One-off tests for simple issues.

It might be just my personal lack of experience of using git submodules, but I'd really prefer to clone repos programmatically and replace submodules with a configuration in the code. In src/execute/run-prettier.ts, we already have something similar:

const targetRepositoryGlobPatternMap: Map<string, string> = new Map([
  ["typescript-eslint", "./**/*.{ts,js,json,md}"],
  ["eslint-plugin-vue", "./**/*.js"],
  ["babel", "./{packages,codemods,eslint}/**/*.js"],
  ["excalidraw", "./**/*.{css,scss,json,md,html,yml}"],
  ["prettier", "."],
  ["vega-lite", "./**/*.ts"],
]);
const targetRepositoryIgnorePathMap: Map<string, string> = new Map([
  ["excalidraw", ".eslintignore"],
  ["babel", ".eslintignore"],
  ["vega-lite", ".eslintignore"],
]);

I propose to make something like this:

const projects = {
  "typescript-eslint": {
    "url": "[email protected]:typescript-eslint/typescript-eslint.git",
    "commit": "a123bcde",
    "glob": "./**/*.{ts,js,json,md}",
    "ignore": "..."
  },
  "babel": {
    // ...
  },
  // and so on
};

Thoughts?

see prettier/prettier#7957 (comment)

run #9992

Cache target repos by https://github.com/actions/cache

prettier/prettier#12349

This is a test for GitHub Actions

prettier/prettier#12087

prettier/prettier#10537

As I noticed in #19 (comment), the wrong diff is shown. The formatted projects go through the three states:

(1) initial state
(2) formatted by the "original" Prettier
(3) formatted by the "alternative" Prettier

expected: a diff between (2) and (3)
actual: a diff between (1) and (3)

This is a test for GitHub Actions trigger.

prettier/prettier#9549

I want take a look main branch without prettier/prettier#10335

This is a demo!

Why not

execa('yarn add prettier@../prettier')
execa('WHAT_EVER_THIS_REPOSITORY_USED_TO_RUN_PRETTIER')

prettier/prettier#11103