Pre-commit hooks

sqlsift is fast enough to run on every commit.

Plain Git hook

Check the staged version of each changed query file, not the working copy, by piping it through stdin:

#!/usr/bin/env bash
# .git/hooks/pre-commit (or .githooks/pre-commit with `git config core.hooksPath .githooks`)
set -euo pipefail

status=0
while IFS= read -r file; do
  git show ":$file" | sqlsift check --stdin-filename "$file" - || status=1
done < <(git diff --cached --name-only --diff-filter=ACM -- '*.sql' '*.ts' '*.tsx' '*.js' '*.jsx' '*.mts' '*.cts' '*.mjs' '*.cjs' '*.vue' '*.svelte')
exit $status

The pathspec covers .sql files and the files sqlsift checks for SQL in tagged template literals (TypeScript, JavaScript, Vue and Svelte), so a commit that only changes embedded SQL is still checked. If your project has no embedded SQL, keep just '*.sql', otherwise the hook runs on every commit that touches those files for nothing.

This uses the schema settings from sqlsift.toml. When a schema or migration file is staged, consider checking every query (sqlsift check) instead, since any of them may be affected.

pre-commit framework

With pre-commit, a local hook that checks all query files works well, because a run is fast:

# .pre-commit-config.yaml
repos:
  - repo: local
    hooks:
      - id: sqlsift
        name: sqlsift
        entry: npx --yes sqlsift-cli check
        language: system
        files: \.(sql|ts|tsx|js|jsx|mts|cts|mjs|cjs|vue|svelte)$
        pass_filenames: false

pass_filenames: false makes sqlsift check the files from sqlsift.toml, so a schema change re-checks every query.

As with the plain hook, files also matches the extensions of files with embedded SQL, so the hook runs when only those change. Drop them if your project has none, and make sure the files in sqlsift.toml include them (for example src/**/*.ts) so they are actually checked.

lefthook / husky

Any hook manager works the same way: run sqlsift check (or npx sqlsift-cli check) and let it read sqlsift.toml.