Deprecation notice

  • has been deprecated due to breaking changes in the API. Please migrate to


  • all:
    • The remaining uses of gtransport.Dial have been removed.
    • The genproto dependency has been updated to a version that makes use of new protoreflect library. For more information on these protobuf changes please see the following post from the official Go blog:
  • internal:
    • Started generation of datastore admin v1 client.
    • Updated protofuf version used for generation to 3.12.X.
    • Update the release levels for several APIs.
    • Generate clients with protoc-gen-go@v1.4.1.
  • monitoring:
    • Re-enable generation of monitoring/apiv3 under v2 directory (see deprecation notice above).
  • profiler:
    • Fixed flakiness in tests.
  • Various updates to autogenerated clients.


  • all:
    • Update module dependency to v0.21.0.
  • errorreporting:
    • Add exported SetGoogleClientInfo wrappers to manual file.
  • expr/v1alpha1:
    • Deprecate client. This client will be removed in a future release.
  • internal:
    • Fix possible data race in TestTracer.
    • Pin versions of tools used for generation.
    • Correct the release levels for BigQuery APIs.
    • Start generation osconfig v1.
  • longrunning:
    • Add exported SetGoogleClientInfo wrappers to manual file.
  • monitoring:
    • Stop generation of monitoring/apiv3 because of incoming breaking change.
  • trace:
    • Add exported SetGoogleClientInfo wrappers to manual file.
  • Various updates to autogenerated clients.


  • secretmanager:
    • add IAM helper
  • profiler:
    • try all us-west1 zones for integration tests
  • internal:
    • add config to generate webrisk v1
    • add repo and commit to buildcop invocation
    • add recaptchaenterprise v1 generation config
    • update microgenerator to v0.12.5
    • add datacatalog client
    • start generating security center settings v1beta
    • start generating osconfig agentendpoint v1
    • setup generation for bigquery/connection/v1beta1
  • all:
    • increase continous testing timeout to 45m
    • various updates to autogenerated clients.


  • Various updates to autogenerated clients.


  • all:
    • remove unused from mod file
    • update links to
  • compute/metadata:
    • use defaultClient when http.Client is nil
    • remove subscribeClient
  • iam:
    • add support for v3 policy and IAM conditions
  • Various updates to autogenerated clients.


  • all: most clients now use transport/grpc.DialPool rather than Dial (see #1777 for outliers).
    • Connection pooling now does not use the deprecated (and soon to be removed) gRPC load balancer API.
  • profiler: remove symbolization (drops support for go1.10)
  • Various updates to autogenerated clients.


  • internal/gapicgen: multiple improvements related to library generation.
  • compute/metadata: unset ResponseHeaderTimeout in defaultClient
  • docs: fix link to KMS in
  • Various updates to autogenerated clients.


  • secretmanager:
    • add IAM helper for generic resource IAM handle
  • cloudbuild:
    • migrate to microgen in a major version
  • Various updates to autogenerated clients.


  • profiler:
    • Support disabling CPU profile collection.
    • Log when a profile creation attempt begins.
  • compute/metadata:
    • Fix panic on malformed URLs.
    • InstanceName returns actual instance name.
  • Various updates to autogenerated clients.


  • functions/metadata:
    • Handle string resources in JSON unmarshaller.
  • Various updates to autogenerated clients.


  • Various updates to autogenerated clients


This release drops support for Go 1.9 and Go 1.10: we continue to officially support Go 1.11, Go 1.12, and Go 1.13.

  • Various updates to autogenerated clients.
  • Add cloudbuild/apiv1 client.


This is an empty release that was created solely to aid in storage's module carve-out. See:


This is an empty release that was created solely to aid in spanner's module carve-out. See:


This is an empty release that was created solely to aid in firestore's module carve-out. See:


  • spanner:
    • Retry “Session not found” for read-only transactions.
    • Retry aborted PDMLs.
  • spanner/spannertest:
    • Fix a bug that was causing 0X-prefixed number to be parsed incorrectly.
  • storage:
    • Add HMACKeyOptions.
    • Remove *REGIONAL from StorageClass documentation. Using MULTI_REGIONAL, DURABLE_REDUCED_AVAILABILITY, and REGIONAL are no longer best practice StorageClasses but they are still acceptable values.
  • trace:
    • Remove Package has been marked OBSOLETE for several years: it is now no longer provided. If you relied on this package, please vendor it or switch to using (which obsoleted it).


This is an empty release that was created solely to aid in pubsub's module carve-out. See:


  • compute/metadata:
    • Add Email method.
  • storage:
    • Fix duplicated retry logic.
    • Add ReaderObjectAttrs.StartOffset.
    • Support reading last N bytes of a file when a negative range is given, such as obj.NewRangeReader(ctx, -10, -1).
    • Add HMACKey listing functionality.
  • spanner/spannertest:
    • Support primary keys with no columns.
    • Fix MinInt64 parsing.
    • Implement deletion of key ranges.
    • Handle reads during a read-write transaction.
    • Handle returning DATE values.
  • pubsub:
    • Fix Ack/Modack request size calculation.
  • logging:
    • Add auto-detection of monitored resources on GAE Standard.


This is an empty release that was created solely to aid in bigtable's module carve-out. See:


This is an empty release that was created solely to aid in bigquery's module carve-out. See:


This is an empty release that was created solely to aid in datastore's module carve-out. See:


  • datastore:
    • Interface elements whose underlying types are supported, are now supported.
    • Reduce time to initial retry from 1s to 100ms.
  • firestore:
    • Add Increment transformation.
  • storage:
    • Allow emulator with STORAGE_EMULATOR_HOST.
    • Add methods for HMAC key management.
  • pubsub:
    • Add PublishCount and PublishLatency measurements.
    • Add DefaultPublishViews and DefaultSubscribeViews for convenience of importing all views.
    • Add add Subscription.PushConfig.AuthenticationMethod.
  • spanner:
    • Allow emulator usage with SPANNER_EMULATOR_HOST.
    • Add, a spanner emulator.
    • Add which contains types and a parser for the Cloud Spanner SQL dialect.
  • asset:
    • Add apiv1p2beta1 client.


This is an empty release that was created solely to aid in logging's module carve-out. See:


  • bigtable:
    • Add an admin method to update an instance and clusters.
    • Fix bttest regex matching behavior for alternations (things like |a).
    • Expose BlockAllFilter filter.
  • bigquery:
    • Add Routines API support.
  • storage:
    • Add read-only Bucket.LocationType.
  • logging:
    • Add TraceSampled to Entry.
    • Fix to properly extract {Trace, Span}Id from X-Cloud-Trace-Context.
  • pubsub:
    • Add Cloud Key Management to TopicConfig.
    • Change ExpirationPolicy to optional.Duration.
  • automl:
    • Add apiv1beta1 client.
  • iam:
    • Fix compilation problem with iam/credentials/apiv1.


  • bigtable:
    • Check results from PredicateFilter in bttest, which fixes certain false matches.
  • profiler:
    • debugLog checks user defined logging options before logging.
  • spanner:
    • PartitionedUpdates respect query parameters.
    • StartInstance allows specifying cloud API access scopes.
  • bigquery:
    • Use empty slice instead of nil for ValueSaver, fixing an issue with zero-length, repeated, nested fields causing panics.
  • firestore:
    • Return same number of snapshots as doc refs (in the form of duplicate records) during GetAll.
  • replay:
    • Change references to IPv4 addresses to localhost, making replay compatible with IPv6.


  • all:
    • Update to protobuf-golang v1.3.1.
  • datastore:
    • Attempt to decode GAE-encoded keys if initial decoding attempt fails.
    • Support integer time conversion.
  • pubsub:
    • Add PublishSettings.BundlerByteLimit. If users receive pubsub.ErrOverflow, this value should be adjusted higher.
    • Use IPv6 compatible target in testutil.
  • bigtable:
    • Fix Latin-1 regexp filters in bttest, allowing \C.
    • Expose PassAllFilter.
  • profiler:
    • Add log messages for slow path in start.
    • Fix start to allow retry until success.
  • firestore:
    • Add admin client.
  • containeranalysis:
    • Add apiv1 client.
  • grafeas:
    • Add apiv1 client.


  • bigtable:
    • Implement DeleteInstance in bttest.
    • Return an error on invalid ReadRowsRequest.RowRange key ranges in bttest.
  • bigquery:
    • Move RequirePartitionFilter outside of TimePartioning.
    • Expose models API.
  • firestore:
    • Allow array values in create and update calls.
    • Add CollectionGroup method.
  • pubsub:
    • Add ExpirationPolicy to Subscription.
  • storage:
    • Add V4 signing.
  • rpcreplay:
    • Match streams by first sent request. This further improves rpcreplay's ability to distinguish streams.
  • httpreplay:
    • Set up Man-In-The-Middle config only once. This should improve proxy creation when multiple proxies are used in a single process.
    • Remove error on empty Content-Type, allowing requests with no Content-Type header but a non-empty body.
  • all:
    • Fix an edge case bug in auto-generated library pagination by properly propagating pagetoken.


This update includes a substantial reduction in our transitive dependency list by way of updating to opencensus@v0.21.0.

  • spanner:
    • Error implements GRPCStatus, allowing status.Convert.
  • bigtable:
    • Fix a bug in bttest that prevents single column queries returning results that match other filters.
    • Remove verbose retry logging.
  • logging:
    • Ensure RequestUrl has proper UTF-8, removing the need for users to wrap and rune replace manually.
  • recaptchaenterprise:
    • Add v1beta1 client.
  • phishingprotection:
    • Add v1beta1 client.


This patch releases re-builds the go.sum. This was not possible in the previous release.

  • firestore:
    • Add sentinel value DetectProjectID for auto-detecting project ID.
    • Add OpenCensus tracing for public methods.
    • Marked stable. All future changes come with a backwards compatibility guarantee.
    • Removed firestore/apiv1beta1. All users relying on this low-level library should migrate to firestore/apiv1. Note that most users should use the high-level firestore package instead.
  • pubsub:
    • Allow large messages in synchronous pull case.
    • Cap bundler byte limit. This should prevent OOM conditions when there are a very large number of message publishes occurring.
  • storage:
    • Add ETag to BucketAttrs and ObjectAttrs.
  • datastore:
    • Removed some non-sensical OpenCensus traces.
  • webrisk:
    • Add v1 client.
  • asset:
    • Add v1 client.
  • cloudtasks:
    • Add v2 client.


This patch release removes from the transitive dependency list, resolving go get -u problems.

Note: this release intentionally has a broken go.sum. Please use v0.37.4.


This patch release is mostly intended to bring in v0.3.0 of, which fixes a GCF deployment issue.

Note: we had to-date accidentally marked Redis as stable. In this release, we've fixed it by downgrading its documentation to alpha, as it is in other languages and docs.

  • all:
    • Document context in generated libraries.


Small go.mod version bumps to bring in v0.2.0 of, which introduces a new oauth2 url.


  • spanner:
    • Add BatchDML method.
    • Reduced initial time between retries.
  • bigquery:
    • Produce better error messages for InferSchema.
    • Add logical type control for avro loads.
    • Add support for the GEOGRAPHY type.
  • datastore:
    • Add sentinel value DetectProjectID for auto-detecting project ID.
    • Allow flatten tag on struct pointers.
    • Fixed a bug that caused queries to panic with invalid queries. Instead they will now return an error.
  • profiler:
    • Add ability to override GCE zone and instance.
  • pubsub:
    • BEHAVIOR CHANGE: Refactor error code retry logic. RPCs should now more consistently retry specific error codes based on whether they're idempotent or non-idempotent.
  • httpreplay: Fixed a bug when a non-GET request had a zero-length body causing the Content-Length header to be dropped.
  • iot:
    • Add new apiv1 client.
  • securitycenter:
    • Add new apiv1 client.
  • cloudscheduler:
    • Add new apiv1 client.


  • spanner:
    • Reduce minimum retry backoff from 1s to 100ms. This makes time between retries much faster and should improve latency.
  • storage:
    • Add support for Bucket Policy Only.
  • kms:
    • Add ResourceIAM helper method.
    • Deprecate KeyRingIAM and CryptoKeyIAM. Please use ResourceIAM.
  • firestore:
    • Switch from v1beta1 API to v1 API.
    • Allow emulator with FIRESTORE_EMULATOR_HOST.
  • bigquery:
    • Add NumLongTermBytes to Table.
    • Add TotalBytesProcessedAccuracy to QueryStatistics.
  • irm:
    • Add new v1alpha2 client.
  • talent:
    • Add new v4beta1 client.
  • rpcreplay:
    • Fix connection to work with grpc >= 1.17.
    • It is now required for an actual gRPC server to be running for Dial to succeed.


  • spanner:
    • Adds OpenCensus views back to public API.


  • all:
    • Add go.mod and go.sum.
    • Switch usage of gax-go to gax-go/v2.
  • bigquery:
    • Fix bug where time partitioning could not be removed from a table.
    • Fix panic that occurred with empty query parameters.
  • bttest:
    • Fix bug where deleted rows were returned by ReadRows.
  • bigtable/emulator:
    • Configure max message size to 256 MiB.
  • firestore:
    • Allow non-transactional queries in transactions.
    • Allow StartAt/EndBefore on direct children at any depth.
    • QuerySnapshotIterator.Stop may be called in an error state.
    • Fix bug the prevented reset of transaction write state in between retries.
  • functions/metadata:
    • Make Metadata.Resource a pointer.
  • logging:
    • Make SpanID available in logging.Entry.
  • metadata:
    • Wrap !200 error code in a typed err.
  • profiler:
    • Add function to check if function name is within a particular file in the profile.
    • Set parent field in create profile request.
    • Return kubernetes client to start cluster, so client can be used to poll cluster.
    • Add function for checking if filename is in profile.
  • pubsub:
    • Fix bug where messages expired without an initial modack in synchronous=true mode.
    • Receive does not retry ResourceExhausted errors.
  • spanner:
    • client.Close now cancels existing requests and should be much faster for large amounts of sessions.
    • Correctly allow MinOpened sessions to be spun up.


  • functions/metadata:
    • Switch to using JSON in context.
    • Make Resource a value.
  • vision: Fix ProductSearch return type.
  • datastore: Add an example for how to handle MultiError.


  • compute: Removes an erroneously added go.mod.
  • logging: Populate source location in fromLogEntry.


  • bttest:
    • Add support for apply_label_transformer.
  • expr:
    • Add expr library.
  • firestore:
    • Support retrieval of missing documents.
  • kms:
    • Add IAM methods.
  • pubsub:
    • Clarify extension documentation.
  • scheduler:
    • Add v1beta1 client.
  • vision:
    • Add product search helper.
    • Add new product search client.


Note: This release is the last to support Go 1.6 and 1.8.

  • bigquery:
    • Add support for removing an expiration.
    • Ignore NeverExpire in Table.Create.
    • Validate table expiration time.
  • cbt:
    • Add note about not supporting arbitrary bytes.
  • datastore:
    • Align key checks.
  • firestore:
    • Return an error when using Start/End without providing values.
  • pubsub:
    • Add pstest Close method.
    • Clarify MaxExtension documentation.
  • securitycenter:
    • Add v1beta1 client.
  • spanner:
    • Allow nil in mutations.
    • Improve doc of SessionPoolConfig.MaxOpened.
    • Increase session deletion timeout from 5s to 15s.


  • bigtable:
    • Group mutations across multiple requests.
  • bigquery:
    • Link to bigquery troubleshooting errors page in bigquery.Error comment.
  • cbt:
    • Fix go generate command.
    • Document usage of both maxage + maxversions.
  • datastore:
    • Passing nil keys results in ErrInvalidKey.
  • firestore:
    • Clarify what Document.DataTo does with untouched struct fields.
  • profile:
    • Validate service name in agent.
  • pubsub:
    • Fix deadlock with pstest and ctx.Cancel.
    • Fix a possible deadlock in pstest.
  • trace:
    • Update doc URL with new fragment.

Special thanks to @fastest963 for going above and beyond helping us to debug hard-to-reproduce Pub/Sub issues.



  • bigtable:
    • Add retry to all idempotent RPCs.
    • cbt supports complex GC policies.
    • Emulator supports arbitrary bytes in regex filters.
  • firestore: Add ArrayUnion and ArrayRemove.
  • logging: Add the ContextFunc option to supply the context used for asynchronous RPCs.
  • profiler: Ignore NotDefinedError when fetching the instance name
  • pubsub:
    • BEHAVIOR CHANGE: Receive doesn't retry if an RPC returns codes.Cancelled.
    • BEHAVIOR CHANGE: Receive retries on Unavailable intead of returning.
    • Fix deadlock.
    • Restore Ack/Nack/Modacks metrics.
    • Improve context handling in iterator.
    • Implement synchronous mode for Receive.
    • pstest: add Pull.
  • spanner: Add a metric for the number of sessions currently opened.
  • storage:
    • Canceling the context releases all resources.
    • Add additional RetentionPolicy attributes.
  • vision/apiv1: Add LocalizeObjects method.


  • bigtable:
    • Emulator returns Unimplemented for snapshot RPCs.
  • bigquery:
    • Support zero-length repeated, nested fields.
  • cloud assets:
    • Add v1beta client.
  • datastore:
    • Don't nil out transaction ID on retry.
  • firestore:
    • BREAKING CHANGE: When watching a query with Query.Snapshots, QuerySnapshotIterator.Next returns a QuerySnapshot which contains read time, result size, change list and the DocumentIterator (previously, QuerySnapshotIterator.Next returned just the DocumentIterator). See:
    • Add array-contains operator.
  • IAM:
    • Add iam/credentials/apiv1 client.
  • pubsub:
    • Canceling the context passed to Subscription.Receive causes Receive to return when processing finishes on all messages currently in progress, even if new messages are arriving.
  • redis:
    • Add redis/apiv1 client.
  • storage:
    • Add Reader.Attrs.
    • Deprecate several Reader getter methods: please use Reader.Attrs for these instead.
    • Add ObjectHandle.Bucket and ObjectHandle.Object methods.


  • bigquery:
    • Allow modification of encryption configuration and partitioning options to a table via the Update call.
    • Add a SchemaFromJSON function that converts a JSON table schema.
  • bigtable:
    • Restore cbt count functionality.
  • containeranalysis:
    • Add v1beta client.
  • spanner:
    • Fix a case where an iterator might not be closed correctly.
  • storage:


  • bigquery:
  • bigtable:
    • Add RowSampleFilter.
    • emulator: BREAKING BEHAVIOR CHANGE: Regexps in row, family, column and value filters must match the entire target string to succeed. Previously, the emulator was succeeding on partial matches. NOTE: As of this release, this change only affects the emulator when run from this repo (bigtable/cmd/emulator/cbtemulator.go). The version launched from gcloud will be updated in a subsequent gcloud release.
  • dataproc: Add apiv1beta2 client.
  • datastore: Save non-nil pointer fields on omitempty.
  • logging: populate Entry.Trace from the HTTP X-Cloud-Trace-Context header.
  • logging/logadmin: Support writer_identity and include_children.
  • pubsub:
    • Support labels on topics and subscriptions.
    • Support message storage policy for topics.
    • Use the distribution of ack times to determine when to extend ack deadlines. The only user-visible effect of this change should be that programs that call only Subscription.Receive need no IAM permissions other than Pub/Sub Subscriber.
  • storage:
    • Support predefined ACLs.
    • Support additional ACL fields other than Entity and Role.
    • Support bucket websites.
    • Support bucket logging.


  • Added Code of Conduct
  • bigtable:
    • cbt: Support a GC policy of “never”.
  • errorreporting:
    • Support User.
    • Close now calls Flush.
    • Use OnError (previously ignored).
    • Pass through the RPC error as-is to OnError.
  • httpreplay: A tool for recording and replaying HTTP requests (for the bigquery and storage clients in this repo).
  • kms: v1 client added
  • logging: add SourceLocation to Entry.
  • storage: improve CRC checking on read.


  • bigquery: Support for the NUMERIC type.
  • bigtable:
    • cbt: Optionally specify columns for read/lookup
    • Support instance-level administration.
  • oslogin: New client for the OS Login API.
  • pubsub:
    • The package is now stable. There will be no further breaking changes.
    • Internal changes to improve Subscription.Receive behavior.
  • storage: Support updating bucket lifecycle config.
  • spanner: Support struct-typed parameter bindings.
  • texttospeech: New client for the Text-to-Speech API.


  • bigquery: Add DDL stats to query statistics.
  • bigtable:
    • cbt: Add cells-per-column limit for row lookup.
    • cbt: Make it possible to combine read filters.
  • dlp: v2beta2 client removed. Use the v2 client instead.
  • firestore, spanner: Fix compilation errors due to protobuf changes.


  • bigtable:

    • cbt: Support cells per column limit for row read.
    • bttest: Correctly handle empty RowSet.
    • Fix ReadModifyWrite operation in emulator.
    • Fix API path in GetCluster.
  • bigquery:

    • BEHAVIOR CHANGE: Retry on 503 status code.
    • Add dataset.DeleteWithContents.
    • Add SchemaUpdateOptions for query jobs.
    • Add Timeline to QueryStatistics.
    • Add more stats to ExplainQueryStage.
    • Support Parquet data format.
  • datastore:

    • Support omitempty for times.
  • dlp:

    • BREAKING CHANGE: Remove v1beta1 client. Please migrate to the v2 client, which is now out of beta.
    • Add v2 client.
  • firestore:

    • BEHAVIOR CHANGE: Treat set({}, MergeAll) as valid.
  • iam:

    • Support JWT signing via SignJwt callopt.
  • profiler:

    • BEHAVIOR CHANGE: PollForSerialOutput returns an error when context.Done.
    • BEHAVIOR CHANGE: Increase the initial backoff to 1 minute.
    • Avoid returning empty serial port output.
  • pubsub:

    • BEHAVIOR CHANGE: Don't backoff during next retryable error once stream is healthy.
    • BEHAVIOR CHANGE: Don't backoff on EOF.
    • pstest: Support Acknowledge and ModifyAckDeadline RPCs.
  • redis:

    • Add v1 beta Redis client.
  • spanner:

    • Support SessionLabels.
  • speech:

    • Add api v1 beta1 client.
  • storage:

    • BEHAVIOR CHANGE: Retry reads when retryable error occurs.
    • Fix delete of object in requester-pays bucket.
    • Support KMS integration.


  • bigquery:

    • Add OpenCensus tracing.
  • firestore:

    • BREAKING CHANGE: If a document does not exist, return a DocumentSnapshot whose Exists method returns false. DocumentRef.Get and Transaction.Get return the non-nil DocumentSnapshot in addition to a NotFound error. DocumentRef.GetAll and Transaction.GetAll return a non-nil DocumentSnapshot instead of nil.
    • Add DocumentIterator.Stop. Call Stop whenever you are done with a DocumentIterator.
    • Added Query.Snapshots and DocumentRef.Snapshots, which provide realtime notification of updates. See
    • Canceling an RPC now always returns a grpc.Status with codes.Canceled.
  • spanner:

    • Add CommitTimestamp, which supports inserting the commit timestamp of a transaction into a column.


  • bigquery: Support SchemaUpdateOptions for load jobs.

  • bigtable:

    • Add SampleRowKeys.
    • cbt: Support union, intersection GCPolicy.
    • Retry admin RPCS.
    • Add trace spans to retries.
  • datastore: Add OpenCensus tracing.

  • firestore:

    • Fix queries involving Null and NaN.
    • Allow Timestamp protobuffers for time values.
  • logging: Add a WriteTimeout option.

  • spanner: Support Batch API.

  • storage: Add OpenCensus tracing.


  • bigquery:

    • Support customer-managed encryption keys.
  • bigtable:

    • Improved emulator support.
    • Support GetCluster.
  • datastore:

    • Add general mutations.
    • Support pointer struct fields.
    • Support transaction options.
  • firestore:

    • Add Transaction.GetAll.
    • Support document cursors.
  • logging:

    • Support concurrent RPCs to the service.
    • Support per-entry resources.
  • profiler:

    • Add config options to disable heap and thread profiling.
    • Read the project ID from $GOOGLE_CLOUD_PROJECT when it's set.
  • pubsub:

    • BEHAVIOR CHANGE: Release flow control after ack/nack (instead of after the callback returns).
    • Add SubscriptionInProject.
    • Add OpenCensus instrumentation for streaming pull.
  • storage:

    • Support CORS.


  • bigquery:

    • Marked stable.
    • Schema inference of nullable fields supported.
    • Added TimePartitioning to QueryConfig.
  • firestore: Data provided to DocumentRef.Set with a Merge option can contain Delete sentinels.

  • logging: Clients can accept parent resources other than projects.

  • pubsub:

    • pubsub/pstest: A lighweight fake for pubsub. Experimental; feedback welcome.
    • Support updating more subscription metadata: AckDeadline, RetainAckedMessages and RetentionDuration.
  • oslogin/apiv1beta: New client for the Cloud OS Login API.

  • rpcreplay: A package for recording and replaying gRPC traffic.

  • spanner:

    • Add a ReadWithOptions that supports a row limit, as well as an index.
    • Support query plan and execution statistics.
    • Added OpenCensus support.
  • storage: Clarify checksum validation for gzipped files (it is not validated when the file is served uncompressed).


  • firestore BREAKING CHANGES:
    • Remove UpdateMap and UpdateStruct; rename UpdatePaths to Update. Change docref.UpdateMap(ctx, map[string]interface{}{"a.b", 1}) to docref.Update(ctx, []firestore.Update{{Path: "a.b", Value: 1}})

      Change docref.UpdateStruct(ctx, []string{"Field"}, aStruct) to docref.Update(ctx, []firestore.Update{{Path: "Field", Value: aStruct.Field}})

    • Rename MergePaths to Merge; require args to be FieldPaths

    • A value stored as an integer can be read into a floating-point field, and vice versa.

  • bigtable/cmd/cbt:
    • Support deleting a column.
    • Add regex option for row read.
  • spanner: Mark stable.
  • storage:
    • Add Reader.ContentEncoding method.
    • Fix handling of SignedURL headers.
  • bigquery:
    • If Uploader.Put is called with no rows, it returns nil without making a call.
    • Schema inference supports the “nullable” option in struct tags for non-required fields.
    • TimePartitioning supports “Field”.


  • Other bigquery changes:

    • JobIterator.Next returns *Job; removed JobInfo (BREAKING CHANGE).
    • UseStandardSQL is deprecated; set UseLegacySQL to true if you need Legacy SQL.
    • Uploader.Put will generate a random insert ID if you do not provide one.
    • Support time partitioning for load jobs.
    • Support dry-run queries.
    • A Job remembers its last retrieved status.
    • Support retrieving job configuration.
    • Support labels for jobs and tables.
    • Support dataset access lists.
    • Improve support for external data sources, including data from Bigtable and Google Sheets, and tables with external data.
    • Support updating a table's view configuration.
    • Fix uploading civil times with nanoseconds.
  • storage:

    • Support PubSub notifications.
    • Support Requester Pays buckets.
  • profiler: Support goroutine and mutex profile types.


  • firestore: beta release. See the announcement.

  • errorreporting: The existing package has been redesigned.

  • errors: This package has been removed. Use errorreporting.


  • bigquery BREAKING CHANGES:

    • Standard SQL is the default for queries and views.
    • Table.Create takes TableMetadata as a second argument, instead of options.
    • Dataset.Create takes DatasetMetadata as a second argument.
    • DatasetMetadata field ID renamed to FullID
    • TableMetadata field ID renamed to FullID
  • Other bigquery changes:

    • The client will append a random suffix to a provided job ID if you set AddJobIDSuffix to true in a job config.
    • Listing jobs is supported.
    • Better retry logic.
  • vision, language, speech: clients are now stable

  • monitoring: client is now beta

  • profiler:

    • Rename InstanceName to Instance, ZoneName to Zone
    • Auto-detect service name and version on AppEngine.


  • bigquery: UseLegacySQL options for CreateTable and QueryConfig. Use these options to continue using Legacy SQL after the client switches its default to Standard SQL.

  • bigquery: Support for updating dataset labels.

  • bigquery: Set DatasetIterator.ProjectID to list datasets in a project other than the client's. DatasetsInProject is no longer needed and is deprecated.

  • bigtable: Fail ListInstances when any zones fail.

  • spanner: support decoding of slices of basic types (e.g. []string, []int64, etc.)

  • logging/logadmin: UpdateSink no longer creates a sink if it is missing (actually a change to the underlying service, not the client)

  • profiler: Service and ServiceVersion replace Target in Config.


  • pubsub: Subscription.Receive now uses streaming pull.

  • pubsub: add Client.TopicInProject to access topics in a different project than the client.

  • errors: renamed errorreporting. The errors package will be removed shortly.

  • datastore: improved retry behavior.

  • bigquery: support updates to dataset metadata, with etags.

  • bigquery: add etag support to Table.Update (BREAKING: etag argument added).

  • bigquery: generate all job IDs on the client.

  • storage: support bucket lifecycle configurations.


  • Clients for spanner, pubsub and video are now in beta.

  • New client for DLP.

  • spanner: performance and testing improvements.

  • storage: requester-pays buckets are supported.

  • storage, profiler, bigtable, bigquery: bug fixes and other minor improvements.

  • pubsub: bug fixes and other minor improvements


  • pubsub: Subscription.ModifyPushConfig replaced with Subscription.Update.

  • pubsub: Subscription.Receive now runs concurrently for higher throughput.

  • vision: is deprecated. Use instead.

  • translation: now stable.

  • trace: several changes to the surface. See the link below.

Code changes required from v0.9.0

  • pubsub: Replace

    sub.ModifyPushConfig(ctx, pubsub.PushConfig{Endpoint: ""})


    sub.Update(ctx, pubsub.SubscriptionConfigToUpdate{
        PushConfig: &pubsub.PushConfig{Endpoint: ""},
  • trace: traceGRPCServerInterceptor will be provided from *trace.Client. Given an initialized *trace.Client named tc, instead of

    s := grpc.NewServer(grpc.UnaryInterceptor(trace.GRPCServerInterceptor(tc)))


    s := grpc.NewServer(grpc.UnaryInterceptor(tc.GRPCServerInterceptor()))
  • trace trace.GRPCClientInterceptor will also provided from *trace.Client. Instead of

    conn, err := grpc.Dial(srv.Addr, grpc.WithUnaryInterceptor(trace.GRPCClientInterceptor()))


    conn, err := grpc.Dial(srv.Addr, grpc.WithUnaryInterceptor(tc.GRPCClientInterceptor()))
  • trace: We removed the deprecated trace.EnableGRPCTracing. Use the gRPC interceptor as a dial option as shown below when initializing Cloud package clients:

    c, err := pubsub.NewClient(ctx, "project-id", option.WithGRPCDialOption(grpc.WithUnaryInterceptor(tc.GRPCClientInterceptor())))
    if err != nil {


  • Breaking changes to some autogenerated clients.
  • rpcreplay package added.


  • profiler package added.
  • storage:
    • Retry Objects.Insert call.
    • Add ProgressFunc to WRiter.
  • pubsub: breaking changes:
    • Publish is now asynchronous (announcement).
    • Subscription.Pull replaced by Subscription.Receive, which takes a callback (announcement).
    • Message.Done replaced with Message.Ack and Message.Nack.


  • Release of a client library for Spanner. See the blog post. Note that although the Spanner service is beta, the Go client library is alpha.


  • Beta release of BigQuery, DataStore, Logging and Storage. See the blog post.

  • bigquery:

    • struct support. Read a row directly into a struct with RowIterator.Next, and upload a row directly from a struct with Uploader.Put. You can also use field tags. See the [package documentation][cloud-bigquery-ref] for details.

    • The ValueList type was removed. It is no longer necessary. Instead of

    var v ValueList
    ... it.Next(&v) ..


    var v []Value
    ... it.Next(&v) ...
    • Previously, repeatedly calling RowIterator.Next on the same []Value or ValueList would append to the slice. Now each call resets the size to zero first.

    • Schema inference will infer the SQL type BYTES for a struct field of type []byte. Previously it inferred STRING.

    • The types uint, uint64 and uintptr are no longer supported in schema inference. BigQuery's integer type is INT64, and those types may hold values that are not correctly represented in a 64-bit signed integer.


  • bigquery:
    • The SQL types DATE, TIME and DATETIME are now supported. They correspond to the Date, Time and DateTime types in the new package.
    • Support for query parameters.
    • Support deleting a dataset.
    • Values from INTEGER columns will now be returned as int64, not int. This will avoid errors arising from large values on 32-bit systems.
  • datastore:
    • Nested Go structs encoded as Entity values, instead of a flattened list of the embedded struct's fields. This means that you may now have twice-nested slices, eg.
      type State struct {
        Cities  []struct{
          Populations []int
      See the announcement for more details.
    • Contexts no longer hold namespaces; instead you must set a key's namespace explicitly. Also, key functions have been changed and renamed.
    • The WithNamespace function has been removed. To specify a namespace in a Query, use the Query.Namespace method:
      q := datastore.NewQuery("Kind").Namespace("ns")
    • All the fields of Key are exported. That means you can construct any Key with a struct literal:
      k := &Key{Kind: "Kind",  ID: 37, Namespace: "ns"}
    • As a result of the above, the Key methods Kind, ID, d.Name, Parent, SetParent and Namespace have been removed.
    • NewIncompleteKey has been removed, replaced by IncompleteKey. Replace
      NewIncompleteKey(ctx, kind, parent)
      IncompleteKey(kind, parent)
      and if you do use namespaces, make sure you set the namespace on the returned key.
    • NewKey has been removed, replaced by NameKey and IDKey. Replace
      NewKey(ctx, kind, name, 0, parent)
      NewKey(ctx, kind, "", id, parent)
      NameKey(kind, name, parent)
      IDKey(kind, id, parent)
      and if you do use namespaces, make sure you set the namespace on the returned key.
    • The Done variable has been removed. Replace datastore.Done with iterator.Done, from the package
    • The Client.Close method will have a return type of error. It will return the result of closing the underlying gRPC connection.
    • See the announcement for more details.


  • bigquery: -NewGCSReference is now a function, not a method on Client.

    • Table.LoaderFrom now accepts a ReaderSource, enabling loading data into a table from a file or any io.Reader.
    • Client.Table and Client.OpenTable have been removed. Replace

      client.OpenTable("project", "dataset", "table")


      client.DatasetInProject("project", "dataset").Table("table")
    • Client.CreateTable has been removed. Replace

      client.CreateTable(ctx, "project", "dataset", "table")


      client.DatasetInProject("project", "dataset").Table("table").Create(ctx)
    • Dataset.ListTables have been replaced with Dataset.Tables. Replace

      tables, err := ds.ListTables(ctx)


      it := ds.Tables(ctx)
      for {
          table, err := it.Next()
          if err == iterator.Done {
          if err != nil {
              // TODO: Handle error.
          // TODO: use table.
    • Client.Read has been replaced with Job.Read, Table.Read and Query.Read. Replace

      it, err := client.Read(ctx, job)


      it, err := job.Read(ctx)

      and similarly for reading from tables or queries.

    • The iterator returned from the Read methods is now named RowIterator. Its behavior is closer to the other iterators in these libraries. It no longer supports the Schema method; see the next item. Replace

      for it.Next(ctx) {
          var vals ValueList
          if err := it.Get(&vals); err != nil {
              // TODO: Handle error.
          // TODO: use vals.
      if err := it.Err(); err != nil {
          // TODO: Handle error.


      for {
          var vals ValueList
          err := it.Next(&vals)
          if err == iterator.Done {
          if err != nil {
              // TODO: Handle error.
          // TODO: use vals.

      Instead of the RecordsPerRequest(n) option, write

      it.PageInfo().MaxSize = n

      Instead of the StartIndex(i) option, write

      it.StartIndex = i
    • ValueLoader.Load now takes a Schema in addition to a slice of Values. Replace

      func (vl *myValueLoader) Load(v []bigquery.Value)


      func (vl *myValueLoader) Load(v []bigquery.Value, s bigquery.Schema)
    • Table.Patch is replace by Table.Update. Replace

      p := table.Patch()
      p.Description("new description")
      metadata, err := p.Apply(ctx)


      metadata, err := table.Update(ctx, bigquery.TableMetadataToUpdate{
          Description: "new description",
    • Client.Copy is replaced by separate methods for each of its four functions. All options have been replaced by struct fields.

      • To load data from Google Cloud Storage into a table, use Table.LoaderFrom.


        client.Copy(ctx, table, gcsRef)



        Instead of passing options to Copy, set fields on the Loader:

        loader := table.LoaderFrom(gcsRef)
        loader.WriteDisposition = bigquery.WriteTruncate
      • To extract data from a table into Google Cloud Storage, use Table.ExtractorTo. Set fields on the returned Extractor instead of passing options.


        client.Copy(ctx, gcsRef, table)


      • To copy data into a table from one or more other tables, use Table.CopierFrom. Set fields on the returned Copier instead of passing options.


        client.Copy(ctx, dstTable, srcTable)


      • To start a query job, create a Query and call its Run method. Set fields on the query instead of passing options.


        client.Copy(ctx, table, query)


    • Table.NewUploader has been renamed to Table.Uploader. Instead of options, configure an Uploader by setting its fields. Replace

      u := table.NewUploader(bigquery.UploadIgnoreUnknownValues())


      u := table.NewUploader(bigquery.UploadIgnoreUnknownValues())
      u.IgnoreUnknownValues = true
  • pubsub: remove pubsub.Done. Use iterator.Done instead, where iterator is the package


  • storage:

    • AdminClient replaced by methods on Client. Replace

      adminClient.CreateBucket(ctx, bucketName, attrs)


      client.Bucket(bucketName).Create(ctx, projectID, attrs)
    • BucketHandle.List replaced by BucketHandle.Objects. Replace

      for query != nil {
          objs, err := bucket.List(d.ctx, query)
          if err != nil { ... }
          query = objs.Next
          for _, obj := range objs.Results {


      iter := bucket.Objects(d.ctx, query)
      for {
          obj, err := iter.Next()
          if err == iterator.Done {
          if err != nil { ... }

      (The iterator package is at

      Replace Query.Cursor with ObjectIterator.PageInfo().Token.

      Replace Query.MaxResults with ObjectIterator.PageInfo().MaxSize.

    • ObjectHandle.CopyTo replaced by ObjectHandle.CopierFrom. Replace

      attrs, err := src.CopyTo(ctx, dst, nil)


      attrs, err := dst.CopierFrom(src).Run(ctx)


      attrs, err := src.CopyTo(ctx, dst, &storage.ObjectAttrs{ContextType: "text/html"})


      c := dst.CopierFrom(src)
      c.ContextType = "text/html"
      attrs, err := c.Run(ctx)
    • ObjectHandle.ComposeFrom replaced by ObjectHandle.ComposerFrom. Replace

      attrs, err := dst.ComposeFrom(ctx, []*storage.ObjectHandle{src1, src2}, nil)


      attrs, err := dst.ComposerFrom(src1, src2).Run(ctx)
    • ObjectHandle.Update's ObjectAttrs argument replaced by ObjectAttrsToUpdate. Replace

      attrs, err := obj.Update(ctx, &storage.ObjectAttrs{ContextType: "text/html"})


      attrs, err := obj.Update(ctx, storage.ObjectAttrsToUpdate{ContextType: "text/html"})
    • ObjectHandle.WithConditions replaced by ObjectHandle.If. Replace

      obj.WithConditions(storage.Generation(gen), storage.IfMetaGenerationMatch(mgen))


      obj.Generation(gen).If(storage.Conditions{MetagenerationMatch: mgen})




      obj.If(storage.Conditions{DoesNotExist: true})
    • storage.Done replaced by iterator.Done (from package

  • Package preview/logging deleted. Use logging instead.


  • Logging client replaced with preview version (see below).

  • New clients for some of Google's Machine Learning APIs: Vision, Speech, and Natural Language.

  • Preview version of a new [Stackdriver Logging][cloud-logging] client in This client uses gRPC as its transport layer, and supports log reading, sinks and metrics. It will replace the current client at shortly.