Admittedly, this is a little funky, but it's the easiest way I can think of at the moment. Because of the multipart spec, we need to a way to check for files, but the file attribute also has to be nulled/None-ed out when we serialize this down to JSON. Because of this, I'm just using aPath instance as a marker since a multipart file can be a file path or function.

Adding more tests shortly..

A few things:

• This looks like it will always callmutatate_multipart, even if there are no binary uploads, and therfore no reason to usemulti-part. Is that right? Maybe only usemulti-part when necessary?
• It feels like this decision to use multipart or not, and the few lines hre of what to do in that case are better abstracted away inside the internals ofself.mutate somehow. That way anything that is uploading binary in a mutation will just have that stuff handled implicitly.

This also feels like something that belongs in a more shared, non-CreateEtchPacket specific spot.

This seem seems akin tohttps://www.npmjs.com/package/extract-files

Our node-anvil library has a decent pattern to follow where it figures out if there files to maphere and then makes a multi-part request if necessaryhere.

The generalextract-files-like functionality is in the function here:https://github.com/anvilco/python-anvil/pull/41/files#diff-615974429dd7f7e34ce2e08e0012206ae6f47d86bcfc9338e6f72546e76436beR14

In this first pass, though, this is mainly to add multipart support oncreateEtchPacket since it's the only one that has underlying model support. Will have a more generic multipart mutation path after this PR.