Support excluding multiline regexes (fix #996)

[devdanzin]

This PR allows using multiline regexes in exclude_lines. It does so by matching regexes against the whole source code, instead of line by line as was done before.

The behavior is equivalent to adding a # pragma: no cover comment to each line matched by the multiline regex. Because of that, if a suite (e.g. if True:) is matched, all statements inside it will also be excluded, just like if there was a no-cover pragma on the suite line.

This is made possible because @nedbat figured out a O(N) algorithm for finding line numbers for the matches. With it, there's a small performance increase for excluding lines compared to the previous code.

Fixes #996.

[nedbat]

Please indent the code strings like this:

Suggested change

	parser = self.parse_text("""\
	def foo():
	print('Hello, world!')
	""", regex)
	parser = self.parse_text("""\
	def foo():
	print('Hello, world!')
	""", regex)

This should match the way other code is included in tests.

[devdanzin]

Thank you! Done.

[nedbat]

This isn't a multi-line regex?

[devdanzin]

Right, that was just a sanity check from when I was sketching this that slipped through. I'll rewrite the test so it covers multiple multi-line matches.

[nedbat]

We should also add a test for the #1803 example.

[nedbat]

Do you want to update the docs, or leave that to me?

[nedbat]

When this pull request is merged, we should update issues #118, #996, #1741, and #1803.

[devdanzin]

Do you want to update the docs, or leave that to me?

No, I'd rather let you do your thing with the docs. But thanks for asking :)

[nedbat]

This is now released as part of coverage 7.6.0.