This is safer than just dropping the backend, as .close() will
block if something else is holding the mutex. Also means we can
drop the extra I18nBackend code.
Media syncing still needs fixing.
This code was an awful hack to provide some semblance of UI
responsiveness while executing DB statements on the main thread.
Instead, we can just run DB statements in a background thread now,
keeping the UI responsive.
We no longer need to worry about pysqlite implicitly beginning
transactions, and can be more explicit about beginning/ending
transactions
save() now also has a trx argument controlling whether a
transaction should be started / left open
Some initial testing with orjson indicates performance varies from
slightly better than pysqlite to about 2x slower depending on the type
of query.
Performance could be improved by building the Python list in rspy
instead of sending back json that needs to be decoded, but it may make
more sense to rewrite the hotspots in Rust instead. More testing is
required in any case.
committing the Protobuf implementation for posterity, but will replace
it with json, as Protobuf measures about 6x slower for some workloads
like 'select * from notes'
AD devs: before updating the sync version to 10, please make sure
the new timezone code has been ported over. The core change to the
scheduler is in _updateCutoff():
9736e4a970/pylib/anki/schedv2.py (L1357)
with the following supporting Python functions:
9736e4a970/pylib/anki/schedv2.py (L1410)
_timing_today() is calculated in Rust (lines up to 92, the rest are
tests):
9736e4a970/rslib/src/sched/cutoff.rs (L20)
The change went through a few iterations before stabilising, so it's
probably easier to refer to the above code than the patches that got
us to that point.
