1353114 - Add Settings and Simple Measurements to main_summary

Status: RESOLVED FIXED

(Data Platform and Tools :: General, enhancement, P1)

Description

[Frank Bertsch [:frank]]

To support heavy users analysis, we need to add the following columns:

["simple_measurements."
 ["active_ticks",
  "main",
  "first_paint",
  "session_restored",
  "uptime",
  "total_time"]]
["settings."
 ["is_default_browser",
  "default_search_engine",
  "locale",
  "blocklist_enabled"]]

Comment 2

[Dave Zeber [:dzeber]]

(In reply to Frank Bertsch [:frank] from comment #0)
> ["settings."
>  ["is_default_browser",
>   "default_search_engine",
>   "locale",
>   "blocklist_enabled"]]

I'm pretty sure these ones are already there? (https://github.com/mozilla/telemetry-batch-view/blob/master/docs/MainSummary.md)

Comment 3

[Frank Bertsch [:frank]]

(In reply to Dave Zeber [:dzeber] from comment #2)
> (In reply to Frank Bertsch [:frank] from comment #0)
> > ["settings."
> >  ["is_default_browser",
> >   "default_search_engine",
> >   "locale",
> >   "blocklist_enabled"]]
> 
> I'm pretty sure these ones are already there?
> (https://github.com/mozilla/telemetry-batch-view/blob/master/docs/
> MainSummary.md)

Yes, just need to add simple measures now :)

Comment 4

[Benjamin Smedberg]

I suggest that total_time and probably uptime are not useful and may be analysis footguns: we should be using subsession_length instead. It's a leftover from pre-main-ping days. We probably ought to remove it from the ping entirely. Did somebody ask for that specifically?

Will the startup columns (main/first_paint/session_restored) only be in the first main ping for each session, and null otherwise?

Comment 5

[Frank Bertsch [:frank]]

(In reply to Benjamin Smedberg [:bsmedberg] from comment #4)
> Did somebody ask for that specifically?

These were asked for by the team working on heavy user analysis (i.e. bcolloran + rweiss). They are looking at long-term heavy-user analysis (e.g. YoY), so they may be interested in more historical data that has that fields, and thus how it compares currently to more appropriate fields like subsession_length,

> Will the startup columns (main/first_paint/session_restored) only be in the
> first main ping for each session, and null otherwise?

I'm not sure. This is client-side behavior that I don't know much about. Georg, any idea?

Comment 6

[Georg Fritzsche [:gfritzsche]]

(In reply to Frank Bertsch [:frank] from comment #5)
> (In reply to Benjamin Smedberg [:bsmedberg] from comment #4)
> > Did somebody ask for that specifically?
> 
> These were asked for by the team working on heavy user analysis (i.e.
> bcolloran + rweiss). They are looking at long-term heavy-user analysis (e.g.
> YoY), so they may be interested in more historical data that has that
> fields, and thus how it compares currently to more appropriate fields like
> subsession_length,

Can you confirm with them if these are actually important over subsession_length etc.?
These are to be dropped when we can get there. We should avoid any further dependencies on them.

> > Will the startup columns (main/first_paint/session_restored) only be in the
> > first main ping for each session, and null otherwise?
> 
> I'm not sure. This is client-side behavior that I don't know much about.
> Georg, any idea?

"main" should always be present.
"firstPaint" and "sessionRestored" should be set "nearly always".
Neither of them can be null. They are either valid or don't exist.

Comment 7

[Benjamin Smedberg]

If we're sending main/firstPaint/sessionRestored for all main pings, we need to add info/subsessionCounter to this dataset, and make sure that people analyzing the data realize that they should only analyze main/firstPaint/sessionRestored when subsessionCounter==1. Otherwise we could significantly overcount the number of starts.

Comment 8

[Frank Bertsch [:frank]]

(In reply to Benjamin Smedberg [:bsmedberg] from comment #7)
> If we're sending main/firstPaint/sessionRestored for all main pings, we need
> to add info/subsessionCounter to this dataset, and make sure that people
> analyzing the data realize that they should only analyze
> main/firstPaint/sessionRestored when subsessionCounter==1. Otherwise we
> could significantly overcount the number of starts.

I'm fine with adding subsessionCounter. I'll update the PR and it will be merged (tentatively) on Monday 5/8.

(In reply to Georg Fritzsche [:gfritzsche] from comment #6)
> Can you confirm with them if these are actually important over
> subsession_length etc.?
> These are to be dropped when we can get there. We should avoid any further
> dependencies on them.

Checking in now with them, Georg. Thanks for the info.

Comment 9

[Frank Bertsch [:frank]]

Will be backfilled with bug 1362161