Use Path in raw.filenames and epochs.filename

[mscheltienne]

Closes #12840

Attempting to use paths inside of raw.filenames and epochs.filename attributes.
Seems to work except for IO 'file-likes' with FIFF. It was a surprise to me that this was even possible, so I'm having a bit of an hard time getting this to work. Ideally, a file-like should not be listed in the filenames attribute which should then be set to None. It isn't informative to display "File-Like" or the repr(...). At the same time, I don't understand how preload works with a file-like object, since it needs the file-like object in raw.filenames to load the segments when preload=False.

[mscheltienne]

I'll pick this up tomorrow, let's see how much stuff breaks. I actually went further than my initial approach where now if _fiff_get_fid receives a str instead of Path, it will issue a warning. It should let me track down and remove a good portion of the existing conversion Path -> str which are left from #11473 and others. Then we can remove the warning and just cast to Path anyway for safety.

[larsoner]

Okay @mscheltienne I finally figured it out, and added some comments along the way. Can you look then mark as ready and merge-when-green if you're also happy?

[larsoner]

Failures on this and other PRs are just scipy/scipy#21623 and can be safely ignored (I think we can trust Windows pip-pre to continue being picky while the Linux pip-pre job fails in the dependencies step)

[mscheltienne]

Looks straightforward 😆 The comments are very clear, thanks for finishing this one!

[sappelhoff]

We are getting an error in mne-bids that may be related to this: https://github.com/mne-tools/mne-bids/actions/runs/11081048651/job/30792323679#step:16:416

It seems to be related to the list of raws that is being created here:

mne-python/mne/io/fiff/raw.py

Lines 96 to 116 in 5492174

	raws = []
	do_check_ext = not _file_like(fname)
	next_fname = fname
	while next_fname is not None:
	raw, next_fname, buffer_size_sec = self._read_raw_file(
	next_fname, allow_maxshield, preload, do_check_ext
	)
	do_check_ext = False
	raws.append(raw)
	if next_fname is not None:
	if not op.exists(next_fname):
	msg = (
	f"Split raw file detected but next file {next_fname} "
	"does not exist. Ensure all files were transferred "
	"properly and that split and original files were not "
	"manually renamed on disk (split files should be "
	"renamed by loading and re-saving with MNE-Python to "
	"preserve proper filename linkage)."
	)
	_on_missing(on_split_missing, msg, name="on_split_missing")
	break

with the eventual error being:

copy.deepcopy(raws[0].info),
E IndexError: list index out of range

[sappelhoff]

@mscheltienne is this expected behavior that I get an error here (on mne main)?

# %%
from pathlib import Path
import numpy as np
import mne

# create a dummy raw
sampling_freq = 1000
times = np.linspace(0, 1, sampling_freq, endpoint=False)
sine = np.sin(20 * np.pi * times)
cosine = np.cos(10 * np.pi * times)
data = np.array([sine, cosine])
info = mne.create_info(
    ch_names=["10 Hz sine", "5 Hz cosine"], ch_types=["misc"] * 2, sfreq=sampling_freq
)
raw = mne.io.RawArray(data, info)

# I want to set raw.filenames
a = Path.cwd() / "IDoNotExist.jpg"

# error: TypeError: 'tuple' object does not support item assignment
raw.filenames = (a,)

# this gives the right error: FileNotFoundError: File <...>/IDoNotExist.jpg not found.
raw.filenames = [a]

# this works correctly
#path_to_file_that_exists = ...
#raw.filenames = [path_to_file_that_exists]

# this works, no errors raised (which is ok, because I am using a private attribute to circumvent checks)
raw._filenames = (a,)