issue_comments

note that xarray.tutorial.open_rasterio will probably need something similar.

thanks for the report!

for those who don't want to click "download", here's the short patch: ```patch

xarray/tutorial.py | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)

--- a/xarray/tutorial.py +++ b/xarray/tutorial.py @@ -139,7 +139,10 @@ def open_dataset( url = f"{base_url}/raw/{version}/{path.name}"

 # retrieve the file

• filepath = pooch.retrieve(url=url, known_hash=None, path=cache_dir)
• fname = pathlib.Path(cache_dir, path).expanduser()
• if not fname.exists():
• fname = None
• filepath = pooch.retrieve(url=url, fname=fname, known_hash=None, path=cache_dir) ds = _open_dataset(filepath, engine=engine, **kws) if not cache: ds = ds.load() ```

note that the "unique" part of the file is the md5sum of the url, so this is not something random: python url = "https://github.com/mapbox/rasterio/raw/1.2.1/tests/data/RGB.byte.tif" filepath = pooch.os_cache("xarray_tutorial_data") / "-".join([ hashlib.md5(url.encode()).hexdigest(), path.name, ])

That said, this is because we use pooch.retrieve instead of pooch.create(...).fetch (which would save as the normal name, possibly in a separate directory if versioning was enabled), and we should probably add (date-based) versions to xarray-data so we can use the registry. Not sure, though.

In the meantime you could use fname even if the file does not already exist (which is actually supposed to be a file name, not a path) to make the patch a bit simpler, or recreate the "unique" filename using the hashed url.