TypeORM Snake Naming Strategy

Sometimes, the best way to solve your own problems is to help someone else.

@rafikidota/serpens ships a TypeORM SnakeNamingStrategy plus three standalone string-case utilities (camelCase, snakeCase, titleCase) you can use independently of TypeORM.

Installation

npm install @rafikidota/serpens typeorm
# or
pnpm add @rafikidota/serpens typeorm

Requirements:

• typeorm ^1.1.0 — peer dependency, bring your own install.
• Node.js 24+ (the version used in CI, see .nvmrc).

The package is published as ESM-first with a dual build: import resolves to dist/index.mjs, require to dist/index.cjs, each with its own type declarations.

Using SnakeNamingStrategy with TypeORM

import { DataSource, DataSourceOptions } from 'typeorm';
import { SnakeNamingStrategy } from '@rafikidota/serpens';

const config: DataSourceOptions = {
    type: 'postgres',
    host: 'localhost',
    port: 5432,
    database: 'postgres',
    username: 'postgres',
    password: 'postgres',
    entities: [__dirname + '/**/*.entity{.ts,.js}'],
    migrations: [__dirname + '/migrations/*{.ts,.js}'],
    synchronize: false,
    namingStrategy: new SnakeNamingStrategy(),
};

export default new DataSource(config);

With this strategy applied, an entity like:

@Entity()
class UserProfile {
  @Column()
  firstName: string;
}

maps to table user_profile, column first_name.

The strategy overrides three members of TypeORM's DefaultNamingStrategy:

| Member | Behaviour | | --- | --- | | tableName | Explicit customName wins, otherwise snakeCase(className) | | columnName | Joins embedded prefixes with the column name, then snake-cases the result | | relationName | snakeCase(propertyName) |

Everything else falls back to DefaultNamingStrategy (index names, foreign keys, join tables, etc.).

String-case utilities

The same conversion helpers used internally by SnakeNamingStrategy are exported for standalone use — no TypeORM required.

import { camelCase, snakeCase, titleCase } from '@rafikidota/serpens';

snakeCase('firstName');           // 'first_name'
snakeCase('UserHTTPServer');       // 'user_http_server'

camelCase('first_name');          // 'firstName'
camelCase('first_name', true);    // 'FirstName'

titleCase('first name');          // 'First Name'

Development

This package uses pnpm (version pinned via packageManager in package.json), Vitest and tsdown.

pnpm install
pnpm typecheck   # tsc --noEmit
pnpm test        # run the test suite (pnpm test:watch for watch mode)
pnpm lint        # lint and auto-fix
pnpm lint:check  # lint without fixing (CI variant)
pnpm format      # prettier --write
pnpm build       # build dual ESM/CJS output to dist/

A husky pre-commit hook runs lint-staged over staged .ts/.json files (prettier, then eslint).

Releasing

CI (GitHub Actions) runs typecheck, lint, test and build on pushes to main/development and on every pull request.

To cut a release:

pnpm version patch         # or minor / major — updates package.json, commits, tags
git push --follow-tags     # pushing the v* tag triggers the publish workflow

Pushing a v* tag runs three jobs:

| Job | Does | | --- | --- | | ci | Reruns the full CI workflow as a reusable workflow | | guard | Fails if the tag doesn't match package.json, then resolves the npm dist-tag — prereleases go to next, stable versions to latest | | publish | Waits for maintainer approval, then publishes |

The publish job runs in the release environment, so it pauses for manual approval with the CI and guard results already visible. It publishes with provenance via OIDC trusted publishing — no npm token secret involved — and only dist/ is included in the tarball.

Use pnpm version rather than editing package.json by hand: it keeps the tag and the version in sync, which is what guard checks. The dist-tag matters because a trusted-publishing token cannot change dist-tags after the fact, so a mistagged prerelease would stay on latest permanently.

Additional Resources

• TypeORM Documentation