👖 👖 👖 Unscientific test result time 👖 👖 👖 We're less than halfway through the full call-by-name migration process, but just from the damage we've inflicted (or repaired?) on the backends - we've pretty dramatically reduced the scheduler initialization time on mainline pants. On

main

- in Pants.toml, I add/remove the JS prettier backend to cause the scheduler to re-run (also to see if a single backend is super noticeable). I then run

time pants --version

Before (Monday's build): ~17.5s Just now: ~11.5s I flipped back and forth between the checkouts a couple times because I didn't think the results would be THAT noticeable - I thought maybe 1-2 seconds by now would have been a win.

I'm super excited about this. seeing similar times on my machine with

time pants --no-pantsd --version

~14.5s (606729c) -> 8.5s (0c821701f5 my latest go pr)

The scheduler re-init has been the bane of my development. I've been maintaining patches that strip out backends and accommodate that in the rest of the repo, just to get down to 6-10 seconds

rust compile times next 😉

https://github.com/pantsbuild/pants/pull/21995 I haven't gone back to this - but we can slash/move a bunch of these deps. Personally, I'm a low dependency person - Rust makes that difficult - but even my Rust projects right now have like 7-8 direct deps and that feels like too many 🤦‍♂️