Support coverage's exclude_lines setting

[rec]

I wasn't accurate in a previous issue: this feature is also gating for my project.

---

In real-world programs, there are significant quantities of code within otherwise tested files which are deliberately not unit tested due to diminishing returns and finite resources.

coverage uses the exclude_lines configuration file setting to match lines and scopes not to be included in coverage, usually with a specific comment - docs.

This is informative when reading code, and it also gives a more useful measure of actual coverage.

This sounds harder than my previous issues, because of the fact it applies to scopes as well as single lines.

---

coverage has a ton of other features, several of which we do use, but IMHO these two are the most important.

Thanks again for a very promising library!

[jaltmayerpizzorno]

Mmm, I wonder how coverage implements these... go through the source matching and collect the set of lines that match, excluding them? That would seem easier than matching in the AST...

[orsinium]

Mmm, I wonder how coverage implements these... go through the source matching and collect the set of lines that match, excluding them?

Yes, it's a regex on the raw line.

That would seem easier than matching in the AST...

It cannot be done using AST because the patterns in exclude_lines may contain comments. Even more, the default exclude_lines excludes the lines with # pragma: no cover comments.

[sigma67]

Semi-related, is there any way to exclude code blocks with pragma: no cover? It seems to only ignore the line that the no cover statement is on

if condition: # pragma: no cover
  print()  # will not be covered

[jaltmayerpizzorno]

Semi-related, is there any way to exclude code blocks with pragma: no cover? It seems to only ignore the line that the no cover statement is on

if condition: # pragma: no cover
  print()  # will not be covered

No, SlipCover doesn't currently support this, sorry.

[LilyFoote]

Being unable to exclude blocks is pretty much a blocker for me - I'd have to comment every line or live with a big reported regression in coverage, neither of which is desirable.

[jaltmayerpizzorno]

I hear you, folks. I will work on this ASAP.

[DanCardin]

Idk if you'd consider it part of this more general feature, but not counting stuff like if TYPE_CHECKING: blocks is also a blocker for me. I do have that as an explicitly exclude_lines for coverage, but (ditto with pragma: no cover, and assert_never) i wish it was just handled by default (and optionally turn-off-able)