Better exceptions

[kushalkolar]

Hi! I have another suggestion for exception handling. I ran into this:

AttributeError                            Traceback (most recent call last)
Cell In[3], line 5
      1 path = "/wrong_path_by_mistake.ome.zarr"
      3 p = parse_url(path)
----> 5 reader = Reader(p)
      7 nodes = list(reader())
      8 image_node = nodes[0]

File ~/repos/ome-zarr-py/ome_zarr/reader.py:569, in Reader.__init__(self, zarr)
    568 def __init__(self, zarr: ZarrLocation) -> None:
--> 569     assert zarr.exists()
    570     self.zarr = zarr
    571     self.seen: list[ZarrLocation] = [zarr]

AttributeError: 'NoneType' object has no attribute 'exists'

I was wondering why the zarr object is None, but the real issue was that the path I fed in does not exist because of a typo. It would be useful to instead handle whether the path is valid earlier.

By the way, somewhat related an exception should be raised/handled rather than using an assert. As far as I know the standard practice is to use assert only in tests (there are a few reasons, one of them is that there are interpreter options that disable assertions).

Happy to do a PR but thought I'd get your input first.

[will-moore]

My proposal is to deprecate parse_url() in #476, since it has a bunch of issues like this and doesn't provide much, especially for writing data.