Adding assignment-unit cache to singleton DB

[JackUrb]

Overview

Assignment.get_units is currently the largest remaining call in the Mephisto profile when running with MephistoSingletonDB This PR looks to remove this chunk by adding a cache for this specific call.

Implementation

Creates a dict in the MephistoSingletonDB to hold a mapping between the given assignment_ids queried and the returned list of Units. Also adds a wrapper around new_unit to ensure that if a new unit is added to a given assignment_id we clear that cached value. Considering we're within a singleton, this is sufficient to ensure that the local units will always be up-to-date.

Testing

In python shell:

>>> from mephisto.abstractions.databases.local_singleton_database import MephistoSingletonDB
>>> db = MephistoSingletonDB()
>>> from mephisto.data_model.task_run import TaskRun
>>> len(TaskRun(db, 1188).get_assignments())
300
>>> TaskRun(db, 1188).get_assignment_statuses()
{'created': 0, 'launched': 2, 'assigned': 0, 'completed': 0, 'accepted': 0, 'mixed': 0, 'rejected': 0, 'soft_rejected': 0, 'expired': 298}
>>> TaskRun(db, 1188).get_assignment_statuses()
{'created': 0, 'launched': 2, 'assigned': 0, 'completed': 0, 'accepted': 0, 'mixed': 0, 'rejected': 0, 'soft_rejected': 0, 'expired': 298}

The second call to get_assignment_statuses() was instant, where the first took about a second.

[mojtaba-komeili]

Does throwing the exception happen in the super class?

[JackUrb]

Yes

[codecov-commenter]

Codecov Report

Merging #441 (be17f06) into master (e01d0ff) will decrease coverage by 0.09%.
The diff coverage is 83.33%.

@@            Coverage Diff             @@
##           master     #441      +/-   ##
==========================================
- Coverage   66.05%   65.95%   -0.10%     
==========================================
  Files          78       78              
  Lines        7105     7118      +13     
==========================================
+ Hits         4693     4695       +2     
- Misses       2412     2423      +11     

Impacted Files	Coverage Δ
...ephisto/abstractions/providers/mturk/mturk_unit.py	17.88% <0.00%> (+0.23%)	⬆️
mephisto/operations/operator.py	56.70% <66.66%> (-0.25%)	⬇️
...abstractions/databases/local_singleton_database.py	98.41% <92.85%> (-1.59%)	⬇️
...tractions/architects/channels/websocket_channel.py	79.06% <0.00%> (-3.49%)	⬇️
mephisto/abstractions/architects/mock_architect.py	88.88% <0.00%> (-2.62%)	⬇️
mephisto/operations/supervisor.py	78.08% <0.00%> (-0.70%)	⬇️
mephisto/data_model/task_run.py	83.33% <0.00%> (-0.53%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e01d0ff...be17f06. Read the comment docs.

[mojtaba-komeili]

Could you add this as hardcoded constant (probably all caps) at the top of the file or in a constants file (if you are using one). Specially having a descriptive name helps someone who may want to debug later see what is readily available for tweaking.

[mojtaba-komeili]

Thanks a lot Jack. Can't wait to run with new changes 😃