Streamline DataFile Metadata Creation/Read/Update

[matthew-a-dunlap]

• Direct API endpoint to GET file level metadata. Using the JSON syntax we use for add/replace.

  • access via file PIDs
  • Access previous versions
  • Access draft version if authorized
  • All attributes

• API endpoint to update metadata without uploading a new file. This will be based off the file /update code.

  • All attributes
  • Restrict file if restrict is passed
  • General tests
    • Tags
    • File hierarchy
  • Should error if you attempt to update when there is already a draft

• Documentation

• Merge develop

[matthew-a-dunlap]

One question I have is what paths do we want?

• GET files/{fileId} fits the same structure as dataset, but to me conveys you are getting actual data not metadata.
• GET files/{fileId}/metadata seems to be more straightforward.

[matthew-a-dunlap]

Working on this story, I had to decide between two JSON "schema" used in our code to expand upon for this story. One is the schema used by most of the dataset methods, the other by the file-update and dataset-addFile commands. Both had some ability to edit file metadata.

I went with the latter one (used by update/add) as that code is more used in practice and amenable to expanding for fileMetadata GET and POST methods. Most important, the existing file-update method already contains the POST functionality needed, it just needs to be altered/forked to not require a file upload.

Before this decision I had started down the road of expanding functionality based on Dataset JSON schema, so today I changed my GET method over to the other schema. This involved fixing the incomplete serializer that had already been created but was unused in our code.

I'm hoping that this part of the work should only take another day or two, as the tests are mostly written and it shouldn't be too hard to break out the metadata part of the update endpoint. We had previously discussed fixing issues with dataset update allowing file MD5 to be changed (the goal was to make it superuser only), but with this new approach that code is not being touched. I think we should move that work to another story.