Hey, thank you so much for your answer - forgot to get back to you.

Absolutely - I’m thinking about that, too: how does one make sure that the analytics or data engineer has explored the data of the model they have built?
Asking for descriptive stats seems like a good way to do so, as even in new tables there aren’t that many “different-from-usual” columns

Thanks for the recommendation!! Currently in the process of implementing dbt-checkpoint in our CI for making sure that our style guide is followed through

I can think of adding a PR template that makes the developer say “yeah, I audited the data”, and that might be a first good practice. It’s just so contingent of the business requirements to make sure that the data can be useful

So glad you did 🐉

well this looks super cool! congrats, will get it for sure :)

little update: uv added ~3s of overhead, so 4s it is.

Point remains, though: `sdf lint` seems an actual contender to displace sqlfluff in the linter category just because of how fast it is

and I say this without wanting to throw any shade to sqlmesh at all. They've done so many amazing design choices.
However, following JS' trend of using Rust (or Zig or whatever) as the tooling language seems the right decision. Speed matters. It seems to unlock other cool stuff along the way too.