We should consider running our test suite with a real mongo database, it would allow better testing of mongo details, for example
aggregate() calls which
mim (mongo in memory) cannot emulate at this point.
Several things would have to change:
wipe_databasemethod to include those database names
mimis already isolated per-process since its in memory)
"mim" is also a bit faster. In my single-process test run it was 70% the time of real mongo. So we'd be sacrificing some speed.
Log in to post a comment.