blob: 97a3c9aa0925262c6744408a65a6c113944511b1 [file] [log] [blame]
Copyright 2017 Google Inc. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
See the License for the specific language governing permissions and
limitations under the License.
Package spanner provides a client for reading and writing to Cloud Spanner
databases. See the packages under admin for clients that operate on databases
and instances.
Note: This package is in beta. Some backwards-incompatible changes may occur.
See for an introduction
to Cloud Spanner and additional help on using this API.
Creating a Client
To start working with this package, create a client that refers to the database
of interest:
ctx := context.Background()
client, err := spanner.NewClient(ctx, "projects/P/instances/I/databases/D")
if err != nil {
// TODO: Handle error.
defer client.Close()
Remember to close the client after use to free up the sessions in the session
Simple Reads and Writes
Two Client methods, Apply and Single, work well for simple reads and writes. As
a quick introduction, here we write a new row to the database and read it back:
_, err := client.Apply(ctx, []*spanner.Mutation{
[]string{"name", "email"},
[]interface{}{"alice", ""})})
if err != nil {
// TODO: Handle error.
row, err := client.Single().ReadRow(ctx, "Users",
spanner.Key{"alice"}, []string{"email"})
if err != nil {
// TODO: Handle error.
All the methods used above are discussed in more detail below.
Every Cloud Spanner row has a unique key, composed of one or more columns.
Construct keys with a literal of type Key:
key1 := spanner.Key{"alice"}
The keys of a Cloud Spanner table are ordered. You can specify ranges of keys
using the KeyRange type:
kr1 := spanner.KeyRange{Start: key1, End: key2}
By default, a KeyRange includes its start key but not its end key. Use
the Kind field to specify other boundary conditions:
// include both keys
kr2 := spanner.KeyRange{Start: key1, End: key2, Kind: spanner.ClosedClosed}
A KeySet represents a set of keys. A single Key or KeyRange can act as a KeySet. Use
the KeySets function to build the union of several KeySets:
ks1 := spanner.KeySets(key1, key2, kr1, kr2)
AllKeys returns a KeySet that refers to all the keys in a table:
ks2 := spanner.AllKeys()
All Cloud Spanner reads and writes occur inside transactions. There are two
types of transactions, read-only and read-write. Read-only transactions cannot
change the database, do not acquire locks, and may access either the current
database state or states in the past. Read-write transactions can read the
database before writing to it, and always apply to the most recent database
Single Reads
The simplest and fastest transaction is a ReadOnlyTransaction that supports a
single read operation. Use Client.Single to create such a transaction. You can
chain the call to Single with a call to a Read method.
When you only want one row whose key you know, use ReadRow. Provide the table
name, key, and the columns you want to read:
row, err := client.Single().ReadRow(ctx, "Accounts", spanner.Key{"alice"}, []string{"balance"})
Read multiple rows with the Read method. It takes a table name, KeySet, and list
of columns:
iter := client.Single().Read(ctx, "Accounts", keyset1, columns)
Read returns a RowIterator. You can call the Do method on the iterator and pass
a callback:
err := iter.Do(func(row *Row) error {
// TODO: use row
return nil
RowIterator also follows the standard pattern for the Google
Cloud Client Libraries:
defer iter.Stop()
for {
row, err := iter.Next()
if err == iterator.Done {
if err != nil {
// TODO: Handle error.
// TODO: use row
Always call Stop when you finish using an iterator this way, whether or not you
iterate to the end. (Failing to call Stop could lead you to exhaust the
database's session quota.)
To read rows with an index, use ReadUsingIndex.
The most general form of reading uses SQL statements. Construct a Statement
with NewStatement, setting any parameters using the Statement's Params map:
stmt := spanner.NewStatement("SELECT First, Last FROM SINGERS WHERE Last >= @start")
stmt.Params["start"] = "Dylan"
You can also construct a Statement directly with a struct literal, providing
your own map of parameters.
Use the Query method to run the statement and obtain an iterator:
iter := client.Single().Query(ctx, stmt)
Once you have a Row, via an iterator or a call to ReadRow, you can extract
column values in several ways. Pass in a pointer to a Go variable of the
appropriate type when you extract a value.
You can extract by column position or name:
err := row.Column(0, &name)
err = row.ColumnByName("balance", &balance)
You can extract all the columns at once:
err = row.Columns(&name, &balance)
Or you can define a Go struct that corresponds to your columns, and extract
into that:
var s struct { Name string; Balance int64 }
err = row.ToStruct(&s)
For Cloud Spanner columns that may contain NULL, use one of the NullXXX types,
like NullString:
var ns spanner.NullString
if err := row.Column(0, &ns); err != nil {
// TODO: Handle error.
if ns.Valid {
} else {
fmt.Println("column is NULL")
Multiple Reads
To perform more than one read in a transaction, use ReadOnlyTransaction:
txn := client.ReadOnlyTransaction()
defer txn.Close()
iter := txn.Query(ctx, stmt1)
// ...
iter = txn.Query(ctx, stmt2)
// ...
You must call Close when you are done with the transaction.
Timestamps and Timestamp Bounds
Cloud Spanner read-only transactions conceptually perform all their reads at a
single moment in time, called the transaction's read timestamp. Once a read has
started, you can call ReadOnlyTransaction's Timestamp method to obtain the read
By default, a transaction will pick the most recent time (a time where all
previously committed transactions are visible) for its reads. This provides the
freshest data, but may involve some delay. You can often get a quicker response
if you are willing to tolerate "stale" data. You can control the read timestamp
selected by a transaction by calling the WithTimestampBound method on the
transaction before using it. For example, to perform a query on data that is at
most one minute stale, use
Query(ctx, stmt)
See the documentation of TimestampBound for more details.
To write values to a Cloud Spanner database, construct a Mutation. The spanner
package has functions for inserting, updating and deleting rows. Except for the
Delete methods, which take a Key or KeyRange, each mutation-building function
comes in three varieties.
One takes lists of columns and values along with the table name:
m1 := spanner.Insert("Users",
[]string{"name", "email"},
[]interface{}{"alice", ""})
One takes a map from column names to values:
m2 := spanner.InsertMap("Users", map[string]interface{}{
"name": "alice",
"email": "",
And the third accepts a struct value, and determines the columns from the
struct field names:
type User struct { Name, Email string }
u := User{Name: "alice", Email: ""}
m3, err := spanner.InsertStruct("Users", u)
To apply a list of mutations to the database, use Apply:
_, err := client.Apply(ctx, []*spanner.Mutation{m1, m2, m3})
If you need to read before writing in a single transaction, use a
ReadWriteTransaction. ReadWriteTransactions may abort and need to be retried.
You pass in a function to ReadWriteTransaction, and the client will handle the
retries automatically. Use the transaction's BufferWrite method to buffer
mutations, which will all be executed at the end of the transaction:
_, err := client.ReadWriteTransaction(ctx, func(ctx context.Context, txn *spanner.ReadWriteTransaction) error {
var balance int64
row, err := txn.ReadRow(ctx, "Accounts", spanner.Key{"alice"}, []string{"balance"})
if err != nil {
// This function will be called again if this is an IsAborted error.
return err
if err := row.Column(0, &balance); err != nil {
return err
if balance <= 10 {
return errors.New("insufficient funds in account")
balance -= 10
m := spanner.Update("Accounts", []string{"user", "balance"}, []interface{}{"alice", balance})
// The buffered mutation will be committed. If the commit
// fails with an IsAborted error, this function will be called
// again.
return nil
This client has been instrumented to use OpenCensus tracing (
To enable tracing, see "Enabling Tracing for a Program" at OpenCensus tracing requires Go 1.8 or higher.
See examples of authorization and authentication at
package spanner // import ""