REF: Split up parsers into smaller files

[phofl]

• closes TST/REF: splitting pandas/io/parsers.py into multiple files #38930
• Ensure all linting tests pass, see here for how to run them

Refactored them into 4 files, similar to described in the issue. If this is accepted I will try to clean this up a bit more through unifying the calls a bit

[jreback]

i would update pandas/io/parsers/__init__.py to import read_csv, read_fwf, read_table, and other public functions here (if any others).

you can leave .readers just import them to __init__.py (then you don't need to change the other modules that import).

otherwise this is a straight move?

[phofl]

Thanks, will do that.

Yes, just moved them around. Will go to refactoring after this

[phofl]

I moved the relevant functions up, I changed the import paths for the helper functions in the test functions

[jreback]

lgtm ping on green

[jreback]

https://github.com/pandas-dev/pandas/pull/39217/checks?check_run_id=1730796224 looks like a real fail

[phofl]

Have to check, this one worked 3 days ago in the pipelines

[phofl]

Hm merged master, let's check if this persists. Can not reproduce this locally

[phofl]

Additionally this worked the commit before when I forgot a few imports. At least a place to start pinning if it persists

[phofl]

Green now

@twoertwein I think you worked with unclosed handles the last few days/weeks? Could you maybe have a look and see if this is familiar?

[phofl]

Failing https://github.com/pandas-dev/pandas/runs/1731152527?check_suite_focus=true too, unrelated here then

[twoertwein]

@twoertwein I think you worked with unclosed handles the last few days/weeks? Could you maybe have a look and see if this is familiar?

I assume that most of the ssl warnings come from boto3 (I honestly gave up on them for now in #39047). Wasn't a test like that recently xfailed (reason s3/boto3 being updated)?

[phofl]

Yes, but this test really failed not only raised a warning,

[twoertwein]

do you mean the OSError: [Errno 98] Address already in use in test_server_and_default_headers test? Is it possible that multiple instances of pytest run on the same machine but they like to use the same port?

or test_warn_if_chunks_have_mismatched_type. I think I have looked at this (or a similar) test when working on #39047. As far as I remember, there shouldn't be a need for any "ssl-stuff" in this test. The only resource that might not be closed is StringIO. Locally (on linux) I can not re-create most of the ResourceWanrings.

[phofl]

Sorry this was not clear enough by myself. I was talking about test_warn_if_chunks_have_mismatched_type. Have the same issue, can not reproduce locally.

[jreback]

might be worth moving from here down to another module (validation.py?) in a followup, e.g. keep readers.py for the actual readers

[jreback]

thanks @phofl