Excellent, I guess this highlights a few problems we have that I would like you to address:
-
We should always be generating strings in our “selection” perf tests. We should fix all our existing specs to do so. I don’t care if we already have data for specs, we can backfill the corrections.
-
We should add a spec that selects a large amount of columns from a table and generates a giant string.
-
We should either always use
sequel_pg
OR optionally use it. I think for simplicity sake we should always use it, cause that is what people are using in production. -
We got to ensure parity between our AR and Sequel specs, otherwise we can not compare them, fixing up that ILIKE is urgent.
Thoughts?
cc @jeremyevans