Retain line ending by default.

[veelo]

If no --end_of_line was specified, find the first line ending in the input and use that for the whole output.

All tests include checking for differences in line endings.

Closes #552.

[WebFreak001]

it seems the enum extension should be automatically supported, can you test it also?

[veelo]

What is the enum extension?

[WebFreak001]

sorry, I meant it looks like --end_of_line=automatic should just work now if we use the std.getopt or std.conv methods with the enum and that it should be documented and tested.

I thought I added a suggestion comment to the review as well but it seems like I didn't.

[veelo]

Oh that. The main reason I didn't do that is because then the set of options would be different from EditorConfig options. Before, we had the lf default, now we don't have a documented default just like EditorConfig. When an EditorConfig has no end_of_line defined then it is up to the editor to decide, and our equivalent is that it is up to dfmt to decide, which now takes a hint from the input. So I see the automatic enum member more as an implementation detail than a user visible feature.

Second reason is that is doesn't make much sense to me to supply the default value for an option on the command line.

I'll add a note to the readme what dfmt does when end_of_line is undefined not set.

[WebFreak001]

for IDE tools / integrations it might make sense to know about the special value, so having it documented would be useful there I think.

[veelo]

for IDE tools / integrations it might make sense to know about the special value, so having it documented would be useful there I think.

But how will these IDEs handle .editorconfig if it contains a value that is not part of the EditorConfig specification? We shouldn't be forking that specification.

By the way I amended the description:

Property Name	Allowed Values	Description
end_of_line	cr, crlf and lf	See EditorConfig documentation. When not set, dfmt adopts the first line ending in the input.

[veelo]

Before now I didn't realize that automatic showed up for dfmt --help. Now it doesn't.

[veelo]

This is good to go as far as I am concerned.

[WebFreak001]

LGTM other than this minor thing, can you also rebase onto master to trigger the CI with latest code?

[veelo]

I am not seeing any CI. Is that just me?