fix: Operator dash user counts, performance #839
Conversation
Don't use the user_pii table for user counts, since it may not be populated. Change this chart to be a count of active enrollments, which people have said is more useful. Also changed the time range for operator dash to 90 days to improve performance of the actors / events graphs.
bmtcril
force-pushed
the
bmtcril/operator_dash_fixes
branch
from
17a05e8 to
0cc17dc
Compare
| -- Need to cast the course key to a string here otherwise the | ||
| -- course_names dictionary throws this: | ||
| -- Key type for complex key at position 0 does not match, expected String, found LowCardinality(String). | ||
| on course_names.course_key = enrollment.course_key::String |
There was a problem hiding this comment.
Doesn't this affect DBT reports too?
There was a problem hiding this comment.
dbt recreates everything with the type of the parent, so they almost all get converted to LowCardinality. LC types get automatically coerced into their parent type (String for us) in almost all cases, there have only been two places I found where it gets confused, both around dictionary joins.
There was a problem hiding this comment.
And custom reports that do the same join such as the fact_*_engagement ones?
There was a problem hiding this comment.
The fact_*_engagement charts are working as-is, but I can't guarantee what other custom reports might run into. That's likely a problem with a number of changes we've been making, though.
Don't use the user_pii table for user counts, since it may not be populated. Change this chart to be a count of active enrollments, which people have said is more useful.
Also changed the time range for operator dash to 90 days to improve performance of the actors / events graphs.
Closes: #788