Reduce import time by removing string and tomli._types imports

[hukkin]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00%. Comparing base (42a570d) to head (764c9ca).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff            @@
##            master      #249   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            4         4           
  Lines          529       525    -4     
  Branches        97        97           
=========================================
- Hits           529       525    -4     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[hauntsaninja]

Looks good!

Note that hiding the Iterable import doesn't do anything, typing imports collections.abc. If you want to get rid of that you'll need to do the TYPE_CHECKING = False trick

[hukkin]

Thanks for the review!

I think similarly the lazy import warnings does nothing here as there's some other module that is importing warnings. I made those changes while experimenting, and thought it may be good to leave them in, as they do no harm, and can in theory have an effect in the future if the imported libraries change.

Unfortunately, since we use typing.NamedTuple (used at runtime) I don't think even the TYPE_CHECKING = False trick will spare us from the need to import typing (and subsequent collections.abc).

I was wondering if the discussion about making TYPE_CHECKING a builtin deserves a revisit. Python 3.14 will make typing module also import annotationlib, so a little more import overhead that is hard to avoid in small (type annotated) CLI tools.

[hauntsaninja]

CPython has been moving towards deferring imports of warnings, so hopefully that one will pay off at some point.

Yeah, could be worth revisiting. If you want something now, in practice, type checkers will recognise if you just define TYPE_CHECKING yourself:

TYPE_CHECKING = False
if TYPE_CHECKING:
    from typing import List

def foo(x: List[int]):
    reveal_type(x)