Introduce golden testing

[Ailrun]

This PR introduces a golden testing facility and restores most of the format tests.

Related comments: #146 (comment)

@alanz, I want to join this project team. Is it possible to join?

[alanz]

LGTM

Are we in favour of this as a concept? I think I am.

[alanz]

Meant to hit the approve button, sorry.

[jneira]

I am in favor too for some cases, formatting could be one of them.
However, one caveat is string files are platform dependent (at least new lines). I guess those tests pass in linux but not sure if it will do in windows (an its "lovely" \r\n\)
If you use unicode the fun could be endless (i hope it will be not the case)

[Ailrun]

@jneira Hm, so documentContents can start to use \r\n even when the initial file is written in \n? I can imagine it depends on both how formatters and HLS work, but still, sound depressing... Anyhow, that case I will add some kinds of normalization before the saves begin.

[jneira]

@Ailrun let me checkout the branch and run them to confirm it, i hope maybe i am wrong 😄

[Ailrun]

OK, I will wait your confirmation then.

[jneira]

Well, tests are passing cause the crlf conversion is already being done, a pleasant surprise:

stylish-haskell
      formats a document:                                                 warning: LF will be replaced by CRLF in C:\TEMP\StyB51A.actual.
The file will have its original line endings in your working directory
OK (21.80s)
      formats a range:                                                    warning: LF will be replaced by CRLF in C:\TEMP\StyC6FE.actual.
The file will have its original line endings in your working directory
OK (20.56s)
    brittany
      formats a document with LF endings:                                 warning: LF will be replaced by CRLF in C:\TEMP\BriE360.actual.
The file will have its original line endings in your working directory
OK (19.45s)
      formats a document with CRLF endings:                               warning: LF will be replaced by CRLF in C:\TEMP\BriF3FB.actual.
The file will have its original line endings in your working directory
OK (21.09s)
      formats a range with LF endings:                                    warning: LF will be replaced by CRLF in C:\TEMP\Bri60D.actual.
The file will have its original line endings in your working directory
OK (20.53s)
      formats a range with CRLF endings:                                  warning: LF will be replaced by CRLF in C:\TEMP\Bri12C0.actual.
The file will have its original line endings in your working directory
OK (19.20s)
    ormolu
      formats correctly:                                                  warning: LF will be replaced by CRLF in C:\TEMP\For1D31.actual.
The file will have its original line endings in your working directory
OK (14.64s)

🎉

[Ailrun]

Great! Let me merge this since it works in Windows too :)

[jneira]

I wonder where the conversion is being done, maybe tasty-golden?

[Ailrun]

@jneira afaik, they don't handle such a conversion. I just googled those error messages and I found they are from git, not from any of Haskell codes. (FYI, the tests use git diff to print diffed outputs)

[jneira]

oh, yeah, the message is exactly that one, i dint know it was using git diff, the las time i used it it asked for the diff utility and i had to install it in windows.
But i am afraid that tests were succesful cause i have auto.crlf git option activated and maybe they might fail with that option disabled. I will try

[Ailrun]

FYI, the tasty-golden does not force users to use a specific tool to compute differences between a golden file and test result (even though their documentation mentions diff), and I believe git diff is more widespread, so I choose that to get the diff result.

auto.crlf might be the case, and if tests do not work without it, I will send a PR for a normalizer of document contents.

[jneira]

thanks for the info, git diff is a nice alternative.
setting auto.crlf to false did not make the tests fail so no need for more changes

[Ailrun]

Great, thank you for checking Windows cases!