Skip unchanged files which is checked

[yanqd0]

The pylint is too slow, especially when the project is large.

pytest-flakes can skip unchanged checked files by pytest cache.
This implementation is learned from that.

[coveralls]

Coverage increased (+0.2%) to 98.381% when pulling 6054701 on yanqd0:patch into ea48694 on carsongee:master.

[coveralls]

Coverage increased (+0.5%) to 98.901% when pulling e48f4d9 on yanqd0:patch into 875d155 on carsongee:master.

[carsongee]

This is really exciting, thanks so much for the contribution and great idea!

So due to earlier performance speedups, we're actually running all the tests in one shot on the session collection finish: https://github.com/carsongee/pytest-pylint/pull/83/files#diff-b8aeb412a793e07c521aa3f51c31e245R207 from the list added here: https://github.com/carsongee/pytest-pylint/pull/83/files#diff-b8aeb412a793e07c521aa3f51c31e245R187 (or a couple lines down if there are ignores). So in this PR as it stands, pylint is still running

I'm thinking everything here is still all setup correct, I think the cleanest way to stop the test from actually running is to add a property to PyLintItem, something like should_run that we populate in __init__, and then rework the section (https://github.com/carsongee/pytest-pylint/pull/83/files#diff-b8aeb412a793e07c521aa3f51c31e245R187), to be something closer to:

item = PyLintItem(path, parent)
if item.should_run:
    session.pylint_files.add(rel_path)
return item

so that we keep all the nice class encapsulation, but don't run the linter.

[yanqd0]

I agree. It is my mistake to ignore the real linting. Thanks for your guide.

The patch is rebased and your solution is implemented.

[carsongee]

Woot woot, thanks!

[carsongee]

This is now released, thanks again!