Fixes the issue where many contained objectives records were not inserted when running the Contained Objectives creation task or remixing a course section.
The cause was a bad assumption of the page's content structure: the logic assumed that objects with the activity-reference type were only at the top level of the model, and they could be at any level.
Technical notes:
The bugfix consists of a recursive function written in PostgreSQL, that receives a jsonb representing a page's content and recursively iterates over the nested structure to find all the activity ids. Another solution could be to implement the same function in Elixir, which would force the query to load all the page contents for a course in memory, and in big courses that could not be optimal. The SQL function not only avoids that, but it also reduces the query's processing time.
Here are the results when running the query for the RealCHEM course locally, before and after the solution. Note that the processing times decrease after the fix, except for the contained-objectives-creation time that obviously increased since we are inserting 800 more rows.
For testing and running the fix in prod, someone with console access should run the mix task again but with the --all flag to remove all the existing contained objectives and create new ones.
NG23-227
Fixes the issue where many contained objectives records were not inserted when running the Contained Objectives creation task or remixing a course section. The cause was a bad assumption of the page's content structure: the logic assumed that objects with the
activity-referencetype were only at the top level of the model, and they could be at any level.Technical notes:
The bugfix consists of a recursive function written in PostgreSQL, that receives a
jsonbrepresenting a page's content and recursively iterates over the nested structure to find all the activity ids. Another solution could be to implement the same function in Elixir, which would force the query to load all the page contents for a course in memory, and in big courses that could not be optimal. The SQL function not only avoids that, but it also reduces the query's processing time.Here are the results when running the query for the RealCHEM course locally, before and after the solution. Note that the processing times decrease after the fix, except for the contained-objectives-creation time that obviously increased since we are inserting 800 more rows.
Without
psqlfunctionbuild_contained_objectives pages: 717.81075ms build_contained_objectives activities: 199.85220800000002ms build_contained_objectives contained_objectives: 0.147708ms contained_objectives_count: 48
With
psqlfunctionbuild_contained_objectives pages: 155.04484200000002ms build_contained_objectives activities: 24.20246ms build_contained_objectives contained_objectives: 1.038736ms contained_objectives_count: 848
Testing notes:
For testing and running the fix in prod, someone with console access should run the mix task again but with the
--allflag to remove all the existing contained objectives and create new ones.mix create_contained_objectives --all