Fetching contributors…
Cannot retrieve contributors at this time
200 lines (137 sloc) 7.16 KB

How to contribute to Knex.js

  • Make changes in the /src directory and run npm run babel (runs once and then quits) or npm run dev (runs once and then watches for changes) to update the files in /lib. npm test will also do this.

  • Before sending a pull request for a feature or bug fix, be sure to have tests. Every pull request that changes the queries should have also integration tests which are ran against real database (in addition to unit tests which checks which kind of queries are being created).

  • Use the same coding style as the rest of the codebase.

  • All pull requests should be made to the master branch.

  • Pull request description should have link to corresponding PR of documentation branch.


Documentation is no longer maintained in knex master repository. All the documentation pull requests should be sent to

Documentation pull requests should not be merged before knex version which has the new documented feature is released.

I would like to add support for new dialect to knex, is it possible?

Currently there are already way too many dialects supported in knex and instead of adding new dialect to central codebase, all the dialects should be moved to separate npm packages out from knex core library with their respective maintainers and test suites.

So if you like to write your own dialect, you can just inherit own dialect from knex base classes and use it by passing dilaect to knex in knex configuration (

// simple dialect overriding sqlite3 dialect to use sqlite3-offline driver
const Knex = require("knex");

const Dialect = require(`knex/lib/dialects/sqlite3/index.js`);
Dialect.prototype._driver = () => require('sqlite3-offline');

const knex = Knex({
  client: Dialect,
  connection: ':memory:'


await knex.schema.createTable('fooobar', (t) => {
await knex('fooobar').insert({ data: 'nomnom' });

console.log("Gimme all the data:", await knex('fooobar'));

What is minimal code to reproduce bug and why I have to provide that when I can just tell whats the problem is

Writing minimal reproduction code for the problem is timeconsuming and some times it also really hard when for example when the original code where the bug happens is written using express or mocha. So why it is necessary for me to commit so much time to it when the problem is in knex? Contributors should be grateful that I reported the bug I found.

The point of runnable code to reproduce the problem is to easily verify that there really is a problem and that the one who did the report did nothing wrong (surprisingly often problem is in the user code). So instead of just description what to do the complete code encourages devs to actually test out that problem exists and start solving it and it saves lots of time.

tl;dr list:

  1. Actually in most of the cases developer already figures out what was the problem when writing the minimal test case or if there was problem how stuff was initilized or how async code was written it is easy to point out the problem.

  2. It motivates developer to actually try out if the bug really exist by not having to figure out from incomplete example environment in which and how bug actually manifests.

  3. There are curently very few people fixing knex issues and if one has to put easily 15-30 minutes time to issue just to see that I cannot reproduce this issue it just wastes development hours that were available for improving knex.

Test case should initialize needed tables, insert needed data and fail...

const knex = require('knex')({
  client: 'pg',
  connection: 'postgres:///knex_test'

async function main() {
  await knex.schema.createTable(...);
  await knex('table').insert({foo: 'bar}');
  await knex.destroy();


Usually issues without reproduction code available are just closed and if the same issue is reported multiple times maybe someone looks into it.

Integration Tests

The Easy Way

By default, Knex runs tests against MySQL (using mysql and mysql2), Postgres, and SQLite. The easiest way to run the tests is by creating the database 'knex_test' and granting permissions to the database's default username:

  • MySQL: root
  • Postgres: postgres

No setup is required for SQLite.

Specifying Databases

You can optionally specify which dialects to test using the DB environment variable. Values should be space separated and can include:

  • mysql
  • mysql2
  • postgres
  • sqlite3
  • maria
  • oracledb
  • mssql
$ DB='postgres mysql' npm test

Custom Configuration

If you'd like to override the database configuration (to use a different host, for example), you can override the path to the default test configuration using the KNEX_TEST environment variable.

$ KNEX_TEST='./path/to/my/config.js' npm test

Creating Postgres User

Depending on your setup you might not have the default postgres user. To create a new user, login to Postgres and use the following queries to add the user. This assumes you've already created the knex_test database.


Once this is done, check it works by attempting to login:

psql -h localhost -U postgres -d knex_test

Running OracleDB tests in docker

Since node-oracledb driver is so hard to install on every platform, oracle tests are actually ran inside docker container. Container has Oracle XE g11, node 8 and node-oracledb driver installed and copies local knex directory in to be able to run the tests.

NODE_VER=10 npm run oracledb:test

You can also manually start shell in the docker image and run build commands manually:

docker run -i -t knex-test-oracledb /bin/bash


cd knex
npm install
npm install oracledb
npm test

Runnin MSSQL SQL Server tests

SQL Server needs to be started as docker container before running tests

# start mssql docker container
npm run mssql:init

# run tests, do changes etc.
npm run mssql:test

# stop mssql container
npm run mssql:destroy

Want to be Collaborator?

There is always room for more collaborators. Be active on resolving github issues / sending pull requests / reviewing code and we will ask you to join.

Etiquette (/ˈɛtᵻkɛt/ or /ˈɛtᵻkɪt/, French: [e.ti.kɛt])

Make pull requests for your changes, do not commit directly to master (release stuff like fixing changelog are ok though).

All the pull requests must be peer reviewed by other collaborator, so don't merge your request before that. If there is no response ping others.

If you are going to add new feature to knex (not just a bugfix) it should be discussed first with others to agree on details.

Join Gitter chat if you feel to chat outside of github issues.