You are probably asking what is the issue with cargo check being slow and wouldn’t that be mostly type checking and borrowing???? Yes summary it is, if you want stop here bye. If you want to be curious of the specifics continue reading.

Testing cargo check on big projects.

I was discussed that cargo check was so slow by @wildwestrom & @jakubDoka. I want to investigate to the root causes of why it’s slow so I closed an example from wildwestom and got started.

Getting a flame graph prepared

I did the basics setup measureme and get a basic flake graph locally:

• Link here

It turns out huh why is analysis taking so long, well it has serveral other components under the hood type_check_crate, etc. But as soon as I got down to the nity-grity it turns out that under coherent_trait impl_trait_header was taking the longest. It got down to metadata_decode_entry_impl_trait_header which is for proc macros I thought this is why cargo check is slow.

Summary

• Proc macros are the cause of cargo check being slow.
• Specifically metadata_decode_entry_impl_trait_header is slow.
• And finally it can be fixed.

I hope you have a great understanding of why `cargo check“ is so SLOWWWWW! STOP USING PROC MACROS.