@cchen: For pageviews and edit counts, what do you think about storing them in separate columns like views_desktop_user, views_mobileweb_spider, views_mobileapp_automated (and other combinations), edits_anon, edits_registered_user, edits_registered_bot?

I think it would be easier to query than if we had complex structures inside just two views & edits columns

Decisions from my chat with @cchen:

• Keep the dimensions from edits & views datasets as dimensions, yielding multiple rows per page per day (different combinations) as that will make addition of more dimensions easier in the future
• Start with 1 month of data and 100 pages (since we would have more than 1 row per page per day) to see how Presto handles it
• Increase volume of data (months & # of pages) iteratively

In T266714#6607288, @mpopov wrote:

Decisions from my chat with @cchen:

• Keep the dimensions from edits & views datasets as dimensions, yielding multiple rows per page per day (different combinations) as that will make addition of more dimensions easier in the future
• Start with 1 month of data and 100 pages (since we would have more than 1 row per page per day) to see how Presto handles it
• Increase volume of data (months & # of pages) iteratively

Oh wait, that doesn't make any sense. Edit counts & view counts are two separate metrics with completely separate dimensions. We have to do this:

separate columns like views_desktop_user, views_mobileweb_spider, views_mobileapp_automated (and other combinations), edits_anon, edits_registered_user, edits_registered_bot

IF we want a single table. Alternatively we could have two tables (view counts & edit counts).

@mpopov the queries in notebook and Superset looks good!

I created a top viewed pages dashboard with the dataset. I am still playing with the topic related dimensions and will add more topic related charts after.