Beware of TSLint's no-unused-variable rule

[cartant]

I'm creating this issue as a reminder to myself and as a note that might be of use to others.

TSLint's no-unused-variable rule still seems to have problems - it has given me grief in the past. If it's enabled, the type information made available to rxjs-tslint-rules seems to change, resulting in the implementation sometimes being unable to determine whether or not an expression is an Observable. This can lead to false positives with the rxjs-no-unused-add rule, etc.

And it's not just rxjs-tslint-rules that's affected. With no-unused-variable enabled, I've seen false positives for TSLint's no-unnecessary-type-assertion rule, too.

[LogvinovLeon]

Did someone investigate that already? I've just spent quite a significant amount of time trying to figure out why no-unnecessary-type-assertion gives me false positives. Didn't expect the other unrelated linter rule to change the types. Will be nice to have that fixed.

[cartant]

@LogvinovLeon I put a fair bit of time into it and raised this issue. IIRC, that situation might have been fixed, but others were still broken. I gave up on that rule entirely, as it had wasted a lot of my time.

Have a look at the number of issues that relate to no-unused-variable before you commit yourself to looking into it.

[cartant]

For anyone who ends up here due to problems with the no-unused-variable rule, I've written an alternative. There is a no-unused-declaration rule in my tslint-etc package. It does much the same thing and includes a fixer (for unused imports).