refactor(tvix/eval): more efficiently intersect attributes

builtins.intersectAttrs is used a _lot_ in nixpkgs eval, for whatever
reason. We previously had a very inefficient implementation that would
allocate for each comparison. It stuck out like a sore thumb in perf
analysis.

This moves to a custom algorithm with two iterators, one for the left
and one for the right side, advancing them along the (borrowed) map
keys until a match is found and allocation is required.

I've not made any effort to reduce the verbosity of this code, I don't
think it's worth it.

On my machine this reduces the mean runtime of evaluating
`nixpkgs.emacs.outPath` by ~8%.

Change-Id: Ie506d82cb8d5f45909628f771a6b73e0eca16b27
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9898
Autosubmit: tazjin <tazjin@tvl.su>
Tested-by: BuildkiteCI
Reviewed-by: flokli <flokli@flokli.de>
This commit is contained in:
Vincent Ambo 2023-11-02 18:15:37 +03:00 committed by clbot
parent b7ea3d7b32
commit d91fcd4f28

View file

@ -5,6 +5,7 @@
use builtin_macros::builtins;
use genawaiter::rc::Gen;
use imbl::OrdMap;
use regex::Regex;
use std::cmp::{self, Ordering};
use std::collections::VecDeque;
@ -484,16 +485,76 @@ mod pure_builtins {
#[builtin("intersectAttrs")]
async fn builtin_intersect_attrs(co: GenCo, x: Value, y: Value) -> Result<Value, ErrorKind> {
let attrs1 = x.to_attrs()?;
let attrs2 = y.to_attrs()?;
let res = attrs2.iter().filter_map(|(k, v)| {
if attrs1.contains(k) {
Some((k.clone(), v.clone()))
} else {
None
let left_set = x.to_attrs()?;
if left_set.is_empty() {
return Ok(Value::attrs(NixAttrs::empty()));
}
let mut left_keys = left_set.keys();
let right_set = y.to_attrs()?;
if right_set.is_empty() {
return Ok(Value::attrs(NixAttrs::empty()));
}
let mut right_keys = right_set.keys();
let mut out: OrdMap<NixString, Value> = OrdMap::new();
// Both iterators have at least one entry
let mut left = left_keys.next().unwrap();
let mut right = right_keys.next().unwrap();
// Calculate the intersection of the attribute sets by simultaneously
// advancing two key iterators, and inserting into the result set from
// the right side when the keys match. Iteration over Nix attribute sets
// is in sorted lexicographical order, so we can advance either iterator
// until it "catches up" with its counterpart.
//
// Only when keys match are the key and value clones actually allocated.
//
// We opted for this implementation over simpler ones because of the
// heavy use of this function in nixpkgs.
loop {
if left == right {
// We know that the key exists in the set, and can
// skip the check instructions.
unsafe {
out.insert(
right.clone(),
right_set.select(right).unwrap_unchecked().clone(),
);
}
left = match left_keys.next() {
Some(x) => x,
None => break,
};
right = match right_keys.next() {
Some(x) => x,
None => break,
};
continue;
}
});
Ok(Value::attrs(NixAttrs::from_iter(res)))
if left < right {
left = match left_keys.next() {
Some(x) => x,
None => break,
};
continue;
}
if right < left {
right = match right_keys.next() {
Some(x) => x,
None => break,
};
continue;
}
}
Ok(Value::attrs(out.into()))
}
#[builtin("isAttrs")]