Transactions overview

This page explains transactions in Spanner and includes sample code for executing transactions.

PostgreSQL interface note: The examples in this topic are intended for GoogleSQL-dialect databases.

Introduction

A transaction in Spanner is a set of reads and writes that execute atomically at a single logical point in time across columns, rows, and tables in a database.

Spanner supports these transaction modes:

• Locking read-write. These transactions rely on pessimistic locking and, if necessary, two-phase commit. Locking read-write transactions may abort, requiring the application to retry.

• Read-only. This transaction type provides guaranteed consistency across several reads, but does not allow writes. By default, read-only transactions execute at a system-chosen timestamp that guarantees external consistency, but they can also be configured to read at a timestamp in the past. Read-only transactions do not need to be committed and do not take locks. In addition, read-only transactions might wait for in-progress writes to complete before executing.

• Partitioned DML. This transaction type executes a Data Manipulation Language (DML) statement as Partitioned DML. Partitioned DML is designed for bulk updates and deletes, particularly periodic cleanup and backfilling. If you need to commit a large number of blind writes, but don't require an atomic transaction, you can bulk modify your Spanner tables using batch write. For more information, see Modify data using batch writes.

This page describes the general properties and semantics of transactions in Spanner and introduces the read-write, read-only, and Partitioned DML transaction interfaces in Spanner.

Note: Because read-only transactions can run at any replica, we recommend that you perform all of your transaction reads in read-only transactions if possible. Use locking read-write transactions only under the scenarios described in the next section.

Read-write transactions

Here are scenarios in which you should use a locking read-write transaction:

• If you do a write that depends on the result of one or more reads, you should do that write and the read(s) in the same read-write transaction.
  
  • Example: double the balance of bank account A. The read of A's balance should be in the same transaction as the write to replace the balance with the doubled value.
  
  
• If you do one or more writes that need to be committed atomically, you should do those writes in the same read-write transaction.
  • Example: transfer $200 from account A to account B. Both of the writes (one to decrease A by $200 and one to increase B by $200) and the reads of initial account balances should be in the same transaction.
  
  
• If you might do one or more writes, depending on the results of one or more reads, you should do those writes and reads in the same read-write transaction, even if the write(s) don't end up executing.
  
  • Example: transfer $200 from bank account A to bank account B if A's current balance is greater than $500. Your transaction should contain a read of A's balance and a conditional statement that contains the writes.

Here is a scenario in which you should not use a locking read-write transaction:

• If you are only doing reads, and you can express your read using a single read method, you should use that single read method or a read-only transaction. Single reads do not lock, unlike read-write transactions.

Properties

A read-write transaction in Spanner executes a set of reads and writes atomically at a single logical point in time. Furthermore, the timestamp at which read-write transactions execute matches wall clock time, and the serialization order matches the timestamp order.

Why use a read-write transaction? Read-write transactions provide the ACID properties of relational databases (In fact, Spanner read-write transactions offer even stronger guarantees than traditional ACID; see the Semantics section below.).

Isolation

The following are isolation properties for read-write and read-only transactions.

Transactions that read and write

Here are the isolation properties that you get after successfully committing a transaction that contains a series of reads (or queries) and writes:

• All reads within the transaction returned values that reflect a consistent snapshot taken at the commit timestamp of the transaction.
• Empty rows or ranges remained so at commit time.
• All writes within the transaction were committed at the transaction's commit timestamp.
• Writes were not visible to any transaction until after the transaction committed.

Certain Spanner client drivers contain transaction retry logic to mask transient errors, which they do by re-running the transaction and validating the client-observed data.

The effect is that all reads and writes appear to have occurred at a single point in time, both from the perspective of the transaction itself and from the perspective of other readers and writers to the Spanner database. In other words, the reads and the writes end up occurring at the same timestamp (see an illustration of this in the Serializability and external consistency section below).

Transactions that only read

The guarantees for a read-write transaction that only reads are similar: all reads within that transaction return data from the same timestamp, even for row non-existence. One difference is that if you read data, and later commit the read-write transaction without any writes, there is no guarantee that the data did not change in the database after the read and before the commit. If you want to know whether data has changed since you read it last, the best approach is to read it again (either in a read-write transaction, or using a strong read.) Also, for efficiency, if you know in advance that you'll only be reading and not writing, you should use a read-only transaction instead of a read-write transaction.

Atomicity, Consistency, Durability

In addition to the Isolation property, Spanner provides Atomicity (if any of the writes in the transaction commit, they all commit), Consistency (the database remains in a consistent state after the transaction) and Durability (committed data stays committed.)

Benefits of these properties

Because of these properties, as an application developer, you can focus on the correctness of each transaction on its own, without worrying about how to protect its execution from other transactions that might execute at the same time.

Interface

The Spanner client libraries provide an interface for executing a body of work in the context of a read-write transaction, with retries for transaction aborts. Here's a bit of context to explain this point: a Spanner transaction may have to be tried multiple times before it commits. For example, if two transactions attempt to work on data at the same time in a way that might cause deadlock, Spanner aborts one of them so that the other transaction can make progress. (More rarely, transient events within Spanner may result in some transactions aborting.) Since transactions are atomic, an aborted transaction has no visible effect on the database. Therefore, transactions should be executed by retrying them until they succeed.

When you use a transaction in a Spanner client library, you define the body of a transaction (i.e., the reads and writes to perform on one or more tables in a database) in the form of a function object. Under the hood, the Spanner client library runs the function repeatedly until the transaction commits or a non-retryable error is encountered.

Example

Assume you added a MarketingBudget column to the Albums table shown in the Schema and Data Model page:

CREATE TABLE Albums (
  SingerId        INT64 NOT NULL,
  AlbumId         INT64 NOT NULL,
  AlbumTitle      STRING(MAX),
  MarketingBudget INT64
) PRIMARY KEY (SingerId, AlbumId);

Your marketing department decides to do a marketing push for the album keyed by Albums (1, 1) and has asked you to move $200,000 from the budget of Albums (2, 2), but only if the money is available in that album's budget. You should use a locking read-write transaction for this operation, because the transaction might do writes depending on the result of a read.

The following shows how to execute a read-write transaction:

C++

void ReadWriteTransaction(google::cloud::spanner::Client client) {
  namespace spanner = ::google::cloud::spanner;
  using ::google::cloud::StatusOr;

  // A helper to read a single album MarketingBudget.
  auto get_current_budget =
      [](spanner::Client client, spanner::Transaction txn,
         std::int64_t singer_id,
         std::int64_t album_id) -> StatusOr<std::int64_t> {
    auto key = spanner::KeySet().AddKey(spanner::MakeKey(singer_id, album_id));
    auto rows = client.Read(std::move(txn), "Albums", std::move(key),
                            {"MarketingBudget"});
    using RowType = std::tuple<std::int64_t>;
    auto row = spanner::GetSingularRow(spanner::StreamOf<RowType>(rows));
    if (!row) return std::move(row).status();
    return std::get<0>(*std::move(row));
  };

  auto commit = client.Commit(
      [&client, &get_current_budget](
          spanner::Transaction const& txn) -> StatusOr<spanner::Mutations> {
        auto b1 = get_current_budget(client, txn, 1, 1);
        if (!b1) return std::move(b1).status();
        auto b2 = get_current_budget(client, txn, 2, 2);
        if (!b2) return std::move(b2).status();
        std::int64_t transfer_amount = 200000;

        return spanner::Mutations{
            spanner::UpdateMutationBuilder(
                "Albums", {"SingerId", "AlbumId", "MarketingBudget"})
                .EmplaceRow(1, 1, *b1 + transfer_amount)
                .EmplaceRow(2, 2, *b2 - transfer_amount)
                .Build()};
      });

  if (!commit) throw std::move(commit).status();
  std::cout << "Transfer was successful [spanner_read_write_transaction]\n";
}

C#

using Google.Cloud.Spanner.Data;
using System;
using System.Threading.Tasks;
using System.Transactions;

public class ReadWriteWithTransactionAsyncSample
{
    public async Task<int> ReadWriteWithTransactionAsync(string projectId, string instanceId, string databaseId)
    {
        // This sample transfers 200,000 from the MarketingBudget
        // field of the second Album to the first Album. Make sure to run
        // the Add Column and Write Data To New Column samples first,
        // in that order.

        string connectionString = $"Data Source=projects/{projectId}/instances/{instanceId}/databases/{databaseId}";

        using TransactionScope scope = new TransactionScope(TransactionScopeAsyncFlowOption.Enabled);
        decimal transferAmount = 200000;
        decimal secondBudget = 0;
        decimal firstBudget = 0;

        using var connection = new SpannerConnection(connectionString);
        using var cmdLookup1 = connection.CreateSelectCommand("SELECT * FROM Albums WHERE SingerId = 2 AND AlbumId = 2");

        using (var reader = await cmdLookup1.ExecuteReaderAsync())
        {
            while (await reader.ReadAsync())
            {
                // Read the second album's budget.
                secondBudget = reader.GetFieldValue<decimal>("MarketingBudget");
                // Confirm second Album's budget is sufficient and
                // if not raise an exception. Raising an exception
                // will automatically roll back the transaction.
                if (secondBudget < transferAmount)
                {
                    throw new Exception($"The second album's budget {secondBudget} is less than the amount to transfer.");
                }
            }
        }

        // Read the first album's budget.
        using var cmdLookup2 = connection.CreateSelectCommand("SELECT * FROM Albums WHERE SingerId = 1 and AlbumId = 1");
        using (var reader = await cmdLookup2.ExecuteReaderAsync())
        {
            while (await reader.ReadAsync())
            {
                firstBudget = reader.GetFieldValue<decimal>("MarketingBudget");
            }
        }

        // Specify update command parameters.
        using var cmdUpdate = connection.CreateUpdateCommand("Albums", new SpannerParameterCollection
        {
            { "SingerId", SpannerDbType.Int64 },
            { "AlbumId", SpannerDbType.Int64 },
            { "MarketingBudget", SpannerDbType.Int64 },
        });

        // Update second album to remove the transfer amount.
        secondBudget -= transferAmount;
        cmdUpdate.Parameters["SingerId"].Value = 2;
        cmdUpdate.Parameters["AlbumId"].Value = 2;
        cmdUpdate.Parameters["MarketingBudget"].Value = secondBudget;
        var rowCount = await cmdUpdate.ExecuteNonQueryAsync();

        // Update first album to add the transfer amount.
        firstBudget += transferAmount;
        cmdUpdate.Parameters["SingerId"].Value = 1;
        cmdUpdate.Parameters["AlbumId"].Value = 1;
        cmdUpdate.Parameters["MarketingBudget"].Value = firstBudget;
        rowCount += await cmdUpdate.ExecuteNonQueryAsync();
        scope.Complete();
        Console.WriteLine("Transaction complete.");
        return rowCount;
    }
}

Go

import (
	"context"
	"fmt"
	"io"

	"cloud.google.com/go/spanner"
)

func writeWithTransaction(w io.Writer, db string) error {
	ctx := context.Background()
	client, err := spanner.NewClient(ctx, db)
	if err != nil {
		return err
	}
	defer client.Close()

	_, err = client.ReadWriteTransaction(ctx, func(ctx context.Context, txn *spanner.ReadWriteTransaction) error {
		getBudget := func(key spanner.Key) (int64, error) {
			row, err := txn.ReadRow(ctx, "Albums", key, []string{"MarketingBudget"})
			if err != nil {
				return 0, err
			}
			var budget int64
			if err := row.Column(0, &budget); err != nil {
				return 0, err
			}
			return budget, nil
		}
		album2Budget, err := getBudget(spanner.Key{2, 2})
		if err != nil {
			return err
		}
		const transferAmt = 200000
		if album2Budget >= transferAmt {
			album1Budget, err := getBudget(spanner.Key{1, 1})
			if err != nil {
				return err
			}
			album1Budget += transferAmt
			album2Budget -= transferAmt
			cols := []string{"SingerId", "AlbumId", "MarketingBudget"}
			txn.BufferWrite([]*spanner.Mutation{
				spanner.Update("Albums", cols, []interface{}{1, 1, album1Budget}),
				spanner.Update("Albums", cols, []interface{}{2, 2, album2Budget}),
			})
			fmt.Fprintf(w, "Moved %d from Album2's MarketingBudget to Album1's.", transferAmt)
		}
		return nil
	})
	return err
}

Java

static void writeWithTransaction(DatabaseClient dbClient) {
  dbClient
      .readWriteTransaction()
      .run(transaction -> {
        // Transfer marketing budget from one album to another. We do it in a transaction to
        // ensure that the transfer is atomic.
        Struct row =
            transaction.readRow("Albums", Key.of(2, 2), Arrays.asList("MarketingBudget"));
        long album2Budget = row.getLong(0);
        // Transaction will only be committed if this condition still holds at the time of
        // commit. Otherwise it will be aborted and the callable will be rerun by the
        // client library.
        long transfer = 200000;
        if (album2Budget >= transfer) {
          long album1Budget =
              transaction
                  .readRow("Albums", Key.of(1, 1), Arrays.asList("MarketingBudget"))
                  .getLong(0);
          album1Budget += transfer;
          album2Budget -= transfer;
          transaction.buffer(
              Mutation.newUpdateBuilder("Albums")
                  .set("SingerId")
                  .to(1)
                  .set("AlbumId")
                  .to(1)
                  .set("MarketingBudget")
                  .to(album1Budget)
                  .build());
          transaction.buffer(
              Mutation.newUpdateBuilder("Albums")
                  .set("SingerId")
                  .to(2)
                  .set("AlbumId")
                  .to(2)
                  .set("MarketingBudget")
                  .to(album2Budget)
                  .build());
        }
        return null;
      });
}

Node.js

// This sample transfers 200,000 from the MarketingBudget field
// of the second Album to the first Album, as long as the second
// Album has enough money in its budget. Make sure to run the
// addColumn and updateData samples first (in that order).

// Imports the Google Cloud client library
const {Spanner} = require('@google-cloud/spanner');

/**
 * TODO(developer): Uncomment the following lines before running the sample.
 */
// const projectId = 'my-project-id';
// const instanceId = 'my-instance';
// const databaseId = 'my-database';

// Creates a client
const spanner = new Spanner({
  projectId: projectId,
});

// Gets a reference to a Cloud Spanner instance and database
const instance = spanner.instance(instanceId);
const database = instance.database(databaseId);

const transferAmount = 200000;

database.runTransaction(async (err, transaction) => {
  if (err) {
    console.error(err);
    return;
  }
  let firstBudget, secondBudget;
  const queryOne = {
    columns: ['MarketingBudget'],
    keys: [[2, 2]], // SingerId: 2, AlbumId: 2
  };

  const queryTwo = {
    columns: ['MarketingBudget'],
    keys: [[1, 1]], // SingerId: 1, AlbumId: 1
  };

  Promise.all([
    // Reads the second album's budget
    transaction.read('Albums', queryOne).then(results => {
      // Gets second album's budget
      const rows = results[0].map(row => row.toJSON());
      secondBudget = rows[0].MarketingBudget;
      console.log(`The second album's marketing budget: ${secondBudget}`);

      // Makes sure the second album's budget is large enough
      if (secondBudget < transferAmount) {
        throw new Error(
          `The second album's budget (${secondBudget}) is less than the transfer amount (${transferAmount}).`
        );
      }
    }),

    // Reads the first album's budget
    transaction.read('Albums', queryTwo).then(results => {
      // Gets first album's budget
      const rows = results[0].map(row => row.toJSON());
      firstBudget = rows[0].MarketingBudget;
      console.log(`The first album's marketing budget: ${firstBudget}`);
    }),
  ])
    .then(() => {
      console.log(firstBudget, secondBudget);
      // Transfers the budgets between the albums
      firstBudget += transferAmount;
      secondBudget -= transferAmount;

      console.log(firstBudget, secondBudget);

      // Updates the database
      // Note: Cloud Spanner interprets Node.js numbers as FLOAT64s, so they
      // must be converted (back) to strings before being inserted as INT64s.
      transaction.update('Albums', [
        {
          SingerId: '1',
          AlbumId: '1',
          MarketingBudget: firstBudget.toString(),
        },
        {
          SingerId: '2',
          AlbumId: '2',
          MarketingBudget: secondBudget.toString(),
        },
      ]);
    })
    .then(() => {
      // Commits the transaction and send the changes to the database
      return transaction.commit();
    })
    .then(() => {
      console.log(
        `Successfully executed read-write transaction to transfer ${transferAmount} from Album 2 to Album 1.`
      );
    })
    .catch(err => {
      console.error('ERROR:', err);
    })
    .then(() => {
      transaction.end();
      // Closes the database when finished
      return database.close();
    });
});

PHP

use Google\Cloud\Spanner\SpannerClient;
use Google\Cloud\Spanner\Transaction;
use UnexpectedValueException;

/**
 * Performs a read-write transaction to update two sample records in the
 * database.
 *
 * This will transfer 200,000 from the `MarketingBudget` field for the second
 * Album to the first Album. If the `MarketingBudget` for the second Album is
 * too low, it will raise an exception.
 *
 * Before running this sample, you will need to run the `update_data` sample
 * to populate the fields.
 * Example:
 * ```
 * read_write_transaction($instanceId, $databaseId);
 * ```
 *
 * @param string $instanceId The Spanner instance ID.
 * @param string $databaseId The Spanner database ID.
 */
function read_write_transaction(string $instanceId, string $databaseId): void
{
    $spanner = new SpannerClient();
    $instance = $spanner->instance($instanceId);
    $database = $instance->database($databaseId);

    $database->runTransaction(function (Transaction $t) use ($spanner) {
        $transferAmount = 200000;

        // Read the second album's budget.
        $secondAlbumKey = [2, 2];
        $secondAlbumKeySet = $spanner->keySet(['keys' => [$secondAlbumKey]]);
        $secondAlbumResult = $t->read(
            'Albums',
            $secondAlbumKeySet,
            ['MarketingBudget'],
            ['limit' => 1]
        );

        $firstRow = $secondAlbumResult->rows()->current();
        $secondAlbumBudget = $firstRow['MarketingBudget'];
        if ($secondAlbumBudget < $transferAmount) {
            // Throwing an exception will automatically roll back the transaction.
            throw new UnexpectedValueException(
                'The second album\'s budget is lower than the transfer amount: ' . $transferAmount
            );
        }

        $firstAlbumKey = [1, 1];
        $firstAlbumKeySet = $spanner->keySet(['keys' => [$firstAlbumKey]]);
        $firstAlbumResult = $t->read(
            'Albums',
            $firstAlbumKeySet,
            ['MarketingBudget'],
            ['limit' => 1]
        );

        // Read the first album's budget.
        $firstRow = $firstAlbumResult->rows()->current();
        $firstAlbumBudget = $firstRow['MarketingBudget'];

        // Update the budgets.
        $secondAlbumBudget -= $transferAmount;
        $firstAlbumBudget += $transferAmount;
        printf('Setting first album\'s budget to %s and the second album\'s ' .
            'budget to %s.' . PHP_EOL, $firstAlbumBudget, $secondAlbumBudget);

        // Update the rows.
        $t->updateBatch('Albums', [
            ['SingerId' => 1, 'AlbumId' => 1, 'MarketingBudget' => $firstAlbumBudget],
            ['SingerId' => 2, 'AlbumId' => 2, 'MarketingBudget' => $secondAlbumBudget],
        ]);

        // Commit the transaction!
        $t->commit();

        print('Transaction complete.' . PHP_EOL);
    });
}

Python

def read_write_transaction(instance_id, database_id):
    """Performs a read-write transaction to update two sample records in the
    database.

    This will transfer 200,000 from the `MarketingBudget` field for the second
    Album to the first Album. If the `MarketingBudget` is too low, it will
    raise an exception.

    Before running this sample, you will need to run the `update_data` sample
    to populate the fields.
    """
    spanner_client = spanner.Client()
    instance = spanner_client.instance(instance_id)
    database = instance.database(database_id)

    def update_albums(transaction):
        # Read the second album budget.
        second_album_keyset = spanner.KeySet(keys=[(2, 2)])
        second_album_result = transaction.read(
            table="Albums",
            columns=("MarketingBudget",),
            keyset=second_album_keyset,
            limit=1,
        )
        second_album_row = list(second_album_result)[0]
        second_album_budget = second_album_row[0]

        transfer_amount = 200000

        if second_album_budget < transfer_amount:
            # Raising an exception will automatically roll back the
            # transaction.
            raise ValueError("The second album doesn't have enough funds to transfer")

        # Read the first album's budget.
        first_album_keyset = spanner.KeySet(keys=[(1, 1)])
        first_album_result = transaction.read(
            table="Albums",
            columns=("MarketingBudget",),
            keyset=first_album_keyset,
            limit=1,
        )
        first_album_row = list(first_album_result)[0]
        first_album_budget = first_album_row[0]

        # Update the budgets.
        second_album_budget -= transfer_amount
        first_album_budget += transfer_amount
        print(
            "Setting first album's budget to {} and the second album's "
            "budget to {}.".format(first_album_budget, second_album_budget)
        )

        # Update the rows.
        transaction.update(
            table="Albums",
            columns=("SingerId", "AlbumId", "MarketingBudget"),
            values=[(1, 1, first_album_budget), (2, 2, second_album_budget)],
        )

    database.run_in_transaction(update_albums)

    print("Transaction complete.")

Ruby

# project_id  = "Your Google Cloud project ID"
# instance_id = "Your Spanner instance ID"
# database_id = "Your Spanner database ID"

require "google/cloud/spanner"

spanner         = Google::Cloud::Spanner.new project: project_id
client          = spanner.client instance_id, database_id
transfer_amount = 200_000

client.transaction do |transaction|
  first_album  = transaction.read("Albums", [:MarketingBudget], keys: [[1, 1]]).rows.first
  second_album = transaction.read("Albums", [:MarketingBudget], keys: [[2, 2]]).rows.first

  raise "The second album does not have enough funds to transfer" if second_album[:MarketingBudget] < transfer_amount

  new_first_album_budget  = first_album[:MarketingBudget] + transfer_amount
  new_second_album_budget = second_album[:MarketingBudget] - transfer_amount

  transaction.update "Albums", [
    { SingerId: 1, AlbumId: 1, MarketingBudget: new_first_album_budget  },
    { SingerId: 2, AlbumId: 2, MarketingBudget: new_second_album_budget }
  ]
end

puts "Transaction complete"

Semantics

Serializability and external consistency

Spanner provides 'serializability', which means that all transactions appear as if they executed in a serial order, even if some of the reads, writes, and other operations of distinct transactions actually occurred in parallel. Spanner assigns commit timestamps that reflect the order of committed transactions to implement this property. In fact, Spanner offers a stronger guarantee than serializability called external consistency: transactions commit in an order that is reflected in their commit timestamps, and these commit timestamps reflect real time so you can compare them to your watch. Reads in a transaction see everything that has been committed before the transaction commits, and writes are seen by everything that starts after the transaction is committed.

For example, consider the execution of two transactions as illustrated in the diagram below:

Transaction Txn1 in blue reads some data A, buffers a write to A, then successfully commits. Transaction Txn2 in green starts after Txn1, reads some data B, then reads the data A. Since Txn2 reads the value of A after Txn1 committed its write to A, Txn2 sees the effect of Txn1's write to A, even though Txn2 started before Txn1 completed.

Even though there is some overlap in time in which Txn1 and Txn2 are both executing, their commit timestamps c1 and c2 respect a linear transaction order, which means that all effects of the reads and writes of Txn1 appear to have occurred at a single point of time (c1), and all effects of the reads and writes of Txn2 appear to have occurred at a single point of time (c2). Furthermore, c1 < c2 (which is guaranteed because both Txn1 and Txn2 committed writes; this is true even if the writes happened on different machines), which respects the order of Txn1 happening before Txn2. (However, if Txn2 only did reads in the transaction, then c1 <= c2).

Reads observe a prefix of the commit history; if a read sees the effect of Txn2, it also sees the effect of Txn1. All transactions that commit successfully have this property.

Note: Changes you make using DML statements are visible to subsequent reads statements in the same transaction. This is different from using mutations, where changes are not visible until the transaction commits. This is because mutations in a transaction are buffered locally and are not sent to the server until commit time.

Read and write guarantees

If a call to run a transaction fails, the read and write guarantees you have depend on what error the underlying commit call failed with.

For example, an error such as "Row Not Found" or "Row Already Exists" means that writing the buffered mutations encountered some error, e.g. a row that the client is trying to update doesn't exist. In that case, the reads are guaranteed consistent, the writes are not applied, and the non-existence of the row is guaranteed to be consistent with the reads as well.

Note: Spanner might try to execute the body of the transaction multiple times under the hood. If an execution attempt fails, the error it returns indicates the condition(s) that occurred and thus, which guarantees you get. However, be aware that any side effects of your transaction body (such as changes to non-Spanner state in your program or other systems) can occur multiple times if Spanner has to retry your transaction multiple times.

Cancelling transaction operations

Asynchronous read operations may be cancelled any time by the user (e.g., when a higher level operation is cancelled or you decide to stop a read based on the initial results received from the read) without affecting any other existing operations within the transaction.

However, even if you have attempted to cancel the read, Spanner does not guarantee that the read is actually cancelled. After you request the cancellation of a read, that read can still successfully complete or fail with some other reason (e.g. Abort). Furthermore, that cancelled read might actually return some results to you, and those possibly incomplete results will be validated as part of the transaction Commit.

Note that unlike reads, cancelling a transaction Commit operation will result in aborting the transaction (unless the transaction has already Committed or failed with another reason).

Performance

Locking

Spanner allows multiple clients to simultaneously interact with the same database. In order to ensure the consistency of multiple concurrent transactions, Spanner uses a combination of shared locks and exclusive locks to control access to the data. When you perform a read as part of a transaction, Spanner acquires shared read locks, which allows other reads to still access the data until your transaction is ready to commit. When your transaction is committing and writes are being applied, the transaction attempts to upgrade to an exclusive lock. It blocks new shared read locks on the data, waits for existing shared read locks to clear, then places an exclusive lock for exclusive access to the data.

Notes about locks:

• Locks are taken at the granularity of row-and-column. If transaction T1 has locked column "A" of row "foo", and transaction T2 wants to write column "B" of row "foo" then there is no conflict.
• Writes to a data item that don't also read the data being written (aka "blind writes") don't conflict with other blind writers of the same item (the commit timestamp of each write determines the order in which it is applied to the database). A consequence of this is that Spanner only needs to upgrade to an exclusive lock if you have read the data you are writing. Otherwise Spanner uses a shared lock called a writer shared lock.
• When performing row lookups inside a read-write transaction, use secondary indexes to limit the rows scanned to a smaller range. This causes Spanner to lock a fewer number of rows in the table, allowing concurrent modification to rows outside of the range.

• Locks should not be used to ensure exclusive access to a resource outside of Spanner. Transactions can be aborted for several reasons by Spanner such as, for example, when allowing data to move around the instance's compute resources. If a transaction is retried, whether explicitly by application code or implicitly by client code such as the Spanner JDBC driver, it is only guaranteed that the locks were held during the attempt that actually committed.

• You can use the Lock statistics introspection tool to investigate lock conflicts in your database.

Deadlock detection

Spanner detects when multiple transactions might be deadlocked, and forces all but one of the transactions to abort. For example, consider the following scenario: transaction Txn1 holds a lock on record A and is waiting for a lock on record B, and Txn2 holds a lock on record B and is waiting for a lock on record A. The only way to make progress in this situation is to abort one of the transactions so it releases its lock, allowing the other transaction to proceed.

Spanner uses the standard "wound-wait" algorithm to handle deadlock detection. Under the hood, Spanner keeps track of the age of each transaction that requests conflicting locks. It also allows older transactions to abort younger transactions (where "older" means that the transaction's earliest read, query, or commit happened sooner).

By giving priority to older transactions, Spanner ensures that every transaction has a chance to acquire locks eventually, after it becomes old enough to have higher priority than other transactions. For example, a transaction that acquires a reader shared lock can be aborted by an older transaction that needs a writer shared lock.

Distributed execution

Spanner can run transactions on data that spans multiple servers. This power comes at a performance cost compared to single-server transactions.

What types of transactions might be distributed? Under the hood, Spanner can divide responsibility for rows in the database across many servers. A row and the corresponding rows in interleaved tables are usually served by the same server, as are two rows in the same table with nearby keys. Spanner can perform transactions across rows on different servers; however, as a rule of thumb, transactions that affect many co-located rows are faster and cheaper than transactions that affect many rows scattered throughout the database, or throughout a large table.

The most efficient transactions in Spanner include only the reads and writes that should be applied atomically. Transactions are fastest when all reads and writes access data in the same part of the key space.

Read-only transactions

In addition to locking read-write transactions, Spanner offers read-only transactions.

Use a read-only transaction when you need to execute more than one read at the same timestamp. If you can express your read using one of Spanner's single read methods, you should use that single read method instead. The performance of using such a single read call should be comparable to the performance of a single read done in a read-only transaction.

If you are reading a large amount of data, consider using partitions to read the data in parallel.

Because read-only transactions don't write, they don't hold locks and they don't block other transactions. Read-only transactions observe a consistent prefix of the transaction commit history, so your application always gets consistent data.

Properties

A Spanner read-only transaction executes a set of reads at a single logical point in time, both from the perspective of the read-only transaction itself and from the perspective of other readers and writers to the Spanner database. This means that read-only transactions always observe a consistent state of the database at a chosen point in the transaction history.

Interface

Spanner provides an interface for executing a body of work in the context of a read-only transaction, with retries for transaction aborts.

Example

The following shows how to use a read-only transaction to get consistent data for two reads at the same timestamp:

C++

void ReadOnlyTransaction(google::cloud::spanner::Client client) {
  namespace spanner = ::google::cloud::spanner;
  auto read_only = spanner::MakeReadOnlyTransaction();

  spanner::SqlStatement select(
      "SELECT SingerId, AlbumId, AlbumTitle FROM Albums");
  using RowType = std::tuple<std::int64_t, std::int64_t, std::string>;

  // Read#1.
  auto rows1 = client.ExecuteQuery(read_only, select);
  std::cout << "Read 1 results\n";
  for (auto& row : spanner::StreamOf<RowType>(rows1)) {
    if (!row) throw std::move(row).status();
    std::cout << "SingerId: " << std::get<0>(*row)
              << " AlbumId: " << std::get<1>(*row)
              << " AlbumTitle: " << std::get<2>(*row) << "\n";
  }
  // Read#2. Even if changes occur in-between the reads the transaction ensures
  // that Read #1 and Read #2 return the same data.
  auto rows2 = client.ExecuteQuery(read_only, select);
  std::cout << "Read 2 results\n";
  for (auto& row : spanner::StreamOf<RowType>(rows2)) {
    if (!row) throw std::move(row).status();
    std::cout << "SingerId: " << std::get<0>(*row)
              << " AlbumId: " << std::get<1>(*row)
              << " AlbumTitle: " << std::get<2>(*row) << "\n";
  }
}

C#

using Google.Cloud.Spanner.Data;
using System;
using System.Collections.Generic;
using System.Threading.Tasks;
using System.Transactions;

public class QueryDataWithTransactionAsyncSample
{
    public class Album
    {
        public int SingerId { get; set; }
        public int AlbumId { get; set; }
        public string AlbumTitle { get; set; }
    }

    public async Task<List<Album>> QueryDataWithTransactionAsync(string projectId, string instanceId, string databaseId)
    {
        string connectionString = $"Data Source=projects/{projectId}/instances/{instanceId}/databases/{databaseId}";

        var albums = new List<Album>();
        using TransactionScope scope = new TransactionScope(TransactionScopeAsyncFlowOption.Enabled);
        using var connection = new SpannerConnection(connectionString);

        // Opens the connection so that the Spanner transaction included in the TransactionScope
        // is read-only TimestampBound.Strong.
        await connection.OpenAsync(SpannerTransactionCreationOptions.ReadOnly, options: null, cancellationToken: default);
        using var cmd = connection.CreateSelectCommand("SELECT SingerId, AlbumId, AlbumTitle FROM Albums");

        // Read #1.
        using (var reader = await cmd.ExecuteReaderAsync())
        {
            while (await reader.ReadAsync())
            {
                Console.WriteLine("SingerId : " + reader.GetFieldValue<string>("SingerId")
                    + " AlbumId : " + reader.GetFieldValue<string>("AlbumId")
                    + " AlbumTitle : " + reader.GetFieldValue<string>("AlbumTitle"));
            }
        }

        // Read #2. Even if changes occur in-between the reads,
        // the transaction ensures that Read #1 and Read #2
        // return the same data.
        using (var reader = await cmd.ExecuteReaderAsync())
        {
            while (await reader.ReadAsync())
            {
                albums.Add(new Album
                {
                    AlbumId = reader.GetFieldValue<int>("AlbumId"),
                    SingerId = reader.GetFieldValue<int>("SingerId"),
                    AlbumTitle = reader.GetFieldValue<string>("AlbumTitle")
                });
            }
        }
        scope.Complete();
        Console.WriteLine("Transaction complete.");
        return albums;
    }
}

Go

import (
	"context"
	"fmt"
	"io"

	"cloud.google.com/go/spanner"
	"google.golang.org/api/iterator"
)

func readOnlyTransaction(w io.Writer, db string) error {
	ctx := context.Background()
	client, err := spanner.NewClient(ctx, db)
	if err != nil {
		return err
	}
	defer client.Close()

	ro := client.ReadOnlyTransaction()
	defer ro.Close()
	stmt := spanner.Statement{SQL: `SELECT SingerId, AlbumId, AlbumTitle FROM Albums`}
	iter := ro.Query(ctx, stmt)
	defer iter.Stop()
	for {
		row, err := iter.Next()
		if err == iterator.Done {
			break
		}
		if err != nil {
			return err
		}
		var singerID int64
		var albumID int64
		var albumTitle string
		if err := row.Columns(&singerID, &albumID, &albumTitle); err != nil {
			return err
		}
		fmt.Fprintf(w, "%d %d %s\n", singerID, albumID, albumTitle)
	}

	iter = ro.Read(ctx, "Albums", spanner.AllKeys(), []string{"SingerId", "AlbumId", "AlbumTitle"})
	defer iter.Stop()
	for {
		row, err := iter.Next()
		if err == iterator.Done {
			return nil
		}
		if err != nil {
			return err
		}
		var singerID int64
		var albumID int64
		var albumTitle string
		if err := row.Columns(&singerID, &albumID, &albumTitle); err != nil {
			return err
		}
		fmt.Fprintf(w, "%d %d %s\n", singerID, albumID, albumTitle)
	}
}

Java

static void readOnlyTransaction(DatabaseClient dbClient) {
  // ReadOnlyTransaction must be closed by calling close() on it to release resources held by it.
  // We use a try-with-resource block to automatically do so.
  try (ReadOnlyTransaction transaction = dbClient.readOnlyTransaction()) {
    ResultSet queryResultSet =
        transaction.executeQuery(
            Statement.of("SELECT SingerId, AlbumId, AlbumTitle FROM Albums"));
    while (queryResultSet.next()) {
      System.out.printf(
          "%d %d %s\n",
          queryResultSet.getLong(0), queryResultSet.getLong(1), queryResultSet.getString(2));
    }
    try (ResultSet readResultSet =
        transaction.read(
            "Albums", KeySet.all(), Arrays.asList("SingerId", "AlbumId", "AlbumTitle"))) {
      while (readResultSet.next()) {
        System.out.printf(
            "%d %d %s\n",
            readResultSet.getLong(0), readResultSet.getLong(1), readResultSet.getString(2));
      }
    }
  }
}

Node.js

// Imports the Google Cloud client library
const {Spanner} = require('@google-cloud/spanner');

/**
 * TODO(developer): Uncomment the following lines before running the sample.
 */
// const projectId = 'my-project-id';
// const instanceId = 'my-instance';
// const databaseId = 'my-database';

// Creates a client
const spanner = new Spanner({
  projectId: projectId,
});

// Gets a reference to a Cloud Spanner instance and database
const instance = spanner.instance(instanceId);
const database = instance.database(databaseId);

// Gets a transaction object that captures the database state
// at a specific point in time
database.getSnapshot(async (err, transaction) => {
  if (err) {
    console.error(err);
    return;
  }
  const queryOne = 'SELECT SingerId, AlbumId, AlbumTitle FROM Albums';

  try {
    // Read #1, using SQL
    const [qOneRows] = await transaction.run(queryOne);

    qOneRows.forEach(row => {
      const json = row.toJSON();
      console.log(
        `SingerId: ${json.SingerId}, AlbumId: ${json.AlbumId}, AlbumTitle: ${json.AlbumTitle}`
      );
    });

    const queryTwo = {
      columns: ['SingerId', 'AlbumId', 'AlbumTitle'],
    };

    // Read #2, using the `read` method. Even if changes occur
    // in-between the reads, the transaction ensures that both
    // return the same data.
    const [qTwoRows] = await transaction.read('Albums', queryTwo);

    qTwoRows.forEach(row => {
      const json = row.toJSON();
      console.log(
        `SingerId: ${json.SingerId}, AlbumId: ${json.AlbumId}, AlbumTitle: ${json.AlbumTitle}`
      );
    });

    console.log('Successfully executed read-only transaction.');
  } catch (err) {
    console.error('ERROR:', err);
  } finally {
    transaction.end();
    // Close the database when finished.
    await database.close();
  }
});

PHP

use Google\Cloud\Spanner\SpannerClient;

/**
 * Reads data inside of a read-only transaction.
 *
 * Within the read-only transaction, or "snapshot", the application sees
 * consistent view of the database at a particular timestamp.
 * Example:
 * ```
 * read_only_transaction($instanceId, $databaseId);
 * ```
 *
 * @param string $instanceId The Spanner instance ID.
 * @param string $databaseId The Spanner database ID.
 */
function read_only_transaction(string $instanceId, string $databaseId): void
{
    $spanner = new SpannerClient();
    $instance = $spanner->instance($instanceId);
    $database = $instance->database($databaseId);

    $snapshot = $database->snapshot();
    $results = $snapshot->execute(
        'SELECT SingerId, AlbumId, AlbumTitle FROM Albums'
    );
    print('Results from the first read:' . PHP_EOL);
    foreach ($results as $row) {
        printf('SingerId: %s, AlbumId: %s, AlbumTitle: %s' . PHP_EOL,
            $row['SingerId'], $row['AlbumId'], $row['AlbumTitle']);
    }

    // Perform another read using the `read` method. Even if the data
    // is updated in-between the reads, the snapshot ensures that both
    // return the same data.
    $keySet = $spanner->keySet(['all' => true]);
    $results = $database->read(
        'Albums',
        $keySet,
        ['SingerId', 'AlbumId', 'AlbumTitle']
    );

    print('Results from the second read:' . PHP_EOL);
    foreach ($results->rows() as $row) {
        printf('SingerId: %s, AlbumId: %s, AlbumTitle: %s' . PHP_EOL,
            $row['SingerId'], $row['AlbumId'], $row['AlbumTitle']);
    }
}

Python

def read_only_transaction(instance_id, database_id):
    """Reads data inside of a read-only transaction.

    Within the read-only transaction, or "snapshot", the application sees
    consistent view of the database at a particular timestamp.
    """
    spanner_client = spanner.Client()
    instance = spanner_client.instance(instance_id)
    database = instance.database(database_id)

    with database.snapshot(multi_use=True) as snapshot:
        # Read using SQL.
        results = snapshot.execute_sql(
            "SELECT SingerId, AlbumId, AlbumTitle FROM Albums"
        )

        print("Results from first read:")
        for row in results:
            print("SingerId: {}, AlbumId: {}, AlbumTitle: {}".format(*row))

        # Perform another read using the `read` method. Even if the data
        # is updated in-between the reads, the snapshot ensures that both
        # return the same data.
        keyset = spanner.KeySet(all_=True)
        results = snapshot.read(
            table="Albums", columns=("SingerId", "AlbumId", "AlbumTitle"), keyset=keyset
        )

        print("Results from second read:")
        for row in results:
            print("SingerId: {}, AlbumId: {}, AlbumTitle: {}".format(*row))

Ruby

# project_id  = "Your Google Cloud project ID"
# instance_id = "Your Spanner instance ID"
# database_id = "Your Spanner database ID"

require "google/cloud/spanner"

spanner = Google::Cloud::Spanner.new project: project_id
client  = spanner.client instance_id, database_id

client.snapshot do |snapshot|
  snapshot.execute("SELECT SingerId, AlbumId, AlbumTitle FROM Albums").rows.each do |row|
    puts "#{row[:AlbumId]} #{row[:AlbumTitle]} #{row[:SingerId]}"
  end

  # Even if changes occur in-between the reads, the transaction ensures that
  # both return the same data.
  snapshot.read("Albums", [:AlbumId, :AlbumTitle, :SingerId]).rows.each do |row|
    puts "#{row[:AlbumId]} #{row[:AlbumTitle]} #{row[:SingerId]}"
  end
end

Partitioned DML transactions

Using Partitioned Data Manipulation Language (Partitioned DML), you can execute large-scale UPDATE and DELETE statements without running into transaction limits or locking an entire table. Spanner partitions the key space and executes the DML statements on each partition in a separate read-write transaction.

You run DML statements in read-write transactions that you explicitly create in your code. For more information, see Using DML.

Properties

You can execute only one Partitioned DML statement at a time, whether you are using a client library method or the Google Cloud CLI.

Partitioned transactions do not support commit or rollback. Spanner executes and applies the DML statement immediately. If you cancel the operation, or the operation fails, then Spanner cancels all the executing partitions and doesn't start any of the remaining partitions. Spanner does not rollback any partitions that have already executed.

Interface

Spanner provides an interface for executing a single Partitioned DML statement.

Examples

The following code example updates the MarketingBudget column of the Albums table.

C++

You use the ExecutePartitionedDml() function to execute a Partitioned DML statement.

void DmlPartitionedUpdate(google::cloud::spanner::Client client) {
  namespace spanner = ::google::cloud::spanner;
  auto result = client.ExecutePartitionedDml(
      spanner::SqlStatement("UPDATE Albums SET MarketingBudget = 100000"
                            "  WHERE SingerId > 1"));
  if (!result) throw std::move(result).status();
  std::cout << "Updated at least " << result->row_count_lower_bound
            << " row(s) [spanner_dml_partitioned_update]\n";
}

C#

You use the ExecutePartitionedUpdateAsync() method to execute a Partitioned DML statement.

using Google.Cloud.Spanner.Data;
using System;
using System.Threading.Tasks;

public class UpdateUsingPartitionedDmlCoreAsyncSample
{
    public async Task<long> UpdateUsingPartitionedDmlCoreAsync(string projectId, string instanceId, string databaseId)
    {
        string connectionString = $"Data Source=projects/{projectId}/instances/{instanceId}/databases/{databaseId}";

        using var connection = new SpannerConnection(connectionString);
        await connection.OpenAsync();

        using var cmd = connection.CreateDmlCommand("UPDATE Albums SET MarketingBudget = 100000 WHERE SingerId > 1");
        long rowCount = await cmd.ExecutePartitionedUpdateAsync();

        Console.WriteLine($"{rowCount} row(s) updated...");
        return rowCount;
    }
}

Go

You use the PartitionedUpdate() method to execute a Partitioned DML statement.

import (
	"context"
	"fmt"
	"io"

	"cloud.google.com/go/spanner"
)

func updateUsingPartitionedDML(w io.Writer, db string) error {
	ctx := context.Background()
	client, err := spanner.NewClient(ctx, db)
	if err != nil {
		return err
	}
	defer client.Close()

	stmt := spanner.Statement{SQL: "UPDATE Albums SET MarketingBudget = 100000 WHERE SingerId > 1"}
	rowCount, err := client.PartitionedUpdate(ctx, stmt)
	if err != nil {
		return err
	}
	fmt.Fprintf(w, "%d record(s) updated.\n", rowCount)
	return nil
}

Java

You use the executePartitionedUpdate() method to execute a Partitioned DML statement.

static void updateUsingPartitionedDml(DatabaseClient dbClient) {
  String sql = "UPDATE Albums SET MarketingBudget = 100000 WHERE SingerId > 1";
  long rowCount = dbClient.executePartitionedUpdate(Statement.of(sql));
  System.out.printf("%d records updated.\n", rowCount);
}

Node.js

You use the runPartitionedUpdate() method to execute a Partitioned DML statement.

// Imports the Google Cloud client library
const {Spanner} = require('@google-cloud/spanner');

/**
 * TODO(developer): Uncomment the following lines before running the sample.
 */
// const projectId = 'my-project-id';
// const instanceId = 'my-instance';
// const databaseId = 'my-database';

// Creates a client
const spanner = new Spanner({
  projectId: projectId,
});

// Gets a reference to a Cloud Spanner instance and database
const instance = spanner.instance(instanceId);
const database = instance.database(databaseId);

try {
  const [rowCount] = await database.runPartitionedUpdate({
    sql: 'UPDATE Albums SET MarketingBudget = 100000 WHERE SingerId > 1',
  });
  console.log(`Successfully updated ${rowCount} records.`);
} catch (err) {
  console.error('ERROR:', err);
} finally {
  // Close the database when finished.
  database.close();
}

PHP

You use the executePartitionedUpdate() method to execute a Partitioned DML statement.

use Google\Cloud\Spanner\SpannerClient;

/**
 * Updates sample data in the database by partition with a DML statement.
 *
 * This updates the `MarketingBudget` column which must be created before
 * running this sample. You can add the column by running the `add_column`
 * sample or by running this DDL statement against your database:
 *
 *     ALTER TABLE Albums ADD COLUMN MarketingBudget INT64
 *
 * Example:
 * ```
 * update_data($instanceId, $databaseId);
 * ```
 *
 * @param string $instanceId The Spanner instance ID.
 * @param string $databaseId The Spanner database ID.
 */
function update_data_with_partitioned_dml(string $instanceId, string $databaseId): void
{
    $spanner = new SpannerClient();
    $instance = $spanner->instance($instanceId);
    $database = $instance->database($databaseId);

    $rowCount = $database->executePartitionedUpdate(
        'UPDATE Albums SET MarketingBudget = 100000 WHERE SingerId > 1'
    );

    printf('Updated %d row(s).' . PHP_EOL, $rowCount);
}

Python

You use the execute_partitioned_dml() method to execute a Partitioned DML statement.

# instance_id = "your-spanner-instance"
# database_id = "your-spanner-db-id"

spanner_client = spanner.Client()
instance = spanner_client.instance(instance_id)
database = instance.database(database_id)

row_ct = database.execute_partitioned_dml(
    "UPDATE Albums SET MarketingBudget = 100000 WHERE SingerId > 1"
)

print("{} records updated.".format(row_ct))

Ruby

You use the execute_partitioned_update() method to execute a Partitioned DML statement.

# project_id  = "Your Google Cloud project ID"
# instance_id = "Your Spanner instance ID"
# database_id = "Your Spanner database ID"

require "google/cloud/spanner"

spanner = Google::Cloud::Spanner.new project: project_id
client  = spanner.client instance_id, database_id

row_count = client.execute_partition_update(
  "UPDATE Albums SET MarketingBudget = 100000 WHERE SingerId > 1"
)

puts "#{row_count} records updated."

The following code example deletes rows from the Singers table, based on the SingerId column.

C++

void DmlPartitionedDelete(google::cloud::spanner::Client client) {
  namespace spanner = ::google::cloud::spanner;
  auto result = client.ExecutePartitionedDml(
      spanner::SqlStatement("DELETE FROM Singers WHERE SingerId > 10"));
  if (!result) throw std::move(result).status();
  std::cout << "Deleted at least " << result->row_count_lower_bound
            << " row(s) [spanner_dml_partitioned_delete]\n";
}

C#

using Google.Cloud.Spanner.Data;
using System;
using System.Threading.Tasks;

public class DeleteUsingPartitionedDmlCoreAsyncSample
{
    public async Task<long> DeleteUsingPartitionedDmlCoreAsync(string projectId, string instanceId, string databaseId)
    {
        string connectionString = $"Data Source=projects/{projectId}/instances/{instanceId}/databases/{databaseId}";

        using var connection = new SpannerConnection(connectionString);
        await connection.OpenAsync();

        using var cmd = connection.CreateDmlCommand("DELETE FROM Singers WHERE SingerId > 10");
        long rowCount = await cmd.ExecutePartitionedUpdateAsync();

        Console.WriteLine($"{rowCount} row(s) deleted...");
        return rowCount;
    }
}

Go

import (
	"context"
	"fmt"
	"io"

	"cloud.google.com/go/spanner"
)

func deleteUsingPartitionedDML(w io.Writer, db string) error {
	ctx := context.Background()
	client, err := spanner.NewClient(ctx, db)
	if err != nil {
		return err
	}
	defer client.Close()

	stmt := spanner.Statement{SQL: "DELETE FROM Singers WHERE SingerId > 10"}
	rowCount, err := client.PartitionedUpdate(ctx, stmt)
	if err != nil {
		return err

	}
	fmt.Fprintf(w, "%d record(s) deleted.", rowCount)
	return nil
}

Java

static void deleteUsingPartitionedDml(DatabaseClient dbClient) {
  String sql = "DELETE FROM Singers WHERE SingerId > 10";
  long rowCount = dbClient.executePartitionedUpdate(Statement.of(sql));
  System.out.printf("%d records deleted.\n", rowCount);
}

Node.js

// Imports the Google Cloud client library
const {Spanner} = require('@google-cloud/spanner');

/**
 * TODO(developer): Uncomment the following lines before running the sample.
 */
// const projectId = 'my-project-id';
// const instanceId = 'my-instance';
// const databaseId = 'my-database';

// Creates a client
const spanner = new Spanner({
  projectId: projectId,
});

// Gets a reference to a Cloud Spanner instance and database
const instance = spanner.instance(instanceId);
const database = instance.database(databaseId);

try {
  const [rowCount] = await database.runPartitionedUpdate({
    sql: 'DELETE FROM Singers WHERE SingerId > 10',
  });
  console.log(`Successfully deleted ${rowCount} records.`);
} catch (err) {
  console.error('ERROR:', err);
} finally {
  // Close the database when finished.
  database.close();
}

PHP

use Google\Cloud\Spanner\SpannerClient;

/**
 * Delete sample data in the database by partition with a DML statement.
 *
 * This updates the `MarketingBudget` column which must be created before
 * running this sample. You can add the column by running the `add_column`
 * sample or by running this DDL statement against your database:
 *
 *     ALTER TABLE Albums ADD COLUMN MarketingBudget INT64
 *
 * Example:
 * ```
 * update_data($instanceId, $databaseId);
 * ```
 *
 * @param string $instanceId The Spanner instance ID.
 * @param string $databaseId The Spanner database ID.
 */
function delete_data_with_partitioned_dml(string $instanceId, string $databaseId): void
{
    $spanner = new SpannerClient();
    $instance = $spanner->instance($instanceId);
    $database = $instance->database($databaseId);

    $rowCount = $database->executePartitionedUpdate(
        'DELETE FROM Singers WHERE SingerId > 10'
    );

    printf('Deleted %d row(s).' . PHP_EOL, $rowCount);
}

Python

# instance_id = "your-spanner-instance"
# database_id = "your-spanner-db-id"
spanner_client = spanner.Client()
instance = spanner_client.instance(instance_id)
database = instance.database(database_id)

row_ct = database.execute_partitioned_dml("DELETE FROM Singers WHERE SingerId > 10")

print("{} record(s) deleted.".format(row_ct))

Ruby

# project_id  = "Your Google Cloud project ID"
# instance_id = "Your Spanner instance ID"
# database_id = "Your Spanner database ID"

require "google/cloud/spanner"

spanner = Google::Cloud::Spanner.new project: project_id
client  = spanner.client instance_id, database_id

row_count = client.execute_partition_update(
  "DELETE FROM Singers WHERE SingerId > 10"
)

puts "#{row_count} records deleted."