Make watchgod 10-100x faster

[fabioz]

I was considering watchgod for doing some file-watching and while reviewing the code I found one issue that I think is critical: watchgod is not taking full advantage of the scandir result!

Instead of doing a new os.stat for each file it should reuse what scandir already did (saving thousands of stat calls depending on the scenario).

I have some different targets, so, I won't be really use watchgod, but as I think this is really critical I decided to provide a pull request ;)

[fabioz]

As a note, it seems I have a typo in the commit (watchdog instead of watchgod), please rebase the fix to provide a proper commit message ;)

[samuelcolvin]

Thanks, this is a really good catch. I thought I looked at this when first creating the library, but either the docs/functionality have changed or I completely missed this.

Long term I want to switch to something like #25, but this is definitely a worth while fix until I get around to that.

As a note, it seems I have a typo in the commit (watchdog instead of watchgod), please rebase the fix to provide a proper commit message ;)

No problem, as noted in the readme my choice of name was definitely a mistake.

[codecov]

Codecov Report

Merging #75 (a9051a5) into master (26a1219) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master       #75   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            5         5           
  Lines          280       280           
  Branches        29        29           
=========================================
  Hits           280       280           

[samuelcolvin]

thanks so much, I'll run through other PRs and make a new release today.