The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

I think webdataset splits the file name and extension using the first dot no ?

https://github.com/webdataset/webdataset/blob/945b251a872ec0d337be8f9ea17a9c5b0d017ff3/webdataset/tariterators.py#L226

links to this function that splits on first dot

def base_plus_ext(path):
    """Split off all file extensions.

    Returns base, allext.

    Args:
        path: path with extensions

    Returns:
        path with all extensions removed
    """
    match = re.match(r"^((?:.*/|)[^.]+)[.]([^/]*)$", path)
    if not match:
        return None, None
    return match.group(1), match.group(2)

So maybe the original issue is actually due to one of the files containing a dot in its file name that is not for the extension

>>> base_plus_ext("15_Cohen_1-s2.0-S0929664620300449-gr3_lrg-b.png")
('15_Cohen_1-s2', '0-S0929664620300449-gr3_lrg-b.png')

Thanks for your review, @lhoestq.

I was not aware that webdataset requires filenames without dots in their basenames.

I they can have dots for the extension (that becomes the column name) but not in the key used to group files into samples