Update WikiPageviews and spec for new API behavior

[ragesoss]

We have a handful of specs in wiki_pageviews_spec.rb that recently started failing. It looks like this is because the behavior of the API has changed when handling requests for which no data is available, so our code needs to be updated to handle the current behavior.

The easiest way to gets started will be to run that spec, then figure out what changes need to be made to make the specs pass.

[gabina]

I'm working on this

[gabina]

These are the three failing tests:

1) WikiPageviews.views_for_article for an unviewed article returns an empty hash
     Failure/Error: raise PageviewApiError, response
     
     WikiPageviews::PageviewApiError:
       {"detail":"The date(s) you used are valid, but we either do not have data for those date(s), or the project you asked for is not loaded yet. Please check documentation for more information.","method":"get","status":404,"title":"Not Found","type":"about:blank","uri":"/metrics/pageviews/per-article/fr.wikisource/all-access/user/Voyages,_aventures_et_combats%2FChapitre_18/daily/2017040100/2017050100"}

2) WikiPageviews.average_views_for_article for an article that does not exist returns 0
   Failure/Error: raise PageviewApiError, response
   
   WikiPageviews::PageviewApiError:
     {"detail":"The date(s) you used are valid, but we either do not have data for those date(s), or the project you asked for is not loaded yet. Please check documentation for more information.","method":"get","status":404,"title":"Not Found","type":"about:blank","uri":"/metrics/pageviews/per-article/en.wikipedia/all-access/user/THIS_IS_NOT_A_REAL_ARTICLE/daily/2023082900/2023101700"}

3) WikiPageviews.average_views_for_article for an article that exist but has no view data returns 0
   Failure/Error: raise PageviewApiError, response
   
   WikiPageviews::PageviewApiError:
     {"detail":"The date(s) you used are valid, but we either do not have data for those date(s), or the project you asked for is not loaded yet. Please check documentation for more information.","method":"get","status":404,"title":"Not Found","type":"about:blank","uri":"/metrics/pageviews/per-article/fr.wikisource/all-access/user/Voyages,_aventures_et_combats%2FChapitre_18/daily/2023082900/2023101700"}

You can make one of the requests for the failing tests using this link.

As it's described on the issue, it looks like the pageviews API changed the behavior and started to response something like this when the article doesn't exist or the article doesn't have views in the given period:

{
    "detail": "The date(s) you used are valid, but we either do not have data for those date(s), or the project you asked for is not loaded yet. Please check documentation for more information.",
    "method": "get",
    "status": 404,
    "title": "Not Found",
    "type": "about:blank",
    "uri": "/metrics/pageviews/per-article/fr.wikisource/all-access/user/Voyages,_aventures_et_combats%2FChapitre_18/daily/2017040100/2017050100"
}

I wasn't able to find any documentation about the specific about:blank type in the docs.

I think a good approach to fix the tests could be to modify parse_results in the WikiPageviews class (link to method) to avoid raising PageviewApiError for these cases. @ragesoss do you mind checking this approach? If you're ok with this, I'll do a PR.

  def parse_results(response)
    return unless response
    data = Utils.parse_json(response)
    return data['items'] if data['items']
    # As of October 2017, the data type is https://www.mediawiki.org/wiki/HyperSwitch/errors/not_found
    return no_results if %r{errors/not_found}.match?(data['type'])
    return no_results if no_data_available_response?(data) # this is the new case
    raise PageviewApiError, response
  end

  # As of October 2023, we started to see 404 not found responses with about:blank type
  # and a specific detail when handling requests for which no data is available
  def no_data_available_response?(response)
    no_data_avialable_detail = 'The date(s) you used are valid, but we either do not have data '\
                               'for those date(s), or the project you asked for is not loaded yet.'\
                               ' Please check documentation for more information.'
    response['status'] == 404 && response['detail'] == no_data_avialable_detail
  end

[ragesoss]

Yes, that sounds like the right strategy @gabina.