What is the logic in the rule sorting?

[wolfv]

Hi @mlschroe!

We have been hard at work porting your wonderful library to Rust (for extra memory safety and because it's much easier to read for ourselves).

We have been debugging the remaining differences between libsolv and our port and one difference is the order the rule evaluation.

I was wondering if the unifyrules_sortcmp is supposed to do an optimization for the order of rule evaluation or if it's just to remove duplicates?

We saw that in pubgrub there is an optimization to pick packages with fewest versions first (https://pubgrub-rs-guide.netlify.app/pubgrub_crate/strategy.html).

PS: if you are interested in the port, you can check it out in this PR - conda/rattler#236. It doesn't implement everything that libsolv can do (only what we need for conda packages so far). We are interested in expanding that a bit further though :) Would love to hear from you!

[mlschroe]

It's just used to remove duplicated rules.

[wolfv]

Thanks for the reply! I guess that will make it tricky to get exactly the same results as libsolv in all cases but let's see!

I am closing this issue – have a nice weekend!