[BUG]: Android controller endpoint doesn't seem to support fetching multiple versions of structures

[BenHenning]

Describe the bug

The Android asset fetching controller (/android_data endpoint) doesn't seem to correctly return multiple structure versions when requested.

Steps To Reproduce

Using curl:

curl "https://<base_oppia_url>/android_data?activity_type=subtopic&activities_data=%5B%7B%22id%22%3A%22C4fqwrvqWpRm-4%22%2C%22version%22%3A1%7D%2C%7B%22id%22%3A%22C4fqwrvqWpRm-4%22%2C%22version%22%3A2%7D%2C%7B%22id%22%3A%22C4fqwrvqWpRm-4%22%2C%22version%22%3A3%7D%2C%7B%22id%22%3A%22C4fqwrvqWpRm-4%22%2C%22version%22%3A4%7D%2C%7B%22id%22%3A%22C4fqwrvqWpRm-4%22%2C%22version%22%3A5%7D%2C%7B%22id%22%3A%22C4fqwrvqWpRm-4%22%2C%22version%22%3A6%7D%2C%7B%22id%22%3A%22C4fqwrvqWpRm-4%22%2C%22version%22%3A7%7D%2C%7B%22id%22%3A%22C4fqwrvqWpRm-4%22%2C%22version%22%3A8%7D%2C%7B%22id%22%3A%22C4fqwrvqWpRm-4%22%2C%22version%22%3A9%7D%5D" -H "X-ApiKey: <redacted_api_key>"

The above request is the following JSON string encoded:

"[{\"id\":\"C4fqwrvqWpRm-4\",\"version\":1},{\"id\":\"C4fqwrvqWpRm-4\",\"version\":2},{\"id\":\"C4fqwrvqWpRm-4\",\"version\":3},{\"id\":\"C4fqwrvqWpRm-4\",\"version\":4},{\"id\":\"C4fqwrvqWpRm-4\",\"version\":5},{\"id\":\"C4fqwrvqWpRm-4\",\"version\":6},{\"id\":\"C4fqwrvqWpRm-4\",\"version\":7},{\"id\":\"C4fqwrvqWpRm-4\",\"version\":8},{\"id\":\"C4fqwrvqWpRm-4\",\"version\":9}]"

Or, in formatted JSON:

[
  {
    "id":"C4fqwrvqWpRm-4",
    "version":1
  },
  {
    "id":"C4fqwrvqWpRm-4",
    "version":2
  },
  {
    "id":"C4fqwrvqWpRm-4",
    "version":3
  },
  {
    "id":"C4fqwrvqWpRm-4",
    "version":4
  },
  {
    "id":"C4fqwrvqWpRm-4",
    "version":5
  },
  {
    "id":"C4fqwrvqWpRm-4",
    "version":6
  },
  {
    "id":"C4fqwrvqWpRm-4",
    "version":7
  },
  {
    "id":"C4fqwrvqWpRm-4",
    "version":8
  },
  {
    "id":"C4fqwrvqWpRm-4",
    "version":9
  }
]

As shown, the request is trying to fetch versions 1-9 of subtopic C4fqwrvqWpRm-4 (subtopic 4 of topic C4fqwrvqWpRm).

However, the above curl request returns the following from the server:

{
  "C4fqwrvqWpRm-4":{
    "id":"C4fqwrvqWpRm-4",
    "topic_id":"C4fqwrvqWpRm",
    <pagedata...>
    "page_contents_schema_version":4,
    "language_code":"en",
    "version":9
  }
}

As seen, the response is only returning version 9.

Expected Behavior

The response should return all versions of the requested subtopic, not just the latest (or potentially last--I didn't verify this). I have verified that these other versions both exist on the server that was tested (backup migration server) and that they fetch through the standard subtopic controller.

Screenshots/Videos

No response

What device are you using?

Desktop

Operating System

Linux

What browsers are you seeing the problem on?

Other

Browser version

N/A

Additional context

No response

[seanlip]

I confirmed, through manual testing, that only the dict corresponding to the last item in the list is returned.

Here is the reason: https://github.com/oppia/oppia/blob/develop/core/controllers/android.py#L184

The handler seems to be assuming that activity_data versions don't repeat for the same ID. Therefore it is overwriting activities[activity_data['id']] with a new version each time. Here, activity_data['id'] is the same for all versions you're requesting, so that's why the return dict only has one element.

@BenHenning This should be fairly easy to fix, though I would advise not gating on it. What do you want the format of the return dict to be (here and in the other cases)?

[BenHenning]

Nice catch @seanlip.

I think that, in order to make the API more predictable, we should do away with the dict-like response and instead return a list of 1:1 responses for the requested ID/version pairs, in the same order as the request. Failures can be represented using null.

For the request structure:

[
  {
    "id": "<value>",
    "version": <version>
  },
  ...
]

We should have a corresponding response structure:

[
  {
    "id": "<value>",
    "version": <version>,
    "payload": {...}
  },
  ...
]

If a particular ID/version request cannot be fulfilled, the payload should be 'null'. The above format would allow for duplicate requests, but I think we should detect that and fail with a 400 in that case (since the script should make an effort not to request duplicates).

Do you think the above would work?

[seanlip]

Yup this structure seems fine to me. Thanks!