Support for fork-ts-checker-webpack-plugin / thread-loader

[johnnyreilly]

Would you be interested in a PR that adds support for fork-ts-checker-webpack-plugin and thread-loader? I mention it as I'm about to start a new project and was planning to use this as my starting point. For now I'm planning to eject post create to add in fork-ts-checker-webpack-plugin and thread-loader as I like faster builds where possible. But I can see how this might generally be useful to users of this fine project. I'd be happy to submit a PR if that's interesting to you?

https://medium.com/webpack/typescript-webpack-super-pursuit-mode-83cc568dea79

FWIW I think support would simply be added by editing these 3 files:

• https://github.com/wmonk/create-react-app-typescript/blob/master/packages/react-scripts/config/webpack.config.dev.js
• https://github.com/wmonk/create-react-app-typescript/blob/master/packages/react-scripts/config/webpack.config.dev.js
• https://github.com/wmonk/create-react-app-typescript/blob/master/packages/react-scripts/package.json

Also, if you did decide this was interesting it also opens up a choice about tslint. The fork-ts-checker-webpack-plugin allows the option to run tslint for you as well as type checking. So you could (if you chose) remove usage of tslint-loader and just have fork-ts-checker-webpack-plugin do that job too. Or not - these are all choices 😄

[DorianGrey]

Since I've had some good experience using fork-ts-checker-webpack-plugin even for larger projects, I'd vote this up.
However, I'm not sure about the benefit of using thread-loader and cache-loader here:

• On smaller projects, this setup would create an unnecessary overhead - a bit of an overkill.
• On larger projects, it would be useful for speeding up transpilation times, but type checking is the more (time-) expensive part. Since the current create-react-app setup enforces using fork-ts-checker-webpack-plugin in sync mode, a rebuild will complete (and thus reload) only after both transpilation and type checking completed successfully. As a result, there is only a limited benefit of using a parallel build and caching.

€dit: Besides, this change would superseed PR #140, and even perform way better.