package google.bigtable.v2
Get desktop application:
View/edit binary Protocol Buffers messages
service Bigtable
bigtable.proto:57Service for reading from and writing to existing Bigtable tables.
rpc CheckAndMutateRow (, )
bigtable.proto:193Mutates a row atomically based on the output of a predicate Reader filter.
message
bigtable.proto:709Request message for Bigtable.CheckAndMutateRow.
string table_name = 1
Optional. The unique name of the table to which the conditional mutation should be applied. Values are of the form `projects/<project>/instances/<instance>/tables/<table>`.
string authorized_view_name = 9
Optional. The unique name of the AuthorizedView to which the conditional mutation should be applied. Values are of the form `projects/<project>/instances/<instance>/tables/<table>/authorizedViews/<authorized_view>`.
string app_profile_id = 7
This value specifies routing for replication. If not specified, the "default" application profile will be used.
bytes row_key = 2
Required. The key of the row to which the conditional mutation should be applied.
optional predicate_filter = 6
The filter to be applied to the contents of the specified row. Depending on whether or not any results are yielded, either `true_mutations` or `false_mutations` will be executed. If unset, checks that the row contains any values at all.
repeated true_mutations = 4
Changes to be atomically applied to the specified row if `predicate_filter` yields at least one cell when applied to `row_key`. Entries are applied in order, meaning that earlier mutations can be masked by later ones. Must contain at least one entry if `false_mutations` is empty, and at most 100000.
repeated false_mutations = 5
Changes to be atomically applied to the specified row if `predicate_filter` does not yield any cells when applied to `row_key`. Entries are applied in order, meaning that earlier mutations can be masked by later ones. Must contain at least one entry if `true_mutations` is empty, and at most 100000.
message
bigtable.proto:764Response message for Bigtable.CheckAndMutateRow.
bool predicate_matched = 1
Whether or not the request's `predicate_filter` yielded any results for the specified row.
rpc ExecuteQuery (, stream )
bigtable.proto:316Executes a SQL query against a particular Bigtable instance.
message
bigtable.proto:1083Request message for Bigtable.ExecuteQuery
string instance_name = 1
Required. The unique name of the instance against which the query should be executed. Values are of the form `projects/<project>/instances/<instance>`
string app_profile_id = 2
Optional. This value specifies routing for replication. If not specified, the `default` application profile will be used.
string query = 3
Required. The query string. Exactly one of `query` and `prepared_query` is required. Setting both or neither is an `INVALID_ARGUMENT`.
bytes prepared_query = 9
A prepared query that was returned from `PrepareQueryResponse`. Exactly one of `query` and `prepared_query` is required. Setting both or neither is an `INVALID_ARGUMENT`. Setting this field also places restrictions on several other fields: - `data_format` must be empty. - `validate_only` must be false. - `params` must match the `param_types` set in the `PrepareQueryRequest`.
oneof data_format
Requested data format for the response. If `prepared_query` is set, then the `data_format` is fixed by the `PrepareQueryRequest`, and a non-empty `data_format` in the `ExecuteQueryRequest` will be rejected with `INVALID_ARGUMENT`.
proto_format = 4
Protocol buffer format as described by ProtoSchema and ProtoRows messages.
bytes resume_token = 8
Optional. If this request is resuming a previously interrupted query execution, `resume_token` should be copied from the last PartialResultSet yielded before the interruption. Doing this enables the query execution to resume where the last one left off. The rest of the request parameters must exactly match the request that yielded this token. Otherwise the request will fail.
map<string, > params = 7
Required. params contains string type keys and Bigtable type values that bind to placeholders in the query string. In query string, a parameter placeholder consists of the `@` character followed by the parameter name (for example, `@firstName`) in the query string. For example, if `params["firstName"] = bytes_value: "foo" type {bytes_type {}}` then `@firstName` will be replaced with googlesql bytes value "foo" in the query string during query evaluation. If `Value.kind` is not set, the value is treated as a NULL value of the given type. For example, if `params["firstName"] = type {string_type {}}` then `@firstName` will be replaced with googlesql null string. If `query` is set, any empty `Value.type` in the map will be rejected with `INVALID_ARGUMENT`. If `prepared_query` is set, any empty `Value.type` in the map will be inferred from the `param_types` in the `PrepareQueryRequest`. Any non-empty `Value.type` must match the corresponding `param_types` entry, or be rejected with `INVALID_ARGUMENT`.
message
bigtable.proto:1162Response message for Bigtable.ExecuteQuery
oneof response
The first response streamed from the server is of type `ResultSetMetadata` and includes information about the columns and types of the result set. From there on, we stream `PartialResultSet` messages with no additional information. `PartialResultSet` will contain `resume_token` to restart the response if query interrupts. In case of resumption with `resume_token`, the server will not resend the ResultSetMetadata.
metadata = 1
Structure of rows in this response stream. The first (and only the first) response streamed from the server will be of this type.
results = 2
A partial result set with row data potentially including additional instructions on how recent past and future partial responses should be interpreted.
rpc GenerateInitialChangeStreamPartitions (, stream )
bigtable.proto:273Returns the current list of partitions that make up the table's change stream. The union of partitions will cover the entire keyspace. Partitions can be read with `ReadChangeStream`. NOTE: This API is only intended to be used by Apache Beam BigtableIO.
message
bigtable.proto:840NOTE: This API is intended to be used by Apache Beam BigtableIO. Request message for Bigtable.GenerateInitialChangeStreamPartitions.
string table_name = 1
Required. The unique name of the table from which to get change stream partitions. Values are of the form `projects/<project>/instances/<instance>/tables/<table>`. Change streaming must be enabled on the table.
string app_profile_id = 2
This value specifies routing for replication. If not specified, the "default" application profile will be used. Single cluster routing must be configured on the profile.
message
bigtable.proto:860NOTE: This API is intended to be used by Apache Beam BigtableIO. Response message for Bigtable.GenerateInitialChangeStreamPartitions.
optional partition = 1
A partition of the change stream.
rpc MutateRow (, )
bigtable.proto:140Mutates a row atomically. Cells already present in the row are left unchanged unless explicitly changed by `mutation`.
message
bigtable.proto:564Request message for Bigtable.MutateRow.
string table_name = 1
Optional. The unique name of the table to which the mutation should be applied. Values are of the form `projects/<project>/instances/<instance>/tables/<table>`.
string authorized_view_name = 6
Optional. The unique name of the AuthorizedView to which the mutation should be applied. Values are of the form `projects/<project>/instances/<instance>/tables/<table>/authorizedViews/<authorized_view>`.
string app_profile_id = 4
This value specifies routing for replication. If not specified, the "default" application profile will be used.
bytes row_key = 2
Required. The key of the row to which the mutation should be applied.
repeated mutations = 3
Required. Changes to be atomically applied to the specified row. Entries are applied in order, meaning that earlier mutations can be masked by later ones. Must contain at least one entry and at most 100000.
optional idempotency = 8
If set consistently across retries, prevents this mutation from being double applied to aggregate column families within a 15m window.
message
bigtable.proto:607Response message for Bigtable.MutateRow.
(message has no fields)
rpc MutateRows (, stream )
bigtable.proto:168Mutates multiple rows in a batch. Each individual row is mutated atomically as in MutateRow, but the entire batch is not executed atomically.
message
bigtable.proto:610Request message for BigtableService.MutateRows.
string table_name = 1
Optional. The unique name of the table to which the mutations should be applied. Values are of the form `projects/<project>/instances/<instance>/tables/<table>`.
string authorized_view_name = 5
Optional. The unique name of the AuthorizedView to which the mutations should be applied. Values are of the form `projects/<project>/instances/<instance>/tables/<table>/authorizedViews/<authorized_view>`.
string app_profile_id = 3
This value specifies routing for replication. If not specified, the "default" application profile will be used.
repeated entries = 2
Required. The row keys and corresponding mutations to be applied in bulk. Each entry is applied as an atomic mutation, but the entries may be applied in arbitrary order (even between entries for the same row). At least one entry must be specified, and in total the entries can contain at most 100000 mutations.
message
bigtable.proto:663Response message for BigtableService.MutateRows.
repeated entries = 1
One or more results for Entries from the batch request.
optional rate_limit_info = 3
Information about how client should limit the rate (QPS). Primirily used by supported official Cloud Bigtable clients. If unset, the rate limit info is not provided by the server.
rpc PingAndWarm (, )
bigtable.proto:222Warm up associated instance metadata for this connection. This call is not required but may be useful for connection keep-alive.
message
bigtable.proto:771Request message for client connection keep-alive and warming.
string name = 1
Required. The unique name of the instance to check permissions for as well as respond. Values are of the form `projects/<project>/instances/<instance>`.
string app_profile_id = 2
This value specifies routing for replication. If not specified, the "default" application profile will be used.
message
bigtable.proto:788Response message for Bigtable.PingAndWarm connection keepalive and warming.
(message has no fields)
rpc PrepareQuery (, )
bigtable.proto:299Prepares a GoogleSQL query for execution on a particular Bigtable instance.
message
bigtable.proto:1182Request message for Bigtable.PrepareQuery
string instance_name = 1
Required. The unique name of the instance against which the query should be executed. Values are of the form `projects/<project>/instances/<instance>`
string app_profile_id = 2
Optional. This value specifies routing for preparing the query. Note that this `app_profile_id` is only used for preparing the query. The actual query execution will use the app profile specified in the `ExecuteQueryRequest`. If not specified, the `default` application profile will be used.
string query = 3
Required. The query string.
oneof data_format
Required. Requested data format for the response. Note that the selected data format is binding for all `ExecuteQuery` rpcs that use the prepared query.
proto_format = 4
Protocol buffer format as described by ProtoSchema and ProtoRows messages.
map<string, > param_types = 6
Required. `param_types` is a map of parameter identifier strings to their `Type`s. In query string, a parameter placeholder consists of the `@` character followed by the parameter name (for example, `@firstName`) in the query string. For example, if param_types["firstName"] = Bytes then @firstName will be a query parameter of type Bytes. The specific `Value` to be used for the query execution must be sent in `ExecuteQueryRequest` in the `params` map.
message
bigtable.proto:1226Response message for Bigtable.PrepareQueryResponse
optional metadata = 1
Structure of rows in the response stream of `ExecuteQueryResponse` for the returned `prepared_query`.
bytes prepared_query = 2
A serialized prepared query. Clients should treat this as an opaque blob of bytes to send in `ExecuteQueryRequest`.
optional valid_until = 3
The time at which the prepared query token becomes invalid. A token may become invalid early due to changes in the data being read, but it provides a guideline to refresh query plans asynchronously.
rpc ReadChangeStream (, stream )
bigtable.proto:288Reads changes from a table's change stream. Changes will reflect both user-initiated mutations and mutations that are caused by garbage collection. NOTE: This API is only intended to be used by Apache Beam BigtableIO.
message
bigtable.proto:867NOTE: This API is intended to be used by Apache Beam BigtableIO. Request message for Bigtable.ReadChangeStream.
string table_name = 1
Required. The unique name of the table from which to read a change stream. Values are of the form `projects/<project>/instances/<instance>/tables/<table>`. Change streaming must be enabled on the table.
string app_profile_id = 2
This value specifies routing for replication. If not specified, the "default" application profile will be used. Single cluster routing must be configured on the profile.
optional partition = 3
The partition to read changes from.
oneof start_from
Options for describing where we want to start reading from the stream.
start_time = 4
Start reading the stream at the specified timestamp. This timestamp must be within the change stream retention period, less than or equal to the current time, and after change stream creation, whichever is greater. This value is inclusive and will be truncated to microsecond granularity.
continuation_tokens = 6
Tokens that describe how to resume reading a stream where reading previously left off. If specified, changes will be read starting at the the position. Tokens are delivered on the stream as part of `Heartbeat` and `CloseStream` messages. If a single token is provided, the token's partition must exactly match the request's partition. If multiple tokens are provided, as in the case of a partition merge, the union of the token partitions must exactly cover the request's partition. Otherwise, INVALID_ARGUMENT will be returned.
optional end_time = 5
If specified, OK will be returned when the stream advances beyond this time. Otherwise, changes will be continuously delivered on the stream. This value is inclusive and will be truncated to microsecond granularity.
optional heartbeat_duration = 7
If specified, the duration between `Heartbeat` messages on the stream. Otherwise, defaults to 5 seconds.
message
bigtable.proto:920NOTE: This API is intended to be used by Apache Beam BigtableIO. Response message for Bigtable.ReadChangeStream.
oneof stream_record
The data or control message on the stream.
data_change = 1
A mutation to the partition.
heartbeat = 2
A periodic heartbeat message.
close_stream = 3
An indication that the stream should be closed.
rpc ReadModifyWriteRow (, )
bigtable.proto:243Modifies a row atomically on the server. The method reads the latest existing timestamp and value from the specified columns and writes a new entry based on pre-defined read/modify/write rules. The new value for the timestamp is the greater of the existing timestamp or the current server time. The method returns the new contents of all modified cells.
message
bigtable.proto:791Request message for Bigtable.ReadModifyWriteRow.
string table_name = 1
Optional. The unique name of the table to which the read/modify/write rules should be applied. Values are of the form `projects/<project>/instances/<instance>/tables/<table>`.
string authorized_view_name = 6
Optional. The unique name of the AuthorizedView to which the read/modify/write rules should be applied. Values are of the form `projects/<project>/instances/<instance>/tables/<table>/authorizedViews/<authorized_view>`.
string app_profile_id = 4
This value specifies routing for replication. If not specified, the "default" application profile will be used.
bytes row_key = 2
Required. The key of the row to which the read/modify/write rules should be applied.
repeated rules = 3
Required. Rules specifying how the specified row's contents are to be transformed into writes. Entries are applied in order, meaning that earlier rules will affect the results of later ones. At least one entry must be specified, and there can be at most 100000 rules.
message
bigtable.proto:833Response message for Bigtable.ReadModifyWriteRow.
optional row = 1
A Row containing the new contents of all cells modified by the request.
rpc ReadRows (, stream )
bigtable.proto:72Streams back the contents of all requested rows in key order, optionally applying the same Reader filter to each. Depending on their size, rows and cells may be broken up across multiple responses, but atomicity of each row will still be preserved. See the ReadRowsResponse documentation for details.
message
bigtable.proto:334Request message for Bigtable.ReadRows.
string table_name = 1
Optional. The unique name of the table from which to read. Values are of the form `projects/<project>/instances/<instance>/tables/<table>`.
string authorized_view_name = 9
Optional. The unique name of the AuthorizedView from which to read. Values are of the form `projects/<project>/instances/<instance>/tables/<table>/authorizedViews/<authorized_view>`.
string materialized_view_name = 11
Optional. The unique name of the MaterializedView from which to read. Values are of the form `projects/<project>/instances/<instance>/materializedViews/<materialized_view>`.
string app_profile_id = 5
This value specifies routing for replication. If not specified, the "default" application profile will be used.
optional rows = 2
The row keys and/or ranges to read sequentially. If not specified, reads from all rows.
optional filter = 3
The filter to apply to the contents of the specified row(s). If unset, reads the entirety of each row.
int64 rows_limit = 4
The read will stop after committing to N rows' worth of results. The default (zero) is to return all results.
ReadRowsRequest.RequestStatsView request_stats_view = 6
The view into RequestStats, as described above.
bool reversed = 7
Experimental API - Please note that this API is currently experimental and can change in the future. Return rows in lexiographical descending order of the row keys. The row contents will not be affected by this flag. Example result set: [ {key: "k2", "f:col1": "v1", "f:col2": "v1"}, {key: "k1", "f:col1": "v2", "f:col2": "v2"} ]
message
bigtable.proto:419Response message for Bigtable.ReadRows.
repeated chunks = 1
A collection of a row's contents as part of the read request.
bytes last_scanned_row_key = 2
Optionally the server might return the row key of the last row it has scanned. The client can use this to construct a more efficient retry request if needed: any row keys or portions of ranges less than this row key can be dropped from the request. This is primarily useful for cases where the server has read a lot of data that was filtered out since the last committed row key, allowing the client to skip that work on a retry.
optional request_stats = 3
If requested, return enhanced query performance statistics. The field request_stats is empty in a streamed response unless the ReadRowsResponse message contains request_stats in the last message of the stream. Always returned when requested, even when the read request returns an empty response.
rpc SampleRowKeys (, stream )
bigtable.proto:108Returns a sample of row keys in the table. The returned row keys will delimit contiguous sections of the table of approximately equal size, which can be used to break up the data for distributed tasks like mapreduces.
message
bigtable.proto:505Request message for Bigtable.SampleRowKeys.
string table_name = 1
Optional. The unique name of the table from which to sample row keys. Values are of the form `projects/<project>/instances/<instance>/tables/<table>`.
string authorized_view_name = 4
Optional. The unique name of the AuthorizedView from which to sample row keys. Values are of the form `projects/<project>/instances/<instance>/tables/<table>/authorizedViews/<authorized_view>`.
string materialized_view_name = 5
Optional. The unique name of the MaterializedView from which to read. Values are of the form `projects/<project>/instances/<instance>/materializedViews/<materialized_view>`.
string app_profile_id = 2
This value specifies routing for replication. If not specified, the "default" application profile will be used.
message
bigtable.proto:546Response message for Bigtable.SampleRowKeys.
bytes row_key = 1
Sorted streamed sequence of sample row keys in the table. The table might have contents before the first row key in the list and after the last one, but a key containing the empty string indicates "end of table" and will be the last response given, if present. Note that row keys in this list may not have ever been written to or read from, and users should therefore not make any assumptions about the row key structure that are specific to their use case.
int64 offset_bytes = 2
Approximate total storage space used by all rows in the table which precede `row_key`. Buffering the contents of all rows between two subsequent samples would require space roughly equal to the difference in their `offset_bytes` fields.
message
data.proto:160`ArrayValue` is an ordered list of `Value`.
Used in:
repeated values = 1
The ordered elements in the array.
message
data.proto:75Specifies (some of) the contents of a single row/column/timestamp of a table.
Used in:
int64 timestamp_micros = 1
The cell's stored timestamp, which also uniquely identifies it within its column. Values are always expressed in microseconds, but individual tables may set a coarser granularity to further restrict the allowed values. For example, a table which specifies millisecond granularity will only allow values of `timestamp_micros` which are multiples of 1000.
bytes value = 2
The value stored in the cell. May contain any byte string, including the empty string, up to 100MiB in length.
repeated string labels = 3
Labels applied to the cell by a [RowFilter][google.bigtable.v2.RowFilter].
message
data.proto:62Specifies (some of) the contents of a single row/column intersection of a table.
Used in:
bytes qualifier = 1
The unique key which identifies this column within its family. This is the same key that's used to identify the column in, for example, a RowFilter which sets its `column_qualifier_regex_filter` field. May contain any byte string, including the empty string, up to 16kiB in length.
repeated cells = 2
Must not be empty. Sorted in order of decreasing "timestamp_micros".
message
data.proto:692Describes a column in a Bigtable Query Language result set.
Used in:
string name = 1
The name of the column.
optional type = 2
The type of the column.
message
data.proto:201Specifies a contiguous range of columns within a single column family. The range spans from <column_family>:<start_qualifier> to <column_family>:<end_qualifier>, where both bounds can be either inclusive or exclusive.
Used in:
string family_name = 1
The name of the column family within which this range falls.
oneof start_qualifier
The column qualifier at which to start the range (within `column_family`). If neither field is set, interpreted as the empty string, inclusive.
bytes start_qualifier_closed = 2
Used when giving an inclusive lower bound for the range.
bytes start_qualifier_open = 3
Used when giving an exclusive lower bound for the range.
oneof end_qualifier
The column qualifier at which to end the range (within `column_family`). If neither field is set, interpreted as the infinite string, exclusive.
bytes end_qualifier_closed = 4
Used when giving an inclusive upper bound for the range.
bytes end_qualifier_open = 5
Used when giving an exclusive upper bound for the range.
message
data.proto:47Specifies (some of) the contents of a single row/column family intersection of a table.
Used in:
string name = 1
The unique key which identifies this family within its row. This is the same key that's used to identify the family in, for example, a RowFilter which sets its "family_name_regex_filter" field. Must match `[-_.a-zA-Z0-9]+`, except that AggregatingRowProcessors may produce cells in a sentinel family with an empty name. Must be no greater than 64 characters in length.
repeated columns = 2
Must not be empty. Sorted in order of increasing "qualifier".
message
request_stats.proto:87FullReadStatsView captures all known information about a read.
Used in:
optional read_iteration_stats = 1
Iteration stats describe how efficient the read is, e.g. comparing rows seen vs. rows returned or cells seen vs cells returned can provide an indication of read efficiency (the higher the ratio of seen to retuned the better).
optional request_latency_stats = 2
Request latency stats describe the time taken to complete a request, from the server side.
message
data.proto:844Parameters on mutations where clients want to ensure idempotency (i.e. at-most-once semantics). This is currently only needed for certain aggregate types.
Used in: ,
bytes token = 1
Unique token used to identify replays of this mutation. Must be at least 8 bytes long.
optional start_time = 2
Client-assigned timestamp when the mutation's first attempt was sent. Used to reject mutations that arrive after idempotency protection may have expired. May cause spurious rejections if clock skew is too high. Leave unset or zero to always accept the mutation, at the risk of double counting if the protection for previous attempts has expired.
message
bigtable.proto:612A mutation for a given row.
Used in:
bytes row_key = 1
The key of the row to which the `mutations` should be applied.
repeated mutations = 2
Required. Changes to be atomically applied to the specified row. Mutations are applied in order, meaning that earlier mutations can be masked by later ones. You must specify at least one mutation.
optional idempotency = 3
If set consistently across retries, prevents this mutation from being double applied to aggregate column families within a 15m window.
message
bigtable.proto:665The result of applying a passed mutation in the original request.
Used in:
int64 index = 1
The index into the original request's `entries` list of the Entry for which a result is being reported.
optional status = 2
The result of the request Entry identified by `index`. Depending on how requests are batched during execution, it is possible for one Entry to fail due to an error with another Entry. In the event that this occurs, the same error will be reported for both entries.
message
data.proto:521Specifies a particular change to be made to the contents of a row.
Used in: , , ,
oneof mutation
Which of the possible Mutation types to apply.
set_cell = 1
Set a cell's value.
add_to_cell = 5
Incrementally updates an `Aggregate` cell.
merge_to_cell = 6
Merges accumulated state to an `Aggregate` cell.
delete_from_column = 2
Deletes cells from a column.
delete_from_family = 3
Deletes cells from a column family.
delete_from_row = 4
Deletes cells from the entire row.
message
data.proto:544A Mutation which incrementally updates a cell in an `Aggregate` family.
Used in:
string family_name = 1
The name of the `Aggregate` family into which new data should be added. This must be a family with a `value_type` of `Aggregate`. Format: `[-_.a-zA-Z0-9]+`
optional column_qualifier = 2
The qualifier of the column into which new data should be added. This must be a `raw_value`.
optional timestamp = 3
The timestamp of the cell to which new data should be added. This must be a `raw_timestamp_micros` that matches the table's `granularity`.
optional input = 4
The input value to be accumulated into the specified cell. This must be compatible with the family's `value_type.input_type`.
message
data.proto:587A Mutation which deletes cells from the specified column, optionally restricting the deletions to a given timestamp range.
Used in:
string family_name = 1
The name of the family from which cells should be deleted. Must match `[-_.a-zA-Z0-9]+`
bytes column_qualifier = 2
The qualifier of the column from which cells should be deleted. Can be any byte string, including the empty string.
optional time_range = 3
The range of timestamps within which cells should be deleted.
message
data.proto:601A Mutation which deletes all cells from the specified column family.
Used in:
string family_name = 1
The name of the family from which cells should be deleted. Must match `[-_.a-zA-Z0-9]+`
message
data.proto:608A Mutation which deletes all cells from the containing row.
Used in:
(message has no fields)
message
data.proto:565A Mutation which merges accumulated state into a cell in an `Aggregate` family.
Used in:
string family_name = 1
The name of the `Aggregate` family into which new data should be added. This must be a family with a `value_type` of `Aggregate`. Format: `[-_.a-zA-Z0-9]+`
optional column_qualifier = 2
The qualifier of the column into which new data should be added. This must be a `raw_value`.
optional timestamp = 3
The timestamp of the cell to which new data should be added. This must be a `raw_timestamp_micros` that matches the table's `granularity`.
optional input = 4
The input value to be merged into the specified cell. This must be compatible with the family's `value_type.state_type`. Merging `NULL` is allowed, but has no effect.
message
data.proto:523A Mutation which sets the value of the specified cell.
Used in:
string family_name = 1
The name of the family into which new data should be written. Must match `[-_.a-zA-Z0-9]+`
bytes column_qualifier = 2
The qualifier of the column into which new data should be written. Can be any byte string, including the empty string.
int64 timestamp_micros = 3
The timestamp of the cell into which new data should be written. Use -1 for current Bigtable server time. Otherwise, the client should set this value itself, noting that the default value is a timestamp of zero if the field is left unspecified. Values must match the granularity of the table (e.g. micros, millis).
bytes value = 4
The value to be written into the specified cell.
message
data.proto:771A partial result set from the streaming query API. Cloud Bigtable clients buffer partial results received in this message until a `resume_token` is received. The pseudocode below describes how to buffer and parse a stream of `PartialResultSet` messages. Having: - queue of row results waiting to be returned `queue` - extensible buffer of bytes `buffer` - a place to keep track of the most recent `resume_token` for each PartialResultSet `p` received { if p.reset { ensure `queue` is empty ensure `buffer` is empty } if p.estimated_batch_size != 0 { (optional) ensure `buffer` is sized to at least `p.estimated_batch_size` } if `p.proto_rows_batch` is set { append `p.proto_rows_batch.bytes` to `buffer` } if p.batch_checksum is set and `buffer` is not empty { validate the checksum matches the contents of `buffer` (see comments on `batch_checksum`) parse `buffer` as `ProtoRows` message, clearing `buffer` add parsed rows to end of `queue` } if p.resume_token is set { release results in `queue` save `p.resume_token` in `resume_token` } }
Used in:
oneof partial_rows
Some rows of the result set in one of the supported formats. Multiple `PartialResultSet` messages may be sent to represent a complete response. The client should buffer data constructed from the fields in `partial_rows` until a non-empty `resume_token` is received. Each sub-message documents the appropriate way to combine results.
proto_rows_batch = 3
Partial rows in serialized ProtoRows format.
optional uint32 batch_checksum = 6
CRC32C checksum of concatenated `partial_rows` data for the current batch. When present, the buffered data from `partial_rows` forms a complete parseable message of the appropriate type. The client should mark the end of a parseable message and prepare to receive a new one starting from the next `PartialResultSet` message. Clients must verify the checksum of the serialized batch before yielding it to the caller. This does NOT mean the values can be yielded to the callers since a `resume_token` is required to safely do so. If `resume_token` is non-empty and any data has been received since the last one, this field is guaranteed to be non-empty. In other words, clients may assume that a batch will never cross a `resume_token` boundary.
bytes resume_token = 5
An opaque token sent by the server to allow query resumption and signal that the buffered values constructed from received `partial_rows` can be yielded to the caller. Clients can provide this token in a subsequent request to resume the result stream from the current point. When `resume_token` is non-empty, the buffered values received from `partial_rows` since the last non-empty `resume_token` can be yielded to the callers, provided that the client keeps the value of `resume_token` and uses it on subsequent retries. A `resume_token` may be sent without information in `partial_rows` to checkpoint the progress of a sparse query. Any previous `partial_rows` data should still be yielded in this case, and the new `resume_token` should be saved for future retries as normal. A `resume_token` will only be sent on a boundary where there is either no ongoing result batch, or `batch_checksum` is also populated. The server will also send a sentinel `resume_token` when last batch of `partial_rows` is sent. If the client retries the ExecuteQueryRequest with the sentinel `resume_token`, the server will emit it again without any data in `partial_rows`, then return OK.
bool reset = 7
If `true`, any data buffered since the last non-empty `resume_token` must be discarded before the other parts of this message, if any, are handled.
int32 estimated_batch_size = 4
Estimated size of the buffer required to hold the next batch of results. This value will be sent with the first `partial_rows` of a batch. That is, on the first `partial_rows` received in a stream, on the first message after a `batch_checksum` message, and any time `reset` is true. The client can use this estimate to allocate a buffer for the next batch of results. This helps minimize the number of allocations required, though the buffer size may still need to be increased if the estimate is too low.
message
data.proto:689Protocol buffers format descriptor, as described by Messages ProtoSchema and ProtoRows
Used in: ,
(message has no fields)
message
data.proto:721Rows represented in proto format. This should be constructed by concatenating the `batch_data` from each of the relevant `ProtoRowsBatch` messages and parsing the result as a `ProtoRows` message.
repeated values = 2
A proto rows message consists of a list of values. Every N complete values defines a row, where N is equal to the number of entries in the `metadata.proto_schema.columns` value received in the first response.
message
data.proto:729A part of a serialized `ProtoRows` message.
Used in:
bytes batch_data = 1
Part of a serialized `ProtoRows` message. A complete, parseable ProtoRows message is constructed by concatenating `batch_data` from multiple `ProtoRowsBatch` messages. The `PartialResultSet` that contains the last part has `complete_batch` set to `true`.
message
data.proto:701ResultSet schema in proto format
Used in:
repeated columns = 1
The columns in the result set.
message
bigtable.proto:687Information about how client should adjust the load to Bigtable.
Used in:
optional period = 1
Time that clients should wait before adjusting the target rate again. If clients adjust rate too frequently, the impact of the previous adjustment may not have been taken into account and may over-throttle or under-throttle. If clients adjust rate too slowly, they will not be responsive to load changes on server side, and may over-throttle or under-throttle.
double factor = 2
If it has been at least one `period` since the last load adjustment, the client should multiply the current load by this value to get the new target load. For example, if the current load is 100 and `factor` is 0.8, the new target load should be 80. After adjusting, the client should ignore `factor` until another `period` has passed. The client can measure its load using any unit that's comparable over time. For example, QPS can be used as long as each request involves a similar amount of work.
message
bigtable.proto:1055A message indicating that the client should stop reading from the stream. If status is OK and `continuation_tokens` & `new_partitions` are empty, the stream has finished (for example if there was an `end_time` specified). If `continuation_tokens` & `new_partitions` are present, then a change in partitioning requires the client to open a new stream for each token to resume reading. Example: [B, D) ends | v new_partitions: [A, C) [C, E) continuation_tokens.partitions: [B,C) [C,D) ^---^ ^---^ ^ ^ | | | StreamContinuationToken 2 | StreamContinuationToken 1 To read the new partition [A,C), supply the continuation tokens whose ranges cover the new partition, for example ContinuationToken[A,B) & ContinuationToken[B,C).
Used in:
optional status = 1
The status of the stream.
repeated continuation_tokens = 2
If non-empty, contains the information needed to resume reading their associated partitions.
repeated new_partitions = 3
If non-empty, contains the new partitions to start reading from, which are related to but not necessarily identical to the partitions for the above `continuation_tokens`.
message
bigtable.proto:954A message corresponding to one or more mutations to the partition being streamed. A single logical `DataChange` message may also be split across a sequence of multiple individual messages. Messages other than the first in a sequence will only have the `type` and `chunks` fields populated, with the final message in the sequence also containing `done` set to true.
Used in:
DataChange.Type type = 1
The type of the mutation.
string source_cluster_id = 2
The cluster where the mutation was applied. Not set when `type` is `GARBAGE_COLLECTION`.
bytes row_key = 3
The row key for all mutations that are part of this `DataChange`. If the `DataChange` is chunked across multiple messages, then this field will only be set for the first message.
optional commit_timestamp = 4
The timestamp at which the mutation was applied on the Bigtable server.
int32 tiebreaker = 5
A value that lets stream consumers reconstruct Bigtable's conflict resolution semantics. https://cloud.google.com/bigtable/docs/writes#conflict-resolution In the event that the same row key, column family, column qualifier, timestamp are modified on different clusters at the same `commit_timestamp`, the mutation with the larger `tiebreaker` will be the one chosen for the eventually consistent state of the system.
repeated chunks = 6
The mutations associated with this change to the partition. May contain complete mutations or chunks of a multi-message chunked `DataChange` record.
bool done = 8
When true, indicates that the entire `DataChange` has been read and the client can safely process the message.
string token = 9
An encoded position for this stream's partition to restart reading from. This token is for the StreamPartition from the request.
optional estimated_low_watermark = 10
An estimate of the commit timestamp that is usually lower than or equal to any timestamp for a record that will be delivered in the future on the stream. It is possible that, under particular circumstances that a future record has a timestamp that is lower than a previously seen timestamp. For an example usage see https://beam.apache.org/documentation/basics/#watermarks
The type of mutation.
Used in:
TYPE_UNSPECIFIED = 0
The type is unspecified.
USER = 1
A user-initiated mutation.
GARBAGE_COLLECTION = 2
A system-initiated mutation as part of garbage collection. https://cloud.google.com/bigtable/docs/garbage-collection
CONTINUATION = 3
This is a continuation of a multi-message change.
message
bigtable.proto:1019A periodic message with information that can be used to checkpoint the state of a stream.
Used in:
optional continuation_token = 1
A token that can be provided to a subsequent `ReadChangeStream` call to pick up reading at the current stream position.
optional estimated_low_watermark = 2
An estimate of the commit timestamp that is usually lower than or equal to any timestamp for a record that will be delivered in the future on the stream. It is possible that, under particular circumstances that a future record has a timestamp that is lower than a previously seen timestamp. For an example usage see https://beam.apache.org/documentation/basics/#watermarks
message
bigtable.proto:922A partial or complete mutation.
Used in:
optional chunk_info = 1
If set, then the mutation is a `SetCell` with a chunked value across multiple messages.
optional mutation = 2
If this is a continuation of a chunked message (`chunked_value_offset` > 0), ignore all fields except the `SetCell`'s value and merge it with the previous message by concatenating the value fields.
message
bigtable.proto:926Information about the chunking of this mutation. Only `SetCell` mutations can be chunked, and all chunks for a `SetCell` will be delivered contiguously with no other mutation types interleaved.
Used in:
int32 chunked_value_size = 1
The total value size of all the chunks that make up the `SetCell`.
int32 chunked_value_offset = 2
The byte offset of this chunk into the total value size of the mutation.
bool last_chunk = 3
When true, this is the last chunk of a chunked `SetCell`.
message
request_stats.proto:46ReadIterationStats captures information about the iteration of rows or cells over the course of a read, e.g. how many results were scanned in a read operation versus the results returned.
Used in:
int64 rows_seen_count = 1
The rows seen (scanned) as part of the request. This includes the count of rows returned, as captured below.
int64 rows_returned_count = 2
The rows returned as part of the request.
int64 cells_seen_count = 3
The cells seen (scanned) as part of the request. This includes the count of cells returned, as captured below.
int64 cells_returned_count = 4
The cells returned as part of the request.
message
data.proto:634Specifies an atomic read/modify/write operation on the latest value of the specified column.
Used in:
string family_name = 1
The name of the family to which the read/modify/write should be applied. Must match `[-_.a-zA-Z0-9]+`
bytes column_qualifier = 2
The qualifier of the column to which the read/modify/write should be applied. Can be any byte string, including the empty string.
oneof rule
The rule used to determine the column's new latest value from its current latest value.
bytes append_value = 3
Rule specifying that `append_value` be appended to the existing value. If the targeted cell is unset, it will be treated as containing the empty string.
int64 increment_amount = 4
Rule specifying that `increment_amount` be added to the existing value. If the targeted cell is unset, it will be treated as containing a zero. Otherwise, the targeted cell must contain an 8-byte value (interpreted as a 64-bit big-endian signed integer), or the entire request will fail.
The desired view into RequestStats that should be returned in the response. See also: RequestStats message.
Used in:
REQUEST_STATS_VIEW_UNSPECIFIED = 0
The default / unset value. The API will default to the NONE option below.
REQUEST_STATS_NONE = 1
Do not include any RequestStats in the response. This will leave the RequestStats embedded message unset in the response.
REQUEST_STATS_FULL = 2
Include the full set of available RequestStats in the response, applicable to this read.
message
bigtable.proto:422Specifies a piece of a row's contents returned as part of the read response stream.
Used in:
bytes row_key = 1
The row key for this chunk of data. If the row key is empty, this CellChunk is a continuation of the same row as the previous CellChunk in the response stream, even if that CellChunk was in a previous ReadRowsResponse message.
optional family_name = 2
The column family name for this chunk of data. If this message is not present this CellChunk is a continuation of the same column family as the previous CellChunk. The empty string can occur as a column family name in a response so clients must check explicitly for the presence of this message, not just for `family_name.value` being non-empty.
optional qualifier = 3
The column qualifier for this chunk of data. If this message is not present, this CellChunk is a continuation of the same column as the previous CellChunk. Column qualifiers may be empty so clients must check for the presence of this message, not just for `qualifier.value` being non-empty.
int64 timestamp_micros = 4
The cell's stored timestamp, which also uniquely identifies it within its column. Values are always expressed in microseconds, but individual tables may set a coarser granularity to further restrict the allowed values. For example, a table which specifies millisecond granularity will only allow values of `timestamp_micros` which are multiples of 1000. Timestamps are only set in the first CellChunk per cell (for cells split into multiple chunks).
repeated string labels = 5
Labels applied to the cell by a [RowFilter][google.bigtable.v2.RowFilter]. Labels are only set on the first CellChunk per cell.
bytes value = 6
The value stored in the cell. Cell values can be split across multiple CellChunks. In that case only the value field will be set in CellChunks after the first: the timestamp and labels will only be present in the first CellChunk, even if the first CellChunk came in a previous ReadRowsResponse.
int32 value_size = 7
If this CellChunk is part of a chunked cell value and this is not the final chunk of that cell, value_size will be set to the total length of the cell value. The client can use this size to pre-allocate memory to hold the full cell value.
oneof row_status
Signals to the client concerning previous CellChunks received.
bool reset_row = 8
Indicates that the client should drop all previous chunks for `row_key`, as it will be re-read from the beginning.
bool commit_row = 9
Indicates that the client can safely process all previous chunks for `row_key`, as its data has been fully read.
message
request_stats.proto:65RequestLatencyStats provides a measurement of the latency of the request as it interacts with different systems over its lifetime, e.g. how long the request took to execute within a frontend server.
Used in:
optional frontend_server_latency = 1
The latency measured by the frontend server handling this request, from when the request was received, to when this value is sent back in the response. For more context on the component that is measuring this latency, see: https://cloud.google.com/bigtable/docs/overview Note: This value may be slightly shorter than the value reported into aggregate latency metrics in Monitoring for this request (https://cloud.google.com/bigtable/docs/monitoring-instance) as this value needs to be sent in the response before the latency measurement including that transmission is finalized. Note: This value includes the end-to-end latency of contacting nodes in the targeted cluster, e.g. measuring from when the first byte arrives at the frontend server, to when this value is sent back as the last value in the response, including any latency incurred by contacting nodes, waiting for results from nodes, and finally sending results from nodes back to the caller.
message
request_stats.proto:102RequestStats is the container for additional information pertaining to a single request, helpful for evaluating the performance of the sent request. Currently, the following method is supported: google.bigtable.v2.ReadRows
Used in:
oneof stats_view
Information pertaining to each request type received. The type is chosen based on the requested view. See the messages above for additional context.
full_read_stats_view = 1
Available with the ReadRowsRequest.RequestStatsView.REQUEST_STATS_FULL view, see package google.bigtable.v2.
message
data.proto:707Describes the structure of a Bigtable result set.
Used in: ,
oneof schema
The schema of the ResultSet, contains ordered list of column names with types
proto_schema = 1
Schema in proto format
message
data.proto:34Specifies the complete (requested) contents of a single row of a table. Rows which exceed 256MiB in size cannot be read in full.
Used in:
bytes key = 1
The unique key which identifies this row within its table. This is the same key that's used to identify the row in, for example, a MutateRowRequest. May contain any non-empty byte string up to 4KiB in length.
repeated families = 2
May be empty, but only if the entire row is empty. The mutual ordering of column families is not specified.
message
data.proto:291Takes a row as input and produces an alternate view of the row based on specified rules. For example, a RowFilter might trim down a row to include just the cells from columns matching a given regular expression, or might return all the cells of a row but not their values. More complicated filters can be composed out of these components to express requests such as, "within every column of a particular family, give just the two most recent cells which are older than timestamp X." There are two broad categories of RowFilters (true filters and transformers), as well as two ways to compose simple filters into more complex ones (chains and interleaves). They work as follows: * True filters alter the input row by excluding some of its cells wholesale from the output row. An example of a true filter is the `value_regex_filter`, which excludes cells whose values don't match the specified pattern. All regex true filters use RE2 syntax (https://github.com/google/re2/wiki/Syntax) in raw byte mode (RE2::Latin1), and are evaluated as full matches. An important point to keep in mind is that `RE2(.)` is equivalent by default to `RE2([^\n])`, meaning that it does not match newlines. When attempting to match an arbitrary byte, you should therefore use the escape sequence `\C`, which may need to be further escaped as `\\C` in your client language. * Transformers alter the input row by changing the values of some of its cells in the output, without excluding them completely. Currently, the only supported transformer is the `strip_value_transformer`, which replaces every cell's value with the empty string. * Chains and interleaves are described in more detail in the RowFilter.Chain and RowFilter.Interleave documentation. The total serialized size of a RowFilter message must not exceed 20480 bytes, and RowFilters may not be nested within each other (in Chains or Interleaves) to a depth of more than 20.
Used in: , , , ,
oneof filter
Which of the possible RowFilter types to apply. If none are set, this RowFilter returns all cells in the input row.
chain = 1
Applies several RowFilters to the data in sequence, progressively narrowing the results.
interleave = 2
Applies several RowFilters to the data in parallel and combines the results.
condition = 3
Applies one of two possible RowFilters to the data based on the output of a predicate RowFilter.
bool sink = 16
ADVANCED USE ONLY. Hook for introspection into the RowFilter. Outputs all cells directly to the output of the read rather than to any parent filter. Consider the following example: Chain( FamilyRegex("A"), Interleave( All(), Chain(Label("foo"), Sink()) ), QualifierRegex("B") ) A,A,1,w A,B,2,x B,B,4,z | FamilyRegex("A") | A,A,1,w A,B,2,x | +------------+-------------+ | | All() Label(foo) | | A,A,1,w A,A,1,w,labels:[foo] A,B,2,x A,B,2,x,labels:[foo] | | | Sink() --------------+ | | | +------------+ x------+ A,A,1,w,labels:[foo] | A,B,2,x,labels:[foo] A,A,1,w | A,B,2,x | | | QualifierRegex("B") | | | A,B,2,x | | | +--------------------------------+ | A,A,1,w,labels:[foo] A,B,2,x,labels:[foo] // could be switched A,B,2,x // could be switched Despite being excluded by the qualifier filter, a copy of every cell that reaches the sink is present in the final result. As with an [Interleave][google.bigtable.v2.RowFilter.Interleave], duplicate cells are possible, and appear in an unspecified mutual order. In this case we have a duplicate with column "A:B" and timestamp 2, because one copy passed through the all filter while the other was passed through the label and sink. Note that one copy has label "foo", while the other does not. Cannot be used within the `predicate_filter`, `true_filter`, or `false_filter` of a [Condition][google.bigtable.v2.RowFilter.Condition].
bool pass_all_filter = 17
Matches all cells, regardless of input. Functionally equivalent to leaving `filter` unset, but included for completeness.
bool block_all_filter = 18
Does not match any cells, regardless of input. Useful for temporarily disabling just part of a filter.
bytes row_key_regex_filter = 4
Matches only cells from rows whose keys satisfy the given RE2 regex. In other words, passes through the entire row when the key matches, and otherwise produces an empty row. Note that, since row keys can contain arbitrary bytes, the `\C` escape sequence must be used if a true wildcard is desired. The `.` character will not match the new line character `\n`, which may be present in a binary key.
double row_sample_filter = 14
Matches all cells from a row with probability p, and matches no cells from the row with probability 1-p.
string family_name_regex_filter = 5
Matches only cells from columns whose families satisfy the given RE2 regex. For technical reasons, the regex must not contain the `:` character, even if it is not being used as a literal. Note that, since column families cannot contain the new line character `\n`, it is sufficient to use `.` as a full wildcard when matching column family names.
bytes column_qualifier_regex_filter = 6
Matches only cells from columns whose qualifiers satisfy the given RE2 regex. Note that, since column qualifiers can contain arbitrary bytes, the `\C` escape sequence must be used if a true wildcard is desired. The `.` character will not match the new line character `\n`, which may be present in a binary qualifier.
column_range_filter = 7
Matches only cells from columns within the given range.
timestamp_range_filter = 8
Matches only cells with timestamps within the given range.
bytes value_regex_filter = 9
Matches only cells with values that satisfy the given regular expression. Note that, since cell values can contain arbitrary bytes, the `\C` escape sequence must be used if a true wildcard is desired. The `.` character will not match the new line character `\n`, which may be present in a binary value.
value_range_filter = 15
Matches only cells with values that fall within the given range.
int32 cells_per_row_offset_filter = 10
Skips the first N cells of each row, matching all subsequent cells. If duplicate cells are present, as is possible when using an Interleave, each copy of the cell is counted separately.
int32 cells_per_row_limit_filter = 11
Matches only the first N cells of each row. If duplicate cells are present, as is possible when using an Interleave, each copy of the cell is counted separately.
int32 cells_per_column_limit_filter = 12
Matches only the most recent N cells within each column. For example, if N=2, this filter would match column `foo:bar` at timestamps 10 and 9, skip all earlier cells in `foo:bar`, and then begin matching again in column `foo:bar2`. If duplicate cells are present, as is possible when using an Interleave, each copy of the cell is counted separately.
bool strip_value_transformer = 13
Replaces each cell's value with the empty string.
string apply_label_transformer = 19
Applies the given label to all cells in the output row. This allows the client to determine which results were produced from which part of the filter. Values must be at most 15 characters in length, and match the RE2 pattern `[a-z0-9\\-]+` Due to a technical limitation, it is not currently possible to apply multiple labels to a cell. As a result, a Chain may have no more than one sub-filter which contains a `apply_label_transformer`. It is okay for an Interleave to contain multiple `apply_label_transformers`, as they will be applied to separate copies of the input. This may be relaxed in the future.
message
data.proto:293A RowFilter which sends rows through several RowFilters in sequence.
Used in:
repeated filters = 1
The elements of "filters" are chained together to process the input row: in row -> f(0) -> intermediate row -> f(1) -> ... -> f(N) -> out row The full chain is executed atomically.
message
data.proto:338A RowFilter which evaluates one of two possible RowFilters, depending on whether or not a predicate RowFilter outputs any cells from the input row. IMPORTANT NOTE: The predicate filter does not execute atomically with the true and false filters, which may lead to inconsistent or unexpected results. Additionally, Condition filters have poor performance, especially when filters are set for the false condition.
Used in:
optional predicate_filter = 1
If `predicate_filter` outputs any cells, then `true_filter` will be evaluated on the input row. Otherwise, `false_filter` will be evaluated.
optional true_filter = 2
The filter to apply to the input row if `predicate_filter` returns any results. If not provided, no results will be returned in the true case.
optional false_filter = 3
The filter to apply to the input row if `predicate_filter` does not return any results. If not provided, no results will be returned in the false case.
message
data.proto:302A RowFilter which sends each row to each of several component RowFilters and interleaves the results.
Used in:
repeated filters = 1
The elements of "filters" all process a copy of the input row, and the results are pooled, sorted, and combined into a single output row. If multiple cells are produced with the same column and timestamp, they will all appear in the output row in an unspecified mutual order. Consider the following example, with three filters: input row | ----------------------------------------------------- | | | f(0) f(1) f(2) | | | 1: foo,bar,10,x foo,bar,10,z far,bar,7,a 2: foo,blah,11,z far,blah,5,x far,blah,5,x | | | ----------------------------------------------------- | 1: foo,bar,10,z // could have switched with #2 2: foo,bar,10,x // could have switched with #1 3: foo,blah,11,z 4: far,bar,7,a 5: far,blah,5,x // identical to #6 6: far,blah,5,x // identical to #5 All interleaved filters are executed atomically.
message
data.proto:166Specifies a contiguous range of rows.
Used in: ,
oneof start_key
The row key at which to start the range. If neither field is set, interpreted as the empty string, inclusive.
bytes start_key_closed = 1
Used when giving an inclusive lower bound for the range.
bytes start_key_open = 2
Used when giving an exclusive lower bound for the range.
oneof end_key
The row key at which to end the range. If neither field is set, interpreted as the infinite row key, exclusive.
bytes end_key_open = 3
Used when giving an exclusive upper bound for the range.
bytes end_key_closed = 4
Used when giving an inclusive upper bound for the range.
message
data.proto:189Specifies a non-contiguous set of rows.
Used in:
repeated bytes row_keys = 1
Single rows included in the set.
repeated row_ranges = 2
Contiguous row ranges included in the set.
message
data.proto:679NOTE: This API is intended to be used by Apache Beam BigtableIO. The information required to continue reading the data from a `StreamPartition` from where a previous read left off.
Used in: , ,
optional partition = 1
The partition that this token applies to.
string token = 2
An encoded position in the stream to restart reading from.
message
data.proto:671NOTE: This API is intended to be used by Apache Beam BigtableIO. The information required to continue reading the data from multiple `StreamPartitions` from where a previous read left off.
Used in:
repeated tokens = 1
List of continuation tokens.
message
data.proto:662NOTE: This API is intended to be used by Apache Beam BigtableIO. A partition of a change stream.
Used in: , , ,
optional row_range = 1
The row range covered by this partition and is specified by [`start_key_closed`, `end_key_open`).
message
data.proto:227Specified a contiguous range of microsecond timestamps.
Used in: ,
int64 start_timestamp_micros = 1
Inclusive lower bound. If left empty, interpreted as 0.
int64 end_timestamp_micros = 2
Exclusive upper bound. If left empty, interpreted as infinity.
message
types.proto:51`Type` represents the type of data that is written to, read from, or stored in Bigtable. It is heavily based on the GoogleSQL standard to help maintain familiarity and consistency across products and features. For compatibility with Bigtable's existing untyped APIs, each `Type` includes an `Encoding` which describes how to convert to or from the underlying data. Each encoding can operate in one of two modes: - Sorted: In this mode, Bigtable guarantees that `Encode(X) <= Encode(Y)` if and only if `X <= Y`. This is useful anywhere sort order is important, for example when encoding keys. - Distinct: In this mode, Bigtable guarantees that if `X != Y` then `Encode(X) != Encode(Y)`. However, the converse is not guaranteed. For example, both `{'foo': '1', 'bar': '2'}` and `{'bar': '2', 'foo': '1'}` are valid encodings of the same JSON value. The API clearly documents which mode is used wherever an encoding can be configured. Each encoding also documents which values are supported in which modes. For example, when encoding INT64 as a numeric STRING, negative numbers cannot be encoded in sorted mode. This is because `INT64(1) > INT64(-1)`, but `STRING("-00001") > STRING("00001")`.
Used in: , , , , , ,
oneof kind
The kind of type that this represents.
bytes_type = 1
Bytes
string_type = 2
String
int64_type = 5
Int64
float32_type = 12
Float32
float64_type = 9
Float64
bool_type = 8
Bool
timestamp_type = 10
Timestamp
date_type = 11
Date
aggregate_type = 6
Aggregate
struct_type = 7
Struct
array_type = 3
Array
map_type = 4
Map
proto_type = 13
Proto
enum_type = 14
Enum
message
types.proto:371A value that combines incremental updates into a summarized value. Data is never directly written or read using type `Aggregate`. Writes provide either the `input_type` or `state_type`, and reads always return the `state_type` .
Used in:
optional input_type = 1
Type of the inputs that are accumulated by this `Aggregate`. Use `AddInput` mutations to accumulate new inputs.
optional state_type = 2
Output only. Type that holds the internal accumulator state for the `Aggregate`. This is a function of the `input_type` and `aggregator` chosen.
oneof aggregator
Which aggregator function to use. The configured types must match.
sum = 4
Sum aggregator.
hllpp_unique_count = 5
HyperLogLogPlusPlusUniqueCount aggregator.
max = 6
Max aggregator.
min = 7
Min aggregator.
message
types.proto:394Computes an approximate unique count over the input values. When using raw data as input, be careful to use a consistent encoding. Otherwise the same value encoded differently could count more than once, or two distinct values could count as identical. Input: Any, or omit for Raw State: TBD Special state conversions: `Int64` (the unique count estimate)
Used in:
(message has no fields)
message
types.proto:380Computes the max of the input values. Allowed input: `Int64` State: same as input
Used in:
(message has no fields)
message
types.proto:385Computes the min of the input values. Allowed input: `Int64` State: same as input
Used in:
(message has no fields)
message
types.proto:375Computes the sum of the input values. Allowed input: `Int64` State: same as input
Used in:
(message has no fields)
message
types.proto:345An ordered list of elements of a given type. Values of type `Array` are stored in `Value.array_value`.
Used in:
optional element_type = 1
The type of the elements in the array. This must not be `Array`.
message
types.proto:178bool Values of type `Bool` are stored in `Value.bool_value`.
Used in:
(message has no fields)
message
types.proto:54Bytes Values of type `Bytes` are stored in `Value.bytes_value`.
Used in: ,
optional encoding = 1
The encoding to use when converting to or from lower level types.
message
types.proto:56Rules used to convert to or from lower level types.
Used in:
oneof encoding
Which encoding to use.
raw = 1
Use `Raw` encoding.
message
types.proto:62Leaves the value as-is. Sorted mode: all values are supported. Distinct mode: all values are supported.
Used in:
bool escape_nulls = 1
If set, allows NULL values to be encoded as the empty string "". The actual empty string, or any value which only contains the null byte `0x00`, has one more null byte appended.
message
types.proto:211Date Values of type `Date` are stored in `Value.date_value`.
Used in:
(message has no fields)
message
types.proto:334A protobuf enum type. Values of type `Enum` are stored in `Value.int_value`.
Used in:
string schema_bundle_id = 1
The ID of the schema bundle that this enum is defined in.
string enum_name = 2
The fully qualified name of the protobuf enum message, including package. In the format of "foo.bar.EnumMessage".
message
types.proto:182Float32 Values of type `Float32` are stored in `Value.float_value`.
Used in:
(message has no fields)
message
types.proto:186Float64 Values of type `Float64` are stored in `Value.float_value`.
Used in:
(message has no fields)
message
types.proto:135Int64 Values of type `Int64` are stored in `Value.int_value`.
Used in:
optional encoding = 1
The encoding to use when converting to or from lower level types.
message
types.proto:137Rules used to convert to or from lower level types.
Used in: ,
oneof encoding
Which encoding to use.
big_endian_bytes = 1
Use `BigEndianBytes` encoding.
ordered_code_bytes = 2
Use `OrderedCodeBytes` encoding.
message
types.proto:149Encodes the value as an 8-byte big-endian two's complement value. Sorted mode: non-negative values are supported. Distinct mode: all values are supported. Compatible with: - BigQuery `BINARY` encoding - HBase `Bytes.toBytes` - Java `ByteBuffer.putLong()` with `ByteOrder.BIG_ENDIAN`
Used in:
optional bytes_type = 1
Deprecated: ignored if set.
message
types.proto:160Encodes the value in a variable length binary format of up to 10 bytes. Values that are closer to zero use fewer bytes. Sorted mode: all values are supported. Distinct mode: all values are supported.
Used in:
(message has no fields)
message
types.proto:357A mapping of keys to values of a given type. Values of type `Map` are stored in a `Value.array_value` where each entry is another `Value.array_value` with two elements (the key and the value, in that order). Normally encoded Map values won't have repeated keys, however, clients are expected to handle the case in which they do. If the same key appears multiple times, the _last_ value takes precedence.
Used in:
optional key_type = 1
The type of a map key. Only `Bytes`, `String`, and `Int64` are allowed as key types.
optional value_type = 2
The type of the values in a map.
message
types.proto:323A protobuf message type. Values of type `Proto` are stored in `Value.bytes_value`.
Used in:
string schema_bundle_id = 1
The ID of the schema bundle that this proto is defined in.
string message_name = 2
The fully qualified name of the protobuf message, including package. In the format of "foo.bar.Message".
message
types.proto:83String Values of type `String` are stored in `Value.string_value`.
Used in:
optional encoding = 1
The encoding to use when converting to or from lower level types.
message
types.proto:85Rules used to convert to or from lower level types.
Used in:
oneof encoding
Which encoding to use.
utf8_raw = 1
Deprecated: if set, converts to an empty `utf8_bytes`.
utf8_bytes = 2
Use `Utf8Bytes` encoding.
message
types.proto:104UTF-8 encoding. Sorted mode: - All values are supported. - Code point order is preserved. Distinct mode: all values are supported. Compatible with: - BigQuery `TEXT` encoding - HBase `Bytes.toBytes` - Java `String#getBytes(StandardCharsets.UTF_8)`
Used in:
string null_escape_char = 1
Single-character escape sequence used to support NULL values. If set, allows NULL values to be encoded as the empty string "". The actual empty string, or any value where every character equals `null_escape_char`, has one more `null_escape_char` appended. If `null_escape_char` is set and does not equal the ASCII null character `0x00`, then the encoding will not support sorted mode. .
message
types.proto:87Deprecated: prefer the equivalent `Utf8Bytes`.
Used in:
(message has no fields)
message
types.proto:217A structured data value, consisting of fields which map to dynamically typed values. Values of type `Struct` are stored in `Value.array_value` where entries are in the same order and number as `field_types`.
Used in:
repeated fields = 1
The names and types of the fields in this struct.
optional encoding = 2
The encoding to use when converting to or from lower level types.
message
types.proto:229Rules used to convert to or from lower level types.
Used in:
oneof encoding
Which encoding to use.
singleton = 1
Use `Singleton` encoding.
delimited_bytes = 2
Use `DelimitedBytes` encoding.
ordered_code_bytes = 3
User `OrderedCodeBytes` encoding.
message
types.proto:250Fields are encoded independently and concatenated with a configurable `delimiter` in between. A struct with no fields defined is encoded as a single `delimiter`. Sorted mode: - Fields are encoded in sorted mode. - Encoded field values must not contain any bytes <= `delimiter[0]` - Element-wise order is preserved: `A < B` if `A[0] < B[0]`, or if `A[0] == B[0] && A[1] < B[1]`, etc. Strict prefixes sort first. Distinct mode: - Fields are encoded in distinct mode. - Encoded field values must not contain `delimiter[0]`.
Used in:
bytes delimiter = 1
Byte sequence used to delimit concatenated fields. The delimiter must contain at least 1 character and at most 50 characters.
message
types.proto:299Fields are encoded independently and concatenated with the fixed byte pair `{0x00, 0x01}` in between. Any null `(0x00)` byte in an encoded field is replaced by the fixed byte pair `{0x00, 0xFF}`. Fields that encode to the empty string "" have special handling: - If *every* field encodes to "", or if the STRUCT has no fields defined, then the STRUCT is encoded as the fixed byte pair `{0x00, 0x00}`. - Otherwise, the STRUCT only encodes until the last non-empty field, omitting any trailing empty fields. Any empty fields that aren't omitted are replaced with the fixed byte pair `{0x00, 0x00}`. Examples: ``` - STRUCT() -> "\00\00" - STRUCT("") -> "\00\00" - STRUCT("", "") -> "\00\00" - STRUCT("", "B") -> "\00\00" + "\00\01" + "B" - STRUCT("A", "") -> "A" - STRUCT("", "B", "") -> "\00\00" + "\00\01" + "B" - STRUCT("A", "", "C") -> "A" + "\00\01" + "\00\00" + "\00\01" + "C" ``` Since null bytes are always escaped, this encoding can cause size blowup for encodings like `Int64.BigEndianBytes` that are likely to produce many such bytes. Sorted mode: - Fields are encoded in sorted mode. - All values supported by the field encodings are allowed - Element-wise order is preserved: `A < B` if `A[0] < B[0]`, or if `A[0] == B[0] && A[1] < B[1]`, etc. Strict prefixes sort first. Distinct mode: - Fields are encoded in distinct mode. - All values supported by the field encodings are allowed.
Used in:
(message has no fields)
message
types.proto:232Uses the encoding of `fields[0].type` as-is. Only valid if `fields.size == 1`.
Used in:
(message has no fields)
message
types.proto:219A struct field and its type.
Used in:
string field_name = 1
The field name (optional). Fields without a `field_name` are considered anonymous and cannot be referenced by name.
optional type = 2
The type of values in this field.
message
types.proto:190Timestamp Values of type `Timestamp` are stored in `Value.timestamp_value`.
Used in:
optional encoding = 1
The encoding to use when converting to or from lower level types.
message
types.proto:192Rules used to convert to or from lower level types.
Used in:
oneof encoding
Which encoding to use.
unix_micros_int64 = 1
Encodes the number of microseconds since the Unix epoch using the given `Int64` encoding. Values must be microsecond-aligned. Compatible with: - Java `Instant.truncatedTo()` with `ChronoUnit.MICROS`
message
data.proto:97`Value` represents a dynamically typed value. The typed fields in `Value` are used as a transport encoding for the actual value (which may be of a more complex type). See the documentation of the `Type` message for more details.
Used in: , , , ,
optional type = 7
The verified `Type` of this `Value`, if it cannot be inferred. Read results will never specify the encoding for `type` since the value will already have been decoded by the server. Furthermore, the `type` will be omitted entirely if it can be inferred from a previous response. The exact semantics for inferring `type` will vary, and are therefore documented separately for each read method. When using composite types (Struct, Array, Map) only the outermost `Value` will specify the `type`. This top-level `type` will define the types for any nested `Struct' fields, `Array` elements, or `Map` key/value pairs. If a nested `Value` provides a `type` on write, the request will be rejected with INVALID_ARGUMENT.
oneof kind
Options for transporting values within the protobuf type system. A given `kind` may support more than one `type` and vice versa. On write, this is roughly analogous to a GoogleSQL literal. The value is `NULL` if none of the fields in `kind` is set. If `type` is also omitted on write, we will infer it based on the schema.
bytes raw_value = 8
Represents a raw byte sequence with no type information. The `type` field must be omitted.
int64 raw_timestamp_micros = 9
Represents a raw cell timestamp with no type information. The `type` field must be omitted.
bytes bytes_value = 2
Represents a typed value transported as a byte sequence.
string string_value = 3
Represents a typed value transported as a string.
int64 int_value = 6
Represents a typed value transported as an integer.
bool bool_value = 10
Represents a typed value transported as a boolean.
double float_value = 11
Represents a typed value transported as a floating point number. Does not support NaN or infinities.
timestamp_value = 12
Represents a typed value transported as a timestamp.
date_value = 13
Represents a typed value transported as a date.
array_value = 4
Represents a typed value transported as a sequence of values. To differentiate between `Struct`, `Array`, and `Map`, the outermost `Value` must provide an explicit `type` on write. This `type` will apply recursively to the nested `Struct` fields, `Array` elements, or `Map` key/value pairs, which *must not* supply their own `type`.
message
data.proto:236Specifies a contiguous range of raw byte values.
Used in:
oneof start_value
The value at which to start the range. If neither field is set, interpreted as the empty string, inclusive.
bytes start_value_closed = 1
Used when giving an inclusive lower bound for the range.
bytes start_value_open = 2
Used when giving an exclusive lower bound for the range.
oneof end_value
The value at which to end the range. If neither field is set, interpreted as the infinite string, exclusive.
bytes end_value_closed = 3
Used when giving an inclusive upper bound for the range.
bytes end_value_open = 4
Used when giving an exclusive upper bound for the range.