Thanks for your work on the library.

I have some questions / suggestions that I want to voice.

Regarding the dialects, have you considered "hard-coding" them in the library? Usually, one might want to avoid the silent import technique employed to do that here. If you hard-coded the 5 most common dialects, it would probably cover >90% of the use cases, without any side effects.

Further, for Postgres, would it not be a good idea to, at leas under the hood always do the Prepared interpolation technique, so that the variables that can be interpolated on the server is done there. Interpolating on the client side is, as you probably know, more risky, and adds one more layer of software that could have SQL injection bugs.

best regards
Rob