ORM is lazy?

[FalloutZone]

Hello all! I just came across this project and I think it looks very cool

One thing tho - is the ORM going to be lazy out of the box? Because as many of us know, the bottle neck for Django isn't python when the project gets bigger, it's that dammed ORM and it's naturally lazy loaded nature. People start using it it like a state machine rather than a database and the whole thing slows down to a crawl. No amount of rust is going to save you from that.

Will the ORM be lazy by default?
Can you turn off lazy loading easily when the project is created, or after the fact?
Will the admin panel rely on laziness?

Thanks a bunch! And I am excited to try this out

[m4tx]

Hello!

These are all valid concerns that you have. Partially because of how async works in Rust, and partially because we're trying to do things a bit more explicitly than Django's ORM, all operations must be explicitly executed using a Database object, e.g.:

use cot::db::query;

let link = query!(Link, $slug == LimitedString::new("cot").unwrap())
    .get(request.db())
    .await?;

(example taken from the guide)

Hopefully, this will make people think twice before executing anything—basically the idea is that if you take this request.db() object a lot in your code, probably you're doing something wrong (as each one probably translates into a database query).

On the side, some places are obviously non-optimal because of "laziness". For example, there is no way to prefetch ForeignKeys at the moment and you need to call ForeignKey::get on each one of them (potentially executing O(n) queries instead of O(1)), but it's only because the ORM is still immature. Eventually, we'd like to have something similar to Django's select_related.

So yeah, the ORM is not lazy by default, we should need no switch between lazy and non-lazy versions, and the admin panel should not rely on laziness. Let me know if there's anything more on your mind, as this is an important problem and it would be nice to get it right from the beginning.

[FalloutZone]

Thank you for the insight! Yes, if the library can be a "Pit of Success" that directs developers to doing the right thing most of the time, that would be a big win.

Developer: "Let's use rust. The payoff will be performance and infrastructure savings."
5 years later ...
Boss: "Product is complaining about performance, and the database instances are costing us $50,000 per month."
Developer: "On the bright side, we only need two API instances to service peak hours." :D