Package google.cloud.bigquery.storage.v1

Stay organized with collections Save and categorize content based on your preferences.

Index

BigQueryRead

BigQuery Read API.

The Read API can be used to read data from BigQuery.

CreateReadSession

rpc CreateReadSession(CreateReadSessionRequest) returns (ReadSession)

Creates a new read session. A read session divides the contents of a BigQuery table into one or more streams, which can then be used to read data from the table. The read session also specifies properties of the data to be read, such as a list of columns or a push-down filter describing the rows to be returned.

A particular row can be read by at most one stream. When the caller has reached the end of each stream in the session, then all the data in the table has been read.

Data is assigned to each stream such that roughly the same number of rows can be read from each stream. Because the server-side unit for assigning data is collections of rows, the API does not guarantee that each stream will return the same number or rows. Additionally, the limits are enforced based on the number of pre-filtered rows, so some filters can lead to lopsided assignments.

Read sessions automatically expire 6 hours after they are created and do not require manual clean-up by the caller.

ReadRows

rpc ReadRows(ReadRowsRequest) returns (ReadRowsResponse)

Reads rows from the stream in the format prescribed by the ReadSession. Each response contains one or more table rows, up to a maximum of 100 MiB per response; read requests which attempt to read individual rows larger than 100 MiB will fail.

Each request also returns a set of stream statistics reflecting the current state of the stream.

SplitReadStream

rpc SplitReadStream(SplitReadStreamRequest) returns (SplitReadStreamResponse)

Splits a given ReadStream into two ReadStream objects. These ReadStream objects are referred to as the primary and the residual streams of the split. The original ReadStream can still be read from in the same manner as before. Both of the returned ReadStream objects can also be read from, and the rows returned by both child streams will be the same as the rows read from the original stream.

Moreover, the two child streams will be allocated back-to-back in the original ReadStream. Concretely, it is guaranteed that for streams original, primary, and residual, that original[0-j] = primary[0-j] and original[j-n] = residual[0-m] once the streams have been read to completion.

BigQueryWrite

BigQuery Write API.

The Write API can be used to write data to BigQuery.

For supplementary information about the Write API, see: https://cloud.google.com/bigquery/docs/write-api

AppendRows

rpc AppendRows(AppendRowsRequest) returns (AppendRowsResponse)

Appends data to the given stream.

If offset is specified, the offset is checked against the end of stream. The server returns OUT_OF_RANGE in AppendRowsResponse if an attempt is made to append to an offset beyond the current end of the stream or ALREADY_EXISTS if user provides an offset that has already been written to. User can retry with adjusted offset within the same RPC connection. If offset is not specified, append happens at the end of the stream.

The response contains an optional offset at which the append happened. No offset information will be returned for appends to a default stream.

Responses are received in the same order in which requests are sent. There will be one response for each successful inserted request. Responses may optionally embed error information if the originating AppendRequest was not successfully processed.

The specifics of when successfully appended data is made visible to the table are governed by the type of stream:

  • For COMMITTED streams (which includes the default stream), data is visible immediately upon successful append.

  • For BUFFERED streams, data is made visible via a subsequent FlushRows rpc which advances a cursor to a newer offset in the stream.

  • For PENDING streams, data is not made visible until the stream itself is finalized (via the FinalizeWriteStream rpc), and the stream is explicitly committed via the BatchCommitWriteStreams rpc.

BatchCommitWriteStreams

rpc BatchCommitWriteStreams(BatchCommitWriteStreamsRequest) returns (BatchCommitWriteStreamsResponse)

Atomically commits a group of PENDING streams that belong to the same parent table.

Streams must be finalized before commit and cannot be committed multiple times. Once a stream is committed, data in the stream becomes available for read operations.

CreateWriteStream

rpc CreateWriteStream(CreateWriteStreamRequest) returns (WriteStream)

Creates a write stream to the given table. Additionally, every table has a special stream named '_default' to which data can be written. This stream doesn't need to be created using CreateWriteStream. It is a stream that can be used simultaneously by any number of clients. Data written to this stream is considered committed as soon as an acknowledgement is received.

FinalizeWriteStream

rpc FinalizeWriteStream(FinalizeWriteStreamRequest) returns (FinalizeWriteStreamResponse)

Finalize a write stream so that no new data can be appended to the stream. Finalize is not supported on the '_default' stream.

FlushRows

rpc FlushRows(FlushRowsRequest) returns (FlushRowsResponse)

Flushes rows to a BUFFERED stream.

If users are appending rows to BUFFERED stream, flush operation is required in order for the rows to become available for reading. A Flush operation flushes up to any previously flushed offset in a BUFFERED stream, to the offset specified in the request.

Flush is not supported on the _default stream, since it is not BUFFERED.

GetWriteStream

rpc GetWriteStream(GetWriteStreamRequest) returns (WriteStream)

Gets information about a write stream.

AppendRowsRequest

Request message for AppendRows.

Due to the nature of AppendRows being a bidirectional streaming RPC, certain parts of the AppendRowsRequest need only be specified for the first request sent each time the gRPC network connection is opened/reopened.

The size of a single AppendRowsRequest must be less than 10 MB in size. Requests larger than this return an error, typically INVALID_ARGUMENT.

Fields
write_stream

string

Required. The write_stream identifies the target of the append operation, and only needs to be specified as part of the first request on the gRPC connection. If provided for subsequent requests, it must match the value of the first request.

For explicitly created write streams, the format is:

  • projects/{project}/datasets/{dataset}/tables/{table}/streams/{id}

For the special default stream, the format is:

  • projects/{project}/datasets/{dataset}/tables/{table}/streams/_default.

Authorization requires the following IAM permission on the specified resource writeStream:

  • bigquery.tables.updateData
offset

Int64Value

If present, the write is only performed if the next append offset is same as the provided value. If not present, the write is performed at the current end of stream. Specifying a value for this field is not allowed when calling AppendRows for the '_default' stream.

trace_id

string

Id set by client to annotate its identity. Only initial request setting is respected.

Union field rows. Input rows. The writer_schema field must be specified at the initial request and currently, it will be ignored if specified in following requests. Following requests must have data in the same format as the initial request. rows can be only one of the following:
proto_rows

ProtoData

Rows in proto format.

ProtoData

ProtoData contains the data rows and schema when constructing append requests.

Fields
writer_schema

ProtoSchema

Proto schema used to serialize the data. This value only needs to be provided as part of the first request on a gRPC network connection, and will be ignored for subsequent requests on the connection.

rows

ProtoRows

Serialized row data in protobuf message format. Currently, the backend expects the serialized rows to adhere to proto2 semantics when appending rows, particularly with respect to how default values are encoded.

AppendRowsResponse

Response message for AppendRows.

Fields
updated_schema

TableSchema

If backend detects a schema update, pass it to user so that user can use it to input new type of message. It will be empty when no schema updates have occurred.

row_errors[]

RowError

If a request failed due to corrupted rows, no rows in the batch will be appended. The API will return row level error info, so that the caller can remove the bad rows and retry the request.

write_stream

string

The target of the append operation. Matches the write_stream in the corresponding request.

Union field response.

response can be only one of the following:

append_result

AppendResult

Result if the append is successful.

error

Status

Error returned when problems were encountered. If present, it indicates rows were not accepted into the system. Users can retry or continue with other append requests within the same connection.

Additional information about error signalling:

ALREADY_EXISTS: Happens when an append specified an offset, and the backend already has received data at this offset. Typically encountered in retry scenarios, and can be ignored.

OUT_OF_RANGE: Returned when the specified offset in the stream is beyond the current end of the stream.

INVALID_ARGUMENT: Indicates a malformed request or data.

ABORTED: Request processing is aborted because of prior failures. The request can be retried if previous failure is addressed.

INTERNAL: Indicates server side error(s) that can be retried.

AppendResult

AppendResult is returned for successful append requests.

Fields
offset

Int64Value

The row offset at which the last append occurred. The offset will not be set if appending using default streams.

ArrowRecordBatch

Arrow RecordBatch.

Fields
serialized_record_batch

bytes

IPC-serialized Arrow RecordBatch.

row_count
(deprecated)

int64

[Deprecated] The count of rows in serialized_record_batch. Please use the format-independent ReadRowsResponse.row_count instead.

ArrowSchema

Arrow schema as specified in https://arrow.apache.org/docs/python/api/datatypes.html and serialized to bytes using IPC: https://arrow.apache.org/docs/format/Columnar.html#serialization-and-interprocess-communication-ipc

See code samples on how this message can be deserialized.

Fields
serialized_schema

bytes

IPC serialized Arrow schema.

ArrowSerializationOptions

Contains options specific to Arrow Serialization.

Fields
buffer_compression

CompressionCodec

The compression codec to use for Arrow buffers in serialized record batches.

CompressionCodec

Compression codec's supported by Arrow.

Enums
COMPRESSION_UNSPECIFIED If unspecified no compression will be used.
LZ4_FRAME LZ4 Frame (https://github.com/lz4/lz4/blob/dev/doc/lz4_Frame_format.md)
ZSTD Zstandard compression.

AvroRows

Avro rows.

Fields
serialized_binary_rows

bytes

Binary serialized rows in a block.

row_count
(deprecated)

int64

[Deprecated] The count of rows in the returning block. Please use the format-independent ReadRowsResponse.row_count instead.

AvroSchema

Avro schema.

Fields
schema

string

Json serialized schema, as described at https://avro.apache.org/docs/1.8.1/spec.html.

AvroSerializationOptions

Contains options specific to Avro Serialization.

Fields
enable_display_name_attribute

bool

Enable displayName attribute in Avro schema.

The Avro specification requires field names to be alphanumeric. By default, in cases when column names do not conform to these requirements (e.g. non-ascii unicode codepoints) and Avro is requested as an output format, the CreateReadSession call will fail.

Setting this field to true, populates avro field names with a placeholder value and populates a "displayName" attribute for every avro field with the original column name.

BatchCommitWriteStreamsRequest

Request message for BatchCommitWriteStreams.

Fields
parent

string

Required. Parent table that all the streams should belong to, in the form of projects/{project}/datasets/{dataset}/tables/{table}.

Authorization requires the following IAM permission on the specified resource parent:

  • bigquery.tables.updateData
write_streams[]

string

Required. The group of streams that will be committed atomically.

BatchCommitWriteStreamsResponse

Response message for BatchCommitWriteStreams.

Fields
commit_time

Timestamp

The time at which streams were committed in microseconds granularity. This field will only exist when there are no stream errors. Note if this field is not set, it means the commit was not successful.

stream_errors[]

StorageError

Stream level error if commit failed. Only streams with error will be in the list. If empty, there is no error and all streams are committed successfully. If non empty, certain streams have errors and ZERO stream is committed due to atomicity guarantee.

CreateReadSessionRequest

Request message for CreateReadSession.

Fields
parent

string

Required. The request project that owns the session, in the form of projects/{project_id}.

Authorization requires the following IAM permission on the specified resource parent:

  • bigquery.readsessions.create
read_session

ReadSession

Required. Session to be created.

Authorization requires the following IAM permission on the specified resource readSession:

  • bigquery.tables.getData
max_stream_count

int32

Max initial number of streams. If unset or zero, the server will provide a value of streams so as to produce reasonable throughput. Must be non-negative. The number of streams may be lower than the requested number, depending on the amount parallelism that is reasonable for the table. There is a default system max limit of 1,000.

This must be greater than or equal to preferred_min_stream_count. Typically, clients should either leave this unset to let the system to determine an upper bound OR set this a size for the maximum "units of work" it can gracefully handle.

preferred_min_stream_count

int32

The minimum preferred stream count. This parameter can be used to inform the service that there is a desired lower bound on the number of streams. This is typically a target parallelism of the client (e.g. a Spark cluster with N-workers would set this to a low multiple of N to ensure good cluster utilization).

The system will make a best effort to provide at least this number of streams, but in some cases might provide less.

CreateWriteStreamRequest

Request message for CreateWriteStream.

Fields
parent

string

Required. Reference to the table to which the stream belongs, in the format of projects/{project}/datasets/{dataset}/tables/{table}.

Authorization requires the following IAM permission on the specified resource parent:

  • bigquery.tables.updateData
write_stream

WriteStream

Required. Stream to be created.

DataFormat

Data format for input or output data.

Enums
DATA_FORMAT_UNSPECIFIED Data format is unspecified.
AVRO Avro is a standard open source row based file format. See https://avro.apache.org/ for more details.
ARROW Arrow is a standard open source column-based message format. See https://arrow.apache.org/ for more details.

FinalizeWriteStreamRequest

Request message for invoking FinalizeWriteStream.

Fields
name

string

Required. Name of the stream to finalize, in the form of projects/{project}/datasets/{dataset}/tables/{table}/streams/{stream}.

Authorization requires the following IAM permission on the specified resource name:

  • bigquery.tables.updateData

FinalizeWriteStreamResponse

Response message for FinalizeWriteStream.

Fields
row_count

int64

Number of rows in the finalized stream.

FlushRowsRequest

Request message for FlushRows.

Fields
write_stream

string

Required. The stream that is the target of the flush operation.

Authorization requires the following IAM permission on the specified resource writeStream:

  • bigquery.tables.updateData
offset

Int64Value

Ending offset of the flush operation. Rows before this offset(including this offset) will be flushed.

FlushRowsResponse

Respond message for FlushRows.

Fields
offset

int64

The rows before this offset (including this offset) are flushed.

GetWriteStreamRequest

Request message for GetWriteStreamRequest.

Fields
name

string

Required. Name of the stream to get, in the form of projects/{project}/datasets/{dataset}/tables/{table}/streams/{stream}.

Authorization requires the following IAM permission on the specified resource name:

  • bigquery.tables.get
view

WriteStreamView

Indicates whether to get full or partial view of the WriteStream. If not set, view returned will be basic.

ProtoRows

Fields
serialized_rows[]

bytes

A sequence of rows serialized as a Protocol Buffer.

See https://developers.google.com/protocol-buffers/docs/overview for more information on deserializing this field.

ProtoSchema

ProtoSchema describes the schema of the serialized protocol buffer data rows.

Fields
proto_descriptor

ProtoDescriptor

Descriptor for input message. The provided descriptor must be self contained, such that data rows sent can be fully decoded using only the single descriptor. For data rows that are compositions of multiple independent messages, this means the descriptor may need to be transformed to only use nested types: https://developers.google.com/protocol-buffers/docs/proto#nested

For additional information for how proto types and values map onto BigQuery see: https://cloud.google.com/bigquery/docs/write-api#data_type_conversions

ReadRowsRequest

Request message for ReadRows.

Fields
read_stream

string

Required. Stream to read rows from.

Authorization requires the following IAM permission on the specified resource readStream:

  • bigquery.readsessions.getData
offset

int64

The offset requested must be less than the last row read from Read. Requesting a larger offset is undefined. If not specified, start reading from offset zero.

ReadRowsResponse

Response from calling ReadRows may include row data, progress and throttling information.

Fields
row_count

int64

Number of serialized rows in the rows block.

stats

StreamStats

Statistics for the stream.

throttle_state

ThrottleState

Throttling state. If unset, the latest response still describes the current throttling status.

Union field rows. Row data is returned in format specified during session creation. rows can be only one of the following:
avro_rows

AvroRows

Serialized row data in AVRO format.

arrow_record_batch

ArrowRecordBatch

Serialized row data in Arrow RecordBatch format.

Union field schema. The schema for the read. If read_options.selected_fields is set, the schema may be different from the table schema as it will only contain the selected fields. This schema is equivalent to the one returned by CreateSession. This field is only populated in the first ReadRowsResponse RPC. schema can be only one of the following:
avro_schema

AvroSchema

Output only. Avro schema.

arrow_schema

ArrowSchema

Output only. Arrow schema.

ReadSession

Information about the ReadSession.

Fields
name

string

Output only. Unique identifier for the session, in the form projects/{project_id}/locations/{location}/sessions/{session_id}.

expire_time

Timestamp

Output only. Time at which the session becomes invalid. After this time, subsequent requests to read this Session will return errors. The expire_time is automatically assigned and currently cannot be specified or updated.

data_format

DataFormat

Immutable. Data format of the output data. DATA_FORMAT_UNSPECIFIED not supported.

table

string

Immutable. Table that this ReadSession is reading from, in the form projects/{project_id}/datasets/{dataset_id}/tables/{table_id}

Authorization requires one or more of the following IAM permissions on the specified resource table:

  • bigquery.tables.get
  • bigquery.tables.getData
table_modifiers

TableModifiers

Optional. Any modifiers which are applied when reading from the specified table.

read_options

TableReadOptions

Optional. Read options for this session (e.g. column selection, filters).

streams[]

ReadStream

Output only. A list of streams created with the session.

At least one stream is created with the session. In the future, larger request_stream_count values may result in this list being unpopulated, in that case, the user will need to use a List method to get the streams instead, which is not yet available.

estimated_total_bytes_scanned

int64

Output only. An estimate on the number of bytes this session will scan when all streams are completely consumed. This estimate is based on metadata from the table which might be incomplete or stale.

estimated_row_count

int64

Output only. An estimate on the number of rows present in this session's streams. This estimate is based on metadata from the table which might be incomplete or stale.

trace_id

string

Optional. ID set by client to annotate a session identity. This does not need to be strictly unique, but instead the same ID should be used to group logically connected sessions (e.g. All using the same ID for all sessions needed to complete a Spark SQL query is reasonable).

Maximum length is 256 bytes.

Union field schema. The schema for the read. If read_options.selected_fields is set, the schema may be different from the table schema as it will only contain the selected fields. schema can be only one of the following:
avro_schema

AvroSchema

Output only. Avro schema.

arrow_schema

ArrowSchema

Output only. Arrow schema.

TableModifiers

Additional attributes when reading a table.

Fields
snapshot_time

Timestamp

The snapshot time of the table. If not set, interpreted as now.

TableReadOptions

Options dictating how we read a table.

Fields
selected_fields[]

string

Optional. The names of the fields in the table to be returned. If no field names are specified, then all fields in the table are returned.

Nested fields -- the child elements of a STRUCT field -- can be selected individually using their fully-qualified names, and will be returned as record fields containing only the selected nested fields. If a STRUCT field is specified in the selected fields list, all of the child elements will be returned.

As an example, consider a table with the following schema:

{ "name": "struct_field", "type": "RECORD", "mode": "NULLABLE", "fields": [ { "name": "string_field1", "type": "STRING", . "mode": "NULLABLE" }, { "name": "string_field2", "type": "STRING", "mode": "NULLABLE" } ] }

Specifying "struct_field" in the selected fields list will result in a read session schema with the following logical structure:

struct_field { string_field1 string_field2 }

Specifying "struct_field.string_field1" in the selected fields list will result in a read session schema with the following logical structure:

struct_field { string_field1 }

The order of the fields in the read session schema is derived from the table schema and does not correspond to the order in which the fields are specified in this list.

row_restriction

string

SQL text filtering statement, similar to a WHERE clause in a query. Aggregates are not supported.

Examples: "int_field > 5" "date_field = CAST('2014-9-27' as DATE)" "nullable_field is not NULL" "st_equals(geo_field, st_geofromtext("POINT(2, 2)"))" "numeric_field BETWEEN 1.0 AND 5.0"

Restricted to a maximum length for 1 MB.

Union field output_format_serialization_options.

output_format_serialization_options can be only one of the following:

arrow_serialization_options

ArrowSerializationOptions

Optional. Options specific to the Apache Arrow output format.

avro_serialization_options

AvroSerializationOptions

Optional. Options specific to the Apache Avro output format

ReadStream

Information about a single stream that gets data out of the storage system. Most of the information about ReadStream instances is aggregated, making ReadStream lightweight.

Fields
name

string

Output only. Name of the stream, in the form projects/{project_id}/locations/{location}/sessions/{session_id}/streams/{stream_id}.

RowError

The message that presents row level error info in a request.

Fields
index

int64

Index of the malformed row in the request.

code

RowErrorCode

Structured error reason for a row error.

message

string

Description of the issue encountered when processing the row.

RowErrorCode

Error code for RowError.

Enums
ROW_ERROR_CODE_UNSPECIFIED Default error.
FIELDS_ERROR One or more fields in the row has errors.

SplitReadStreamRequest

Request message for SplitReadStream.

Fields
name

string

Required. Name of the stream to split.

Authorization requires the following IAM permission on the specified resource name:

  • bigquery.readsessions.update
fraction

double

A value in the range (0.0, 1.0) that specifies the fractional point at which the original stream should be split. The actual split point is evaluated on pre-filtered rows, so if a filter is provided, then there is no guarantee that the division of the rows between the new child streams will be proportional to this fractional value. Additionally, because the server-side unit for assigning data is collections of rows, this fraction will always map to a data storage boundary on the server side.

SplitReadStreamResponse

Response message for SplitReadStream.

Fields
primary_stream

ReadStream

Primary stream, which contains the beginning portion of |original_stream|. An empty value indicates that the original stream can no longer be split.

remainder_stream

ReadStream

Remainder stream, which contains the tail of |original_stream|. An empty value indicates that the original stream can no longer be split.

StorageError

Structured custom BigQuery Storage error message. The error can be attached as error details in the returned rpc Status. In particular, the use of error codes allows more structured error handling, and reduces the need to evaluate unstructured error text strings.

Fields
code

StorageErrorCode

BigQuery Storage specific error code.

entity

string

Name of the failed entity.

error_message

string

Message that describes the error.

StorageErrorCode

Error code for StorageError.

Enums
STORAGE_ERROR_CODE_UNSPECIFIED Default error.