File Formats#

CSV reader#

struct ConvertOptions#

Public Functions

Status Validate() const#: Test that all set options are valid.

Public Members

bool check_utf8 = true#: Whether to check UTF8 validity of string columns.

std::unordered_map<std::string, std::shared_ptr<DataType>> column_types#: Optional per-column types (disabling type inference on those columns)

std::vector<std::string> null_values#: Recognized spellings for null values.

std::vector<std::string> true_values#: Recognized spellings for boolean true values.

std::vector<std::string> false_values#: Recognized spellings for boolean false values.

bool strings_can_be_null = false#

Whether string / binary columns can have null values.

If true, then strings in “null_values” are considered null for string columns. If false, then all strings are valid string values.

bool quoted_strings_can_be_null = true#

Whether quoted values can be null.

If true, then strings in “null_values” are also considered null when they appear quoted in the CSV file. Otherwise, quoted values are never considered null.

bool auto_dict_encode = false#

Whether to try to automatically dict-encode string / binary data.

If true, then when type inference detects a string or binary column, it is dict-encoded up to auto_dict_max_cardinality distinct values (per chunk), after which it switches to regular encoding.

This setting is ignored for non-inferred columns (those in column_types).

char decimal_point = '.'#: Decimal point character for floating-point and decimal data.

std::vector<std::string> include_columns#

If non-empty, indicates the names of columns from the CSV file that should be actually read and converted (in the vector’s order).

Columns not in this vector will be ignored.

bool include_missing_columns = false#

If false, columns in include_columns but not in the CSV file will error out.

If true, columns in include_columns but not in the CSV file will produce a column of nulls (whose type is selected using column_types, or null by default) This option is ignored if include_columns is empty.

std::vector<std::shared_ptr<TimestampParser>> timestamp_parsers#

User-defined timestamp parsers, using the virtual parser interface in arrow/util/value_parsing.h.

More than one parser can be specified, and the CSV conversion logic will try parsing values starting from the beginning of this vector. If no parsers are specified, we use the default built-in ISO-8601 parser.

Public Static Functions

static ConvertOptions Defaults()#: Create conversion options with default values, including conventional values for null_values, true_values and false_values

struct ParseOptions#

Public Functions

Status Validate() const#: Test that all set options are valid.

Public Members

char delimiter = ','#: Field delimiter.

bool quoting = true#: Whether quoting is used.

char quote_char = '"'#: Quoting character (if quoting is true)

bool double_quote = true#: Whether a quote inside a value is double-quoted.

bool escaping = false#: Whether escaping is used.

char escape_char = kDefaultEscapeChar#: Escaping character (if escaping is true)

bool newlines_in_values = false#: Whether values are allowed to contain CR (0x0d) and LF (0x0a) characters.

bool ignore_empty_lines = true#

Whether empty lines are ignored.

If false, an empty line represents a single empty value (assuming a one-column CSV file).

InvalidRowHandler invalid_row_handler#: A handler function for rows which do not have the correct number of columns.

Public Static Functions

static ParseOptions Defaults()#: Create parsing options with default values.

struct ReadOptions#

Public Functions

Status Validate() const#: Test that all set options are valid.

Public Members

bool use_threads = true#: Whether to use the global CPU thread pool.

int32_t block_size = 1 << 20#

Block size we request from the IO layer.

This will determine multi-threading granularity as well as the size of individual record batches. Minimum valid value for block size is 1

int32_t skip_rows = 0#: Number of header rows to skip (not including the row of column names, if any)

int32_t skip_rows_after_names = 0#: Number of rows to skip after the column names are read, if any.

std::vector<std::string> column_names#

Column names for the target table.

If empty, fall back on autogenerate_column_names.

bool autogenerate_column_names = false#

Whether to autogenerate column names if column_names is empty.

If true, column names will be of the form “f0”, “f1”… If false, column names will be read from the first CSV row after skip_rows.

Public Static Functions

static ReadOptions Defaults()#: Create read options with default values.

class TableReader#

A class that reads an entire CSV file into a Arrow Table.

Public Functions

virtual Result<std::shared_ptr<Table>> Read() = 0#: Read the entire CSV file and convert it to a Arrow Table.

virtual Future<std::shared_ptr<Table>> ReadAsync() = 0#: Read the entire CSV file and convert it to a Arrow Table.

Public Static Functions

static Result<std::shared_ptr<TableReader>> Make(io::IOContext io_context, std::shared_ptr<io::InputStream> input, const ReadOptions&, const ParseOptions&, const ConvertOptions&)#: Create a TableReader instance.

class StreamingReader : public arrow::RecordBatchReader #

A class that reads a CSV file incrementally.

Caveats:

For now, this is always single-threaded (regardless of ReadOptions::use_threads.
Type inference is done on the first block and types are frozen afterwards; to make sure the right data types are inferred, either set ReadOptions::block_size to a large enough value, or use ConvertOptions::column_types to set the desired data types explicitly.

Public Functions

virtual int64_t bytes_read() const = 0#

Return the number of bytes which have been read and processed.

The returned number includes CSV bytes which the StreamingReader has finished processing, but not bytes for which some processing (e.g. CSV parsing or conversion to Arrow layout) is still ongoing.

Furthermore, the following rules apply:

bytes skipped by ReadOptions.skip_rows are counted as being read before any records are returned.
bytes read while parsing the header are counted as being read before any records are returned.
bytes skipped by ReadOptions.skip_rows_after_names are counted after the first batch is returned.

Public Static Functions

static Future<std::shared_ptr<StreamingReader>> MakeAsync(io::IOContext io_context, std::shared_ptr<io::InputStream> input, arrow::internal::Executor *cpu_executor, const ReadOptions&, const ParseOptions&, const ConvertOptions&)#

Create a StreamingReader instance.

This involves some I/O as the first batch must be loaded during the creation process so it is returned as a future

Currently, the StreamingReader is not async-reentrant and does not do any fan-out parsing (see ARROW-11889)

CSV writer#

struct WriteOptions#

Public Functions

Status Validate() const#: Test that all set options are valid.

Public Members

bool include_header = true#: Whether to write an initial header line with column names.

int32_t batch_size = 1024#

Maximum number of rows processed at a time.

The CSV writer converts and writes data in batches of N rows. This number can impact performance.

char delimiter = ','#: Field delimiter.

std::string null_string#: The string to write for null values. Quotes are not allowed in this string.

io::IOContext io_context#: IO context for writing.

std::string eol = "\n"#: The end of line character to use for ending rows.

QuotingStyle quoting_style = QuotingStyle::Needed#: Quoting style.

Public Static Functions

static WriteOptions Defaults()#: Create write options with default values.

Status WriteCSV(const Table &table, const WriteOptions &options, arrow::io::OutputStream *output)#

Convert table to CSV and write the result to output.

Experimental

Status WriteCSV(const RecordBatch &batch, const WriteOptions &options, arrow::io::OutputStream *output)#

Convert batch to CSV and write the result to output.

Experimental

Status WriteCSV(const std::shared_ptr<RecordBatchReader> &reader, const WriteOptions &options, arrow::io::OutputStream *output)#

Convert batches read through a RecordBatchReader to CSV and write the results to output.

Experimental

Result<std::shared_ptr<ipc::RecordBatchWriter>> MakeCSVWriter(std::shared_ptr<io::OutputStream> sink, const std::shared_ptr<Schema> &schema, const WriteOptions &options = WriteOptions::Defaults())#

Create a new CSV writer.

User is responsible for closing the actual OutputStream.

Parameters:

sink – [in] output stream to write to
schema – [in] the schema of the record batches to be written
options – [in] options for serialization

Returns:

Result<std::shared_ptr<RecordBatchWriter>>

Result<std::shared_ptr<ipc::RecordBatchWriter>> MakeCSVWriter(io::OutputStream *sink, const std::shared_ptr<Schema> &schema, const WriteOptions &options = WriteOptions::Defaults())#

Create a new CSV writer.

Parameters:

sink – [in] output stream to write to (does not take ownership)
schema – [in] the schema of the record batches to be written
options – [in] options for serialization

Returns:

Result<std::shared_ptr<RecordBatchWriter>>

Line-separated JSON#

enum class arrow::json::UnexpectedFieldBehavior : char#

Values:

enumerator Ignore#: Unexpected JSON fields are ignored.

enumerator Error#: Unexpected JSON fields error out.

enumerator InferType#: Unexpected JSON fields are type-inferred and included in the output.

struct ReadOptions#

Public Members

bool use_threads = true#: Whether to use the global CPU thread pool.

int32_t block_size = 1 << 20#: Block size we request from the IO layer; also determines the size of chunks when use_threads is true.

Public Static Functions

static ReadOptions Defaults()#: Create read options with default values.

struct ParseOptions#

Public Members

std::shared_ptr<Schema> explicit_schema#: Optional explicit schema (disables type inference on those fields)

bool newlines_in_values = false#

Whether objects may be printed across multiple lines (for example pretty-printed)

If true, parsing may be slower.

UnexpectedFieldBehavior unexpected_field_behavior = UnexpectedFieldBehavior::InferType #: How JSON fields outside of explicit_schema (if given) are treated.

Public Static Functions

static ParseOptions Defaults()#: Create parsing options with default values.

class TableReader#

A class that reads an entire JSON file into a Arrow Table.

The file is expected to consist of individual line-separated JSON objects

Public Functions

virtual Result<std::shared_ptr<Table>> Read() = 0#: Read the entire JSON file and convert it to a Arrow Table.

Public Static Functions

static Result<std::shared_ptr<TableReader>> Make(MemoryPool *pool, std::shared_ptr<io::InputStream> input, const ReadOptions&, const ParseOptions&)#: Create a TableReader instance.

class StreamingReader : public arrow::RecordBatchReader #

A class that reads a JSON file incrementally.

JSON data is read from a stream in fixed-size blocks (configurable with ReadOptions::block_size). Each block is converted to a RecordBatch. Yielded batches have a consistent schema but may differ in row count.

The supplied ParseOptions are used to determine a schema, based either on a provided explicit schema or inferred from the first non-empty block. Afterwards, the target schema is frozen. If UnexpectedFieldBehavior::InferType is specified, unexpected fields will only be inferred for the first block. Afterwards they’ll be treated as errors.

If ReadOptions::use_threads is true, each block’s parsing/decoding task will be parallelized on the given cpu_executor (with readahead corresponding to the executor’s capacity). If an executor isn’t provided, the global thread pool will be used.

If ReadOptions::use_threads is false, computations will be run on the calling thread and cpu_executor will be ignored.

Public Functions

virtual Future<std::shared_ptr<RecordBatch>> ReadNextAsync() = 0#

Read the next RecordBatch asynchronously This function is async-reentrant (but not synchronously reentrant).

However, if threading is disabled, this will block until completion.

virtual int64_t bytes_processed() const = 0#: Get the number of bytes which have been successfully converted to record batches and consumed.

Public Static Functions

static Result<std::shared_ptr<StreamingReader>> Make(std::shared_ptr<io::InputStream> stream, const ReadOptions &read_options, const ParseOptions &parse_options, const io::IOContext &io_context = io::default_io_context(), ::arrow::internal::Executor *cpu_executor = NULLPTR)#

Create a StreamingReader from an InputStream Blocks until the initial batch is loaded.

Parameters:

stream – [in] JSON source stream
read_options – [in] Options for reading
parse_options – [in] Options for chunking, parsing, and conversion
io_context – [in] Context for IO operations (optional)
cpu_executor – [in] Executor for computation tasks (optional)

Returns:

The initialized reader

static Future<std::shared_ptr<StreamingReader>> MakeAsync(std::shared_ptr<io::InputStream> stream, const ReadOptions &read_options, const ParseOptions &parse_options, const io::IOContext &io_context = io::default_io_context(), ::arrow::internal::Executor *cpu_executor = NULLPTR)#

Create a StreamingReader from an InputStream asynchronously Returned future completes after loading the first batch.

Parameters:

stream – [in] JSON source stream
read_options – [in] Options for reading
parse_options – [in] Options for chunking, parsing, and conversion
io_context – [in] Context for IO operations (optional)
cpu_executor – [in] Executor for computation tasks (optional)

Returns:

Future for the initialized reader

Parquet reader#

class ReaderProperties#

Public Functions

inline bool is_buffered_stream_enabled() const#

Buffered stream reading allows the user to control the memory usage of parquet readers.

This ensure that all RandomAccessFile::ReadAt calls are wrapped in a buffered reader that uses a fix sized buffer (of size buffer_size()) instead of the full size of the ReadAt.

The primary reason for this control knobs is for resource control and not performance.

inline void enable_buffered_stream()#: Enable buffered stream reading.

inline void disable_buffered_stream()#: Disable buffered stream reading.

inline int64_t buffer_size() const#: Return the size of the buffered stream buffer.

inline void set_buffer_size(int64_t size)#: Set the size of the buffered stream buffer in bytes.

inline int32_t thrift_string_size_limit() const#

Return the size limit on thrift strings.

This limit helps prevent space and time bombs in files, but may need to be increased in order to read files with especially large headers.

inline void set_thrift_string_size_limit(int32_t size)#: Set the size limit on thrift strings.

inline int32_t thrift_container_size_limit() const#

Return the size limit on thrift containers.

This limit helps prevent space and time bombs in files, but may need to be increased in order to read files with especially large headers.

inline void set_thrift_container_size_limit(int32_t size)#: Set the size limit on thrift containers.

inline void file_decryption_properties(std::shared_ptr<FileDecryptionProperties> decryption)#: Set the decryption properties.

inline const std::shared_ptr<FileDecryptionProperties> &file_decryption_properties() const#: Return the decryption properties.

class ArrowReaderProperties#

EXPERIMENTAL: Properties for configuring FileReader behavior.

Public Functions

inline void set_use_threads(bool use_threads)#

Set whether to use the IO thread pool to parse columns in parallel.

Default is false.

inline bool use_threads() const#: Return whether will use multiple threads.

inline void set_read_dictionary(int column_index, bool read_dict)#

Set whether to read a particular column as dictionary encoded.

If the file metadata contains a serialized Arrow schema, then …

This is only supported for columns with a Parquet physical type of BYTE_ARRAY, such as string or binary types.

inline bool read_dictionary(int column_index) const#: Return whether the column at the index will be read as dictionary.

inline void set_batch_size(int64_t batch_size)#

Set the maximum number of rows to read into a record batch.

Will only be fewer rows when there are no more rows in the file. Note that some APIs such as ReadTable may ignore this setting.

inline int64_t batch_size() const#

Return the batch size in rows.

Note that some APIs such as ReadTable may ignore this setting.

inline void set_pre_buffer(bool pre_buffer)#

Enable read coalescing (default false).

When enabled, the Arrow reader will pre-buffer necessary regions of the file in-memory. This is intended to improve performance on high-latency filesystems (e.g. Amazon S3).

inline bool pre_buffer() const#: Return whether read coalescing is enabled.

inline void set_cache_options(::arrow::io::CacheOptions options)#

Set options for read coalescing.

This can be used to tune the implementation for characteristics of different filesystems.

inline const ::arrow::io::CacheOptions &cache_options() const#: Return the options for read coalescing.

inline void set_io_context(const ::arrow::io::IOContext &ctx)#: Set execution context for read coalescing.

inline const ::arrow::io::IOContext &io_context() const#: Return the execution context used for read coalescing.

inline void set_coerce_int96_timestamp_unit(::arrow::TimeUnit::type unit)#: Set timestamp unit to use for deprecated INT96-encoded timestamps (default is NANO).

inline void set_arrow_extensions_enabled(bool extensions_enabled)#

Enable Parquet-supported Arrow extension types.

When enabled, Parquet logical types will be mapped to their corresponding Arrow extension types at read time, if such exist. Currently only arrow::extension::json() extension type is supported. Columns whose LogicalType is JSON will be interpreted as arrow::extension::json(), with storage type inferred from the serialized Arrow schema if present, or utf8 by default.

class ParquetFileReader#

Public Functions

std::shared_ptr<PageIndexReader> GetPageIndexReader()#

Returns the PageIndexReader.

Only one instance is ever created.

If the file does not have the page index, nullptr may be returned. Because it pays to check existence of page index in the file, it is possible to return a non null value even if page index does not exist. It is the caller’s responsibility to check the return value and follow-up calls to PageIndexReader.

WARNING: The returned PageIndexReader must not outlive the ParquetFileReader. Initialize GetPageIndexReader() is not thread-safety.

BloomFilterReader &GetBloomFilterReader()#

Returns the BloomFilterReader.

Only one instance is ever created.

WARNING: The returned BloomFilterReader must not outlive the ParquetFileReader. Initialize GetBloomFilterReader() is not thread-safety.

void PreBuffer(const std::vector<int> &row_groups, const std::vector<int> &column_indices, const ::arrow::io::IOContext &ctx, const ::arrow::io::CacheOptions &options)#

Pre-buffer the specified column indices in all row groups.

Readers can optionally call this to cache the necessary slices of the file in-memory before deserialization. Arrow readers can automatically do this via an option. This is intended to increase performance when reading from high-latency filesystems (e.g. Amazon S3).

After calling this, creating readers for row groups/column indices that were not buffered may fail. Creating multiple readers for the a subset of the buffered regions is acceptable. This may be called again to buffer a different set of row groups/columns.

If memory usage is a concern, note that data will remain buffered in memory until either PreBuffer() is called again, or the reader itself is destructed. Reading - and buffering - only one row group at a time may be useful.

This method may throw.

::arrow::Future WhenBuffered(const std::vector<int> &row_groups, const std::vector<int> &column_indices) const#

Wait for the specified row groups and column indices to be pre-buffered.

After the returned Future completes, reading the specified row groups/columns will not block.

PreBuffer must be called first. This method does not throw.

struct Contents#

class FileReader#

Arrow read adapter class for deserializing Parquet files as Arrow row batches.

This interfaces caters for different use cases and thus provides different interfaces. In its most simplistic form, we cater for a user that wants to read the whole Parquet at once with the FileReader::ReadTable method.

More advanced users that also want to implement parallelism on top of each single Parquet files should do this on the RowGroup level. For this, they can call FileReader::RowGroup(i)->ReadTable to receive only the specified RowGroup as a table.

In the most advanced situation, where a consumer wants to independently read RowGroups in parallel and consume each column individually, they can call FileReader::RowGroup(i)->Column(j)->Read and receive an arrow::Column instance.

Finally, one can also get a stream of record batches using FileReader::GetRecordBatchReader(). This can internally decode columns in parallel if use_threads was enabled in the ArrowReaderProperties.

The parquet format supports an optional integer field_id which can be assigned to a field. Arrow will convert these field IDs to a metadata key named PARQUET:field_id on the appropriate field.

Public Functions

virtual ::arrow::Status GetSchema(std::shared_ptr<::arrow::Schema> *out) = 0#: Return arrow schema for all the columns.

virtual ::arrow::Status ReadColumn(int i, std::shared_ptr<::arrow::ChunkedArray> *out) = 0#

Read column as a whole into a chunked array.

The index i refers the index of the top level schema field, which may be nested or flat - e.g.

0 foo.bar foo.bar.baz foo.qux 1 foo2 2 foo3

i=0 will read the entire foo struct, i=1 the foo2 primitive column etc

virtual ::arrow::Status GetRecordBatchReader(std::unique_ptr<::arrow::RecordBatchReader> *out) = 0#: Return a RecordBatchReader of all row groups and columns.

virtual ::arrow::Status GetRecordBatchReader(const std::vector<int> &row_group_indices, std::unique_ptr<::arrow::RecordBatchReader> *out) = 0#

Return a RecordBatchReader of row groups selected from row_group_indices.

Note that the ordering in row_group_indices matters. FileReaders must outlive their RecordBatchReaders.

Returns:: error Status if row_group_indices contains an invalid index

virtual ::arrow::Status GetRecordBatchReader(const std::vector<int> &row_group_indices, const std::vector<int> &column_indices, std::unique_ptr<::arrow::RecordBatchReader> *out) = 0#

Return a RecordBatchReader of row groups selected from row_group_indices, whose columns are selected by column_indices.

Note that the ordering in row_group_indices and column_indices matter. FileReaders must outlive their RecordBatchReaders.

Returns:: error Status if either row_group_indices or column_indices contains an invalid index

::arrow::Status GetRecordBatchReader(const std::vector<int> &row_group_indices, const std::vector<int> &column_indices, std::shared_ptr<::arrow::RecordBatchReader> *out)#

Return a RecordBatchReader of row groups selected from row_group_indices, whose columns are selected by column_indices.

Note that the ordering in row_group_indices and column_indices matter. FileReaders must outlive their RecordBatchReaders.

Parameters:

row_group_indices – which row groups to read (order determines read order).
column_indices – which columns to read (order determines output schema).
out – [out] record batch stream from parquet data.

Returns:

error Status if either row_group_indices or column_indices contains an invalid index

virtual ::arrow::Result<std::function<::arrow::Future<std::shared_ptr<::arrow::RecordBatch>>()>> GetRecordBatchGenerator(std::shared_ptr<FileReader> reader, const std::vector<int> row_group_indices, const std::vector<int> column_indices, ::arrow::internal::Executor *cpu_executor = NULLPTR, int64_t rows_to_readahead = 0) = 0#

Return a generator of record batches.

The FileReader must outlive the generator, so this requires that you pass in a shared_ptr.

Returns:: error Result if either row_group_indices or column_indices contains an invalid index

virtual ::arrow::Status ReadTable(std::shared_ptr<::arrow::Table> *out) = 0#: Read all columns into a Table.

virtual ::arrow::Status ReadTable(const std::vector<int> &column_indices, std::shared_ptr<::arrow::Table> *out) = 0#

Read the given columns into a Table.

The indicated column indices are relative to the internal representation of the parquet table. For instance : 0 foo.bar foo.bar.baz 0 foo.bar.baz2 1 foo.qux 2 1 foo2 3 2 foo3 4

i=0 will read foo.bar.baz, i=1 will read only foo.bar.baz2 and so on. Only leaf fields have indices; foo itself doesn’t have an index. To get the index for a particular leaf field, one can use manifest().schema_fields to get the top level fields, and then walk the tree to identify the relevant leaf fields and access its column_index. To get the total number of leaf fields, use FileMetadata.num_columns().

virtual ::arrow::Status ScanContents(std::vector<int> columns, const int32_t column_batch_size, int64_t *num_rows) = 0#: Scan file contents with one thread, return number of rows.

virtual std::shared_ptr<RowGroupReader> RowGroup(int row_group_index) = 0#: Return a reader for the RowGroup, this object must not outlive the FileReader.

virtual int num_row_groups() const = 0#: The number of row groups in the file.

virtual void set_use_threads(bool use_threads) = 0#

Set whether to use multiple threads during reads of multiple columns.

By default only one thread is used.

virtual void set_batch_size(int64_t batch_size) = 0#: Set number of records to read per batch for the RecordBatchReader.

Public Static Functions

static ::arrow::Status Make(::arrow::MemoryPool *pool, std::unique_ptr<ParquetFileReader> reader, const ArrowReaderProperties &properties, std::unique_ptr<FileReader> *out)#: Factory function to create a FileReader from a ParquetFileReader and properties.

static ::arrow::Status Make(::arrow::MemoryPool *pool, std::unique_ptr<ParquetFileReader> reader, std::unique_ptr<FileReader> *out)#: Factory function to create a FileReader from a ParquetFileReader.

class FileReaderBuilder#

Experimental helper class for bindings (like Python) that struggle either with std::move or C++ exceptions.

Public Functions

::arrow::Status Open(std::shared_ptr<::arrow::io::RandomAccessFile> file, const ReaderProperties &properties = default_reader_properties(), std::shared_ptr<FileMetaData> metadata = NULLPTR)#: Create FileReaderBuilder from Arrow file and optional properties / metadata.

::arrow::Status OpenFile(const std::string &path, bool memory_map = false, const ReaderProperties &props = default_reader_properties(), std::shared_ptr<FileMetaData> metadata = NULLPTR)#: Create FileReaderBuilder from file path and optional properties / metadata.

FileReaderBuilder *memory_pool(::arrow::MemoryPool *pool)#: Set Arrow MemoryPool for memory allocation.

FileReaderBuilder *properties(const ArrowReaderProperties &arg_properties)#: Set Arrow reader properties.

::arrow::Status Build(std::unique_ptr<FileReader> *out)#: Build FileReader instance.

::arrow::Status OpenFile(std::shared_ptr<::arrow::io::RandomAccessFile>, ::arrow::MemoryPool *allocator, std::unique_ptr<FileReader> *reader)#

Build FileReader from Arrow file and MemoryPool.

Advanced settings are supported through the FileReaderBuilder class.

class StreamReader#

A class for reading Parquet files using an output stream type API.

The values given must be of the correct type i.e. the type must match the file schema exactly otherwise a ParquetException will be thrown.

The user must explicitly advance to the next row using the EndRow() function or EndRow input manipulator.

Required and optional fields are supported:

Required fields are read using operator>>(T)
Optional fields are read with operator>>(std::optional<T>)

Note that operator>>(std::optional<T>) can be used to read required fields.

Similarly operator>>(T) can be used to read optional fields. However, if the value is not present then a ParquetException will be raised.

Currently there is no support for repeated fields.

Public Functions

void EndRow()#

Terminate current row and advance to next one.

Throws:: ParquetException – if all columns in the row were not read or skipped.

int64_t SkipColumns(int64_t num_columns_to_skip)#

Skip the data in the next columns.

If the number of columns exceeds the columns remaining on the current row then skipping is terminated - it does not continue skipping columns on the next row. Skipping of columns still requires the use ‘EndRow’ even if all remaining columns were skipped.

Returns:: Number of columns actually skipped.

int64_t SkipRows(int64_t num_rows_to_skip)#

Skip the data in the next rows.

Skipping of rows is not allowed if reading of data for the current row is not finished. Skipping of rows will be terminated if the end of file is reached.

Returns:: Number of rows actually skipped.

Parquet writer#

class WriterProperties#

class Builder#

Public Functions

inline Builder *memory_pool(MemoryPool *pool)#: Specify the memory pool for the writer. Default default_memory_pool.

inline Builder *enable_dictionary()#

Enable dictionary encoding in general for all columns.

Default enabled.

inline Builder *disable_dictionary()#

Disable dictionary encoding in general for all columns.

Default enabled.

inline Builder *enable_dictionary(const std::string &path)#

Enable dictionary encoding for column specified by path.

Default enabled.

inline Builder *enable_dictionary(const std::shared_ptr<schema::ColumnPath> &path)#

Enable dictionary encoding for column specified by path.

Default enabled.

inline Builder *disable_dictionary(const std::string &path)#

Disable dictionary encoding for column specified by path.

Default enabled.

inline Builder *disable_dictionary(const std::shared_ptr<schema::ColumnPath> &path)#

Disable dictionary encoding for column specified by path.

Default enabled.

inline Builder *dictionary_pagesize_limit(int64_t dictionary_psize_limit)#: Specify the dictionary page size limit per row group. Default 1MB.

inline Builder *write_batch_size(int64_t write_batch_size)#

Specify the write batch size while writing batches of Arrow values into Parquet.

Default 1024.

inline Builder *max_row_group_length(int64_t max_row_group_length)#

Specify the max number of rows to put in a single row group.

Default 1Mi rows.

inline Builder *data_pagesize(int64_t pg_size)#

Specify the data page size.

Default 1MB.

inline Builder *data_page_version(ParquetDataPageVersion data_page_version)#

Specify the data page version.

Default V1.

inline Builder *version(ParquetVersion::type version)#

Specify the Parquet file version.

Default PARQUET_2_6.

inline Builder *encoding(Encoding::type encoding_type)#

Define the encoding that is used when we don’t utilise dictionary encoding.

This either apply if dictionary encoding is disabled or if we fallback as the dictionary grew too large.

inline Builder *encoding(const std::string &path, Encoding::type encoding_type)#

Define the encoding that is used when we don’t utilise dictionary encoding.

This either apply if dictionary encoding is disabled or if we fallback as the dictionary grew too large.

inline Builder *encoding(const std::shared_ptr<schema::ColumnPath> &path, Encoding::type encoding_type)#

Define the encoding that is used when we don’t utilise dictionary encoding.

This either apply if dictionary encoding is disabled or if we fallback as the dictionary grew too large.

inline Builder *compression(Compression::type codec)#

Specify compression codec in general for all columns.

Default UNCOMPRESSED.

inline Builder *max_statistics_size(size_t max_stats_sz)#

Specify max statistics size to store min max value.

Default 4KB.

inline Builder *compression(const std::string &path, Compression::type codec)#

Specify compression codec for the column specified by path.

Default UNCOMPRESSED.

inline Builder *compression(const std::shared_ptr<schema::ColumnPath> &path, Compression::type codec)#

Specify compression codec for the column specified by path.

Default UNCOMPRESSED.

inline Builder *compression_level(int compression_level)#

Specify the default compression level for the compressor in every column.

In case a column does not have an explicitly specified compression level, the default one would be used.

The provided compression level is compressor specific. The user would have to familiarize oneself with the available levels for the selected compressor. If the compressor does not allow for selecting different compression levels, calling this function would not have any effect. Parquet and Arrow do not validate the passed compression level. If no level is selected by the user or if the special std::numeric_limits<int>::min() value is passed, then Arrow selects the compression level.

If other compressor-specific options need to be set in addition to the compression level, use the codec_options method.

inline Builder *compression_level(const std::string &path, int compression_level)#

Specify a compression level for the compressor for the column described by path.

The provided compression level is compressor specific. The user would have to familiarize oneself with the available levels for the selected compressor. If the compressor does not allow for selecting different compression levels, calling this function would not have any effect. Parquet and Arrow do not validate the passed compression level. If no level is selected by the user or if the special std::numeric_limits<int>::min() value is passed, then Arrow selects the compression level.

inline Builder *compression_level(const std::shared_ptr<schema::ColumnPath> &path, int compression_level)#

Specify a compression level for the compressor for the column described by path.

The provided compression level is compressor specific. The user would have to familiarize oneself with the available levels for the selected compressor. If the compressor does not allow for selecting different compression levels, calling this function would not have any effect. Parquet and Arrow do not validate the passed compression level. If no level is selected by the user or if the special std::numeric_limits<int>::min() value is passed, then Arrow selects the compression level.

inline Builder *codec_options(const std::shared_ptr<::arrow::util::CodecOptions> &codec_options)#

Specify the default codec options for the compressor in every column.

The codec options allow configuring the compression level as well as other codec-specific options.

inline Builder *codec_options(const std::string &path, const std::shared_ptr<::arrow::util::CodecOptions> &codec_options)#: Specify the codec options for the compressor for the column described by path.

inline Builder *codec_options(const std::shared_ptr<schema::ColumnPath> &path, const std::shared_ptr<::arrow::util::CodecOptions> &codec_options)#: Specify the codec options for the compressor for the column described by path.

inline Builder *encryption(std::shared_ptr<FileEncryptionProperties> file_encryption_properties)#

Define the file encryption properties.

Default NULL.

inline Builder *enable_statistics()#

Enable statistics in general.

Default enabled.

inline Builder *disable_statistics()#

Disable statistics in general.

Default enabled.

inline Builder *enable_statistics(const std::string &path)#

Enable statistics for the column specified by path.

Default enabled.

inline Builder *enable_statistics(const std::shared_ptr<schema::ColumnPath> &path)#

Enable statistics for the column specified by path.

Default enabled.

inline Builder *set_sorting_columns(std::vector<SortingColumn> sorting_columns)#

Define the sorting columns.

Default empty.

If sorting columns are set, user should ensure that records are sorted by sorting columns. Otherwise, the storing data will be inconsistent with sorting_columns metadata.

inline Builder *disable_statistics(const std::string &path)#

Disable statistics for the column specified by path.

Default enabled.

inline Builder *disable_statistics(const std::shared_ptr<schema::ColumnPath> &path)#

Disable statistics for the column specified by path.

Default enabled.

inline Builder *enable_store_decimal_as_integer()#

Allow decimals with 1 <= precision <= 18 to be stored as integers.

In Parquet, DECIMAL can be stored in any of the following physical types:

int32: for 1 <= precision <= 9.
int64: for 10 <= precision <= 18.
fixed_len_byte_array: precision is limited by the array size. Length n can store <= floor(log_10(2^(8*n - 1) - 1)) base-10 digits.
binary: precision is unlimited. The minimum number of bytes to store the unscaled value is used.

By default, this is DISABLED and all decimal types annotate fixed_len_byte_array.

When enabled, the C++ writer will use following physical types to store decimals:

int32: for 1 <= precision <= 9.
int64: for 10 <= precision <= 18.
fixed_len_byte_array: for precision > 18.

As a consequence, decimal columns stored in integer types are more compact.

inline Builder *disable_store_decimal_as_integer()#

Disable decimal logical type with 1 <= precision <= 18 to be stored as integer physical type.

Default disabled.

inline Builder *enable_write_page_index()#

Enable writing page index in general for all columns.

Default disabled.

Writing statistics to the page index disables the old method of writing statistics to each data page header. The page index makes filtering more efficient than the page header, as it gathers all the statistics for a Parquet file in a single place, avoiding scattered I/O.

Please check the link below for more details: apache/parquet-format

inline Builder *disable_write_page_index()#: Disable writing page index in general for all columns. Default disabled.

inline Builder *enable_write_page_index(const std::string &path)#: Enable writing page index for column specified by path. Default disabled.

inline Builder *enable_write_page_index(const std::shared_ptr<schema::ColumnPath> &path)#: Enable writing page index for column specified by path. Default disabled.

inline Builder *disable_write_page_index(const std::string &path)#: Disable writing page index for column specified by path. Default disabled.

inline Builder *disable_write_page_index(const std::shared_ptr<schema::ColumnPath> &path)#: Disable writing page index for column specified by path. Default disabled.

inline std::shared_ptr<WriterProperties> build()#

Build the WriterProperties with the builder parameters.

Returns:: The WriterProperties defined by the builder.

class ArrowWriterProperties#

Public Functions

inline bool compliant_nested_types() const#

Enable nested type naming according to the parquet specification.

Older versions of arrow wrote out field names for nested lists based on the name of the field. According to the parquet specification they should always be “element”.

inline EngineVersion engine_version() const#

The underlying engine version to use when writing Arrow data.

V2 is currently the latest V1 is considered deprecated but left in place in case there are bugs detected in V2.

inline bool use_threads() const#: Returns whether the writer will use multiple threads to write columns in parallel in the buffered row group mode.

::arrow::internal::Executor *executor() const#: Returns the executor used to write columns in parallel.

class Builder#

Public Functions

inline Builder *disable_deprecated_int96_timestamps()#: Disable writing legacy int96 timestamps (default disabled).

inline Builder *enable_deprecated_int96_timestamps()#

Enable writing legacy int96 timestamps (default disabled).

May be turned on to write timestamps compatible with older Parquet writers. This takes precedent over coerce_timestamps.

inline Builder *coerce_timestamps(::arrow::TimeUnit::type unit)#

Coerce all timestamps to the specified time unit.

Parameters:: unit – time unit to truncate to. For Parquet versions 1.0 and 2.4, nanoseconds are casted to microseconds.

inline Builder *allow_truncated_timestamps()#

Allow loss of data when truncating timestamps.

This is disallowed by default and an error will be returned.

inline Builder *disallow_truncated_timestamps()#: Disallow loss of data when truncating timestamps (default).

inline Builder *store_schema()#: EXPERIMENTAL: Write binary serialized Arrow schema to the file, to enable certain read options (like “read_dictionary”) to be set automatically.

inline Builder *enable_compliant_nested_types()#

When enabled, will not preserve Arrow field names for list types.

Instead of using the field names Arrow uses for the values array of list types (default “item”), will use “element”, as is specified in the Parquet spec.

This is enabled by default.

inline Builder *disable_compliant_nested_types()#: Preserve Arrow list field name.

inline Builder *set_engine_version(EngineVersion version)#: Set the version of the Parquet writer engine.

inline Builder *set_use_threads(bool use_threads)#

Set whether to use multiple threads to write columns in parallel in the buffered row group mode.

WARNING: If writing multiple files in parallel in the same executor, deadlock may occur if use_threads is true. Please disable it in this case.

Default is false.

inline Builder *set_executor(::arrow::internal::Executor *executor)#

Set the executor to write columns in parallel in the buffered row group mode.

Default is nullptr and the default cpu executor will be used.

inline std::shared_ptr<ArrowWriterProperties> build()#: Create the final properties.

class FileWriter#

Iterative FileWriter class.

For basic usage, can write a Table at a time, creating one or more row groups per write call.

For advanced usage, can write column-by-column: Start a new RowGroup or Chunk with NewRowGroup, then write column-by-column the whole column chunk.

If PARQUET:field_id is present as a metadata key on a field, and the corresponding value is a nonnegative integer, then it will be used as the field_id in the parquet file.

Public Functions

::arrow::Result<std::unique_ptr<FileWriter>> Open(const ::arrow::Schema &schema, MemoryPool *pool, std::shared_ptr<::arrow::io::OutputStream> sink, std::shared_ptr<WriterProperties> properties = default_writer_properties(), std::shared_ptr<ArrowWriterProperties> arrow_properties = default_arrow_writer_properties())#

Try to create an Arrow to Parquet file writer.

Since: 11.0.0

Parameters:

schema – schema of data that will be passed.
pool – memory pool to use.
sink – output stream to write Parquet data.
properties – general Parquet writer properties.
arrow_properties – Arrow-specific writer properties.

virtual std::shared_ptr<::arrow::Schema> schema() const = 0#: Return the Arrow schema to be written to.

virtual ::arrow::Status WriteTable(const ::arrow::Table &table, int64_t chunk_size = DEFAULT_MAX_ROW_GROUP_LENGTH) = 0#

Write a Table to Parquet.

Parameters:

table – Arrow table to write.
chunk_size – maximum number of rows to write per row group.

virtual ::arrow::Status NewRowGroup(int64_t chunk_size) = 0#

Start a new row group.

Returns an error if not all columns have been written.

Parameters:: chunk_size – the number of rows in the next row group.

virtual ::arrow::Status WriteColumnChunk(const ::arrow::Array &data) = 0#: Write ColumnChunk in row group using an array.

virtual ::arrow::Status WriteColumnChunk(const std::shared_ptr<::arrow::ChunkedArray> &data, int64_t offset, int64_t size) = 0#: Write ColumnChunk in row group using slice of a ChunkedArray.

virtual ::arrow::Status WriteColumnChunk(const std::shared_ptr<::arrow::ChunkedArray> &data) = 0#: Write ColumnChunk in a row group using a ChunkedArray.

virtual ::arrow::Status NewBufferedRowGroup() = 0#

Start a new buffered row group.

Returns an error if not all columns have been written.

virtual ::arrow::Status WriteRecordBatch(const ::arrow::RecordBatch &batch) = 0#

Write a RecordBatch into the buffered row group.

Multiple RecordBatches can be written into the same row group through this method.

WriterProperties.max_row_group_length() is respected and a new row group will be created if the current row group exceeds the limit.

Batches get flushed to the output stream once NewBufferedRowGroup() or Close() is called.

WARNING: If you are writing multiple files in parallel in the same executor, deadlock may occur if ArrowWriterProperties::use_threads is set to true to write columns in parallel. Please disable use_threads option in this case.

virtual ::arrow::Status Close() = 0#: Write the footer and close the file.

virtual ::arrow::Status AddKeyValueMetadata(const std::shared_ptr<const ::arrow::KeyValueMetadata> &key_value_metadata) = 0#

Add key-value metadata to the file.

WARNING: If store_schema is enabled, ARROW:schema would be stored in the key-value metadata. Overwriting this key would result in store_schema being unusable during read.

Note

This will overwrite any existing metadata with the same key.

Parameters:: key_value_metadata – [in] the metadata to add.
Returns:: Error if Close() has been called.

virtual const std::shared_ptr<FileMetaData> metadata() const = 0#: Return the file metadata, only available after calling Close().

::arrow::Status parquet::arrow::WriteTable(const ::arrow::Table &table, MemoryPool *pool, std::shared_ptr<::arrow::io::OutputStream> sink, int64_t chunk_size = DEFAULT_MAX_ROW_GROUP_LENGTH, std::shared_ptr<WriterProperties> properties = default_writer_properties(), std::shared_ptr<ArrowWriterProperties> arrow_properties = default_arrow_writer_properties())#

Write a Table to Parquet.

This writes one table in a single shot. To write a Parquet file with multiple tables iteratively, see parquet::arrow::FileWriter.

Parameters:

table – Table to write.
pool – memory pool to use.
sink – output stream to write Parquet data.
chunk_size – maximum number of rows to write per row group.
properties – general Parquet writer properties.
arrow_properties – Arrow-specific writer properties.

class StreamWriter#

A class for writing Parquet files using an output stream type API.

The values given must be of the correct type i.e. the type must match the file schema exactly otherwise a ParquetException will be thrown.

The user must explicitly indicate the end of the row using the EndRow() function or EndRow output manipulator.

A maximum row group size can be configured, the default size is 512MB. Alternatively the row group size can be set to zero and the user can create new row groups by calling the EndRowGroup() function or using the EndRowGroup output manipulator.

Required and optional fields are supported:

Required fields are written using operator<<(T)
Optional fields are written using operator<<(std::optional<T>).

Note that operator<<(T) can be used to write optional fields.

Similarly, operator<<(std::optional<T>) can be used to write required fields. However if the optional parameter does not have a value (i.e. it is nullopt) then a ParquetException will be raised.

Currently there is no support for repeated fields.

Public Functions

StreamWriter &operator<<(bool v)#

Output operators for required fields.

These can also be used for optional fields when a value must be set.

template<int N> inline StreamWriter &operator<<(const char (&v)[N])#: Output operators for fixed length strings.

StreamWriter &operator<<(const char *v)#: Output operators for variable length strings.

template<typename T> inline StreamWriter &operator<<(const optional<T> &v)#: Output operator for optional fields.

int64_t SkipColumns(int num_columns_to_skip)#

Skip the next N columns of optional data.

If there are less than N columns remaining then the excess columns are ignored.

Throws:: ParquetException – if there is an attempt to skip any required column.
Returns:: Number of columns actually skipped.

void EndRow()#

Terminate the current row and advance to next one.

Throws:: ParquetException – if all columns in the row were not written or skipped.

void EndRowGroup()#: Terminate the current row group and create new one.

struct FixedStringView#

Helper class to write fixed length strings.

This is useful as the standard string view (such as std::string_view) is for variable length data.

ORC#

class ORCFileReader#

Read an Arrow Table or RecordBatch from an ORC file.

Public Functions

Result<std::shared_ptr<Schema>> ReadSchema()#

Return the schema read from the ORC file.

Returns:: the returned Schema object

Result<std::shared_ptr<Table>> Read()#

Read the file as a Table.