Community Surveys and Scoring

[jctanner]

Community scoring in old galaxy was based on a computed average of survey responses from the UX users. On each role or collection, they could rate aspects of the content on a 1-5 scale or a yes/no (which translated to 1 or 5). A user could change their answers at will, but could not delete them. After the POST call to provide their responses, a new score for the content would be computed and displayed in the UX.

There was also a scoring component that derived from ansible-lint outputs, which was called "quality score". This PR does not attempt to re-implement that function.

The new endpoints are in api/v1 for multiple reasons:

• _ui/v1 is unsupported / unstable and flagged for migration to api/v3
• api/v3 is no longer truly a galaxy_ng controlled api but is instead a redirect to pulp.
• this feature is not yet ready for being a full on generic thing that is useful to satellite or other pulp stakeholders
• we need to iterate quickly on it for the community site

Example survey payload ...

{'docs': 3,
 'does_what_it_says': 4,
 'ease_of_use': 3,
 'used_in_production': 1,
 'works_as_is': 0}

For a role, the payload should be POST'ed to /api/v1/surveys/roles/<roleid>/ For a collection, the payload should be POST'ed to /api/v1/surveys/collections/<namespace>/<name>/ ... because we have no way of getting the pulp_id for a collection from the API right now.

Computed scores come from these endpoints ...

(venv) [jtanner@p1 galaxy_ng]$ curl -s http://localhost:5001/api/v1/scores/roles/ | jq .
{
  "count": 1,
  "next": null,
  "previous": null,
  "results": [
    {
      "id": 7,
      "created": "2023-11-29T03:29:06.708629Z",
      "modified": "2023-11-29T03:29:10.572142Z",
      "role": 7,
      "namespace": "jctannerTEST",
      "name": "role1",
      "score": "1.98"
    }
  ]
}

(venv) [jtanner@p1 galaxy_ng]$ curl -s http://localhost:5001/api/v1/scores/collections/ | jq .
{
  "count": 2,
  "next": null,
  "previous": null,
  "results": [
    {
      "id": 1,
      "created": "2023-11-29T02:53:09.562222Z",
      "modified": "2023-11-29T02:53:13.397804Z",
      "collection": "018c18fe-59a2-7cd1-8574-ea51f2105133",
      "namespace": "autohubtest2",
      "name": "krdqdlyw",
      "score": "2.40"
    },
    {
      "id": 2,
      "created": "2023-11-29T03:29:31.486404Z",
      "modified": "2023-11-29T03:29:35.316854Z",
      "collection": "018c191f-a4b8-74cd-9389-a597da785d66",
      "namespace": "autohubtest2",
      "name": "adozvbxh",
      "score": "1.63"
    }
  ]
}

Scores have namespace and name filtering ...

(venv) [jtanner@p1 galaxy_ng]$ curl -s 'http://localhost:5001/api/v1/scores/collections/?namespace=autohubtest2&name=adozvbxh' | jq .                                                                                                         
{
  "count": 1,
  "next": null,
  "previous": null,
  "results": [
    {
      "id": 2,
      "created": "2023-11-29T03:29:31.486404Z",
      "modified": "2023-11-29T03:29:35.316854Z",
      "collection": "018c191f-a4b8-74cd-9389-a597da785d66",
      "namespace": "autohubtest2",
      "name": "adozvbxh",
      "score": "1.63"
    }
  ]
}

Both the role and the collection scores can be filtered by their roleid or collectionid, if known.

[rochacbruno]

Would be nice to add a data migration reading the existing static data on https://galaxy.ansible.com/static/scores/role.json and https://galaxy.ansible.com/static/scores/collection.json and adding to the new datamodels.

[jctanner]

@rochacbruno these models require the full survey data rather than the final computed scores. I've dumped everthing required and made scripts to import the old data here: https://github.com/jctanner/galaxy-hacking/blob/main/beta.surveys/load_data.py