node-postgres.md

PGAdapter - node-postgres Connection Options

PGAdapter supports the node-postgres driver version 8.8.0 and higher.

Usage

First start PGAdapter:

export GOOGLE_APPLICATION_CREDENTIALS=/path/to/credentials.json
docker pull gcr.io/cloud-spanner-pg-adapter/pgadapter
docker run \
  -d -p 5432:5432 \
  -v ${GOOGLE_APPLICATION_CREDENTIALS}:${GOOGLE_APPLICATION_CREDENTIALS}:ro \
  -e GOOGLE_APPLICATION_CREDENTIALS \
  gcr.io/cloud-spanner-pg-adapter/pgadapter \
  -p my-project -i my-instance \
  -x

Then connect to PGAdapter using TCP like this:

const { Client } = require('pg');
const client = new Client({
  host: 'localhost',
  port: 5432,
  database: 'my-database',
});
await client.connect();
const res = await client.query("select 'Hello world!' as hello");
console.log(res.rows[0].hello);
await client.end();

You can also connect to PGAdapter using Unix Domain Sockets if PGAdapter is running on the same host as the client application, or the /tmp directory in the Docker container has been mapped to a directory on the local machine:

export GOOGLE_APPLICATION_CREDENTIALS=/path/to/credentials.json
docker pull gcr.io/cloud-spanner-pg-adapter/pgadapter
docker run \
  -d -p 5432:5432 \
  -v /tmp:/tmp
  -v ${GOOGLE_APPLICATION_CREDENTIALS}:${GOOGLE_APPLICATION_CREDENTIALS}:ro \
  -e GOOGLE_APPLICATION_CREDENTIALS \
  gcr.io/cloud-spanner-pg-adapter/pgadapter \
  -p my-project -i my-instance \
  -x

const { Client } = require('pg');
const client = new Client({
  host: '/tmp',
  port: 5432,
  database: 'my-database',
});
await client.connect();
const res = await client.query("select 'Hello world!' as hello");
console.log(res.rows[0].hello);
await client.end();

Running PGAdapter

This example uses the pre-built Docker image to run PGAdapter. See README for more options for how to run PGAdapter.

Integer Values

node-postgres by default returns int8 / bigint values as strings. This is because Javascript does not support 64-bit integers. You can override this behavior and instruct node-postgres to return int8 / bigint values as 32-bit Javascript integers by adding this configuration to your code:

var types = require('pg').types

// '20' on the following line is the PostgreSQL type code (OID) for `int8`.
types.setTypeParser(20, function(val) {
  return parseInt(val, 10)
})

See https://github.com/brianc/node-pg-types for more information.

Performance Considerations

The following will give you the best possible performance when using node-postgres with PGAdapter.

Unix Domain Sockets

Use Unix Domain Socket connections for the lowest possible latency when PGAdapter and the client application are running on the same host. See https://node-postgres.com/features/connecting for more information on connection options for node-postgres.

Batching

Use the batching options that are available as SQL commands in PGAdapter to batch DDL or DML statements. PGAdapter will combine DML and DDL statements that are executed in a batch into a single request on Cloud Spanner. This can significantly reduce the overhead of executing multiple DML or DDL statements.

Example for DML statements:

  const sql = "insert into test (id, value) values ($1, $2)";
// This will start a DML batch for this client. All subsequent
// DML statements will be cached locally until RUN BATCH is executed.
await client.query("start batch dml");
await client.query({text: sql, values: [1, 'One']});
await client.query({text: sql, values: [2, 'Two']});
await client.query({text: sql, values: [3, 'Three']});
// This will send the DML statements to Cloud Spanner as one batch.
const res = await client.query("run batch");
console.log(res);

Example for DDL statements:

// This will start a DDL batch for this client. All subsequent
// DDL statements will be cached locally until RUN BATCH is executed.
await client.query("start batch ddl");
await client.query("create table my_table1 (key varchar primary key, value varchar)");
await client.query("create table my_table2 (key varchar primary key, value varchar)");
await client.query("create index my_index1 on my_table1 (value)");
// This will send the DDL statements to Cloud Spanner as one batch.
const res = await client.query("run batch");
console.log(res);

Limitations

• node-postgres always uses untyped query parameters encoded as strings. Cloud Spanner requires query parameters to be typed. PGAdapter therefore infers the types of the query parameters the first time it sees a given SQL string by analyzing the SQL statement and caching the result on the connection. This means that the first execution of a SQL statement with query parameters on a connection will be slightly slower than the following executions.