Support for data URI scheme?

[joouha]

Would you be interested in supporting data URIs in universal_pathlib?

A data URI consists of a scheme and a path, with no authority part, query string, or fragment. The optional media type, the optional base64 indicator, and the data are all parts of the URI path.

Most methods would be unimplemented, as the "path" is not a real hierarchical path 😃

I think the only useful method would be open, which would return the data encoded in the path.
So a DataPath wouldn't require fsspec.

[normanrz]

What would be the use case for this? Why not just use a bytes buffer?

[joouha]

My use case is for loading resources from web pages, such as images or CSS style-sheets. A resource link is typically a web address or a data URI:

<img src="http://example.com/image.png">
<img src="data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAUAAAAFCAYAAACNbyblAAAAHElEQVQI12P4//8/w38GIAXDIBKE0DHxgljNBAAO9TXL0Y4OHwAAAABJRU5ErkJggg==">

A DataPath class would allow universal_pathlib to be used to load both of these resources without having to handle data URIs separately - making it even more universal.

I currently have a DataPath implementation, which I register in upath.core.registry to make this work.

I was wondering if you'd consider including non-fsspec backed Path classes as part of this library?

[ap--]

I currently don't see how this is a good use case for upath. If I understand correctly, the only useful method would be .read_bytes(), and none of the others would make sense in this example.

Is there another example with data uris that requires more functionality for path handling?

[normanrz]

I agree with @ap--. If you have a DataPath implementation and can register that with upath in your application, that is great! I don't think it is a great fit for this library because there is limited added-value in terms of path handling.

[joouha]

Fair enough

[ap--]

@joouha this might soon be supported ootb with release of the next fsspec version. see: fsspec/filesystem_spec#1415

We'll have to add some tests, since I believe that most of the pathlib interface will just be non-functional. But the read methods might just work!

Cheers,
Andreas 😃

[joouha]

Ooh! That's awesome news 😃 😃 😃

Thanks Andreas