From dff2f3c3c637fd5c3b30ed0cf26fac75c22973ac Mon Sep 17 00:00:00 2001 From: Andrew Lamb Date: Thu, 18 Jul 2024 06:03:39 -0400 Subject: [PATCH] Minor: Clarify which parquet options are used for reading/writing (#11511) --- datafusion/common/src/config.rs | 63 ++++++++++--------- .../common/src/file_options/parquet_writer.rs | 1 + .../test_files/information_schema.slt | 52 +++++++-------- docs/source/user-guide/configs.md | 52 +++++++-------- 4 files changed, 87 insertions(+), 81 deletions(-) diff --git a/datafusion/common/src/config.rs b/datafusion/common/src/config.rs index 880f0119ce0d..b46b002baac0 100644 --- a/datafusion/common/src/config.rs +++ b/datafusion/common/src/config.rs @@ -315,93 +315,96 @@ config_namespace! { } config_namespace! { - /// Options related to parquet files + /// Options for reading and writing parquet files /// /// See also: [`SessionConfig`] /// /// [`SessionConfig`]: https://docs.rs/datafusion/latest/datafusion/prelude/struct.SessionConfig.html pub struct ParquetOptions { - /// If true, reads the Parquet data page level metadata (the + // The following options affect reading parquet files + + /// (reading) If true, reads the Parquet data page level metadata (the /// Page Index), if present, to reduce the I/O and number of /// rows decoded. pub enable_page_index: bool, default = true - /// If true, the parquet reader attempts to skip entire row groups based + /// (reading) If true, the parquet reader attempts to skip entire row groups based /// on the predicate in the query and the metadata (min/max values) stored in /// the parquet file pub pruning: bool, default = true - /// If true, the parquet reader skip the optional embedded metadata that may be in + /// (reading) If true, the parquet reader skip the optional embedded metadata that may be in /// the file Schema. This setting can help avoid schema conflicts when querying /// multiple parquet files with schemas containing compatible types but different metadata pub skip_metadata: bool, default = true - /// If specified, the parquet reader will try and fetch the last `size_hint` + /// (reading) If specified, the parquet reader will try and fetch the last `size_hint` /// bytes of the parquet file optimistically. If not specified, two reads are required: /// One read to fetch the 8-byte parquet footer and /// another to fetch the metadata length encoded in the footer pub metadata_size_hint: Option, default = None - /// If true, filter expressions are be applied during the parquet decoding operation to + /// (reading) If true, filter expressions are be applied during the parquet decoding operation to /// reduce the number of rows decoded. This optimization is sometimes called "late materialization". pub pushdown_filters: bool, default = false - /// If true, filter expressions evaluated during the parquet decoding operation + /// (reading) If true, filter expressions evaluated during the parquet decoding operation /// will be reordered heuristically to minimize the cost of evaluation. If false, /// the filters are applied in the same order as written in the query pub reorder_filters: bool, default = false - // The following map to parquet::file::properties::WriterProperties + // The following options affect writing to parquet files + // and map to parquet::file::properties::WriterProperties - /// Sets best effort maximum size of data page in bytes + /// (writing) Sets best effort maximum size of data page in bytes pub data_pagesize_limit: usize, default = 1024 * 1024 - /// Sets write_batch_size in bytes + /// (writing) Sets write_batch_size in bytes pub write_batch_size: usize, default = 1024 - /// Sets parquet writer version + /// (writing) Sets parquet writer version /// valid values are "1.0" and "2.0" pub writer_version: String, default = "1.0".into() - /// Sets default parquet compression codec + /// (writing) Sets default parquet compression codec. /// Valid values are: uncompressed, snappy, gzip(level), /// lzo, brotli(level), lz4, zstd(level), and lz4_raw. /// These values are not case sensitive. If NULL, uses /// default parquet writer setting pub compression: Option, default = Some("zstd(3)".into()) - /// Sets if dictionary encoding is enabled. If NULL, uses + /// (writing) Sets if dictionary encoding is enabled. If NULL, uses /// default parquet writer setting pub dictionary_enabled: Option, default = None - /// Sets best effort maximum dictionary page size, in bytes + /// (writing) Sets best effort maximum dictionary page size, in bytes pub dictionary_page_size_limit: usize, default = 1024 * 1024 - /// Sets if statistics are enabled for any column + /// (writing) Sets if statistics are enabled for any column /// Valid values are: "none", "chunk", and "page" /// These values are not case sensitive. If NULL, uses /// default parquet writer setting pub statistics_enabled: Option, default = None - /// Sets max statistics size for any column. If NULL, uses + /// (writing) Sets max statistics size for any column. If NULL, uses /// default parquet writer setting pub max_statistics_size: Option, default = None - /// Target maximum number of rows in each row group (defaults to 1M + /// (writing) Target maximum number of rows in each row group (defaults to 1M /// rows). Writing larger row groups requires more memory to write, but /// can get better compression and be faster to read. pub max_row_group_size: usize, default = 1024 * 1024 - /// Sets "created by" property + /// (writing) Sets "created by" property pub created_by: String, default = concat!("datafusion version ", env!("CARGO_PKG_VERSION")).into() - /// Sets column index truncate length + /// (writing) Sets column index truncate length pub column_index_truncate_length: Option, default = None - /// Sets best effort maximum number of rows in data page + /// (writing) Sets best effort maximum number of rows in data page pub data_page_row_count_limit: usize, default = usize::MAX - /// Sets default encoding for any column + /// (writing) Sets default encoding for any column. /// Valid values are: plain, plain_dictionary, rle, /// bit_packed, delta_binary_packed, delta_length_byte_array, /// delta_byte_array, rle_dictionary, and byte_stream_split. @@ -409,27 +412,27 @@ config_namespace! { /// default parquet writer setting pub encoding: Option, default = None - /// Use any available bloom filters when reading parquet files + /// (writing) Use any available bloom filters when reading parquet files pub bloom_filter_on_read: bool, default = true - /// Write bloom filters for all columns when creating parquet files + /// (writing) Write bloom filters for all columns when creating parquet files pub bloom_filter_on_write: bool, default = false - /// Sets bloom filter false positive probability. If NULL, uses + /// (writing) Sets bloom filter false positive probability. If NULL, uses /// default parquet writer setting pub bloom_filter_fpp: Option, default = None - /// Sets bloom filter number of distinct values. If NULL, uses + /// (writing) Sets bloom filter number of distinct values. If NULL, uses /// default parquet writer setting pub bloom_filter_ndv: Option, default = None - /// Controls whether DataFusion will attempt to speed up writing + /// (writing) Controls whether DataFusion will attempt to speed up writing /// parquet files by serializing them in parallel. Each column /// in each row group in each output file are serialized in parallel /// leveraging a maximum possible core count of n_files*n_row_groups*n_columns. pub allow_single_file_parallelism: bool, default = true - /// By default parallel parquet writer is tuned for minimum + /// (writing) By default parallel parquet writer is tuned for minimum /// memory usage in a streaming execution plan. You may see /// a performance benefit when writing large parquet files /// by increasing maximum_parallel_row_group_writers and @@ -440,7 +443,7 @@ config_namespace! { /// data frame. pub maximum_parallel_row_group_writers: usize, default = 1 - /// By default parallel parquet writer is tuned for minimum + /// (writing) By default parallel parquet writer is tuned for minimum /// memory usage in a streaming execution plan. You may see /// a performance benefit when writing large parquet files /// by increasing maximum_parallel_row_group_writers and @@ -450,7 +453,6 @@ config_namespace! { /// writing out already in-memory data, such as from a cached /// data frame. pub maximum_buffered_record_batches_per_stream: usize, default = 2 - } } @@ -1534,6 +1536,9 @@ macro_rules! config_namespace_with_hashmap { } config_namespace_with_hashmap! { + /// Options controlling parquet format for individual columns. + /// + /// See [`ParquetOptions`] for more details pub struct ColumnOptions { /// Sets if bloom filter is enabled for the column path. pub bloom_filter_enabled: Option, default = None diff --git a/datafusion/common/src/file_options/parquet_writer.rs b/datafusion/common/src/file_options/parquet_writer.rs index dd4bb8ce505e..abe7db2009a2 100644 --- a/datafusion/common/src/file_options/parquet_writer.rs +++ b/datafusion/common/src/file_options/parquet_writer.rs @@ -35,6 +35,7 @@ use parquet::{ /// Options for writing parquet files #[derive(Clone, Debug)] pub struct ParquetWriterOptions { + /// parquet-rs writer properties pub writer_options: WriterProperties, } diff --git a/datafusion/sqllogictest/test_files/information_schema.slt b/datafusion/sqllogictest/test_files/information_schema.slt index 95bea1223a9c..f7b755b01911 100644 --- a/datafusion/sqllogictest/test_files/information_schema.slt +++ b/datafusion/sqllogictest/test_files/information_schema.slt @@ -262,32 +262,32 @@ datafusion.execution.listing_table_ignore_subdirectory true Should sub directori datafusion.execution.max_buffered_batches_per_output_file 2 This is the maximum number of RecordBatches buffered for each output file being worked. Higher values can potentially give faster write performance at the cost of higher peak memory consumption datafusion.execution.meta_fetch_concurrency 32 Number of files to read in parallel when inferring schema and statistics datafusion.execution.minimum_parallel_output_files 4 Guarantees a minimum level of output files running in parallel. RecordBatches will be distributed in round robin fashion to each parallel writer. Each writer is closed and a new file opened once soft_max_rows_per_output_file is reached. -datafusion.execution.parquet.allow_single_file_parallelism true Controls whether DataFusion will attempt to speed up writing parquet files by serializing them in parallel. Each column in each row group in each output file are serialized in parallel leveraging a maximum possible core count of n_files*n_row_groups*n_columns. -datafusion.execution.parquet.bloom_filter_fpp NULL Sets bloom filter false positive probability. If NULL, uses default parquet writer setting -datafusion.execution.parquet.bloom_filter_ndv NULL Sets bloom filter number of distinct values. If NULL, uses default parquet writer setting -datafusion.execution.parquet.bloom_filter_on_read true Use any available bloom filters when reading parquet files -datafusion.execution.parquet.bloom_filter_on_write false Write bloom filters for all columns when creating parquet files -datafusion.execution.parquet.column_index_truncate_length NULL Sets column index truncate length -datafusion.execution.parquet.compression zstd(3) Sets default parquet compression codec Valid values are: uncompressed, snappy, gzip(level), lzo, brotli(level), lz4, zstd(level), and lz4_raw. These values are not case sensitive. If NULL, uses default parquet writer setting -datafusion.execution.parquet.created_by datafusion Sets "created by" property -datafusion.execution.parquet.data_page_row_count_limit 18446744073709551615 Sets best effort maximum number of rows in data page -datafusion.execution.parquet.data_pagesize_limit 1048576 Sets best effort maximum size of data page in bytes -datafusion.execution.parquet.dictionary_enabled NULL Sets if dictionary encoding is enabled. If NULL, uses default parquet writer setting -datafusion.execution.parquet.dictionary_page_size_limit 1048576 Sets best effort maximum dictionary page size, in bytes -datafusion.execution.parquet.enable_page_index true If true, reads the Parquet data page level metadata (the Page Index), if present, to reduce the I/O and number of rows decoded. -datafusion.execution.parquet.encoding NULL Sets default encoding for any column Valid values are: plain, plain_dictionary, rle, bit_packed, delta_binary_packed, delta_length_byte_array, delta_byte_array, rle_dictionary, and byte_stream_split. These values are not case sensitive. If NULL, uses default parquet writer setting -datafusion.execution.parquet.max_row_group_size 1048576 Target maximum number of rows in each row group (defaults to 1M rows). Writing larger row groups requires more memory to write, but can get better compression and be faster to read. -datafusion.execution.parquet.max_statistics_size NULL Sets max statistics size for any column. If NULL, uses default parquet writer setting -datafusion.execution.parquet.maximum_buffered_record_batches_per_stream 2 By default parallel parquet writer is tuned for minimum memory usage in a streaming execution plan. You may see a performance benefit when writing large parquet files by increasing maximum_parallel_row_group_writers and maximum_buffered_record_batches_per_stream if your system has idle cores and can tolerate additional memory usage. Boosting these values is likely worthwhile when writing out already in-memory data, such as from a cached data frame. -datafusion.execution.parquet.maximum_parallel_row_group_writers 1 By default parallel parquet writer is tuned for minimum memory usage in a streaming execution plan. You may see a performance benefit when writing large parquet files by increasing maximum_parallel_row_group_writers and maximum_buffered_record_batches_per_stream if your system has idle cores and can tolerate additional memory usage. Boosting these values is likely worthwhile when writing out already in-memory data, such as from a cached data frame. -datafusion.execution.parquet.metadata_size_hint NULL If specified, the parquet reader will try and fetch the last `size_hint` bytes of the parquet file optimistically. If not specified, two reads are required: One read to fetch the 8-byte parquet footer and another to fetch the metadata length encoded in the footer -datafusion.execution.parquet.pruning true If true, the parquet reader attempts to skip entire row groups based on the predicate in the query and the metadata (min/max values) stored in the parquet file -datafusion.execution.parquet.pushdown_filters false If true, filter expressions are be applied during the parquet decoding operation to reduce the number of rows decoded. This optimization is sometimes called "late materialization". -datafusion.execution.parquet.reorder_filters false If true, filter expressions evaluated during the parquet decoding operation will be reordered heuristically to minimize the cost of evaluation. If false, the filters are applied in the same order as written in the query -datafusion.execution.parquet.skip_metadata true If true, the parquet reader skip the optional embedded metadata that may be in the file Schema. This setting can help avoid schema conflicts when querying multiple parquet files with schemas containing compatible types but different metadata -datafusion.execution.parquet.statistics_enabled NULL Sets if statistics are enabled for any column Valid values are: "none", "chunk", and "page" These values are not case sensitive. If NULL, uses default parquet writer setting -datafusion.execution.parquet.write_batch_size 1024 Sets write_batch_size in bytes -datafusion.execution.parquet.writer_version 1.0 Sets parquet writer version valid values are "1.0" and "2.0" +datafusion.execution.parquet.allow_single_file_parallelism true (writing) Controls whether DataFusion will attempt to speed up writing parquet files by serializing them in parallel. Each column in each row group in each output file are serialized in parallel leveraging a maximum possible core count of n_files*n_row_groups*n_columns. +datafusion.execution.parquet.bloom_filter_fpp NULL (writing) Sets bloom filter false positive probability. If NULL, uses default parquet writer setting +datafusion.execution.parquet.bloom_filter_ndv NULL (writing) Sets bloom filter number of distinct values. If NULL, uses default parquet writer setting +datafusion.execution.parquet.bloom_filter_on_read true (writing) Use any available bloom filters when reading parquet files +datafusion.execution.parquet.bloom_filter_on_write false (writing) Write bloom filters for all columns when creating parquet files +datafusion.execution.parquet.column_index_truncate_length NULL (writing) Sets column index truncate length +datafusion.execution.parquet.compression zstd(3) (writing) Sets default parquet compression codec. Valid values are: uncompressed, snappy, gzip(level), lzo, brotli(level), lz4, zstd(level), and lz4_raw. These values are not case sensitive. If NULL, uses default parquet writer setting +datafusion.execution.parquet.created_by datafusion (writing) Sets "created by" property +datafusion.execution.parquet.data_page_row_count_limit 18446744073709551615 (writing) Sets best effort maximum number of rows in data page +datafusion.execution.parquet.data_pagesize_limit 1048576 (writing) Sets best effort maximum size of data page in bytes +datafusion.execution.parquet.dictionary_enabled NULL (writing) Sets if dictionary encoding is enabled. If NULL, uses default parquet writer setting +datafusion.execution.parquet.dictionary_page_size_limit 1048576 (writing) Sets best effort maximum dictionary page size, in bytes +datafusion.execution.parquet.enable_page_index true (reading) If true, reads the Parquet data page level metadata (the Page Index), if present, to reduce the I/O and number of rows decoded. +datafusion.execution.parquet.encoding NULL (writing) Sets default encoding for any column. Valid values are: plain, plain_dictionary, rle, bit_packed, delta_binary_packed, delta_length_byte_array, delta_byte_array, rle_dictionary, and byte_stream_split. These values are not case sensitive. If NULL, uses default parquet writer setting +datafusion.execution.parquet.max_row_group_size 1048576 (writing) Target maximum number of rows in each row group (defaults to 1M rows). Writing larger row groups requires more memory to write, but can get better compression and be faster to read. +datafusion.execution.parquet.max_statistics_size NULL (writing) Sets max statistics size for any column. If NULL, uses default parquet writer setting +datafusion.execution.parquet.maximum_buffered_record_batches_per_stream 2 (writing) By default parallel parquet writer is tuned for minimum memory usage in a streaming execution plan. You may see a performance benefit when writing large parquet files by increasing maximum_parallel_row_group_writers and maximum_buffered_record_batches_per_stream if your system has idle cores and can tolerate additional memory usage. Boosting these values is likely worthwhile when writing out already in-memory data, such as from a cached data frame. +datafusion.execution.parquet.maximum_parallel_row_group_writers 1 (writing) By default parallel parquet writer is tuned for minimum memory usage in a streaming execution plan. You may see a performance benefit when writing large parquet files by increasing maximum_parallel_row_group_writers and maximum_buffered_record_batches_per_stream if your system has idle cores and can tolerate additional memory usage. Boosting these values is likely worthwhile when writing out already in-memory data, such as from a cached data frame. +datafusion.execution.parquet.metadata_size_hint NULL (reading) If specified, the parquet reader will try and fetch the last `size_hint` bytes of the parquet file optimistically. If not specified, two reads are required: One read to fetch the 8-byte parquet footer and another to fetch the metadata length encoded in the footer +datafusion.execution.parquet.pruning true (reading) If true, the parquet reader attempts to skip entire row groups based on the predicate in the query and the metadata (min/max values) stored in the parquet file +datafusion.execution.parquet.pushdown_filters false (reading) If true, filter expressions are be applied during the parquet decoding operation to reduce the number of rows decoded. This optimization is sometimes called "late materialization". +datafusion.execution.parquet.reorder_filters false (reading) If true, filter expressions evaluated during the parquet decoding operation will be reordered heuristically to minimize the cost of evaluation. If false, the filters are applied in the same order as written in the query +datafusion.execution.parquet.skip_metadata true (reading) If true, the parquet reader skip the optional embedded metadata that may be in the file Schema. This setting can help avoid schema conflicts when querying multiple parquet files with schemas containing compatible types but different metadata +datafusion.execution.parquet.statistics_enabled NULL (writing) Sets if statistics are enabled for any column Valid values are: "none", "chunk", and "page" These values are not case sensitive. If NULL, uses default parquet writer setting +datafusion.execution.parquet.write_batch_size 1024 (writing) Sets write_batch_size in bytes +datafusion.execution.parquet.writer_version 1.0 (writing) Sets parquet writer version valid values are "1.0" and "2.0" datafusion.execution.planning_concurrency 13 Fan-out during initial physical planning. This is mostly use to plan `UNION` children in parallel. Defaults to the number of CPU cores on the system datafusion.execution.soft_max_rows_per_output_file 50000000 Target number of rows in output files when writing multiple. This is a soft max, so it can be exceeded slightly. There also will be one file smaller than the limit if the total number of rows written is not roughly divisible by the soft max datafusion.execution.sort_in_place_threshold_bytes 1048576 When sorting, below what size should data be concatenated and sorted in a single RecordBatch rather than sorted in batches and merged. diff --git a/docs/source/user-guide/configs.md b/docs/source/user-guide/configs.md index 5130b0a56d0e..8d3ecbc98544 100644 --- a/docs/source/user-guide/configs.md +++ b/docs/source/user-guide/configs.md @@ -49,32 +49,32 @@ Environment variables are read during `SessionConfig` initialisation so they mus | datafusion.execution.collect_statistics | false | Should DataFusion collect statistics after listing files | | datafusion.execution.target_partitions | 0 | Number of partitions for query execution. Increasing partitions can increase concurrency. Defaults to the number of CPU cores on the system | | datafusion.execution.time_zone | +00:00 | The default time zone Some functions, e.g. `EXTRACT(HOUR from SOME_TIME)`, shift the underlying datetime according to this time zone, and then extract the hour | -| datafusion.execution.parquet.enable_page_index | true | If true, reads the Parquet data page level metadata (the Page Index), if present, to reduce the I/O and number of rows decoded. | -| datafusion.execution.parquet.pruning | true | If true, the parquet reader attempts to skip entire row groups based on the predicate in the query and the metadata (min/max values) stored in the parquet file | -| datafusion.execution.parquet.skip_metadata | true | If true, the parquet reader skip the optional embedded metadata that may be in the file Schema. This setting can help avoid schema conflicts when querying multiple parquet files with schemas containing compatible types but different metadata | -| datafusion.execution.parquet.metadata_size_hint | NULL | If specified, the parquet reader will try and fetch the last `size_hint` bytes of the parquet file optimistically. If not specified, two reads are required: One read to fetch the 8-byte parquet footer and another to fetch the metadata length encoded in the footer | -| datafusion.execution.parquet.pushdown_filters | false | If true, filter expressions are be applied during the parquet decoding operation to reduce the number of rows decoded. This optimization is sometimes called "late materialization". | -| datafusion.execution.parquet.reorder_filters | false | If true, filter expressions evaluated during the parquet decoding operation will be reordered heuristically to minimize the cost of evaluation. If false, the filters are applied in the same order as written in the query | -| datafusion.execution.parquet.data_pagesize_limit | 1048576 | Sets best effort maximum size of data page in bytes | -| datafusion.execution.parquet.write_batch_size | 1024 | Sets write_batch_size in bytes | -| datafusion.execution.parquet.writer_version | 1.0 | Sets parquet writer version valid values are "1.0" and "2.0" | -| datafusion.execution.parquet.compression | zstd(3) | Sets default parquet compression codec Valid values are: uncompressed, snappy, gzip(level), lzo, brotli(level), lz4, zstd(level), and lz4_raw. These values are not case sensitive. If NULL, uses default parquet writer setting | -| datafusion.execution.parquet.dictionary_enabled | NULL | Sets if dictionary encoding is enabled. If NULL, uses default parquet writer setting | -| datafusion.execution.parquet.dictionary_page_size_limit | 1048576 | Sets best effort maximum dictionary page size, in bytes | -| datafusion.execution.parquet.statistics_enabled | NULL | Sets if statistics are enabled for any column Valid values are: "none", "chunk", and "page" These values are not case sensitive. If NULL, uses default parquet writer setting | -| datafusion.execution.parquet.max_statistics_size | NULL | Sets max statistics size for any column. If NULL, uses default parquet writer setting | -| datafusion.execution.parquet.max_row_group_size | 1048576 | Target maximum number of rows in each row group (defaults to 1M rows). Writing larger row groups requires more memory to write, but can get better compression and be faster to read. | -| datafusion.execution.parquet.created_by | datafusion version 40.0.0 | Sets "created by" property | -| datafusion.execution.parquet.column_index_truncate_length | NULL | Sets column index truncate length | -| datafusion.execution.parquet.data_page_row_count_limit | 18446744073709551615 | Sets best effort maximum number of rows in data page | -| datafusion.execution.parquet.encoding | NULL | Sets default encoding for any column Valid values are: plain, plain_dictionary, rle, bit_packed, delta_binary_packed, delta_length_byte_array, delta_byte_array, rle_dictionary, and byte_stream_split. These values are not case sensitive. If NULL, uses default parquet writer setting | -| datafusion.execution.parquet.bloom_filter_on_read | true | Use any available bloom filters when reading parquet files | -| datafusion.execution.parquet.bloom_filter_on_write | false | Write bloom filters for all columns when creating parquet files | -| datafusion.execution.parquet.bloom_filter_fpp | NULL | Sets bloom filter false positive probability. If NULL, uses default parquet writer setting | -| datafusion.execution.parquet.bloom_filter_ndv | NULL | Sets bloom filter number of distinct values. If NULL, uses default parquet writer setting | -| datafusion.execution.parquet.allow_single_file_parallelism | true | Controls whether DataFusion will attempt to speed up writing parquet files by serializing them in parallel. Each column in each row group in each output file are serialized in parallel leveraging a maximum possible core count of n_files*n_row_groups*n_columns. | -| datafusion.execution.parquet.maximum_parallel_row_group_writers | 1 | By default parallel parquet writer is tuned for minimum memory usage in a streaming execution plan. You may see a performance benefit when writing large parquet files by increasing maximum_parallel_row_group_writers and maximum_buffered_record_batches_per_stream if your system has idle cores and can tolerate additional memory usage. Boosting these values is likely worthwhile when writing out already in-memory data, such as from a cached data frame. | -| datafusion.execution.parquet.maximum_buffered_record_batches_per_stream | 2 | By default parallel parquet writer is tuned for minimum memory usage in a streaming execution plan. You may see a performance benefit when writing large parquet files by increasing maximum_parallel_row_group_writers and maximum_buffered_record_batches_per_stream if your system has idle cores and can tolerate additional memory usage. Boosting these values is likely worthwhile when writing out already in-memory data, such as from a cached data frame. | +| datafusion.execution.parquet.enable_page_index | true | (reading) If true, reads the Parquet data page level metadata (the Page Index), if present, to reduce the I/O and number of rows decoded. | +| datafusion.execution.parquet.pruning | true | (reading) If true, the parquet reader attempts to skip entire row groups based on the predicate in the query and the metadata (min/max values) stored in the parquet file | +| datafusion.execution.parquet.skip_metadata | true | (reading) If true, the parquet reader skip the optional embedded metadata that may be in the file Schema. This setting can help avoid schema conflicts when querying multiple parquet files with schemas containing compatible types but different metadata | +| datafusion.execution.parquet.metadata_size_hint | NULL | (reading) If specified, the parquet reader will try and fetch the last `size_hint` bytes of the parquet file optimistically. If not specified, two reads are required: One read to fetch the 8-byte parquet footer and another to fetch the metadata length encoded in the footer | +| datafusion.execution.parquet.pushdown_filters | false | (reading) If true, filter expressions are be applied during the parquet decoding operation to reduce the number of rows decoded. This optimization is sometimes called "late materialization". | +| datafusion.execution.parquet.reorder_filters | false | (reading) If true, filter expressions evaluated during the parquet decoding operation will be reordered heuristically to minimize the cost of evaluation. If false, the filters are applied in the same order as written in the query | +| datafusion.execution.parquet.data_pagesize_limit | 1048576 | (writing) Sets best effort maximum size of data page in bytes | +| datafusion.execution.parquet.write_batch_size | 1024 | (writing) Sets write_batch_size in bytes | +| datafusion.execution.parquet.writer_version | 1.0 | (writing) Sets parquet writer version valid values are "1.0" and "2.0" | +| datafusion.execution.parquet.compression | zstd(3) | (writing) Sets default parquet compression codec. Valid values are: uncompressed, snappy, gzip(level), lzo, brotli(level), lz4, zstd(level), and lz4_raw. These values are not case sensitive. If NULL, uses default parquet writer setting | +| datafusion.execution.parquet.dictionary_enabled | NULL | (writing) Sets if dictionary encoding is enabled. If NULL, uses default parquet writer setting | +| datafusion.execution.parquet.dictionary_page_size_limit | 1048576 | (writing) Sets best effort maximum dictionary page size, in bytes | +| datafusion.execution.parquet.statistics_enabled | NULL | (writing) Sets if statistics are enabled for any column Valid values are: "none", "chunk", and "page" These values are not case sensitive. If NULL, uses default parquet writer setting | +| datafusion.execution.parquet.max_statistics_size | NULL | (writing) Sets max statistics size for any column. If NULL, uses default parquet writer setting | +| datafusion.execution.parquet.max_row_group_size | 1048576 | (writing) Target maximum number of rows in each row group (defaults to 1M rows). Writing larger row groups requires more memory to write, but can get better compression and be faster to read. | +| datafusion.execution.parquet.created_by | datafusion version 40.0.0 | (writing) Sets "created by" property | +| datafusion.execution.parquet.column_index_truncate_length | NULL | (writing) Sets column index truncate length | +| datafusion.execution.parquet.data_page_row_count_limit | 18446744073709551615 | (writing) Sets best effort maximum number of rows in data page | +| datafusion.execution.parquet.encoding | NULL | (writing) Sets default encoding for any column. Valid values are: plain, plain_dictionary, rle, bit_packed, delta_binary_packed, delta_length_byte_array, delta_byte_array, rle_dictionary, and byte_stream_split. These values are not case sensitive. If NULL, uses default parquet writer setting | +| datafusion.execution.parquet.bloom_filter_on_read | true | (writing) Use any available bloom filters when reading parquet files | +| datafusion.execution.parquet.bloom_filter_on_write | false | (writing) Write bloom filters for all columns when creating parquet files | +| datafusion.execution.parquet.bloom_filter_fpp | NULL | (writing) Sets bloom filter false positive probability. If NULL, uses default parquet writer setting | +| datafusion.execution.parquet.bloom_filter_ndv | NULL | (writing) Sets bloom filter number of distinct values. If NULL, uses default parquet writer setting | +| datafusion.execution.parquet.allow_single_file_parallelism | true | (writing) Controls whether DataFusion will attempt to speed up writing parquet files by serializing them in parallel. Each column in each row group in each output file are serialized in parallel leveraging a maximum possible core count of n_files*n_row_groups*n_columns. | +| datafusion.execution.parquet.maximum_parallel_row_group_writers | 1 | (writing) By default parallel parquet writer is tuned for minimum memory usage in a streaming execution plan. You may see a performance benefit when writing large parquet files by increasing maximum_parallel_row_group_writers and maximum_buffered_record_batches_per_stream if your system has idle cores and can tolerate additional memory usage. Boosting these values is likely worthwhile when writing out already in-memory data, such as from a cached data frame. | +| datafusion.execution.parquet.maximum_buffered_record_batches_per_stream | 2 | (writing) By default parallel parquet writer is tuned for minimum memory usage in a streaming execution plan. You may see a performance benefit when writing large parquet files by increasing maximum_parallel_row_group_writers and maximum_buffered_record_batches_per_stream if your system has idle cores and can tolerate additional memory usage. Boosting these values is likely worthwhile when writing out already in-memory data, such as from a cached data frame. | | datafusion.execution.aggregate.scalar_update_factor | 10 | Specifies the threshold for using `ScalarValue`s to update accumulators during high-cardinality aggregations for each input batch. The aggregation is considered high-cardinality if the number of affected groups is greater than or equal to `batch_size / scalar_update_factor`. In such cases, `ScalarValue`s are utilized for updating accumulators, rather than the default batch-slice approach. This can lead to performance improvements. By adjusting the `scalar_update_factor`, you can balance the trade-off between more efficient accumulator updates and the number of groups affected. | | datafusion.execution.planning_concurrency | 0 | Fan-out during initial physical planning. This is mostly use to plan `UNION` children in parallel. Defaults to the number of CPU cores on the system | | datafusion.execution.sort_spill_reservation_bytes | 10485760 | Specifies the reserved memory for each spillable sort operation to facilitate an in-memory merge. When a sort operation spills to disk, the in-memory data must be sorted and merged before being written to a file. This setting reserves a specific amount of memory for that in-memory sort/merge process. Note: This setting is irrelevant if the sort operation cannot spill (i.e., if there's no `DiskManager` configured). |