pg-iterator

TypeScript wrapper for pg-query-stream, which adds the following:

• Produces a safe AsyncIterable, for row-by-row processing, with for await or a library (like RxJs).
• The library is strongly-typed throughout: you can optionally specify row entity types for your queries.
• Auto-connects Pool on the first-row iteration (disconnects on last) - internal connection management.
• Unifies error handling for queries and connections - initial, interrupted or lost / broken.
• Offers one protocol for working with Client or Pool objects.
• Provides workarounds for some issues in pg-query-stream - missing fields, issue #2870, etc.

Installation

$ npm i pg-iterator

Usage

You have the flexibility of using this module with Pool or Client, or a dynamically-determined type, via createQueryIterable function.

Each of the interfaces - QueryIterablePool, QueryIterableClient or createQueryIterable supports strong-type parametrization, for typed row iteration.

See complete examples.

Using Pool

When using an existing Pool object, this library will automatically acquire the connection, create AsyncIterable from a query and release the connection, once the stream has finished.

Class QueryIterablePool implements such functionality:

1import {Pool} from 'pg';
2import {QueryIterablePool} from 'pg-iterator';
3
4const pool = new Pool(/* connection config */);
5
6const q = new QueryIterablePool(pool); // creating our Pool container
7
8const i = q.query('SELECT * FROM users WHERE id = $1', [123]);
9
10for await(const u of i) {
11    console.log(u); // output each row
12}

Using Client

This library can use a connected Client object directly, via QueryIterableClient class:

1import {Pool, Client} from 'pg';
2import {QueryIterableClient} from 'pg-iterator';
3
4const pool = new Pool(/* connection config */);
5const client: Client = await pool.connect();
6
7const q = new QueryIterableClient(client); // creating our Client container
8
9const i = q.query('SELECT * FROM users WHERE id = $1', [123]);
10
11for await(const u of i) {
12    console.log(u); // output each row
13}
14
15// the onus is on you when to release the client and the pool when done:
16// client.release(), pool.end()

Using dynamic driver

When you do not know whether the source is a Pool or Client, you can use function createQueryIterable instead, which will check the type at run-time, and return either QueryIterablePool or QueryIterableClient, which share generic QueryIterable protocol.

Fields information

In every usage scenario, you end up with QueryIterable base interface, which exposes information about columns.

• You can either access it after reading the very first row:

1const q = new QueryIterablePool(pool);
2
3const i = q.query('SELECT * FROM users WHERE id = $1', [123]);
4
5// q.fields is empty at this point
6
7for await(const u of i) {
8    const {fields} = q; // fields details are available at this point
9
10    console.log(u); // output each row
11}

• Or you can use notification event fields instead:

1const q = new QueryIterablePool(pool);
2
3q.on('fields', fields => {
4    // sent with complete list of fields here,
5    // before the first row in the loop below
6});
7
8const i = q.query('SELECT * FROM users WHERE id = $1', [123]);
9
10for await(const u of i) {
11    console.log(u); // output each row
12}

Events

Base interface QueryIterable can emit the following events:

• fields - fields details, as explained above;
• stream - notification of a new stream created;
• complete - notification of completing the current query.

Error handling

This library manages connection and runs queries inside the same row iteration, the only thing that can throw errors:

1const q = new QueryIterablePool(pool);
2
3const i = q.query('SELECT * FROM users WHERE id = $1', [123]);
4
5try {
6    for await(const u of i) {
7        console.log(u); // output each row
8    }
9} catch (err) {
10    // all connection and query errors arrive here
11}

Integration

Most libraries that are based on node-postgres expose Pool and Client interfaces.

For example, pg-promise exposes Pool via Database.$pool, so you can do:

1const q = new QueryIterablePool(db.$pool); // creating Pool container from Database object

And in terms of data consumption, since the data here is AsyncIterable, there are many libraries that can consume and process it.

• Example with RxJs:

1import {from, take} from 'rxjs';
2
3const i = q.query('SELECT * FROM users WHERE id = $1', [123]);
4
5from(i).pipe(take(10)).subscribe(row => {
6    console.log(row); // up to 10 rows
7});

• Example with iter-ops:

1import {pipe, take} from 'iter-ops';
2
3const i = q.query('SELECT * FROM users WHERE id = $1', [123]);
4
5const r = pipe(i, take(10));
6
7for await (const a of r) {
8    console.log(a); // up to 10 rows
9}

Note that if iteration is incomplete because you interrupted the iteration loop, or used some limiting operators (like take above), the connection will remain open indefinitely. In such cases you may want to force-release the connection, by calling method release of QueryIterable manually:

1import {from, take} from 'rxjs';
2
3const q = new QueryIterablePool(pool);
4
5const i = q.query('SELECT * FROM users WHERE id = $1', [123]);
6
7from(i).pipe(take(10)).subscribe({
8    next(row) {
9        console.log(row);
10    },
11    complete() {
12        // since we use "take(10)" above, the iteration may be incomplete,
13        // and the connection will be stuck, so we have to force-release it: 
14        q.release();
15    }
16});

Alternatively, you can wrap QueryIterable + query into a safe Observable creator:

1function fromQuery<T>(qi: QueryIterable<T>, text: string, params?: any[]): Observable<T> {
2    return from(qi.query(text, params)).pipe(finalize(() => {
3        qi.release();
4    }));
5}

See also: complete examples.