No annotation, no coverage

[dmnd]

When calculating coverage for a project, I don't care about coverage in
files that don't have the @flow pragma. While it's nice that Flow
infers types anyway, since it's never going to report an error it
doesn't help me at all and is potentially misleading. For example, my
project is 75% covered with master, but only 34% covered with this
change.

This PR changes flow-coverage-report to ignore coverage in unannotated
files. This could be considered a breaking change or a bugfix - I'm not
sure. You might consider making this a configuration option.

Resolves #85

[rpl]

@dmnd Yeah, I definitely agree.

When @ryan953 contributed the @flow annotation detection, we were discussing which kind of behavior would be reasonable to be the default.

The following can be a reasonable behavior:

• the command exit with a failure exit code if the report includes any file which doesn't have the @flow pragma
• define a new command line option which automatically exclude any file that doesn't include the @flow pragma (also customizable using the configuration, like the other command line options)

@dmnd @ryan953 How that sounds to you?

[dmnd]

That sounds pretty good to me @rpl 👍

[ryan953]

I was thinking that we'd want to add a flag in order to keep backwards compat for people, but either way sounds good. I would use the only-@flow files mode because it's easier to explain that "75% coverage of files with @flow, those other files don't count"

[dmnd]

I'd use "no annotation, no coverage" mode because I have a long term goal of getting to 100% coverage everywhere.

[arjun-menon]

This would be very useful for me as well. Are there plans to merge this soon, @dmnd? It seems like the only thing blocking this is a broken test. I'd happy to fix/add the tests for this.

[dmnd]

@arjun-menon I don't have time to make the changes @rpl suggested and fix the broken test. I've just been using my fork to measure Flow coverage in my project.

I'd really appreciate you taking this over and landing it! ❤️

[rpl]

Closing in favor of #141 (which also include a new test case)