accept pathlib paths wherever a path/file is expected

[flying-sheep]

i’d like pd.read_table(some_path / 'filename.tsv') to work without going through str()

[jreback]

sure, these just need stringifing. want to do a pull-request?

[flying-sheep]

gladly, but we need to define how first:

• stringify everything that isn’t file-like and try treating it as a path?

  (there might be a problem if some custom __str__ function of an accidentally passed object happens to be a valid path)

• test for being a Path subclass?

  this might need backwards-sensible code like

  try:
      from pathlib import Path
  except:
      class Path: pass  # dummy for instance checks
  
  ...
  
  if isinstance(file_or_path, Path):
      file_or_path = str(file_or_path)

[jreback]

here is where all of this path inference is done (well not 100% sure of all but vast majority)

you can use several/both of those strategies (though keep in mind that other pathlib like libs exists, e.g. py.local.path is the big one).

so I think its probably safe to str(filepath_or_buffer) (if this is not a file-like) might just work, but some tests would be gr8.

[flying-sheep]

io has no tests. should i start a module?

[jreback]

all of the io tests are in their respective directory. however for this you can add in pandas/io/tests/test_common.py (ideally you can add for all of the major ones, e.g. excel,sql,hdf5,csv,json,stata)

[flying-sheep]

the tests are a bit uninspired but yeah 😄

[jreback]

closed by #11051