_settings, format_settings, error)) + col_key.insertData(key.data(), key.size()); + } + + if (col_tuple.size() == old_size) + return false; + + col_arr.getOffsets().push_back(col_tuple.size()); + return true; + } + +private: + std::unique_ptr> extract_tree; + JSONExtractInsertSettings insert_settings; +}; + + +template +class JSONExtractRawImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) + { + return std::make_shared(); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings & format_settings, String &) + { + ColumnString & col_str = assert_cast(dest); + auto & chars = col_str.getChars(); + WriteBufferFromVector buf(chars, AppendModeTag()); + jsonElementToString(element, buf, format_settings); + buf.finalize(); + chars.push_back(0); + col_str.getOffsets().push_back(chars.size()); + return true; + } +}; + + +template +class JSONExtractArrayRawImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) + { + return std::make_shared(std::make_shared()); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings & format_settings, String & error) + { + if (!element.isArray()) + return false; + + auto array = element.getArray(); + ColumnArray & col_res = assert_cast(dest); + + for (auto value : array) + JSONExtractRawImpl::insertResultToColumn(col_res.getData(), value, {}, format_settings, error); + + col_res.getOffsets().push_back(col_res.getOffsets().back() + array.size()); + return true; + } +}; + + +template +class JSONExtractKeysAndValuesRawImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) + { + DataTypePtr string_type = std::make_unique(); + DataTypePtr tuple_type = std::make_unique(DataTypes{string_type, string_type}); + return std::make_unique(tuple_type); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings & format_settings, String & error) + { + if (!element.isObject()) + return false; + + auto object = element.getObject(); + + auto & col_arr = assert_cast(dest); + auto & col_tuple = assert_cast(col_arr.getData()); + auto & col_key = assert_cast(col_tuple.getColumn(0)); + auto & col_value = assert_cast(col_tuple.getColumn(1)); + + for (const auto & [key, value] : object) + { + col_key.insertData(key.data(), key.size()); + JSONExtractRawImpl::insertResultToColumn(col_value, value, {}, format_settings, error); + } + + col_arr.getOffsets().push_back(col_arr.getOffsets().back() + object.size()); + return true; + } +}; + +template +class JSONExtractKeysImpl +{ +public: + using Element = typename JSONParser::Element; + + static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) + { + return std::make_unique(std::make_shared()); + } + + static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } + + bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view, const FormatSettings &, String &) + { + if (!element.isObject()) + return false; + + auto object = element.getObject(); + + ColumnArray & col_res = assert_cast(dest); + auto & col_key = assert_cast(col_res.getData()); + + for (const auto & [key, value] : object) + { + col_key.insertData(key.data(), key.size()); + } + + col_res.getOffsets().push_back(col_res.getOffsets().back() + object.size()); + return true; + } +}; + REGISTER_FUNCTION(JSON) { factory.registerFunction>(); diff --git a/src/Functions/FunctionsJSON.h b/src/Functions/FunctionsJSON.h deleted file mode 100644 index 8a2ad457d34..00000000000 --- a/src/Functions/FunctionsJSON.h +++ /dev/null @@ -1,1781 +0,0 @@ -#pragma once - -#include -#include - -#include - -#include -#include -#include - -#include -#include - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include -#include -#include - -#include -#include - - -#include "config.h" - - -namespace DB -{ - -namespace ErrorCodes -{ - extern const int ILLEGAL_TYPE_OF_ARGUMENT; - extern const int ILLEGAL_COLUMN; - extern const int NUMBER_OF_ARGUMENTS_DOESNT_MATCH; -} - -template -concept HasIndexOperator = requires (T t) -{ - t[0]; -}; - -/// Functions to parse JSONs and extract values from it. -/// The first argument of all these functions gets a JSON, -/// after that there are any number of arguments specifying path to a desired part from the JSON's root. -/// For example, -/// select JSONExtractInt('{"a": "hello", "b": [-100, 200.0, 300]}', 'b', 1) = -100 - -class FunctionJSONHelpers -{ -public: - template typename Impl, class JSONParser> - class Executor - { - public: - static ColumnPtr run(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) - { - MutableColumnPtr to{result_type->createColumn()}; - to->reserve(input_rows_count); - - if (arguments.empty()) - throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, "Function {} requires at least one argument", String(Name::name)); - - const auto & first_column = arguments[0]; - if (!isString(first_column.type)) - throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, - "The first argument of function {} should be a string containing JSON, illegal type: " - "{}", String(Name::name), first_column.type->getName()); - - const ColumnPtr & arg_json = first_column.column; - const auto * col_json_const = typeid_cast(arg_json.get()); - const auto * col_json_string - = typeid_cast(col_json_const ? col_json_const->getDataColumnPtr().get() : arg_json.get()); - - if (!col_json_string) - throw Exception(ErrorCodes::ILLEGAL_COLUMN, "Illegal column {}", arg_json->getName()); - - const ColumnString::Chars & chars = col_json_string->getChars(); - const ColumnString::Offsets & offsets = col_json_string->getOffsets(); - - size_t num_index_arguments = Impl::getNumberOfIndexArguments(arguments); - std::vector moves = prepareMoves(Name::name, arguments, 1, num_index_arguments); - - /// Preallocate memory in parser if necessary. - JSONParser parser; - if constexpr (has_member_function_reserve::value) - { - size_t max_size = calculateMaxSize(offsets); - if (max_size) - parser.reserve(max_size); - } - - Impl impl; - - /// prepare() does Impl-specific preparation before handling each row. - if constexpr (has_member_function_prepare::*)(const char *, const ColumnsWithTypeAndName &, const DataTypePtr &)>::value) - impl.prepare(Name::name, arguments, result_type); - - using Element = typename JSONParser::Element; - - Element document; - bool document_ok = false; - if (col_json_const) - { - std::string_view json{reinterpret_cast(chars.data()), offsets[0] - 1}; - document_ok = parser.parse(json, document); - } - - for (const auto i : collections::range(0, input_rows_count)) - { - if (!col_json_const) - { - std::string_view json{reinterpret_cast(&chars[offsets[i - 1]]), offsets[i] - offsets[i - 1] - 1}; - document_ok = parser.parse(json, document); - } - - bool added_to_column = false; - if (document_ok) - { - /// Perform moves. - Element element; - std::string_view last_key; - bool moves_ok = performMoves(arguments, i, document, moves, element, last_key); - - if (moves_ok) - added_to_column = impl.insertResultToColumn(*to, element, last_key); - } - - /// We add default value (=null or zero) if something goes wrong, we don't throw exceptions in these JSON functions. - if (!added_to_column) - to->insertDefault(); - } - return to; - } - }; - -private: - BOOST_TTI_HAS_MEMBER_FUNCTION(reserve) - BOOST_TTI_HAS_MEMBER_FUNCTION(prepare) - - /// Represents a move of a JSON iterator described by a single argument passed to a JSON function. - /// For example, the call JSONExtractInt('{"a": "hello", "b": [-100, 200.0, 300]}', 'b', 1) - /// contains two moves: {MoveType::ConstKey, "b"} and {MoveType::ConstIndex, 1}. - /// Keys and indices can be nonconst, in this case they are calculated for each row. - enum class MoveType : uint8_t - { - Key, - Index, - ConstKey, - ConstIndex, - }; - - struct Move - { - explicit Move(MoveType type_, size_t index_ = 0) : type(type_), index(index_) {} - Move(MoveType type_, const String & key_) : type(type_), key(key_) {} - MoveType type; - size_t index = 0; - String key; - }; - - static std::vector prepareMoves( - const char * function_name, - const ColumnsWithTypeAndName & columns, - size_t first_index_argument, - size_t num_index_arguments) - { - std::vector moves; - moves.reserve(num_index_arguments); - for (const auto i : collections::range(first_index_argument, first_index_argument + num_index_arguments)) - { - const auto & column = columns[i]; - if (!isString(column.type) && !isNativeInteger(column.type)) - throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, - "The argument {} of function {} should be a string specifying key " - "or an integer specifying index, illegal type: {}", - std::to_string(i + 1), String(function_name), column.type->getName()); - - if (column.column && isColumnConst(*column.column)) - { - const auto & column_const = assert_cast(*column.column); - if (isString(column.type)) - moves.emplace_back(MoveType::ConstKey, column_const.getValue()); - else - moves.emplace_back(MoveType::ConstIndex, column_const.getInt(0)); - } - else - { - if (isString(column.type)) - moves.emplace_back(MoveType::Key, ""); - else - moves.emplace_back(MoveType::Index, 0); - } - } - return moves; - } - - - /// Performs moves of types MoveType::Index and MoveType::ConstIndex. - template - static bool performMoves(const ColumnsWithTypeAndName & arguments, size_t row, - const typename JSONParser::Element & document, const std::vector & moves, - typename JSONParser::Element & element, std::string_view & last_key) - { - typename JSONParser::Element res_element = document; - std::string_view key; - - for (size_t j = 0; j != moves.size(); ++j) - { - switch (moves[j].type) - { - case MoveType::ConstIndex: - { - if (!moveToElementByIndex(res_element, static_cast(moves[j].index), key)) - return false; - break; - } - case MoveType::ConstKey: - { - key = moves[j].key; - if (!moveToElementByKey(res_element, key)) - return false; - break; - } - case MoveType::Index: - { - Int64 index = (*arguments[j + 1].column)[row].get(); - if (!moveToElementByIndex(res_element, static_cast(index), key)) - return false; - break; - } - case MoveType::Key: - { - key = arguments[j + 1].column->getDataAt(row).toView(); - if (!moveToElementByKey(res_element, key)) - return false; - break; - } - } - } - - element = res_element; - last_key = key; - return true; - } - - template - static bool moveToElementByIndex(typename JSONParser::Element & element, int index, std::string_view & out_key) - { - if (element.isArray()) - { - auto array = element.getArray(); - if (index >= 0) - --index; - else - index += array.size(); - - if (static_cast(index) >= array.size()) - return false; - element = array[index]; - out_key = {}; - return true; - } - - if constexpr (HasIndexOperator) - { - if (element.isObject()) - { - auto object = element.getObject(); - if (index >= 0) - --index; - else - index += object.size(); - - if (static_cast(index) >= object.size()) - return false; - std::tie(out_key, element) = object[index]; - return true; - } - } - - return {}; - } - - /// Performs moves of types MoveType::Key and MoveType::ConstKey. - template - static bool moveToElementByKey(typename JSONParser::Element & element, std::string_view key) - { - if (!element.isObject()) - return false; - auto object = element.getObject(); - return object.find(key, element); - } - - static size_t calculateMaxSize(const ColumnString::Offsets & offsets) - { - size_t max_size = 0; - for (const auto i : collections::range(0, offsets.size())) - { - size_t size = offsets[i] - offsets[i - 1]; - max_size = std::max(max_size, size); - } - if (max_size) - --max_size; - return max_size; - } - -}; - -template -class JSONExtractImpl; - -template -class JSONExtractKeysAndValuesImpl; - -/** -* Functions JSONExtract and JSONExtractKeysAndValues force the return type - it is specified in the last argument. -* For example - `SELECT JSONExtract(materialize('{"a": 131231, "b": 1234}'), 'b', 'LowCardinality(FixedString(4))')` -* But by default ClickHouse decides on its own whether the return type will be LowCardinality based on the types of -* input arguments. -* And for these specific functions we cannot rely on this mechanism, so these functions have their own implementation - -* just convert all of the LowCardinality input columns to full ones, execute and wrap the resulting column in LowCardinality -* if needed. -*/ -template typename Impl> -constexpr bool functionForcesTheReturnType() -{ - return std::is_same_v, JSONExtractImpl> || std::is_same_v, JSONExtractKeysAndValuesImpl>; -} - -template typename Impl> -class ExecutableFunctionJSON : public IExecutableFunction -{ - -public: - explicit ExecutableFunctionJSON(const NullPresence & null_presence_, bool allow_simdjson_, const DataTypePtr & json_return_type_) - : null_presence(null_presence_), allow_simdjson(allow_simdjson_), json_return_type(json_return_type_) - { - } - - String getName() const override { return Name::name; } - bool useDefaultImplementationForNulls() const override { return false; } - bool useDefaultImplementationForConstants() const override { return true; } - bool useDefaultImplementationForLowCardinalityColumns() const override - { - return !functionForcesTheReturnType(); - } - - ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override - { - if (null_presence.has_null_constant) - return result_type->createColumnConstWithDefaultValue(input_rows_count); - - if constexpr (functionForcesTheReturnType()) - { - ColumnsWithTypeAndName columns_without_low_cardinality = arguments; - - for (auto & column : columns_without_low_cardinality) - { - column.column = recursiveRemoveLowCardinality(column.column); - column.type = recursiveRemoveLowCardinality(column.type); - } - - ColumnsWithTypeAndName temporary_columns = null_presence.has_nullable ? createBlockWithNestedColumns(columns_without_low_cardinality) : columns_without_low_cardinality; - ColumnPtr temporary_result = chooseAndRunJSONParser(temporary_columns, json_return_type, input_rows_count); - - if (null_presence.has_nullable) - temporary_result = wrapInNullable(temporary_result, columns_without_low_cardinality, result_type, input_rows_count); - - if (result_type->lowCardinality()) - temporary_result = recursiveLowCardinalityTypeConversion(temporary_result, json_return_type, result_type); - - return temporary_result; - } - else - { - ColumnsWithTypeAndName temporary_columns = null_presence.has_nullable ? createBlockWithNestedColumns(arguments) : arguments; - ColumnPtr temporary_result = chooseAndRunJSONParser(temporary_columns, json_return_type, input_rows_count); - - if (null_presence.has_nullable) - temporary_result = wrapInNullable(temporary_result, arguments, result_type, input_rows_count); - - if (result_type->lowCardinality()) - temporary_result = recursiveLowCardinalityTypeConversion(temporary_result, json_return_type, result_type); - - return temporary_result; - } - } - -private: - - ColumnPtr - chooseAndRunJSONParser(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const - { -#if USE_SIMDJSON - if (allow_simdjson) - return FunctionJSONHelpers::Executor::run(arguments, result_type, input_rows_count); -#endif - -#if USE_RAPIDJSON - return FunctionJSONHelpers::Executor::run(arguments, result_type, input_rows_count); -#else - return FunctionJSONHelpers::Executor::run(arguments, result_type, input_rows_count); -#endif - } - - NullPresence null_presence; - bool allow_simdjson; - DataTypePtr json_return_type; -}; - - -template typename Impl> -class FunctionBaseFunctionJSON : public IFunctionBase -{ -public: - explicit FunctionBaseFunctionJSON( - const NullPresence & null_presence_, - bool allow_simdjson_, - DataTypes argument_types_, - DataTypePtr return_type_, - DataTypePtr json_return_type_) - : null_presence(null_presence_) - , allow_simdjson(allow_simdjson_) - , argument_types(std::move(argument_types_)) - , return_type(std::move(return_type_)) - , json_return_type(std::move(json_return_type_)) - { - } - - String getName() const override { return Name::name; } - - const DataTypes & getArgumentTypes() const override - { - return argument_types; - } - - const DataTypePtr & getResultType() const override - { - return return_type; - } - - bool isSuitableForShortCircuitArgumentsExecution(const DataTypesWithConstInfo & /*arguments*/) const override { return true; } - - ExecutableFunctionPtr prepare(const ColumnsWithTypeAndName &) const override - { - return std::make_unique>(null_presence, allow_simdjson, json_return_type); - } - -private: - NullPresence null_presence; - bool allow_simdjson; - DataTypes argument_types; - DataTypePtr return_type; - DataTypePtr json_return_type; -}; - -/// We use IFunctionOverloadResolver instead of IFunction to handle non-default NULL processing. -/// Both NULL and JSON NULL should generate NULL value. If any argument is NULL, return NULL. -template typename Impl> -class JSONOverloadResolver : public IFunctionOverloadResolver, WithContext -{ -public: - static constexpr auto name = Name::name; - - String getName() const override { return name; } - - static FunctionOverloadResolverPtr create(ContextPtr context_) - { - return std::make_unique(context_); - } - - explicit JSONOverloadResolver(ContextPtr context_) : WithContext(context_) {} - - bool isVariadic() const override { return true; } - size_t getNumberOfArguments() const override { return 0; } - bool useDefaultImplementationForNulls() const override { return false; } - bool useDefaultImplementationForLowCardinalityColumns() const override - { - return !functionForcesTheReturnType(); - } - - FunctionBasePtr build(const ColumnsWithTypeAndName & arguments) const override - { - bool has_nothing_argument = false; - for (const auto & arg : arguments) - has_nothing_argument |= isNothing(arg.type); - - DataTypePtr json_return_type = Impl::getReturnType(Name::name, createBlockWithNestedColumns(arguments)); - NullPresence null_presence = getNullPresense(arguments); - DataTypePtr return_type; - if (has_nothing_argument) - return_type = std::make_shared(); - else if (null_presence.has_null_constant) - return_type = makeNullable(std::make_shared()); - else if (null_presence.has_nullable) - return_type = makeNullable(json_return_type); - else - return_type = json_return_type; - - /// Top-level LowCardinality columns are processed outside JSON parser. - json_return_type = removeLowCardinality(json_return_type); - - DataTypes argument_types; - argument_types.reserve(arguments.size()); - for (const auto & argument : arguments) - argument_types.emplace_back(argument.type); - return std::make_unique>( - null_presence, getContext()->getSettingsRef().allow_simdjson, argument_types, return_type, json_return_type); - } -}; - -struct NameJSONHas { static constexpr auto name{"JSONHas"}; }; -struct NameIsValidJSON { static constexpr auto name{"isValidJSON"}; }; -struct NameJSONLength { static constexpr auto name{"JSONLength"}; }; -struct NameJSONKey { static constexpr auto name{"JSONKey"}; }; -struct NameJSONType { static constexpr auto name{"JSONType"}; }; -struct NameJSONExtractInt { static constexpr auto name{"JSONExtractInt"}; }; -struct NameJSONExtractUInt { static constexpr auto name{"JSONExtractUInt"}; }; -struct NameJSONExtractFloat { static constexpr auto name{"JSONExtractFloat"}; }; -struct NameJSONExtractBool { static constexpr auto name{"JSONExtractBool"}; }; -struct NameJSONExtractString { static constexpr auto name{"JSONExtractString"}; }; -struct NameJSONExtract { static constexpr auto name{"JSONExtract"}; }; -struct NameJSONExtractKeysAndValues { static constexpr auto name{"JSONExtractKeysAndValues"}; }; -struct NameJSONExtractRaw { static constexpr auto name{"JSONExtractRaw"}; }; -struct NameJSONExtractArrayRaw { static constexpr auto name{"JSONExtractArrayRaw"}; }; -struct NameJSONExtractKeysAndValuesRaw { static constexpr auto name{"JSONExtractKeysAndValuesRaw"}; }; -struct NameJSONExtractKeys { static constexpr auto name{"JSONExtractKeys"}; }; - - -template -class JSONHasImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) { return std::make_shared(); } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - static bool insertResultToColumn(IColumn & dest, const Element &, std::string_view) - { - ColumnVector & col_vec = assert_cast &>(dest); - col_vec.insertValue(1); - return true; - } -}; - - -template -class IsValidJSONImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char * function_name, const ColumnsWithTypeAndName & arguments) - { - if (arguments.size() != 1) - { - /// IsValidJSON() shouldn't get parameters other than JSON. - throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, "Function {} needs exactly one argument", - String(function_name)); - } - return std::make_shared(); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName &) { return 0; } - - static bool insertResultToColumn(IColumn & dest, const Element &, std::string_view) - { - /// This function is called only if JSON is valid. - /// If JSON isn't valid then `FunctionJSON::Executor::run()` adds default value (=zero) to `dest` without calling this function. - ColumnVector & col_vec = assert_cast &>(dest); - col_vec.insertValue(1); - return true; - } -}; - - -template -class JSONLengthImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) - { - return std::make_shared(); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - size_t size; - if (element.isArray()) - size = element.getArray().size(); - else if (element.isObject()) - size = element.getObject().size(); - else - return false; - - ColumnVector & col_vec = assert_cast &>(dest); - col_vec.insertValue(size); - return true; - } -}; - - -template -class JSONKeyImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) - { - return std::make_shared(); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - static bool insertResultToColumn(IColumn & dest, const Element &, std::string_view last_key) - { - if (last_key.empty()) - return false; - ColumnString & col_str = assert_cast(dest); - col_str.insertData(last_key.data(), last_key.size()); - return true; - } -}; - - -template -class JSONTypeImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) - { - static const std::vector> values = { - {"Array", '['}, - {"Object", '{'}, - {"String", '"'}, - {"Int64", 'i'}, - {"UInt64", 'u'}, - {"Double", 'd'}, - {"Bool", 'b'}, - {"Null", 0}, /// the default value for the column. - }; - return std::make_shared>(values); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - UInt8 type; - switch (element.type()) - { - case ElementType::INT64: - type = 'i'; - break; - case ElementType::UINT64: - type = 'u'; - break; - case ElementType::DOUBLE: - type = 'd'; - break; - case ElementType::STRING: - type = '"'; - break; - case ElementType::ARRAY: - type = '['; - break; - case ElementType::OBJECT: - type = '{'; - break; - case ElementType::BOOL: - type = 'b'; - break; - case ElementType::NULL_VALUE: - type = 0; - break; - } - - ColumnVector & col_vec = assert_cast &>(dest); - col_vec.insertValue(type); - return true; - } -}; - - -template -class JSONExtractNumericImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) - { - return std::make_shared>(); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - NumberType value; - - switch (element.type()) - { - case ElementType::DOUBLE: - if constexpr (std::is_floating_point_v) - { - /// We permit inaccurate conversion of double to float. - /// Example: double 0.1 from JSON is not representable in float. - /// But it will be more convenient for user to perform conversion. - value = static_cast(element.getDouble()); - } - else if (!accurate::convertNumeric(element.getDouble(), value)) - return false; - break; - case ElementType::UINT64: - if (!accurate::convertNumeric(element.getUInt64(), value)) - return false; - break; - case ElementType::INT64: - if (!accurate::convertNumeric(element.getInt64(), value)) - return false; - break; - case ElementType::BOOL: - if constexpr (is_integer && convert_bool_to_integer) - { - value = static_cast(element.getBool()); - break; - } - return false; - case ElementType::STRING: - { - auto rb = ReadBufferFromMemory{element.getString()}; - if constexpr (std::is_floating_point_v) - { - if (!tryReadFloatText(value, rb) || !rb.eof()) - return false; - } - else - { - if (tryReadIntText(value, rb) && rb.eof()) - break; - - /// Try to parse float and convert it to integer. - Float64 tmp_float; - rb.position() = rb.buffer().begin(); - if (!tryReadFloatText(tmp_float, rb) || !rb.eof()) - return false; - - if (!accurate::convertNumeric(tmp_float, value)) - return false; - } - break; - } - default: - return false; - } - - if (dest.getDataType() == TypeIndex::LowCardinality) - { - ColumnLowCardinality & col_low = assert_cast(dest); - col_low.insertData(reinterpret_cast(&value), sizeof(value)); - } - else - { - auto & col_vec = assert_cast &>(dest); - col_vec.insertValue(value); - } - return true; - } -}; - - -template -using JSONExtractInt64Impl = JSONExtractNumericImpl; -template -using JSONExtractUInt64Impl = JSONExtractNumericImpl; -template -using JSONExtractFloat64Impl = JSONExtractNumericImpl; - - -template -class JSONExtractBoolImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) - { - return std::make_shared(); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - bool value; - switch (element.type()) - { - case ElementType::BOOL: - value = element.getBool(); - break; - case ElementType::INT64: - value = element.getInt64() != 0; - break; - case ElementType::UINT64: - value = element.getUInt64() != 0; - break; - default: - return false; - } - - auto & col_vec = assert_cast &>(dest); - col_vec.insertValue(static_cast(value)); - return true; - } -}; - -template -class JSONExtractRawImpl; - -template -class JSONExtractStringImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) - { - return std::make_shared(); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - if (element.isNull()) - return false; - - if (!element.isString()) - return JSONExtractRawImpl::insertResultToColumn(dest, element, {}); - - auto str = element.getString(); - - if (dest.getDataType() == TypeIndex::LowCardinality) - { - ColumnLowCardinality & col_low = assert_cast(dest); - col_low.insertData(str.data(), str.size()); - } - else - { - ColumnString & col_str = assert_cast(dest); - col_str.insertData(str.data(), str.size()); - } - return true; - } -}; - -/// Nodes of the extract tree. We need the extract tree to extract from JSON complex values containing array, tuples or nullables. -template -struct JSONExtractTree -{ - using Element = typename JSONParser::Element; - - class Node - { - public: - Node() = default; - virtual ~Node() = default; - virtual bool insertResultToColumn(IColumn &, const Element &) = 0; - }; - - template - class NumericNode : public Node - { - public: - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - return JSONExtractNumericImpl::insertResultToColumn(dest, element, {}); - } - }; - - class LowCardinalityFixedStringNode : public Node - { - public: - explicit LowCardinalityFixedStringNode(const size_t fixed_length_) : fixed_length(fixed_length_) { } - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - // If element is an object we delegate the insertion to JSONExtractRawImpl - if (element.isObject()) - return JSONExtractRawImpl::insertResultToLowCardinalityFixedStringColumn(dest, element, fixed_length); - else if (!element.isString()) - return false; - - auto str = element.getString(); - if (str.size() > fixed_length) - return false; - - // For the non low cardinality case of FixedString, the padding is done in the FixedString Column implementation. - // In order to avoid having to pass the data to a FixedString Column and read it back (which would slow down the execution) - // the data is padded here and written directly to the Low Cardinality Column - if (str.size() == fixed_length) - { - assert_cast(dest).insertData(str.data(), str.size()); - } - else - { - String padded_str(str); - padded_str.resize(fixed_length, '\0'); - - assert_cast(dest).insertData(padded_str.data(), padded_str.size()); - } - return true; - } - - private: - const size_t fixed_length; - }; - - class UUIDNode : public Node - { - public: - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - if (!element.isString()) - return false; - - auto uuid = parseFromString(element.getString()); - if (dest.getDataType() == TypeIndex::LowCardinality) - { - ColumnLowCardinality & col_low = assert_cast(dest); - col_low.insertData(reinterpret_cast(&uuid), sizeof(uuid)); - } - else - { - assert_cast(dest).insert(uuid); - } - return true; - } - }; - - template - class DecimalNode : public Node - { - public: - explicit DecimalNode(DataTypePtr data_type_) : data_type(data_type_) {} - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - const auto * type = assert_cast *>(data_type.get()); - - DecimalType value{}; - - switch (element.type()) - { - case ElementType::DOUBLE: - value = convertToDecimal, DataTypeDecimal>( - element.getDouble(), type->getScale()); - break; - case ElementType::UINT64: - value = convertToDecimal, DataTypeDecimal>( - element.getUInt64(), type->getScale()); - break; - case ElementType::INT64: - value = convertToDecimal, DataTypeDecimal>( - element.getInt64(), type->getScale()); - break; - case ElementType::STRING: { - auto rb = ReadBufferFromMemory{element.getString()}; - if (!SerializationDecimal::tryReadText(value, rb, DecimalUtils::max_precision, type->getScale())) - return false; - break; - } - default: - return false; - } - - assert_cast &>(dest).insertValue(value); - return true; - } - - private: - DataTypePtr data_type; - }; - - class StringNode : public Node - { - public: - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - return JSONExtractStringImpl::insertResultToColumn(dest, element, {}); - } - }; - - class FixedStringNode : public Node - { - public: - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - if (element.isNull()) - return false; - - if (!element.isString()) - return JSONExtractRawImpl::insertResultToFixedStringColumn(dest, element, {}); - - auto str = element.getString(); - auto & col_str = assert_cast(dest); - if (str.size() > col_str.getN()) - return false; - col_str.insertData(str.data(), str.size()); - - return true; - } - }; - - template - class EnumNode : public Node - { - public: - explicit EnumNode(const std::vector> & name_value_pairs_) : name_value_pairs(name_value_pairs_) - { - for (const auto & name_value_pair : name_value_pairs) - { - name_to_value_map.emplace(name_value_pair.first, name_value_pair.second); - only_values.emplace(name_value_pair.second); - } - } - - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - auto & col_vec = assert_cast &>(dest); - - if (element.isInt64()) - { - Type value; - if (!accurate::convertNumeric(element.getInt64(), value) || !only_values.contains(value)) - return false; - col_vec.insertValue(value); - return true; - } - - if (element.isUInt64()) - { - Type value; - if (!accurate::convertNumeric(element.getUInt64(), value) || !only_values.contains(value)) - return false; - col_vec.insertValue(value); - return true; - } - - if (element.isString()) - { - auto value = name_to_value_map.find(element.getString()); - if (value == name_to_value_map.end()) - return false; - col_vec.insertValue(value->second); - return true; - } - - return false; - } - - private: - std::vector> name_value_pairs; - std::unordered_map name_to_value_map; - std::unordered_set only_values; - }; - - class NullableNode : public Node - { - public: - explicit NullableNode(std::unique_ptr nested_) : nested(std::move(nested_)) {} - - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - if (dest.getDataType() == TypeIndex::LowCardinality) - { - /// We do not need to handle nullability in that case - /// because nested node handles LowCardinality columns and will call proper overload of `insertData` - return nested->insertResultToColumn(dest, element); - } - - ColumnNullable & col_null = assert_cast(dest); - if (!nested->insertResultToColumn(col_null.getNestedColumn(), element)) - return false; - col_null.getNullMapColumn().insertValue(0); - return true; - } - - private: - std::unique_ptr nested; - }; - - class ArrayNode : public Node - { - public: - explicit ArrayNode(std::unique_ptr nested_) : nested(std::move(nested_)) {} - - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - if (!element.isArray()) - return false; - - auto array = element.getArray(); - - ColumnArray & col_arr = assert_cast(dest); - auto & data = col_arr.getData(); - size_t old_size = data.size(); - bool were_valid_elements = false; - - for (auto value : array) - { - if (nested->insertResultToColumn(data, value)) - were_valid_elements = true; - else - data.insertDefault(); - } - - if (!were_valid_elements) - { - data.popBack(data.size() - old_size); - return false; - } - - col_arr.getOffsets().push_back(data.size()); - return true; - } - - private: - std::unique_ptr nested; - }; - - class TupleNode : public Node - { - public: - TupleNode(std::vector> nested_, const std::vector & explicit_names_) : nested(std::move(nested_)), explicit_names(explicit_names_) - { - for (size_t i = 0; i != explicit_names.size(); ++i) - name_to_index_map.emplace(explicit_names[i], i); - } - - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - ColumnTuple & tuple = assert_cast(dest); - size_t old_size = dest.size(); - bool were_valid_elements = false; - - auto set_size = [&](size_t size) - { - for (size_t i = 0; i != tuple.tupleSize(); ++i) - { - auto & col = tuple.getColumn(i); - if (col.size() != size) - { - if (col.size() > size) - col.popBack(col.size() - size); - else - while (col.size() < size) - col.insertDefault(); - } - } - }; - - if (element.isArray()) - { - auto array = element.getArray(); - auto it = array.begin(); - - for (size_t index = 0; (index != nested.size()) && (it != array.end()); ++index) - { - if (nested[index]->insertResultToColumn(tuple.getColumn(index), *it++)) - were_valid_elements = true; - else - tuple.getColumn(index).insertDefault(); - } - - set_size(old_size + static_cast(were_valid_elements)); - return were_valid_elements; - } - - if (element.isObject()) - { - auto object = element.getObject(); - if (name_to_index_map.empty()) - { - auto it = object.begin(); - for (size_t index = 0; (index != nested.size()) && (it != object.end()); ++index) - { - if (nested[index]->insertResultToColumn(tuple.getColumn(index), (*it++).second)) - were_valid_elements = true; - else - tuple.getColumn(index).insertDefault(); - } - } - else - { - for (const auto & [key, value] : object) - { - auto index = name_to_index_map.find(key); - if (index != name_to_index_map.end()) - { - if (nested[index->second]->insertResultToColumn(tuple.getColumn(index->second), value)) - were_valid_elements = true; - } - } - } - - set_size(old_size + static_cast(were_valid_elements)); - return were_valid_elements; - } - - return false; - } - - private: - std::vector> nested; - std::vector explicit_names; - std::unordered_map name_to_index_map; - }; - - class MapNode : public Node - { - public: - MapNode(std::unique_ptr key_, std::unique_ptr value_) : key(std::move(key_)), value(std::move(value_)) { } - - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - if (!element.isObject()) - return false; - - ColumnMap & map_col = assert_cast(dest); - auto & offsets = map_col.getNestedColumn().getOffsets(); - auto & tuple_col = map_col.getNestedData(); - auto & key_col = tuple_col.getColumn(0); - auto & value_col = tuple_col.getColumn(1); - size_t old_size = tuple_col.size(); - - auto object = element.getObject(); - auto it = object.begin(); - for (; it != object.end(); ++it) - { - auto pair = *it; - - /// Insert key - key_col.insertData(pair.first.data(), pair.first.size()); - - /// Insert value - if (!value->insertResultToColumn(value_col, pair.second)) - value_col.insertDefault(); - } - - offsets.push_back(old_size + object.size()); - return true; - } - - private: - std::unique_ptr key; - std::unique_ptr value; - }; - - class VariantNode : public Node - { - public: - VariantNode(std::vector> variant_nodes_, std::vector order_) : variant_nodes(std::move(variant_nodes_)), order(std::move(order_)) { } - - bool insertResultToColumn(IColumn & dest, const Element & element) override - { - auto & column_variant = assert_cast(dest); - for (size_t i : order) - { - auto & variant = column_variant.getVariantByGlobalDiscriminator(i); - if (variant_nodes[i]->insertResultToColumn(variant, element)) - { - column_variant.getLocalDiscriminators().push_back(column_variant.localDiscriminatorByGlobal(i)); - column_variant.getOffsets().push_back(variant.size() - 1); - return true; - } - } - - return false; - } - - private: - std::vector> variant_nodes; - /// Order in which we should try variants nodes. - /// For example, String should be always the last one. - std::vector order; - }; - - static std::unique_ptr build(const char * function_name, const DataTypePtr & type) - { - switch (type->getTypeId()) - { - case TypeIndex::UInt8: return std::make_unique>(); - case TypeIndex::UInt16: return std::make_unique>(); - case TypeIndex::UInt32: return std::make_unique>(); - case TypeIndex::UInt64: return std::make_unique>(); - case TypeIndex::UInt128: return std::make_unique>(); - case TypeIndex::UInt256: return std::make_unique>(); - case TypeIndex::Int8: return std::make_unique>(); - case TypeIndex::Int16: return std::make_unique>(); - case TypeIndex::Int32: return std::make_unique>(); - case TypeIndex::Int64: return std::make_unique>(); - case TypeIndex::Int128: return std::make_unique>(); - case TypeIndex::Int256: return std::make_unique>(); - case TypeIndex::Float32: return std::make_unique>(); - case TypeIndex::Float64: return std::make_unique>(); - case TypeIndex::String: return std::make_unique(); - case TypeIndex::FixedString: return std::make_unique(); - case TypeIndex::UUID: return std::make_unique(); - case TypeIndex::LowCardinality: - { - // The low cardinality case is treated in two different ways: - // For FixedString type, an especial class is implemented for inserting the data in the destination column, - // as the string length must be passed in order to check and pad the incoming data. - // For the rest of low cardinality types, the insertion is done in their corresponding class, adapting the data - // as needed for the insertData function of the ColumnLowCardinality. - auto dictionary_type = typeid_cast(type.get())->getDictionaryType(); - if ((*dictionary_type).getTypeId() == TypeIndex::FixedString) - { - auto fixed_length = typeid_cast(dictionary_type.get())->getN(); - return std::make_unique(fixed_length); - } - return build(function_name, dictionary_type); - } - case TypeIndex::Decimal256: return std::make_unique>(type); - case TypeIndex::Decimal128: return std::make_unique>(type); - case TypeIndex::Decimal64: return std::make_unique>(type); - case TypeIndex::Decimal32: return std::make_unique>(type); - case TypeIndex::Enum8: - return std::make_unique>(static_cast(*type).getValues()); - case TypeIndex::Enum16: - return std::make_unique>(static_cast(*type).getValues()); - case TypeIndex::Nullable: - { - return std::make_unique(build(function_name, static_cast(*type).getNestedType())); - } - case TypeIndex::Array: - { - return std::make_unique(build(function_name, static_cast(*type).getNestedType())); - } - case TypeIndex::Tuple: - { - const auto & tuple = static_cast(*type); - const auto & tuple_elements = tuple.getElements(); - std::vector> elements; - elements.reserve(tuple_elements.size()); - for (const auto & tuple_element : tuple_elements) - elements.emplace_back(build(function_name, tuple_element)); - return std::make_unique(std::move(elements), tuple.haveExplicitNames() ? tuple.getElementNames() : Strings{}); - } - case TypeIndex::Map: - { - const auto & map_type = static_cast(*type); - const auto & key_type = map_type.getKeyType(); - if (!isString(removeLowCardinality(key_type))) - throw Exception( - ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, - "Function {} doesn't support the return type schema: {} with key type not String", - String(function_name), - type->getName()); - - const auto & value_type = map_type.getValueType(); - return std::make_unique(build(function_name, key_type), build(function_name, value_type)); - } - case TypeIndex::Variant: - { - const auto & variant_type = static_cast(*type); - const auto & variants = variant_type.getVariants(); - std::vector> variant_nodes; - variant_nodes.reserve(variants.size()); - for (const auto & variant : variants) - variant_nodes.push_back(build(function_name, variant)); - return std::make_unique(std::move(variant_nodes), SerializationVariant::getVariantsDeserializeTextOrder(variants)); - } - default: - throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, - "Function {} doesn't support the return type schema: {}", - String(function_name), type->getName()); - } - } -}; - - -template -class JSONExtractImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char * function_name, const ColumnsWithTypeAndName & arguments) - { - if (arguments.size() < 2) - throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, "Function {} requires at least two arguments", String(function_name)); - - const auto & col = arguments.back(); - const auto * col_type_const = typeid_cast(col.column.get()); - if (!col_type_const || !isString(col.type)) - throw Exception(ErrorCodes::ILLEGAL_COLUMN, - "The last argument of function {} should " - "be a constant string specifying the return data type, illegal value: {}", - String(function_name), col.name); - - return DataTypeFactory::instance().get(col_type_const->getValue()); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 2; } - - void prepare(const char * function_name, const ColumnsWithTypeAndName &, const DataTypePtr & result_type) - { - extract_tree = JSONExtractTree::build(function_name, result_type); - } - - bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - return extract_tree->insertResultToColumn(dest, element); - } - -protected: - std::unique_ptr::Node> extract_tree; -}; - - -template -class JSONExtractKeysAndValuesImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char * function_name, const ColumnsWithTypeAndName & arguments) - { - if (arguments.size() < 2) - throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, "Function {} requires at least two arguments", String(function_name)); - - const auto & col = arguments.back(); - const auto * col_type_const = typeid_cast(col.column.get()); - if (!col_type_const || !isString(col.type)) - throw Exception(ErrorCodes::ILLEGAL_COLUMN, - "The last argument of function {} should " - "be a constant string specifying the values' data type, illegal value: {}", - String(function_name), col.name); - - DataTypePtr key_type = std::make_unique(); - DataTypePtr value_type = DataTypeFactory::instance().get(col_type_const->getValue()); - DataTypePtr tuple_type = std::make_unique(DataTypes{key_type, value_type}); - return std::make_unique(tuple_type); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 2; } - - void prepare(const char * function_name, const ColumnsWithTypeAndName &, const DataTypePtr & result_type) - { - const auto tuple_type = typeid_cast(result_type.get())->getNestedType(); - const auto value_type = typeid_cast(tuple_type.get())->getElements()[1]; - extract_tree = JSONExtractTree::build(function_name, value_type); - } - - bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - if (!element.isObject()) - return false; - - auto object = element.getObject(); - - auto & col_arr = assert_cast(dest); - auto & col_tuple = assert_cast(col_arr.getData()); - size_t old_size = col_tuple.size(); - auto & col_key = assert_cast(col_tuple.getColumn(0)); - auto & col_value = col_tuple.getColumn(1); - - for (const auto & [key, value] : object) - { - if (extract_tree->insertResultToColumn(col_value, value)) - col_key.insertData(key.data(), key.size()); - } - - if (col_tuple.size() == old_size) - return false; - - col_arr.getOffsets().push_back(col_tuple.size()); - return true; - } - -private: - std::unique_ptr::Node> extract_tree; -}; - - -template -class JSONExtractRawImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) - { - return std::make_shared(); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - if (dest.getDataType() == TypeIndex::LowCardinality) - { - ColumnString::Chars chars; - WriteBufferFromVector buf(chars, AppendModeTag()); - traverse(element, buf); - buf.finalize(); - assert_cast(dest).insertData(reinterpret_cast(chars.data()), chars.size()); - } - else - { - ColumnString & col_str = assert_cast(dest); - auto & chars = col_str.getChars(); - WriteBufferFromVector buf(chars, AppendModeTag()); - traverse(element, buf); - buf.finalize(); - chars.push_back(0); - col_str.getOffsets().push_back(chars.size()); - } - return true; - } - - // We use insertResultToFixedStringColumn in case we are inserting raw data in a FixedString column - static bool insertResultToFixedStringColumn(IColumn & dest, const Element & element, std::string_view) - { - ColumnFixedString::Chars chars; - WriteBufferFromVector buf(chars, AppendModeTag()); - traverse(element, buf); - buf.finalize(); - - auto & col_str = assert_cast(dest); - - if (chars.size() > col_str.getN()) - return false; - - chars.resize_fill(col_str.getN()); - col_str.insertData(reinterpret_cast(chars.data()), chars.size()); - - - return true; - } - - // We use insertResultToLowCardinalityFixedStringColumn in case we are inserting raw data in a Low Cardinality FixedString column - static bool insertResultToLowCardinalityFixedStringColumn(IColumn & dest, const Element & element, size_t fixed_length) - { - if (element.getObject().size() > fixed_length) - return false; - - ColumnFixedString::Chars chars; - WriteBufferFromVector buf(chars, AppendModeTag()); - traverse(element, buf); - buf.finalize(); - - if (chars.size() > fixed_length) - return false; - chars.resize_fill(fixed_length); - assert_cast(dest).insertData(reinterpret_cast(chars.data()), chars.size()); - - return true; - } - -private: - static void traverse(const Element & element, WriteBuffer & buf) - { - if (element.isInt64()) - { - writeIntText(element.getInt64(), buf); - return; - } - if (element.isUInt64()) - { - writeIntText(element.getUInt64(), buf); - return; - } - if (element.isDouble()) - { - writeFloatText(element.getDouble(), buf); - return; - } - if (element.isBool()) - { - if (element.getBool()) - writeCString("true", buf); - else - writeCString("false", buf); - return; - } - if (element.isString()) - { - writeJSONString(element.getString(), buf, formatSettings()); - return; - } - if (element.isArray()) - { - writeChar('[', buf); - bool need_comma = false; - for (auto value : element.getArray()) - { - if (std::exchange(need_comma, true)) - writeChar(',', buf); - traverse(value, buf); - } - writeChar(']', buf); - return; - } - if (element.isObject()) - { - writeChar('{', buf); - bool need_comma = false; - for (auto [key, value] : element.getObject()) - { - if (std::exchange(need_comma, true)) - writeChar(',', buf); - writeJSONString(key, buf, formatSettings()); - writeChar(':', buf); - traverse(value, buf); - } - writeChar('}', buf); - return; - } - if (element.isNull()) - { - writeCString("null", buf); - return; - } - } - - static const FormatSettings & formatSettings() - { - static const FormatSettings the_instance = [] - { - FormatSettings settings; - settings.json.escape_forward_slashes = false; - return settings; - }(); - return the_instance; - } -}; - - -template -class JSONExtractArrayRawImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) - { - return std::make_shared(std::make_shared()); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - static bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - if (!element.isArray()) - return false; - - auto array = element.getArray(); - ColumnArray & col_res = assert_cast(dest); - - for (auto value : array) - JSONExtractRawImpl::insertResultToColumn(col_res.getData(), value, {}); - - col_res.getOffsets().push_back(col_res.getOffsets().back() + array.size()); - return true; - } -}; - - -template -class JSONExtractKeysAndValuesRawImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) - { - DataTypePtr string_type = std::make_unique(); - DataTypePtr tuple_type = std::make_unique(DataTypes{string_type, string_type}); - return std::make_unique(tuple_type); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - if (!element.isObject()) - return false; - - auto object = element.getObject(); - - auto & col_arr = assert_cast(dest); - auto & col_tuple = assert_cast(col_arr.getData()); - auto & col_key = assert_cast(col_tuple.getColumn(0)); - auto & col_value = assert_cast(col_tuple.getColumn(1)); - - for (const auto & [key, value] : object) - { - col_key.insertData(key.data(), key.size()); - JSONExtractRawImpl::insertResultToColumn(col_value, value, {}); - } - - col_arr.getOffsets().push_back(col_arr.getOffsets().back() + object.size()); - return true; - } -}; - -template -class JSONExtractKeysImpl -{ -public: - using Element = typename JSONParser::Element; - - static DataTypePtr getReturnType(const char *, const ColumnsWithTypeAndName &) - { - return std::make_unique(std::make_shared()); - } - - static size_t getNumberOfIndexArguments(const ColumnsWithTypeAndName & arguments) { return arguments.size() - 1; } - - bool insertResultToColumn(IColumn & dest, const Element & element, std::string_view) - { - if (!element.isObject()) - return false; - - auto object = element.getObject(); - - ColumnArray & col_res = assert_cast(dest); - auto & col_key = assert_cast(col_res.getData()); - - for (const auto & [key, value] : object) - { - col_key.insertData(key.data(), key.size()); - } - - col_res.getOffsets().push_back(col_res.getOffsets().back() + object.size()); - return true; - } -}; - -} diff --git a/src/Interpreters/InterpreterAlterNamedCollectionQuery.cpp b/src/Interpreters/InterpreterAlterNamedCollectionQuery.cpp index 79a17fd1844..0e83e2039f6 100644 --- a/src/Interpreters/InterpreterAlterNamedCollectionQuery.cpp +++ b/src/Interpreters/InterpreterAlterNamedCollectionQuery.cpp @@ -4,6 +4,7 @@ #include #include #include +#include #include @@ -13,14 +14,16 @@ namespace DB BlockIO InterpreterAlterNamedCollectionQuery::execute() { auto current_context = getContext(); - const auto & query = query_ptr->as(); + + const auto updated_query = removeOnClusterClauseIfNeeded(query_ptr, getContext()); + const auto & query = updated_query->as(); current_context->checkAccess(AccessType::ALTER_NAMED_COLLECTION, query.collection_name); if (!query.cluster.empty()) { DDLQueryOnClusterParams params; - return executeDDLQueryOnCluster(query_ptr, current_context, params); + return executeDDLQueryOnCluster(updated_query, current_context, params); } NamedCollectionFactory::instance().updateFromSQL(query); diff --git a/src/Interpreters/InterpreterCreateNamedCollectionQuery.cpp b/src/Interpreters/InterpreterCreateNamedCollectionQuery.cpp index c71441daa8c..b4920b1729f 100644 --- a/src/Interpreters/InterpreterCreateNamedCollectionQuery.cpp +++ b/src/Interpreters/InterpreterCreateNamedCollectionQuery.cpp @@ -4,6 +4,7 @@ #include #include #include +#include #include @@ -13,14 +14,16 @@ namespace DB BlockIO InterpreterCreateNamedCollectionQuery::execute() { auto current_context = getContext(); - const auto & query = query_ptr->as(); + + const auto updated_query = removeOnClusterClauseIfNeeded(query_ptr, getContext()); + const auto & query = updated_query->as(); current_context->checkAccess(AccessType::CREATE_NAMED_COLLECTION, query.collection_name); if (!query.cluster.empty()) { DDLQueryOnClusterParams params; - return executeDDLQueryOnCluster(query_ptr, current_context, params); + return executeDDLQueryOnCluster(updated_query, current_context, params); } NamedCollectionFactory::instance().createFromSQL(query); diff --git a/src/Interpreters/InterpreterDropNamedCollectionQuery.cpp b/src/Interpreters/InterpreterDropNamedCollectionQuery.cpp index 2edaef1b2f2..6233d21b439 100644 --- a/src/Interpreters/InterpreterDropNamedCollectionQuery.cpp +++ b/src/Interpreters/InterpreterDropNamedCollectionQuery.cpp @@ -4,6 +4,7 @@ #include #include #include +#include #include @@ -13,14 +14,16 @@ namespace DB BlockIO InterpreterDropNamedCollectionQuery::execute() { auto current_context = getContext(); - const auto & query = query_ptr->as(); + + const auto updated_query = removeOnClusterClauseIfNeeded(query_ptr, getContext()); + const auto & query = updated_query->as(); current_context->checkAccess(AccessType::DROP_NAMED_COLLECTION, query.collection_name); if (!query.cluster.empty()) { DDLQueryOnClusterParams params; - return executeDDLQueryOnCluster(query_ptr, current_context, params); + return executeDDLQueryOnCluster(updated_query, current_context, params); } NamedCollectionFactory::instance().removeFromSQL(query); diff --git a/src/Interpreters/OptimizeIfWithConstantConditionVisitor.cpp b/src/Interpreters/OptimizeIfWithConstantConditionVisitor.cpp index 20451fb20ad..48c9988b6fc 100644 --- a/src/Interpreters/OptimizeIfWithConstantConditionVisitor.cpp +++ b/src/Interpreters/OptimizeIfWithConstantConditionVisitor.cpp @@ -73,66 +73,55 @@ static bool tryExtractConstValueFromCondition(const ASTPtr & condition, bool & v return false; } -void OptimizeIfWithConstantConditionVisitor::visit(ASTPtr & current_ast) +void OptimizeIfWithConstantConditionVisitorData::visit(ASTFunction & function_node, ASTPtr & ast) { - if (!current_ast) - return; - checkStackSize(); - for (ASTPtr & child : current_ast->children) + if (function_node.name != "if") + return; + + if (!function_node.arguments) + throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, "Wrong number of arguments for function 'if' (0 instead of 3)"); + + if (function_node.arguments->children.size() != 3) + throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, + "Wrong number of arguments for function 'if' ({} instead of 3)", + function_node.arguments->children.size()); + + const auto * args = function_node.arguments->as(); + + ASTPtr condition_expr = args->children[0]; + ASTPtr then_expr = args->children[1]; + ASTPtr else_expr = args->children[2]; + + bool condition; + if (tryExtractConstValueFromCondition(condition_expr, condition)) { - auto * function_node = child->as(); - if (!function_node || function_node->name != "if") + ASTPtr replace_ast = condition ? then_expr : else_expr; + ASTPtr child_copy = ast; + String replace_alias = replace_ast->tryGetAlias(); + String if_alias = ast->tryGetAlias(); + + if (replace_alias.empty()) { - visit(child); - continue; + replace_ast->setAlias(if_alias); + ast = replace_ast; + } + else + { + /// Only copy of one node is required here. + /// But IAST has only method for deep copy of subtree. + /// This can be a reason of performance degradation in case of deep queries. + ASTPtr replace_ast_deep_copy = replace_ast->clone(); + replace_ast_deep_copy->setAlias(if_alias); + ast = replace_ast_deep_copy; } - if (!function_node->arguments) - throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, "Wrong number of arguments for function 'if' (0 instead of 3)"); - - if (function_node->arguments->children.size() != 3) - throw Exception(ErrorCodes::NUMBER_OF_ARGUMENTS_DOESNT_MATCH, - "Wrong number of arguments for function 'if' ({} instead of 3)", - function_node->arguments->children.size()); - - visit(function_node->arguments); - const auto * args = function_node->arguments->as(); - - ASTPtr condition_expr = args->children[0]; - ASTPtr then_expr = args->children[1]; - ASTPtr else_expr = args->children[2]; - - bool condition; - if (tryExtractConstValueFromCondition(condition_expr, condition)) + if (!if_alias.empty()) { - ASTPtr replace_ast = condition ? then_expr : else_expr; - ASTPtr child_copy = child; - String replace_alias = replace_ast->tryGetAlias(); - String if_alias = child->tryGetAlias(); - - if (replace_alias.empty()) - { - replace_ast->setAlias(if_alias); - child = replace_ast; - } - else - { - /// Only copy of one node is required here. - /// But IAST has only method for deep copy of subtree. - /// This can be a reason of performance degradation in case of deep queries. - ASTPtr replace_ast_deep_copy = replace_ast->clone(); - replace_ast_deep_copy->setAlias(if_alias); - child = replace_ast_deep_copy; - } - - if (!if_alias.empty()) - { - auto alias_it = aliases.find(if_alias); - if (alias_it != aliases.end() && alias_it->second.get() == child_copy.get()) - alias_it->second = child; - } + auto alias_it = aliases.find(if_alias); + if (alias_it != aliases.end() && alias_it->second.get() == child_copy.get()) + alias_it->second = ast; } } } diff --git a/src/Interpreters/OptimizeIfWithConstantConditionVisitor.h b/src/Interpreters/OptimizeIfWithConstantConditionVisitor.h index ad98f92bafd..3b46f90f07c 100644 --- a/src/Interpreters/OptimizeIfWithConstantConditionVisitor.h +++ b/src/Interpreters/OptimizeIfWithConstantConditionVisitor.h @@ -1,23 +1,24 @@ #pragma once #include +#include namespace DB { - -/// It removes Function_if node from AST if condition is constant. -/// TODO: rewrite with InDepthNodeVisitor -class OptimizeIfWithConstantConditionVisitor +struct OptimizeIfWithConstantConditionVisitorData { -public: - explicit OptimizeIfWithConstantConditionVisitor(Aliases & aliases_) + using TypeToVisit = ASTFunction; + + explicit OptimizeIfWithConstantConditionVisitorData(Aliases & aliases_) : aliases(aliases_) {} - void visit(ASTPtr & ast); - + void visit(ASTFunction & function_node, ASTPtr & ast); private: Aliases & aliases; }; +/// It removes Function_if node from AST if condition is constant. +using OptimizeIfWithConstantConditionVisitor = InDepthNodeVisitor, false>; + } diff --git a/src/Interpreters/TreeOptimizer.cpp b/src/Interpreters/TreeOptimizer.cpp index b88d75cd5a2..b872eb94fde 100644 --- a/src/Interpreters/TreeOptimizer.cpp +++ b/src/Interpreters/TreeOptimizer.cpp @@ -577,7 +577,8 @@ void TreeOptimizer::optimizeIf(ASTPtr & query, Aliases & aliases, bool if_chain_ optimizeMultiIfToIf(query); /// Optimize if with constant condition after constants was substituted instead of scalar subqueries. - OptimizeIfWithConstantConditionVisitor(aliases).visit(query); + OptimizeIfWithConstantConditionVisitorData visitor_data(aliases); + OptimizeIfWithConstantConditionVisitor(visitor_data).visit(query); if (if_chain_to_multiif) OptimizeIfChainsVisitor().visit(query); diff --git a/src/Interpreters/removeOnClusterClauseIfNeeded.cpp b/src/Interpreters/removeOnClusterClauseIfNeeded.cpp index 44167fe7242..dd20164925c 100644 --- a/src/Interpreters/removeOnClusterClauseIfNeeded.cpp +++ b/src/Interpreters/removeOnClusterClauseIfNeeded.cpp @@ -15,6 +15,10 @@ #include #include #include +#include +#include +#include +#include namespace DB @@ -38,6 +42,13 @@ static bool isAccessControlQuery(const ASTPtr & query) || query->as(); } +static bool isNamedCollectionQuery(const ASTPtr & query) +{ + return query->as() + || query->as() + || query->as(); +} + ASTPtr removeOnClusterClauseIfNeeded(const ASTPtr & query, ContextPtr context, const WithoutOnClusterASTRewriteParams & params) { auto * query_on_cluster = dynamic_cast(query.get()); @@ -50,7 +61,10 @@ ASTPtr removeOnClusterClauseIfNeeded(const ASTPtr & query, ContextPtr context, c && context->getUserDefinedSQLObjectsStorage().isReplicated()) || (isAccessControlQuery(query) && context->getSettings().ignore_on_cluster_for_replicated_access_entities_queries - && context->getAccessControl().containsStorage(ReplicatedAccessStorage::STORAGE_TYPE))) + && context->getAccessControl().containsStorage(ReplicatedAccessStorage::STORAGE_TYPE)) + || (isNamedCollectionQuery(query) + && context->getSettings().ignore_on_cluster_for_replicated_named_collections_queries + && NamedCollectionFactory::instance().usesReplicatedStorage())) { LOG_DEBUG(getLogger("removeOnClusterClauseIfNeeded"), "ON CLUSTER clause was ignored for query {}", query->getID()); return query_on_cluster->getRewrittenASTWithoutOnCluster(params); diff --git a/src/Storages/ObjectStorage/DataLakes/DeltaLakeMetadata.cpp b/src/Storages/ObjectStorage/DataLakes/DeltaLakeMetadata.cpp index bc64ef15cf1..c896a760597 100644 --- a/src/Storages/ObjectStorage/DataLakes/DeltaLakeMetadata.cpp +++ b/src/Storages/ObjectStorage/DataLakes/DeltaLakeMetadata.cpp @@ -8,6 +8,7 @@ #include #include #include +#include #include #include @@ -30,6 +31,7 @@ #include #include #include +#include #include #include @@ -111,7 +113,7 @@ struct DeltaLakeMetadataImpl std::set result_files; NamesAndTypesList current_schema; DataLakePartitionColumns current_partition_columns; - const auto checkpoint_version = getCheckpointIfExists(result_files); + const auto checkpoint_version = getCheckpointIfExists(result_files, current_schema, current_partition_columns); if (checkpoint_version) { @@ -205,9 +207,32 @@ struct DeltaLakeMetadataImpl Poco::Dynamic::Var json = parser.parse(json_str); Poco::JSON::Object::Ptr object = json.extract(); - // std::ostringstream oss; // STYLE_CHECK_ALLOW_STD_STRING_STREAM - // object->stringify(oss); - // LOG_TEST(log, "Metadata: {}", oss.str()); + std::ostringstream oss; // STYLE_CHECK_ALLOW_STD_STRING_STREAM + object->stringify(oss); + LOG_TEST(log, "Metadata: {}", oss.str()); + + if (object->has("metaData")) + { + const auto metadata_object = object->get("metaData").extract(); + const auto schema_object = metadata_object->getValue("schemaString"); + + Poco::JSON::Parser p; + Poco::Dynamic::Var fields_json = parser.parse(schema_object); + const Poco::JSON::Object::Ptr & fields_object = fields_json.extract(); + + auto current_schema = parseMetadata(fields_object); + if (file_schema.empty()) + { + file_schema = current_schema; + } + else if (file_schema != current_schema) + { + throw Exception(ErrorCodes::NOT_IMPLEMENTED, + "Reading from files with different schema is not possible " + "({} is different from {})", + file_schema.toString(), current_schema.toString()); + } + } if (object->has("add")) { @@ -230,7 +255,12 @@ struct DeltaLakeMetadataImpl const auto value = partition_values->getValue(partition_name); auto name_and_type = file_schema.tryGetByName(partition_name); if (!name_and_type) - throw Exception(ErrorCodes::LOGICAL_ERROR, "No such column in schema: {}", partition_name); + { + throw Exception( + ErrorCodes::LOGICAL_ERROR, + "No such column in schema: {} (schema: {})", + partition_name, file_schema.toNamesAndTypesDescription()); + } auto field = getFieldValue(value, name_and_type->type); current_partition_columns.emplace_back(*name_and_type, field); @@ -246,52 +276,35 @@ struct DeltaLakeMetadataImpl auto path = object->get("remove").extract()->getValue("path"); result.erase(fs::path(configuration->getPath()) / path); } - if (object->has("metaData")) - { - const auto metadata_object = object->get("metaData").extract(); - const auto schema_object = metadata_object->getValue("schemaString"); - - Poco::JSON::Parser p; - Poco::Dynamic::Var fields_json = parser.parse(schema_object); - Poco::JSON::Object::Ptr fields_object = fields_json.extract(); - - const auto fields = fields_object->get("fields").extract(); - NamesAndTypesList current_schema; - for (size_t i = 0; i < fields->size(); ++i) - { - const auto field = fields->getObject(static_cast(i)); - auto column_name = field->getValue("name"); - auto type = field->getValue("type"); - auto is_nullable = field->getValue("nullable"); - - std::string physical_name; - auto schema_metadata_object = field->get("metadata").extract(); - if (schema_metadata_object->has("delta.columnMapping.physicalName")) - physical_name = schema_metadata_object->getValue("delta.columnMapping.physicalName"); - else - physical_name = column_name; - - LOG_TEST(log, "Found column: {}, type: {}, nullable: {}, physical name: {}", - column_name, type, is_nullable, physical_name); - - current_schema.push_back({physical_name, getFieldType(field, "type", is_nullable)}); - } - - if (file_schema.empty()) - { - file_schema = current_schema; - } - else if (file_schema != current_schema) - { - throw Exception(ErrorCodes::NOT_IMPLEMENTED, - "Reading from files with different schema is not possible " - "({} is different from {})", - file_schema.toString(), current_schema.toString()); - } - } } } + NamesAndTypesList parseMetadata(const Poco::JSON::Object::Ptr & metadata_json) + { + NamesAndTypesList schema; + const auto fields = metadata_json->get("fields").extract(); + for (size_t i = 0; i < fields->size(); ++i) + { + const auto field = fields->getObject(static_cast(i)); + auto column_name = field->getValue("name"); + auto type = field->getValue("type"); + auto is_nullable = field->getValue("nullable"); + + std::string physical_name; + auto schema_metadata_object = field->get("metadata").extract(); + if (schema_metadata_object->has("delta.columnMapping.physicalName")) + physical_name = schema_metadata_object->getValue("delta.columnMapping.physicalName"); + else + physical_name = column_name; + + LOG_TEST(log, "Found column: {}, type: {}, nullable: {}, physical name: {}", + column_name, type, is_nullable, physical_name); + + schema.push_back({physical_name, getFieldType(field, "type", is_nullable)}); + } + return schema; + } + DataTypePtr getFieldType(const Poco::JSON::Object::Ptr & field, const String & type_key, bool is_nullable) { if (field->isObject(type_key)) @@ -505,7 +518,10 @@ struct DeltaLakeMetadataImpl throw Exception(ErrorCodes::BAD_ARGUMENTS, "Arrow error: {}", _s.ToString()); \ } while (false) - size_t getCheckpointIfExists(std::set & result) + size_t getCheckpointIfExists( + std::set & result, + NamesAndTypesList & file_schema, + DataLakePartitionColumns & file_partition_columns) { const auto version = readLastCheckpointIfExists(); if (!version) @@ -526,7 +542,8 @@ struct DeltaLakeMetadataImpl auto columns = ParquetSchemaReader(*buf, format_settings).readSchema(); /// Read only columns that we need. - columns.filterColumns(NameSet{"add", "remove"}); + auto filter_column_names = NameSet{"add", "metaData"}; + columns.filterColumns(filter_column_names); Block header; for (const auto & column : columns) header.insert({column.type->createColumn(), column.type, column.name}); @@ -540,9 +557,6 @@ struct DeltaLakeMetadataImpl ArrowMemoryPool::instance(), &reader)); - std::shared_ptr file_schema; - THROW_ARROW_NOT_OK(reader->GetSchema(&file_schema)); - ArrowColumnToCHColumn column_reader( header, "Parquet", format_settings.parquet.allow_missing_columns, @@ -553,29 +567,85 @@ struct DeltaLakeMetadataImpl std::shared_ptr table; THROW_ARROW_NOT_OK(reader->ReadTable(&table)); - Chunk res = column_reader.arrowTableToCHChunk(table, reader->parquet_reader()->metadata()->num_rows()); - const auto & res_columns = res.getColumns(); + Chunk chunk = column_reader.arrowTableToCHChunk(table, reader->parquet_reader()->metadata()->num_rows()); + auto res_block = header.cloneWithColumns(chunk.detachColumns()); + res_block = Nested::flatten(res_block); - if (res_columns.size() != 2) - { - throw Exception( - ErrorCodes::INCORRECT_DATA, - "Unexpected number of columns: {} (having: {}, expected: {})", - res_columns.size(), res.dumpStructure(), header.dumpStructure()); - } + const auto * nullable_path_column = assert_cast(res_block.getByName("add.path").column.get()); + const auto & path_column = assert_cast(nullable_path_column->getNestedColumn()); + + const auto * nullable_schema_column = assert_cast(res_block.getByName("metaData.schemaString").column.get()); + const auto & schema_column = assert_cast(nullable_schema_column->getNestedColumn()); + + auto partition_values_column_raw = res_block.getByName("add.partitionValues").column; + const auto & partition_values_column = assert_cast(*partition_values_column_raw); - const auto * tuple_column = assert_cast(res_columns[0].get()); - const auto & nullable_column = assert_cast(tuple_column->getColumn(0)); - const auto & path_column = assert_cast(nullable_column.getNestedColumn()); for (size_t i = 0; i < path_column.size(); ++i) { - const auto filename = String(path_column.getDataAt(i)); - if (filename.empty()) + const auto metadata = String(schema_column.getDataAt(i)); + if (!metadata.empty()) + { + Poco::JSON::Parser parser; + Poco::Dynamic::Var json = parser.parse(metadata); + const Poco::JSON::Object::Ptr & object = json.extract(); + + auto current_schema = parseMetadata(object); + if (file_schema.empty()) + { + file_schema = current_schema; + LOG_TEST(log, "Processed schema from checkpoint: {}", file_schema.toString()); + } + else if (file_schema != current_schema) + { + throw Exception(ErrorCodes::NOT_IMPLEMENTED, + "Reading from files with different schema is not possible " + "({} is different from {})", + file_schema.toString(), current_schema.toString()); + } + } + } + + for (size_t i = 0; i < path_column.size(); ++i) + { + const auto path = String(path_column.getDataAt(i)); + if (path.empty()) continue; - LOG_TEST(log, "Adding {}", filename); - const auto [_, inserted] = result.insert(std::filesystem::path(configuration->getPath()) / filename); + + auto filename = fs::path(path).filename().string(); + auto it = file_partition_columns.find(filename); + if (it == file_partition_columns.end()) + { + Field map; + partition_values_column.get(i, map); + auto partition_values_map = map.safeGet(); + if (!partition_values_map.empty()) + { + auto & current_partition_columns = file_partition_columns[filename]; + for (const auto & map_value : partition_values_map) + { + const auto tuple = map_value.safeGet(); + const auto partition_name = tuple[0].safeGet(); + auto name_and_type = file_schema.tryGetByName(partition_name); + if (!name_and_type) + { + throw Exception( + ErrorCodes::LOGICAL_ERROR, + "No such column in schema: {} (schema: {})", + partition_name, file_schema.toString()); + } + const auto value = tuple[1].safeGet(); + auto field = getFieldValue(value, name_and_type->type); + current_partition_columns.emplace_back(std::move(name_and_type.value()), std::move(field)); + + LOG_TEST(log, "Partition {} value is {} (for {})", partition_name, value, filename); + } + } + } + + LOG_TEST(log, "Adding {}", path); + const auto [_, inserted] = result.insert(std::filesystem::path(configuration->getPath()) / path); if (!inserted) - throw Exception(ErrorCodes::INCORRECT_DATA, "File already exists {}", filename); + throw Exception(ErrorCodes::INCORRECT_DATA, "File already exists {}", path); } return version; diff --git a/src/Storages/ObjectStorage/DataLakes/IStorageDataLake.h b/src/Storages/ObjectStorage/DataLakes/IStorageDataLake.h index f1217bc9729..c8603fccb86 100644 --- a/src/Storages/ObjectStorage/DataLakes/IStorageDataLake.h +++ b/src/Storages/ObjectStorage/DataLakes/IStorageDataLake.h @@ -41,6 +41,7 @@ public: auto object_storage = base_configuration->createObjectStorage(context, /* is_readonly */true); DataLakeMetadataPtr metadata; NamesAndTypesList schema_from_metadata; + const bool use_schema_from_metadata = columns_.empty(); if (base_configuration->format == "auto") base_configuration->format = "Parquet"; @@ -50,8 +51,9 @@ public: try { metadata = DataLakeMetadata::create(object_storage, base_configuration, context); - schema_from_metadata = metadata->getTableSchema(); configuration->setPaths(metadata->getDataFiles()); + if (use_schema_from_metadata) + schema_from_metadata = metadata->getTableSchema(); } catch (...) { @@ -66,7 +68,7 @@ public: return std::make_shared>( base_configuration, std::move(metadata), configuration, object_storage, context, table_id_, - columns_.empty() ? ColumnsDescription(schema_from_metadata) : columns_, + use_schema_from_metadata ? ColumnsDescription(schema_from_metadata) : columns_, constraints_, comment_, format_settings_); } diff --git a/src/Storages/ObjectStorage/StorageObjectStorageSource.cpp b/src/Storages/ObjectStorage/StorageObjectStorageSource.cpp index 6940f10cb91..a9a7e062076 100644 --- a/src/Storages/ObjectStorage/StorageObjectStorageSource.cpp +++ b/src/Storages/ObjectStorage/StorageObjectStorageSource.cpp @@ -206,23 +206,25 @@ Chunk StorageObjectStorageSource::generate() if (!partition_columns.empty() && chunk_size && chunk.hasColumns()) { auto partition_values = partition_columns.find(filename); - - for (const auto & [name_and_type, value] : partition_values->second) + if (partition_values != partition_columns.end()) { - if (!read_from_format_info.source_header.has(name_and_type.name)) - continue; + for (const auto & [name_and_type, value] : partition_values->second) + { + if (!read_from_format_info.source_header.has(name_and_type.name)) + continue; - const auto column_pos = read_from_format_info.source_header.getPositionByName(name_and_type.name); - auto partition_column = name_and_type.type->createColumnConst(chunk.getNumRows(), value)->convertToFullColumnIfConst(); + const auto column_pos = read_from_format_info.source_header.getPositionByName(name_and_type.name); + auto partition_column = name_and_type.type->createColumnConst(chunk.getNumRows(), value)->convertToFullColumnIfConst(); - /// This column is filled with default value now, remove it. - chunk.erase(column_pos); + /// This column is filled with default value now, remove it. + chunk.erase(column_pos); - /// Add correct values. - if (chunk.hasColumns()) - chunk.addColumn(column_pos, std::move(partition_column)); - else - chunk.addColumn(std::move(partition_column)); + /// Add correct values. + if (column_pos < chunk.getNumColumns()) + chunk.addColumn(column_pos, std::move(partition_column)); + else + chunk.addColumn(std::move(partition_column)); + } } } return chunk; diff --git a/src/Storages/StorageReplicatedMergeTree.cpp b/src/Storages/StorageReplicatedMergeTree.cpp index 546ad90f9e9..9fb0f082c6c 100644 --- a/src/Storages/StorageReplicatedMergeTree.cpp +++ b/src/Storages/StorageReplicatedMergeTree.cpp @@ -5,20 +5,21 @@ #include #include +#include #include #include #include #include +#include #include #include #include #include +#include #include +#include #include #include -#include -#include -#include #include @@ -5272,6 +5273,8 @@ void StorageReplicatedMergeTree::flushAndPrepareForShutdown() if (shutdown_prepared_called.exchange(true)) return; + LOG_TRACE(log, "Start preparing for shutdown"); + try { auto settings_ptr = getSettings(); @@ -5282,7 +5285,11 @@ void StorageReplicatedMergeTree::flushAndPrepareForShutdown() stopBeingLeader(); if (attach_thread) + { attach_thread->shutdown(); + LOG_TRACE(log, "The attach thread is shutdown"); + } + restarting_thread.shutdown(/* part_of_full_shutdown */true); /// Explicitly set the event, because the restarting thread will not set it again @@ -5295,6 +5302,8 @@ void StorageReplicatedMergeTree::flushAndPrepareForShutdown() shutdown_deadline.emplace(std::chrono::system_clock::now()); throw; } + + LOG_TRACE(log, "Finished preparing for shutdown"); } void StorageReplicatedMergeTree::partialShutdown() @@ -5332,6 +5341,8 @@ void StorageReplicatedMergeTree::shutdown(bool) if (shutdown_called.exchange(true)) return; + LOG_TRACE(log, "Shutdown started"); + flushAndPrepareForShutdown(); if (!shutdown_deadline.has_value()) @@ -5374,6 +5385,7 @@ void StorageReplicatedMergeTree::shutdown(bool) /// Wait for all of them std::lock_guard lock(data_parts_exchange_ptr->rwlock); } + LOG_TRACE(log, "Shutdown finished"); } diff --git a/src/Storages/registerStorages.cpp b/src/Storages/registerStorages.cpp index 9f849052071..adc1074b1fe 100644 --- a/src/Storages/registerStorages.cpp +++ b/src/Storages/registerStorages.cpp @@ -35,7 +35,6 @@ void registerStorageFuzzJSON(StorageFactory & factory); void registerStorageS3(StorageFactory & factory); void registerStorageHudi(StorageFactory & factory); void registerStorageS3Queue(StorageFactory & factory); -void registerStorageAzureQueue(StorageFactory & factory); #if USE_PARQUET void registerStorageDeltaLake(StorageFactory & factory); @@ -45,6 +44,10 @@ void registerStorageIceberg(StorageFactory & factory); #endif #endif +#if USE_AZURE_BLOB_STORAGE +void registerStorageAzureQueue(StorageFactory & factory); +#endif + #if USE_HDFS #if USE_HIVE void registerStorageHive(StorageFactory & factory); diff --git a/tests/ci/.mypy.ini b/tests/ci/.mypy.ini index 9bc44025826..f12d27979ce 100644 --- a/tests/ci/.mypy.ini +++ b/tests/ci/.mypy.ini @@ -15,3 +15,4 @@ warn_return_any = True no_implicit_reexport = True strict_equality = True extra_checks = True +ignore_missing_imports = True \ No newline at end of file diff --git a/tests/ci/ci.py b/tests/ci/ci.py index af2f4c0a1fc..fac50d30022 100644 --- a/tests/ci/ci.py +++ b/tests/ci/ci.py @@ -15,7 +15,7 @@ import upload_result_helper from build_check import get_release_or_pr from ci_config import CI from ci_metadata import CiMetadata -from ci_utils import GHActions, normalize_string +from ci_utils import GHActions, normalize_string, Shell from clickhouse_helper import ( CiLogsCredentials, ClickHouseHelper, @@ -53,6 +53,7 @@ from stopwatch import Stopwatch from tee_popen import TeePopen from ci_cache import CiCache from ci_settings import CiSettings +from ci_buddy import CIBuddy from version_helper import get_version_from_repo # pylint: disable=too-many-lines @@ -262,6 +263,8 @@ def check_missing_images_on_dockerhub( def _pre_action(s3, indata, pr_info): + print("Clear dmesg") + Shell.run("sudo dmesg --clear ||:") CommitStatusData.cleanup() JobReport.cleanup() BuildResult.cleanup() @@ -1118,6 +1121,14 @@ def main() -> int: ### POST action: start elif args.post: + if Shell.check( + "sudo dmesg -T | grep -q -e 'Out of memory: Killed process' -e 'oom_reaper: reaped process' -e 'oom-kill:constraint=CONSTRAINT_NONE'" + ): + print("WARNING: OOM while job execution") + CIBuddy(dry_run=not pr_info.is_release).post_error( + "Out Of Memory", job_name=_get_ext_check_name(args.job_name) + ) + job_report = JobReport.load() if JobReport.exist() else None if job_report: ch_helper = ClickHouseHelper() diff --git a/tests/ci/ci_buddy.py b/tests/ci/ci_buddy.py new file mode 100644 index 00000000000..ea690bb602c --- /dev/null +++ b/tests/ci/ci_buddy.py @@ -0,0 +1,88 @@ +import json +import os + +import boto3 +import requests +from botocore.exceptions import ClientError + +from pr_info import PRInfo +from ci_utils import Shell + + +class CIBuddy: + _HEADERS = {"Content-Type": "application/json"} + + def __init__(self, dry_run=False): + self.repo = os.getenv("GITHUB_REPOSITORY", "") + self.dry_run = dry_run + res = self._get_webhooks() + self.test_channel = "" + self.dev_ci_channel = "" + if res: + self.test_channel = json.loads(res)["test_channel"] + self.dev_ci_channel = json.loads(res)["ci_channel"] + self.job_name = os.getenv("CHECK_NAME", "unknown") + pr_info = PRInfo() + self.pr_number = pr_info.number + self.head_ref = pr_info.head_ref + self.commit_url = pr_info.commit_html_url + + @staticmethod + def _get_webhooks(): + name = "ci_buddy_web_hooks" + + session = boto3.Session(region_name="us-east-1") # Replace with your region + ssm_client = session.client("ssm") + json_string = None + try: + response = ssm_client.get_parameter( + Name=name, + WithDecryption=True, # Set to True if the parameter is a SecureString + ) + json_string = response["Parameter"]["Value"] + except ClientError as e: + print(f"An error occurred: {e}") + + return json_string + + def post(self, message, dry_run=None): + if dry_run is None: + dry_run = self.dry_run + print(f"Posting slack message, dry_run [{dry_run}]") + if dry_run: + url = self.test_channel + else: + url = self.dev_ci_channel + data = {"text": message} + try: + requests.post(url, headers=self._HEADERS, data=json.dumps(data), timeout=10) + except Exception as e: + print(f"ERROR: Failed to post message, ex {e}") + + def post_error(self, error_description, job_name="", with_instance_info=True): + instance_id, instance_type = "unknown", "unknown" + if with_instance_info: + instance_id = Shell.run("ec2metadata --instance-id") or instance_id + instance_type = Shell.run("ec2metadata --instance-type") or instance_type + if not job_name: + job_name = os.getenv("CHECK_NAME", "unknown") + line_err = f":red_circle: *Error: {error_description}*\n\n" + line_ghr = f" *Runner:* `{instance_type}`, `{instance_id}`\n" + line_job = f" *Job:* `{job_name}`\n" + line_pr_ = f" *PR:* \n" + line_br_ = f" *Branch:* `{self.head_ref}`, <{self.commit_url}|commit>\n" + message = line_err + message += line_job + if with_instance_info: + message += line_ghr + if self.pr_number > 0: + message += line_pr_ + else: + message += line_br_ + self.post(message) + + +if __name__ == "__main__": + # test + buddy = CIBuddy(dry_run=True) + buddy.post_error("Out of memory") diff --git a/tests/ci/ci_utils.py b/tests/ci/ci_utils.py index e7034d0b104..629f37289a9 100644 --- a/tests/ci/ci_utils.py +++ b/tests/ci/ci_utils.py @@ -1,4 +1,5 @@ import os +import subprocess from contextlib import contextmanager from pathlib import Path from typing import Any, Iterator, List, Union @@ -42,3 +43,43 @@ class GHActions: for line in lines: print(line) print("::endgroup::") + + +class Shell: + @classmethod + def run_strict(cls, command): + subprocess.run( + command + " 2>&1", + shell=True, + stdout=subprocess.PIPE, + stderr=subprocess.PIPE, + text=True, + check=True, + ) + + @classmethod + def run(cls, command): + res = "" + result = subprocess.run( + command, + shell=True, + stdout=subprocess.PIPE, + stderr=subprocess.PIPE, + text=True, + check=False, + ) + if result.returncode == 0: + res = result.stdout + return res.strip() + + @classmethod + def check(cls, command): + result = subprocess.run( + command + " 2>&1", + shell=True, + stdout=subprocess.PIPE, + stderr=subprocess.PIPE, + text=True, + check=False, + ) + return result.returncode == 0 diff --git a/tests/integration/test_named_collections/configs/config.d/named_collections_with_zookeeper.xml b/tests/integration/test_named_collections/configs/config.d/named_collections_with_zookeeper.xml index 2d7946d1587..43d80ee6f69 100644 --- a/tests/integration/test_named_collections/configs/config.d/named_collections_with_zookeeper.xml +++ b/tests/integration/test_named_collections/configs/config.d/named_collections_with_zookeeper.xml @@ -9,4 +9,21 @@ value1 + + + + + true + + node_with_keeper + 9000 + + + node_with_keeper_2 + 9000 + + + true + + diff --git a/tests/integration/test_named_collections/configs/users.d/users.xml b/tests/integration/test_named_collections/configs/users.d/users.xml index 15da914f666..7d4f0543ff1 100644 --- a/tests/integration/test_named_collections/configs/users.d/users.xml +++ b/tests/integration/test_named_collections/configs/users.d/users.xml @@ -1,4 +1,9 @@ + + + 0 + + diff --git a/tests/integration/test_named_collections/test.py b/tests/integration/test_named_collections/test.py index dbc502236c0..32846c79d23 100644 --- a/tests/integration/test_named_collections/test.py +++ b/tests/integration/test_named_collections/test.py @@ -3,6 +3,8 @@ import pytest import os import time from helpers.cluster import ClickHouseCluster +from contextlib import nullcontext as does_not_raise +from helpers.client import QueryRuntimeException SCRIPT_DIR = os.path.dirname(os.path.realpath(__file__)) NAMED_COLLECTIONS_CONFIG = os.path.join( @@ -761,3 +763,32 @@ def test_keeper_storage(cluster): check_dropped(node1) check_dropped(node2) + + +@pytest.mark.parametrize( + "ignore, expected_raise", + [(True, does_not_raise()), (False, pytest.raises(QueryRuntimeException))], +) +def test_keeper_storage_remove_on_cluster(cluster, ignore, expected_raise): + node = cluster.instances["node_with_keeper"] + + replace_in_users_config( + node, + "ignore_on_cluster_for_replicated_named_collections_queries>.", + f"ignore_on_cluster_for_replicated_named_collections_queries>{int(ignore)}", + ) + node.query("SYSTEM RELOAD CONFIG") + + with expected_raise: + node.query( + "DROP NAMED COLLECTION IF EXISTS test_nc ON CLUSTER `replicated_nc_nodes_cluster`" + ) + node.query( + f"CREATE NAMED COLLECTION test_nc ON CLUSTER `replicated_nc_nodes_cluster` AS key1=1, key2=2 OVERRIDABLE" + ) + node.query( + f"ALTER NAMED COLLECTION test_nc ON CLUSTER `replicated_nc_nodes_cluster` SET key2=3" + ) + node.query( + f"DROP NAMED COLLECTION test_nc ON CLUSTER `replicated_nc_nodes_cluster`" + ) diff --git a/tests/integration/test_storage_delta/test.py b/tests/integration/test_storage_delta/test.py index 4cb71895881..d3dd7cfe52a 100644 --- a/tests/integration/test_storage_delta/test.py +++ b/tests/integration/test_storage_delta/test.py @@ -596,19 +596,116 @@ def test_partition_columns(started_cluster): ) assert result == 1 - # instance.query( - # f""" - # DROP TABLE IF EXISTS {TABLE_NAME}; - # CREATE TABLE {TABLE_NAME} (a Int32, b String, c DateTime) - # ENGINE=DeltaLake('http://{started_cluster.minio_ip}:{started_cluster.minio_port}/{bucket}/{result_file}/', 'minio', 'minio123')""" - # ) - # assert ( - # int( - # instance.query( - # f"SELECT count() FROM {TABLE_NAME} WHERE c != toDateTime('2000/01/05')" - # ) - # ) - # == num_rows - 1 - # ) - # instance.query(f"SELECT a, b, c, FROM {TABLE_NAME}") - # assert False + instance.query( + f""" + DROP TABLE IF EXISTS {TABLE_NAME}; + CREATE TABLE {TABLE_NAME} (a Nullable(Int32), b Nullable(String), c Nullable(Date32), d Nullable(Int32), e Nullable(Bool)) + ENGINE=DeltaLake('http://{started_cluster.minio_ip}:{started_cluster.minio_port}/{bucket}/{result_file}/', 'minio', 'minio123')""" + ) + assert ( + """1 test1 2000-01-01 1 false +2 test2 2000-01-02 2 false +3 test3 2000-01-03 3 false +4 test4 2000-01-04 4 false +5 test5 2000-01-05 5 false +6 test6 2000-01-06 6 false +7 test7 2000-01-07 7 false +8 test8 2000-01-08 8 false +9 test9 2000-01-09 9 false""" + == instance.query(f"SELECT * FROM {TABLE_NAME} ORDER BY b").strip() + ) + + assert ( + int( + instance.query( + f"SELECT count() FROM {TABLE_NAME} WHERE c == toDateTime('2000/01/05')" + ) + ) + == 1 + ) + + # Subset of columns should work. + instance.query( + f""" + DROP TABLE IF EXISTS {TABLE_NAME}; + CREATE TABLE {TABLE_NAME} (b Nullable(String), c Nullable(Date32), d Nullable(Int32)) + ENGINE=DeltaLake('http://{started_cluster.minio_ip}:{started_cluster.minio_port}/{bucket}/{result_file}/', 'minio', 'minio123')""" + ) + assert ( + """test1 2000-01-01 1 +test2 2000-01-02 2 +test3 2000-01-03 3 +test4 2000-01-04 4 +test5 2000-01-05 5 +test6 2000-01-06 6 +test7 2000-01-07 7 +test8 2000-01-08 8 +test9 2000-01-09 9""" + == instance.query(f"SELECT * FROM {TABLE_NAME} ORDER BY b").strip() + ) + + for i in range(num_rows + 1, 2 * num_rows + 1): + data = [ + ( + i, + "test" + str(i), + datetime.strptime(f"2000-01-{i}", "%Y-%m-%d"), + i, + False, + ) + ] + df = spark.createDataFrame(data=data, schema=schema) + df.printSchema() + df.write.mode("append").format("delta").partitionBy(partition_columns).save( + f"/{TABLE_NAME}" + ) + + files = upload_directory(minio_client, bucket, f"/{TABLE_NAME}", "") + ok = False + for file in files: + if file.endswith("last_checkpoint"): + ok = True + assert ok + + result = int( + instance.query( + f"""SELECT count() + FROM deltaLake('http://{started_cluster.minio_ip}:{started_cluster.minio_port}/{bucket}/{result_file}/', 'minio', 'minio123') + """ + ) + ) + assert result == num_rows * 2 + + assert ( + """1 test1 2000-01-01 1 false +2 test2 2000-01-02 2 false +3 test3 2000-01-03 3 false +4 test4 2000-01-04 4 false +5 test5 2000-01-05 5 false +6 test6 2000-01-06 6 false +7 test7 2000-01-07 7 false +8 test8 2000-01-08 8 false +9 test9 2000-01-09 9 false +10 test10 2000-01-10 10 false +11 test11 2000-01-11 11 false +12 test12 2000-01-12 12 false +13 test13 2000-01-13 13 false +14 test14 2000-01-14 14 false +15 test15 2000-01-15 15 false +16 test16 2000-01-16 16 false +17 test17 2000-01-17 17 false +18 test18 2000-01-18 18 false""" + == instance.query( + f""" +SELECT * FROM deltaLake('http://{started_cluster.minio_ip}:{started_cluster.minio_port}/{bucket}/{result_file}/', 'minio', 'minio123') ORDER BY c + """ + ).strip() + ) + assert ( + int( + instance.query( + f"SELECT count() FROM {TABLE_NAME} WHERE c == toDateTime('2000/01/15')" + ) + ) + == 1 + ) diff --git a/tests/integration/test_storage_rabbitmq/test.py b/tests/integration/test_storage_rabbitmq/test.py index 3240039ee81..f885a3507ac 100644 --- a/tests/integration/test_storage_rabbitmq/test.py +++ b/tests/integration/test_storage_rabbitmq/test.py @@ -78,13 +78,13 @@ def wait_rabbitmq_to_start(rabbitmq_docker_id, cookie, timeout=180): def kill_rabbitmq(rabbitmq_id): p = subprocess.Popen(("docker", "stop", rabbitmq_id), stdout=subprocess.PIPE) - p.communicate() + p.wait(timeout=60) return p.returncode == 0 def revive_rabbitmq(rabbitmq_id, cookie): p = subprocess.Popen(("docker", "start", rabbitmq_id), stdout=subprocess.PIPE) - p.communicate() + p.wait(timeout=60) wait_rabbitmq_to_start(rabbitmq_id, cookie) diff --git a/tests/queries/0_stateless/02841_not_ready_set_constraints.reference b/tests/queries/0_stateless/02841_not_ready_set_constraints.reference index d81cc0710eb..daaac9e3030 100644 --- a/tests/queries/0_stateless/02841_not_ready_set_constraints.reference +++ b/tests/queries/0_stateless/02841_not_ready_set_constraints.reference @@ -1 +1,2 @@ 42 +42 diff --git a/tests/queries/0_stateless/02841_not_ready_set_constraints.sql b/tests/queries/0_stateless/02841_not_ready_set_constraints.sql index ecdf4d50635..274940f50a3 100644 --- a/tests/queries/0_stateless/02841_not_ready_set_constraints.sql +++ b/tests/queries/0_stateless/02841_not_ready_set_constraints.sql @@ -17,3 +17,27 @@ ENGINE = MergeTree ORDER BY conversation; INSERT INTO t2(conversation) VALUES (42); select * from t2; + +drop table t1; + +INSERT INTO t2(conversation) VALUES (42); -- { serverError UNKNOWN_TABLE } + +drop table t2; + +CREATE TABLE t2 ( + `conversation` UInt64, + CONSTRAINT constraint_conversation CHECK conversation IN (SELECT id FROM t1) +) +ENGINE = MergeTree ORDER BY conversation; + +INSERT INTO t2(conversation) VALUES (42); -- { serverError UNKNOWN_TABLE } + +CREATE TABLE t1 ( + `id` UInt64 +) +ENGINE = MergeTree ORDER BY id; + +INSERT INTO t1(id) VALUES (42); + +INSERT INTO t2(conversation) VALUES (42); +select * from t2; diff --git a/tests/queries/0_stateless/03198_json_extract_more_types.reference b/tests/queries/0_stateless/03198_json_extract_more_types.reference new file mode 100644 index 00000000000..9a6580ff81b --- /dev/null +++ b/tests/queries/0_stateless/03198_json_extract_more_types.reference @@ -0,0 +1,21 @@ +2020-01-01 +2020-01-01 +2020-01-01 00:00:00 +2020-01-01 00:00:00.000000 +127.0.0.1 +2001:db8:85a3::8a2e:370:7334 +42 +42 +42 +42 +42 +42 +42 +42 +42 +42 +Hello +Hello +\0\0\0 +Hello\0\0\0\0\0 +5801c962-1182-458a-89f8-d077da5074f9 diff --git a/tests/queries/0_stateless/03198_json_extract_more_types.sql b/tests/queries/0_stateless/03198_json_extract_more_types.sql new file mode 100644 index 00000000000..28d24bbb271 --- /dev/null +++ b/tests/queries/0_stateless/03198_json_extract_more_types.sql @@ -0,0 +1,29 @@ +set allow_suspicious_low_cardinality_types=1; + +select JSONExtract('{"a" : "2020-01-01"}', 'a', 'Date'); +select JSONExtract('{"a" : "2020-01-01"}', 'a', 'Date32'); +select JSONExtract('{"a" : "2020-01-01 00:00:00"}', 'a', 'DateTime'); +select JSONExtract('{"a" : "2020-01-01 00:00:00.000000"}', 'a', 'DateTime64(6)'); +select JSONExtract('{"a" : "127.0.0.1"}', 'a', 'IPv4'); +select JSONExtract('{"a" : "2001:0db8:85a3:0000:0000:8a2e:0370:7334"}', 'a', 'IPv6'); + + +select JSONExtract('{"a" : 42}', 'a', 'LowCardinality(UInt8)'); +select JSONExtract('{"a" : 42}', 'a', 'LowCardinality(Int8)'); +select JSONExtract('{"a" : 42}', 'a', 'LowCardinality(UInt16)'); +select JSONExtract('{"a" : 42}', 'a', 'LowCardinality(Int16)'); +select JSONExtract('{"a" : 42}', 'a', 'LowCardinality(UInt32)'); +select JSONExtract('{"a" : 42}', 'a', 'LowCardinality(Int32)'); +select JSONExtract('{"a" : 42}', 'a', 'LowCardinality(UInt64)'); +select JSONExtract('{"a" : 42}', 'a', 'LowCardinality(Int64)'); + +select JSONExtract('{"a" : 42}', 'a', 'LowCardinality(Float32)'); +select JSONExtract('{"a" : 42}', 'a', 'LowCardinality(Float32)'); + +select JSONExtract('{"a" : "Hello"}', 'a', 'LowCardinality(String)'); +select JSONExtract('{"a" : "Hello"}', 'a', 'LowCardinality(FixedString(5))'); +select JSONExtract('{"a" : "Hello"}', 'a', 'LowCardinality(FixedString(3))'); +select JSONExtract('{"a" : "Hello"}', 'a', 'LowCardinality(FixedString(10))'); + +select JSONExtract('{"a" : "5801c962-1182-458a-89f8-d077da5074f9"}', 'a', 'LowCardinality(UUID)'); + diff --git a/tests/queries/0_stateless/03199_join_with_materialized_column.reference b/tests/queries/0_stateless/03199_join_with_materialized_column.reference new file mode 100644 index 00000000000..e69de29bb2d diff --git a/tests/queries/0_stateless/03199_join_with_materialized_column.sql b/tests/queries/0_stateless/03199_join_with_materialized_column.sql new file mode 100644 index 00000000000..8c53c5b3e66 --- /dev/null +++ b/tests/queries/0_stateless/03199_join_with_materialized_column.sql @@ -0,0 +1,6 @@ +SET allow_experimental_analyzer = 1; + +DROP TABLE IF EXISTS table_with_materialized; +CREATE TABLE table_with_materialized (col String MATERIALIZED 'A') ENGINE = Memory; +SELECT number FROM numbers(1) AS n, table_with_materialized; +DROP TABLE table_with_materialized; diff --git a/tests/queries/0_stateless/03199_json_extract_dynamic.reference b/tests/queries/0_stateless/03199_json_extract_dynamic.reference new file mode 100644 index 00000000000..759b7763cd1 --- /dev/null +++ b/tests/queries/0_stateless/03199_json_extract_dynamic.reference @@ -0,0 +1,30 @@ +true Bool +42 Int64 +-42 Int64 +18446744073709551615 UInt64 +42.42 Float64 +42 Int64 +-42 Int64 +18446744073709551615 UInt64 +Hello String +2020-01-01 Date +2020-01-01 00:00:00.000000000 DateTime64(9) +[1,2,3] Array(Nullable(Int64)) +['str1','str2','str3'] Array(Nullable(String)) +[[[1],[2,3,4]],[[5,6],[7]]] Array(Array(Array(Nullable(Int64)))) +['2020-01-01 00:00:00.000000000','2020-01-01 00:00:00.000000000'] Array(Nullable(DateTime64(9))) +['2020-01-01','2020-01-01 date'] Array(Nullable(String)) +['2020-01-01','2020-01-01 00:00:00','str'] Array(Nullable(String)) +['2020-01-01','2020-01-01 00:00:00','42'] Array(Nullable(String)) +['str','42'] Array(Nullable(String)) +[42,42.42] Array(Nullable(Float64)) +[42,18446744073709552000,42.42] Array(Nullable(Float64)) +[42,42.42] Array(Nullable(Float64)) +[NULL,NULL] Array(Nullable(String)) +[NULL,42] Array(Nullable(Int64)) +[[NULL],[],[42]] Array(Array(Nullable(Int64))) +[[],[NULL,NULL],[1,NULL,3],[NULL,2,NULL]] Array(Array(Nullable(Int64))) +[[],[NULL,NULL],['1',NULL,'3'],[NULL,'2',NULL],['2020-01-01']] Array(Array(Nullable(String))) +('str',42,[42]) Tuple(Nullable(String), Nullable(Int64), Array(Nullable(Int64))) +[42,18446744073709551615] Array(Nullable(UInt64)) +(-42,18446744073709551615) Tuple(Nullable(Int64), Nullable(UInt64)) diff --git a/tests/queries/0_stateless/03199_json_extract_dynamic.sql b/tests/queries/0_stateless/03199_json_extract_dynamic.sql new file mode 100644 index 00000000000..286949f4d3e --- /dev/null +++ b/tests/queries/0_stateless/03199_json_extract_dynamic.sql @@ -0,0 +1,37 @@ +set input_format_json_try_infer_numbers_from_strings=1; + +select JSONExtract(materialize('{"d" : true}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : 42}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : -42}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : 18446744073709551615}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : 42.42}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : "42"}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : "-42"}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : "18446744073709551615"}'), 'd', 'Dynamic') as d, dynamicType(d); + +select JSONExtract(materialize('{"d" : "Hello"}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : "2020-01-01"}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : "2020-01-01 00:00:00.000"}'), 'd', 'Dynamic') as d, dynamicType(d); + +select JSONExtract(materialize('{"d" : [1, 2, 3]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : ["str1", "str2", "str3"]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : [[[1], [2, 3, 4]], [[5, 6], [7]]]}'), 'd', 'Dynamic') as d, dynamicType(d); + +select JSONExtract(materialize('{"d" : ["2020-01-01", "2020-01-01 00:00:00"]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : ["2020-01-01", "2020-01-01 date"]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : ["2020-01-01", "2020-01-01 00:00:00", "str"]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : ["2020-01-01", "2020-01-01 00:00:00", "42"]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : ["str", "42"]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : [42, 42.42]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : [42, 18446744073709551615, 42.42]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : [42, 42.42]}'), 'd', 'Dynamic') as d, dynamicType(d); + +select JSONExtract(materialize('{"d" : [null, null]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : [null, 42]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : [[null], [], [42]]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"a" : [[], [null, null], ["1", null, "3"], [null, "2", null]]}'), 'a', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"a" : [[], [null, null], ["1", null, "3"], [null, "2", null], ["2020-01-01"]]}'), 'a', 'Dynamic') as d, dynamicType(d); + +select JSONExtract(materialize('{"d" : ["str", 42, [42]]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : [42, 18446744073709551615]}'), 'd', 'Dynamic') as d, dynamicType(d); +select JSONExtract(materialize('{"d" : [-42, 18446744073709551615]}'), 'd', 'Dynamic') as d, dynamicType(d); diff --git a/utils/check-style/aspell-ignore/en/aspell-dict.txt b/utils/check-style/aspell-ignore/en/aspell-dict.txt index 78c4b6bde95..ca2c4ec4192 100644 --- a/utils/check-style/aspell-ignore/en/aspell-dict.txt +++ b/utils/check-style/aspell-ignore/en/aspell-dict.txt @@ -1900,11 +1900,13 @@ kurtosis kurtpop kurtsamp laion +lagInFrame lang laravel largestTriangleThreeBuckets latencies ldap +leadInFrame leftPad leftPadUTF leftUTF