ava-postgres
is a test fixture for AVA that provides you with nearly-instant access to a fresh Postgres database for every test.
ava-postgres
's only dependency is a running instance of Docker.
npm install --save-dev ava-postgres
or
yarn add --dev ava-postgres
ava-postgres
's main export is a factory function, so you'll probably want to create a file like tests/fixtures/get-test-database.ts
:
import { getTestPostgresDatabaseFactory } from "ava-postgres"
export const getTestDatabase = getTestPostgresDatabaseFactory({
// Any tag for the official Postgres Docker image, defaults to "14"
postgresVersion: "14",
})
Then, in your tests, you can use the getTestDatabase()
function to get a fresh database for each test:
import test from "ava"
import { getTestDatabase } from "./fixtures/get-test-database"
test("foo bar", async (t) => {
const { pool } = await getTestDatabase()
await pool.query("SELECT 1")
t.pass()
})
Full list of connection details returned by getTestDatabase
ava-postgres
uses Postgres templates so you only pay the setup cost once. After a template has been created, Postgres can create a new database from it in milliseconds.
If you want to perform common database setup, you can use a hook and pass parameters to the getTestDatabase()
function:
import { getTestPostgresDatabaseFactory } from "ava-postgres"
type GetTestDatabaseParams = {
shouldMigrate?: boolean
shouldSeed?: boolean
}
export const getTestDatabase =
getTestPostgresDatabaseFactory<GetTestDatabaseParams>({
beforeTemplateIsBaked: async ({
connection: { pool },
params: { shouldMigrate, shouldSeed },
}) => {
if (shouldMigrate) {
await pool.query("CREATE TABLE foo (id int)")
}
if (shouldSeed) {
await pool.query("INSERT INTO foo VALUES (1)")
}
},
})
Then, in your tests, you can pass parameters to the getTestDatabase()
function:
import test from "ava"
import { getTestDatabase } from "./fixtures/get-test-database"
test("foo bar", async (t) => {
const { pool } = await getTestDatabase({
shouldMigrate: true,
shouldSeed: true,
})
await pool.query("SELECT * FROM foo")
t.pass()
})
In rare cases, you may want to spawn more than one Postgres container.
Internally, this library uses an AVA "shared worker". A shared worker is a singleton shared with the entire running test suite, and so one ava-postgres
shared worker maps to exactly one Postgres container.
To spawn separate shared workers and thus additional Postgres containers, you have two options:
Specify different version strings for the postgresVersion
option in the factory function:
const getTestPostgresDatabase = getTestPostgresDatabaseFactory({
postgresVersion: "14",
})
Each unique version will map to a unique shared worker.
Set the workerDedupeKey
option in the factory function:
const getTestPostgresDatabase = getTestPostgresDatabaseFactory({
workerDedupeKey: "foo",
})
Each unique key will map to a unique shared worker.
By default, ava-postgres
will create a new database for each test. If you want to share a database between tests, you can use the databaseDedupeKey
option:
import test from "ava"
const getTestPostgresDatabase = getTestPostgresDatabaseFactory({})
test("foo", async (t) => {
const connection1 = await getTestPostgresDatabase(t, null, {
databaseDedupeKey: "foo",
})
const connection2 = await getTestPostgresDatabase(t, null, {
databaseDedupeKey: "foo",
})
t.is(connection1.database, connection2.database)
})
This works across the entire test suite.
Note that if unique parameters are passed to the beforeTemplateIsBaked
(null
in the above example), separate databases will still be created.
In some cases, if you do extensive setup in your beforeTemplateIsBaked
hook, you might want to obtain a separate, additional database within it if your application uses several databases for different purposes. This is possible by using the manuallyBuildAdditionalTemplate()
function passed to your hook callback:
import test from "ava"
const getTestDatabase = getTestPostgresDatabaseFactory<DatabaseParams>({
beforeTemplateIsBaked: async ({
params,
connection: { pool },
manuallyBuildAdditionalTemplate,
}) => {
await pool.query(`CREATE TABLE "bar" ("id" SERIAL PRIMARY KEY)`)
const fooTemplateBuilder = await manuallyBuildAdditionalTemplate()
await fooTemplateBuilder.connection.pool.query(
`CREATE TABLE "foo" ("id" SERIAL PRIMARY KEY)`
)
const { templateName: fooTemplateName } = await fooTemplateBuilder.finish()
return { fooTemplateName }
},
})
test("foo", async (t) => {
const barDatabase = await getTestDatabase({ type: "bar" })
// the "bar" database has the "bar" table...
await t.notThrowsAsync(async () => {
await barDatabase.pool.query(`SELECT * FROM "bar"`)
})
// ...but not the "foo" table...
await t.throwsAsync(async () => {
await barDatabase.pool.query(`SELECT * FROM "foo"`)
})
// ...and we can obtain a separate database with the "foo" table
const fooDatabase = await getTestDatabase.fromTemplate(
t,
barDatabase.beforeTemplateIsBakedResult.fooTemplateName
)
await t.notThrowsAsync(async () => {
await fooDatabase.pool.query(`SELECT * FROM "foo"`)
})
})
ava-postgres
uses testcontainers under the hood to manage the Postgres container.
In some scenarios you might want to mount a SQL script into the container and manually load it using psql
.
You can do this with the bindMounts
option:
const getTestPostgresDatabase = getTestPostgresDatabaseFactory({
container: {
bindMounts: [
{
source: "/path/on/host",
target: "/test.sql",
},
],
},
beforeTemplateIsBaked: async ({
connection: { username, database },
containerExec,
}) => {
const { exitCode } = await containerExec(
`psql -U ${username} -d ${database} -f /test.sql`.split(" ")
)
if (exitCode !== 0) {
throw new Error(`Failed to load test file`)
}
},
})