February 26, 2018


  • 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.

January 18, 2018


  • 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).

December 11, 2017


  • 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”.

October 30, 2017


  • 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.

October 3, 2017


  • firestore: beta release. See the announcement.

  • errorreporting: The existing package has been redesigned.

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

September 28, 2017


  • 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.

September 8, 2017


  • 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.

August 22, 2017


  • 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.

July 31, 2017


  • 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

June 17, 2017


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

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

  • vision: cloud.google.com/go/vision is deprecated. Use cloud.google.com/go/vision/apiv1 instead.

  • translation: now stable.

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

Code changes required from v0.9.0.

March 17, 2017

Breaking Pubsub 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.

February 14, 2017

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.

December 12, 2016

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

Also, BigQuery now supports structs. 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.

December 5, 2016

More changes to BigQuery:

  • 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.

  • The SQL types DATE, TIME and DATETIME are now supported. They correspond to the Date, Time and DateTime types in the new cloud.google.com/go/civil package.

November 17, 2016

Change to BigQuery: values from INTEGER columns will now be returned as int64, not int. This will avoid errors arising from large values on 32-bit systems.

November 8, 2016

New datastore feature: datastore now encodes your nested Go structs 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.

November 8, 2016

Breaking changes to datastore: 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 google.golang.org/api/iterator.

  • 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.

October 27, 2016

Breaking change to bigquery: NewGCSReference is now a function, not a method on Client.

New bigquery feature: Table.LoaderFrom now accepts a ReaderSource, enabling loading data into a table from a file or any io.Reader.

October 21, 2016

Breaking change to pubsub: removed pubsub.Done.

Use iterator.Done instead, where iterator is the package google.golang.org/api/iterator.

October 19, 2016

Breaking changes to cloud.google.com/go/bigquery:

  • 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

October 10, 2016

Breaking changes to cloud.google.com/go/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 google.golang.org/api/iterator.)

    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 google.golang.org/api/iterator).

October 6, 2016

Package preview/logging deleted. Use logging instead.

September 27, 2016

Logging client replaced with preview version (see below).

September 8, 2016

  • 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 cloud.google.com/go/preview/logging. This client uses gRPC as its transport layer, and supports log reading, sinks and metrics. It will replace the current client at cloud.google.com/go/logging shortly.