Set blob Content-Type in AzureBlobFileSystem.put()

[jhamman]

Is it possible to set the Content-Type attribute (or other content settings) of blobs using adlfs?

I would like to do something like this:

fs = AzureBlobFileSystem(...)
fs.put('index.html', 'foo/bar', content_settings={'Content-Type': 'text/html'})

This does seem to be possible when using the azure storage library directly. For example:

blob_client = ...
with open(index.html, "rb") as data:
    blob_client.upload_blob(data, content_settings=ContentSettings(content_type='text/html'))

I suspect this is possible today but the above example does not work so I'm looking for examples and/or documentation to support this sort of workflow.

cc @orianac

[TomAugspurger]

That method take **kwags but silently ignores them. I think we could pass them through to BlobClient.upload_blob at https://github.com/fsspec/adlfs/blob/master/adlfs/spec.py#L1575-L1581.

IMO, it'd be best to just have users use the azure.storage.blob objects rather than trying to wrap them / infer what they mean from something like content_settings={"Content-Type": ...}. So I'd suggest just supporting

fs.put('index.html', 'foo/bar', content_settings=azure.storage.blob.ContentSettings(...))

by passing through kwargs unmodified.

[isidentical]

Not sure if there are any complications to do this on azure blog storage, but just as passing-by note, s3fs does this automatically by guessing the mime type (and supports overriding, if you need a specific content type): https://github.com/fsspec/s3fs/blob/736ee5ae7a16494e14fa12838cd963e1472afb9e/s3fs/core.py#L967-L970

[TomAugspurger]

https://github.com/fsspec/filesystem_spec/issues/916 is an upstream issue for standardizing this. For now, I'm going to pass through kwargs so that we at least have the option to set it using azure.storage.blob.ContentSettings.

[TomAugspurger]

Actually, just passing through kwargs isn't going to do it. They're silently dropped at https://github.com/fsspec/filesystem_spec/blob/504483cec8a1d7f9afb77d9b6b637f40f74b5a81/fsspec/spec.py#L661-L664.

So I'm probably just going to keep working around this for now :/

[TomAugspurger]

Fixed by #392.