New subject: [Wikistats 2.0] [Regular Update] Wrapping up Q2

3 Dec 2016


      We're starting to wrap up the calendar year, here's what we've accomplished
so far with Wikistats.  We're really excited to have some data in our
production Hive database for people to play with.  We worked really hard to
clean up and present an intuitive interface to all of mediawiki history.
The results are captured in the tables mentioned below, which we'll cover
more in an upcoming tech talk.  Documentation for the project is here
https://wikitech.wikimedia.org/wiki/Analytics/Data_Lake.
Our goals so far and progress breakdown:
1. [done] Build pipeline to process and analyze *pageview* data
2. [done] Load pageview data into an *API*
3. [        ] *Sanitize* pageview data with more dimensions for public
consumption
4. [ beta] Build pipeline to process and analyze *editing* data
5. [ beta] Load editing data into an *API*
6. [        ] *Sanitize* editing data for public consumption
7. [        ] *Design* UI to organize dashboards built around new data
8. [        ] Build enough *dashboards* to replace the main functionality
of stats.wikipedia.org
9. [        ] Officially Replace stats.wikipedia.org with *(maybe)
analytics.wikipedia.org
http://analytics.wikipedia.org/*
***. [         ] Bonus: *replace dumps generation* based on the new data
pipelines
4 & 5.  Since our last update, we've finished the pipeline that imports
data from mediawiki databases, cleans it up as best as possible, reshapes
it in a analytics-friendly way, and makes it easily queryable.  I'm marking
these goals as "beta" because we're still tweaking the algorithm for
performance and productionizing the jobs.  This will be completed early
next quarter, but in the meantime we have data for people to play with
internally.  Sadly we haven't sanitized it yet so we can't publish it.  For
those with internal access:
* https://pivot.wikimedia.org/#edit-history-test is the full history across
all wikis.  It's a bit hard to understand how to slice and dice, so we will
host a tech talk and present it at the January metrics meeting if we can.
* In hive, you can access this data in the wmf database, the tables are:
    - wmf.mediawiki_history: denormalized full history with this schema
https://wikitech.wikimedia.org/wiki/Analytics/Data_Lake/Mediawiki_history
    - wmf.mediawiki_page_history: the sequence of states of each wiki page (
schema
https://wikitech.wikimedia.org/wiki/Analytics/Data_Lake/Mediawiki_page_history
)
    - wmf.mediawiki_user_history: the sequence of states of each user
account (schema
https://wikitech.wikimedia.org/wiki/Analytics/Data_Lake/Mediawiki_user_history
)
6.  Sanitizing has not moved forward, as we need DBA time and they've been
overloaded.  We will attempt to restart this effort in Q3.
7.  We have begun the design process, we'll share more about this as we go.
Our goals and planning for next quarter support us finishing 4, 5, 7, and
8, so basically putting a UI on top of the data pipeline we have in place,
and updating it weekly.  We also hope to have good progress on 6, but that
depends on collaboration with the DBA team and is harder than we originally
imagined.
And remember, voice your opinions about important reports in the current
Wikistats here:
https://www.mediawiki.org/wiki/Analytics/Wikistats/DumpReports/Future_per_re...
 (thank you so so much to the many people who already chimed in).