FlexMeasures / flexmeasures / 41821542780

88%

push

github

perf: skip two repeated queries when loading chart data (#2759)

* perf: skip two repeated queries when loading chart data

Part of #2753 (items 1 and 2).

TimedBelief.search asked the database for a materialized view of most recent beliefs once for each sensor, but FlexMeasures has no such view yet. It now passes use_materialized_view=False.

Sensor.as_dict ran its own query for the parent asset. It now reads the parent_asset relationship, which is usually loaded already.

Two tests record the SQL text of the queries and check that both are gone.

Signed-off-by: charanteja25 <charanteja.teja25@gmail.com>

* docs: add a changelog item for the chart data speed-up

Signed-off-by: charanteja25 <charanteja.teja25@gmail.com>

* ci: run the checks again, now that the clock change has passed

Signed-off-by: charanteja25 <charanteja.teja25@gmail.com>

* fix: look up the materialized view once per search, and keep using it if a host created one

Turning the view off for every search would silently stop using a view that a host
created by hand. Now TimedBelief.search looks the view up once, before the loop over
the sensors, and only when it asks for the most recent beliefs, as only that search can
use it. Each sensor then gets the view that was found, or no view at all, so
timely-beliefs does not look it up again for every sensor.

The tests check that a search of several sensors asks the database for the view once,
that a view created by a host is used and finds the same beliefs as without it, and that
a search that does not ask for the most recent beliefs does not look for the view.
The as_dict test now gives the asset a parent and checks the other sensors of that
asset, and the tests order the sensors by id. The changelog entry no longer says that
nothing changes for hosts with a view.

The code and the tests for the view are based on a patch by Ahmad-Wahid in the review of #2759.

Signed-off-by: charanteja25 <charanteja.teja25@gmail.com>

* fix: skip the vi... (continued)

3 of 3 new or added lines in 1 file covered. (100.0%)

23104 of 26346 relevant lines covered (87.69%)

0.88 hits per line