I’ve got a lot of code calling sqlite3.connect() in my program now. Particularly in my test suites. In MySQL and Postgres this can be slow and you end up needing connection pooling. I had a feeling it shouldn’t be in sqlite3 given its simplicity, but I wasn’t sure.
A bit of work with timeit later I can report it’s taking about 18usec to open a database. My connection function is more complex too, doing a select into a “globals” table to test the schema version. Even with all that work it’s 170usec, or < 2ms. That’s not super fast but it’s not a problem if my test suite does it 100 times.
$ python3 -m timeit -s 'import sqlite3' "sqlite3.connect('games.sqlite')" 100000 loops, best of 3: 17.6 usec per loop