Closed nicholasjng closed 5 months ago
For reviewers:
1) Should we explicitly keep allowing {r,w,x}t
modes if the only thing we do is silently coerce them into {r,w,x}b
? We have one such example in the readme code, which would need to be updated.
2) What's your feeling on finally adding an integration test for the data science zoo (see the commit message)?
EDIT: Seems that if we allow text mode, we should instead coerce to r,w,x
instead of their bytemode counterparts. I'll update the PR.
All modified and coverable lines are covered by tests :white_check_mark:
Comparison is base (
4ebee87
) 94.07% compared to head (8d1076d
) 94.10%.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
This is sensible because a) the builtin
fs.open()
was giving wrong type hints ofAbstractBufferedFile
s, which we no longer use, and b) the new lakeFS APIs allow native handling of non-bytemode open(), which the basefs.open()
needs a TextIOWrapper for.No new tests, since all file-system APIs are already being tested. What is missing are integration tests to pandas, polars, etc., since those all rely on
fs.open()
.