Fix jsonData not passed correctly

[JR-1991]

Overview

This PR addresses the issue reported at datalad/datalad-dataverse#320, where metadata sent via NativeApi.replace_datafile wasn’t updating the file metadata correctly. The problem arose because the jsonData payload needs to be sent as form-data, which HTTPX’s json kwarg doesn’t handle properly.

The fix involves sending the payload via the data argument. However, since other endpoints require data to be sent through the json argument, this PR introduces a private method in api.py to determine the correct way to send the data. This method checks for the presence of the jsonData key and adjusts the argument accordingly. If the key isn’t found, the json argument is used by default.

To prevent similar issues in the future, the tests have been updated. The replace_datafile tests now ensure that the metadata is correctly updated.

TLDR

• If jsonData key is present sent through data kwarg
• If not, JSON payload is sent through json
• Extended tests to verify metadata is updated correctly

[mih]

I can confirm that this change makes the previously failing test in datalad-dataverse pass again. Thank you!

[pdurbin]

Nice work, @JR-1991!

[shoeffner]

Please note @adswa's changes in datalad/datalad-dataverse#322 and my alternative fix in #207, which is a PR onto this branch, not the main branch.

[JR-1991]

@shoeffner, thanks for the notice and the PR! Once I return from vacation, I will add a test to check for empty json_str to this PR.

[JR-1991]

I've included a test case to verify whether the behavior is as expected when json_str is set to None in the case of uploading a file. Would either @shoeffner or @pdurbin be available to review the PR?

[pdurbin]

@shoeffner are you interested in "triage" access? I just sent you an invite. You should see it at https://github.com/gdcc 😄

[shoeffner]

I haven't run the tests, so I trust you that it works :) The CI did, though, so this looks good!

This looks good and I just added a few comments as food for thought here and there. In general, I would merge it and then see what we can do in #189 to improve further.
Right now, this looks like a very good fix for the problem at hand.
Thanks!

[JR-1991]

@shoeffner I have updated this PR and specifically the tests to use pyDataverse methods instead of manual requests, as highlighted by you. The PR is now ready for review.

[pdurbin]

I didn't test this but tests are passing (and there are lots of new tests!) and the code makes sense.

[shoeffner]

Sorry for not responding earlier, but it looks good and I am happy it got merged!