Should we convert uploaded file metadata into Python types?

[pbugnion]

Following PRs #2724 and #2666 , the FileUpload.value traitlet looks like:

[{'name': 'example.csv',
  'type': 'text/csv',
  'size': 2548,
  'lastModified': 1536574751152,
  'content': <memory at 0x11d51ab88>}]

It is defined as List(Dict()).

We could consider the following, in increasing order of complexity:

1. change lastModified to be a datetime object. We should check browser compatibility on whether it's always expressed as a timestamp.
2. define a namedtuple or a Python class that exposes the attributes, rather than a dict. This would make the attributes more discoverable.

notifying @jtpio , @jasongrout who have been working on this.

[jasongrout]

define a namedtuple or a Python class that exposes the attributes, rather than a dict. This would make the attributes more discoverable.

Why not have both? Use a Bunch! :)

https://github.com/ipython/traitlets/blob/master/traitlets/utils/bunch.py

(That's what we use for the change argument in traitlets change notifications)

[jtpio]

Both suggestions sound good, especially if we want to stay close to the contents api: https://jupyter-notebook.readthedocs.io/en/stable/extending/contents.html#filesystem-entities

[vidartf]

As an xref, this might tie in with the datetime serialization in #2715. There, it is serialized along the pattern of DatePicker, as a dict of year, month, day, etc. Maybe we should switch to an ISO timestamp everywhere instead?

[pbugnion]

I'll try and address this issue.

[jasongrout]

Fixed in #2767.