Page MenuHomePhabricator

Further improvements to the WMCS edits dashboard
Closed, ResolvedPublic

Description

Here are some suggestions from @Bmueller to further improve the WMCS edits dashboard:

Note: By default 10 wikis are selected

  • Solution for the ambiguity of "allwikis" - alternative labels? A 4th view that only shows "allwiki" to not confuse it with all wikis?

Note: ALLWIKIS changed to TOTAL

Note: We are not going to change the checkbox behavior, as that would require overriding the default UI and the outcome won't be nice

  • ability to filter by month and not only by date in all 3 views

Note: This is currently possible. One can select dates in the same month in the left and right calendar tables (this may not be apparent in the UI though)

  • ability to specify wikis in a selection box like in the pageviews tool for comparison

Note: A nice to have feature in Dashiki.

  • Add small explanation for the different views

Added link to the meta page on the dashboard (commit details)

  • Investigate why the reportupdater didn't output data for tabular and hierarchical view
  • Investigate how the reportupdater can be improved to accept the date without the day of the month

Filed a task here: T245096

Event Timeline

Here is an update on suggested improvements:

  • In the Tabular View, we are now showing data sorted by wmcs edits percentage.
    • Technical details:
      • Patch submitted and merged: T240044.
      • Currently, Dashiki does not support sorting data in the Tabular View by multiple columns; filed a task around it: T240049
  • In the Timeseries View, we are now providing the option to select and compare all wikis.
    • UI changes this entails:
      • 10 popular wikis including TOTAL selected by default in the left menu.
      • ALLWIKIS changed to TOTAL
      • Data from a particular month listed under the last day of that month. e.g., 2019-11 as 2019-11-30. It makes more sense to me to have it this way than listing under the first day of the month as with that it seems like the data corresponds to the previous month.
    • Technical details:
WITH
        slice AS (
            SELECT
                wiki_db,
                SUM(IF(network_origin='wikimedia_labs', edit_count, 0)) AS wmcs_edits,
                SUM(edit_count) AS total_edits,
                ROUND(SUM(IF(network_origin='wikimedia_labs', edit_count, 0)) / SUM(edit_count), 3) AS wmcs_percent
            FROM wmf.editors_daily
            WHERE
                month = '2019-10'
            GROUP BY
                wiki_db
        ),
        total AS (
            SELECT
                last_day('2019-10-01') AS date,
                'TOTAL' AS wiki_db,
                ROUND(SUM(wmcs_edits) / SUM(total_edits), 3) AS wmcs_percent
            FROM slice
        )
    SELECT * FROM total
    UNION ALL
    SELECT
        last_day('2019-10-01') AS date,
        wiki_db,
        wmcs_percent
    FROM slice
    ORDER BY
        wiki_db
    LIMIT 10000
    ;
 SELECT
   last_day('2019-11-01') AS date,
   wiki_db,
   SUM(IF(network_origin='wikimedia_labs', edit_count, 0)) AS wmcs_edits
FROM wmf.editors_daily
WHERE
   month = '2019-11'
GROUP BY
   wiki_db
ORDER BY
   wiki_db
LIMIT 10000
;
  • To be investigated further: ability to filter by month and not only by date in all 3 views (to be further investigated...)

Note to self: once we lock these changes, I will make a patch containing modifications to hive queries.

(nothing more remaining here)