Feature: support only running tests for modified files

[briancappello]

I took a stab at fixing #71 - this solution works for me, and hopefully should be flexible enough for others to make work relatively easily with their directory structures. But feedback is appreciated!

I added docs to the README, hopefully they should be sufficient explanation for how to use it.

Basically how it works:

On startup, run all tests, and watch for events (same as before).
On startup, collect a list of all the valid test filepaths (using py.test --collect-only)
On new events, update (if necessary) the valid test filepaths
On new events, collect the list of modified files from the events, and for each modified file, call an end-user-supplied function to convert the modified file path into its respective test file path. The test file paths are then appended to the args for the call to pytest. If any of the given test file paths are not in the list of valid tests, then run all tests.

Thanks!

edits: clarify explanation of implementation details

[blueyed]

Interesting, but rather complex (from a quick look).
Do you expect the transformation function to figure out what test (file) needs to run for some changed code?

Do you know about pytest-testmon, or pytest-incremental?
You could trigger / use those instead.

[briancappello]

@blueyed Thanks for taking a look!

Do you expect the transformation function to figure out what test (file) needs to run for some changed code?

Yes, I wasn't sure how else to do it (as different projects might have different conventions they follow).

Do you know about pytest-testmon, or pytest-incremental?

I did not! But I do now :) I'll take a look at them when I have some time later this week. Thanks for the suggestions, hopefully they can help simplify the code.