Ignored files are excluded unless --force is passed on the CLI (fixes #813)

[spadgos]

This PR implements the suggestion in #813 -- to exclude ignored files passed on the CLI.

To override this behaviour and lint all files passed in regardless of their ignore-state, I've added a flag --force.

This is still WIP, since there's no documentation of the feature yet, but I didn't want to spend time on that when the interface was not yet agreed upon.

One other thing which I think is missing is some form of alert to users when one or more ignored files were passed in. Without that, it would be quite possible that someone would run eslint my-ignored-file.js expecting it to be linted, and receive no errors.

As a point of reference, when ignored files are passed to git add, it will always exit immediately. I don't think this is useful for eslint though, since the purpose of this feature is to allow a list of files to be passed in and a subset to be linted.

[nzakas]

Yeah, aborting doesn't seem correct. We could provide a warning for each ignored file, pushed onto the front of the messages that are passed to the formatter?

[spadgos]

Yeah, a warning seems to fit nicely. I'll give it a shot.

[spadgos]

Okay, here's an update. Still with no docs or enough tests, but hopefully should be enough to tell whether this is a good interface.

An example: assuming a directory with an ignored file called 'foo.js', the console output below shows three scenarios. First, specifying it on the CLI; second, adding the --force flag; and third, linting the whole directory.

$ eslint foo.js

foo.js
  0:0  warning  File ignored because of your .eslintignore file. Use --force to override

✖ 1 problem

$ echo $?
0

$ eslint --force foo.js

foo.js
  1:0   error  Missing "use strict" statement  strict

✖ 1 problem

$ echo $?
1

$ eslint .
$ echo $?
0

[spadgos]

The tests are failing right now firstly because of a linting problem (ironically enough), but then because I didn't update all the formatters to allow for fatal: false. If what I've done with stylish.js is okay, then I'll port it over to the rest

[nzakas]

Can you add -f as a shortcut?

[spadgos]

Already taken by the shortcut for --formatter. Any other suggestions?

[nzakas]

Dammit! No, never mind :)

[nzakas]

Seems like we may be overloading the meaning of the fatal flag here. If we leave off fatal: false and there's no associated rule, won't this end up as a warning anyway? If not, I think that's a more appropriate approach to implement because who knows if we'd want to have other situations where we augment the messages.

[spadgos]

Okay, I'll just change it to be more defensive in checking for the severity.

[spadgos]

Anything else you need from me on this one? Apart from the question about a shortcut name for --force, I think this one's ready to go.

[nzakas]

LGTM. Sorry for the delay.