Makefile vs JS scripts

[tejasbubane]

See sample js files in :

Makefile:

• Simple elegant bash commands
• Already in place - tiny changes required.
• Needs separate software on windows (like Cygwin etc)

JS Script

• Single language everywhere - assuming contributors already have nodeJS installed
• Not as elegant as Makefile - extremely elaborate for simple task like copying a file: examples

(Please use 👍 to make the change to JS Script and 👎 to stay with Make)

[SleeplessByte]

I would also want to add that:

Makefile:

• there is significant overhead for new contributors who don't know make and don't use bash (think: using guis a lot, are on windows, or use a different non-bourne style shell)
• make commands are easy to get wrong cross-platform, even if a windows user has cygwin / bash installed

[iHiD]

Sounds good to me.

[tejasbubane]

As of now I prefer Makefile for the simplicity. But my opinion is biased since I don't use windows and makefiles work flawlessly on Unix systems.

Should we investigate any JS libraries that ease out the process and reduce the elaborate handrolled scripts? I found one https://github.com/jakejs/jake but did have a proper look yet.

there is significant overhead for new contributors who don't know make and don't use bash
(think: using guis a lot, are on windows, or use a different non-bourne style shell)

Makefiles will not be changed by contributors most of the times. Maybe we should add instructions to install Cygwin for windows contributors?

make commands are easy to get wrong cross-platform, even if a windows user has cygwin / bash
installed

I did not know this. Any specific examples?

[iHiD]

What's the Makefile actually do here?

[SleeplessByte]

What's the Makefile actually do here?

@iHiD Copy a few top-level files, create a temporary directory, copy in all the exercise tests, and rename example.js to the correct filename -- then lints and runs tests. Used by travis at the moment. I prefer having all this in package.json scripts.

As of now I prefer Makefile for the simplicity. But my opinion is biased since I don't use windows and makefiles work flawlessly on Unix systems.

@tejasbubane I work with both which is why I know the pain. Make is really not straightforward on windows, even with cygwin.

I did not know this. Any specific examples?

Yeah, the most simple one is that anytime you need to escape something it's not the same cross shells; similar to why cross-env exists.

Makefiles will not be changed by contributors most of the times. Maybe we should add instructions to install Cygwin for windows contributors?

If we stay with make, I 👍 this proposal -- I think #287 is also very relevant.

Should we investigate any JS libraries that ease out the process and reduce the elaborate handrolled scripts? I found one https://github.com/jakejs/jake but did have a proper look yet.

If we move to JS scripts, then 👍 this proposal!

[tejasbubane]

What's the Makefile actually do here?

@iHiD Ohh right sorry for not clarifying that in the issue.

Makefile does a few things in this repo:

1. Copy package.json from the root directory to each exercise directory. This makes sure all files being served to users are in sync. Contributors doing any changes to package.json are supposed ro run this and commit changes.

2. To be sure contributors making any change to package.json in project root have propagated (and commited) that change to all exercises, we run md5 checksum on top-level and exercise-level package.json files. Soon will be having more files like babel and eslint configs to keep in sync.

3. Create a ./tmp_exercises copy source and test files there. Run tests inside tmp_exercises directory. This allows Jest to run tests in parallel and is way faster than running inside in exercise directory - faster CI builds.

[tejasbubane]

@SleeplessByte Just saw another point in your favour - #676 (comment)

[SleeplessByte]

Ah yes. I also believe we should update CONTRIBUTING.