We used this query to get pageview data in India by project and by access type, for the last three years:
SELECT project, year, access_method, SUM(view_count) as am_total FROM wmf.pageview_hourly WHERE (year >= 2017) AND agent_type='user' AND country_code = 'IN' AND page_id != 0 AND project IN ("bn.wikipedia", "hi.wikipedia", "ml.wikipedia", "pa.wikipedia", "ta.wikipedia", "te.wikipedia", "as.wikipedia", "sa.wikipedia", "kn.wikipedia", "tcy.wikipedia", "gu.wikipedia", "bn.wikipedia", "mr.wikipedia", "sat.wikipedia", "ur.wikipedia", "or.wikipedia", "en.wikipedia") GROUP BY project, year, access_method
The output doesn't align with data previously gathered on wikistats for the same time period.
See the wikistats query:
https://stats.wikimedia.org/v2/#/mr.wikipedia.org/reading/total-page-views/normal|bar|2-year|access~mobile-app|monthly
For example,
via the hive query in Jupyter Lab we see: 14,154 Marathi pageviews on mobile in 2019
via wikistats we see 67k Marathi pageviews on mobile in Sept 2019
We tried again with a simplified query and found a closer match though not yet matching numbers.
SELECT project, year, access_method, SUM(view_count) as am_total FROM wmf.pageview_hourly WHERE year = 2019 AND project = "mr.wikipedia" GROUP BY project, year, access_method
query results:
project year access_method am_total
mr.wikipedia 2019 desktop 27013186
mr.wikipedia 2019 mobile web 75567749
mr.wikipedia 2019 mobile app 509134