Swiftlint only expects files with type ".swift"

[lswith]

I know that swiftlint explicitly expects files of type .swift, but I'd like to pass a file with ".lint-test" instead of ".swift". This is mainly for easy integration with https://github.com/phacility/arcanist.

[SDGGiesbrecht]

I’ve seen several issues where quick bug demonstrations were written as scripts like this:

echo "func doSomething() {}" | swiftlint lint --use-stdin

Does it work to pipe your file to swiftlint like that?

[lswith]

The standard way is to create the .lint-test file. I can override the standard but it seems odd that you can't take any file

…

On Fri, Jul 28, 2017, 4:14 PM Jeremy David Giesbrecht < ***@***.***> wrote: I’ve seen several issues where quick bug demonstrations were written as scripts like this: echo "func doSomething() {}" | swiftlint lint --use-stdin Does it work to pipe your file to swiftlint like that? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1721 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA-K-xCr7BFeimux7CcL_NiZtei_TAuIks5sSXw2gaJpZM4OmIxR> .

[marcelofabri]

I don't think we should change the default behavior, but I think it'd be reasonable to offer a way to override it, perhaps in the configuration file.

[lswith]

You guys are still not at 1.0, so you can do whatever you want haha. Honestly I think specifying exact files is the way to go. Checking if something is a swift file or not shouldn't be the job of the linter. A good example is to simply pass globs to the ‘--path‘ variable. Something like ‘--path **/*.swift‘.

…

On Fri, Jul 28, 2017, 6:05 PM Marcelo Fabri ***@***.***> wrote: I don't think we should change the default behavior, but I think it'd be reasonable to offer a way to override it, perhaps in the configuration file. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1721 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA-K-29EJN8XVM3tDvTmdBlqgRcrtHi3ks5sSZY5gaJpZM4OmIxR> .

[marcelofabri]

Checking if
something is a swift file or not shouldn't be the job of the linter.

I disagree. I think the responsibility of a tool is to provide sensitive defaults. I wouldn't want that, by default, SwiftLint linted Objective-C files for example. Or that we needed to use --path all the times.

A good example is to simply pass globs to the --path variable. Something like --path **/*.swift.

That could also work, but currently I don't think we even support passing a non-Swift file to --path.

[lswith]

If the path argument allowed non swift files and the default was all .swift files I think that would work to both of our cases

…

On Fri, Jul 28, 2017, 6:12 PM Marcelo Fabri ***@***.***> wrote: Checking if something is a swift file or not shouldn't be the job of the linter. I disagree. I think the responsibility of a tool is to provide sensitive defaults. I wouldn't want that, by default, SwiftLint linted Objective-C files for example. Or that we needed to use --path all the times. A good example is to simply pass globs to the --path variable. Something like --path **/*.swift. That could also work, but currently I don't think we even support passing a non-Swift file to --path. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1721 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA-K-3Ef1t-yVmvvDb50cC8NGB5NW-Vuks5sSZgCgaJpZM4OmIxR> .

[SDGGiesbrecht]

@lswith, @marcelofabri

The standard way is to create the .lint-test file.

I do not understand what a .lint-test file is. I ran a Google search for ".lint-test", but it generated no accurate hits. What standard are you referring to?

I added arcanist to the search term list and about the 60th hit was the source code for the unit tests for Arcanist itself.

Those files were the equivalent of SwiftLint’s RuleDescription.(non)TriggeringExamples, and the files (the one’s I saw described CSS) are not pure source, but are repeatedly interrupted with error expectations (analogous to ↓ in triggeringExamples) and end with a JSON configuration to use for the unit test.

.lint-test looks to me like an internal implementation detail of Arcanist’s unit tests.

If the .lint-test files I saw are what we are talking about, then SwiftLint is saving you a headache by refusing to lint them, because SourceKit cannot not even parse them (and in my experience SourceKit does not always fail gracefully).

---

In case this is all a misunderstanding, you do know the normal way to use SwiftLint is to run swiftlint lint in the project root, which outputs a list of violations for every Swift file in the project, right?

[lswith]

ok I'll try an explain better. The arcanist test loader (for writing arcanist integration with swiftlint I might add) expects the test files to be ".lint-test" extension. This file is then separated into 2 parts, one which is the source and one which is the lint error. It's basically a generic way to test all linters effectively. Most linters allow the ability to specify a path and then to take any file at all. If no path is specified, then you can have sensible defaults such as linting the whole directory etc.

There are 2 options, I can either change our generic test loader to make accomodations for swiftlint, which is the only linter which ignores a file even when explicitly pointed to. Simply allowing the --path variable to allow for non-swift files would save me in this circumstance and would help put this linter in line with a whole heap of others out there.

[SDGGiesbrecht]

So then the first and last lines here are running afoul of SwiftLint?

I think it would be wiser if Arcanist’s test loader were to consistently append a meaningful file extension when it creates the temporary files to begin with, but that is just how I would do things.

However, if you just want to get it working easily, I think you could just change line 67 from...

$basename = basename($file);

...to...

$basename = basename($file, ".lint-test");

...based on this example.

Then you just name your Swift test cases according to the pattern MyTestCase.swift.lint-test.

It looks to me like your existing test loader would then spit out temporary files with the .swift extension, and your other linters—which you say are agnostic about file extensions—should not care that the .lint-test extension has been removed from their temporary files.

[lswith]

I'll see what the arcanist maintainers say but its just surprising that swiftlint doesn't support exact file names. Ktlint, eslint, terraform lint, flake8, golint, pylint, rubylint, pep8, csslint, coffeelint, puppetlint, and rubocop all seem to be able to either specify the extension of the files they are looking for with sensible defaults or allow a direct file to be passed in.

[jpsim]

I really don't see why we wouldn't make SwiftLint work when passed single files to swiftlint lint --path not_a_swift_file.txt.

[marcelofabri]

We definitely should change that check from isSwiftFile to just check if it's a file. 👍