lib: memory optimization of RepoPathComponent

Summary: As suggested by Yuya. This is actually a huge win in heap allocations
for repositories with a lot of files; according to the bucket statistics from
mimalloc, on gecko-dev, a repo of ~350,000 files, this reduces the number of
small alloctions (8, 16, or 32 byte size class) from nearly ~4 million to just
over ~200k, a 20x allocation reduction, with a total peak heap size reduction of
over 50MiB.

Signed-off-by: Austin Seipp <[email protected]>
Change-Id: I4014145d740367ecec5be32c6b9e2de8

Cargo.toml

@@ -35,6 +35,7 @@ chrono = { version = "0.4.31", default-features = false, features = [
     "clock",
 ] }
 config = { version = "0.13.2", default-features = false, features = ["toml"] }
+compact_str = "0.7.1"
 criterion = "0.5.1"
 crossterm = { version = "0.26", default-features = false }
 digest = "0.10.7"

lib/Cargo.toml

@@ -26,6 +26,7 @@ byteorder = { workspace = true }
 bytes = { workspace = true }
 chrono = { workspace = true }
 config = { workspace = true }
+compact_str = { workspace = true }
 digest = { workspace = true }
 futures = { workspace = true }
 either = { workspace = true }

lib/src/content_hash.rs

@@ -73,6 +73,12 @@ impl ContentHash for String {
     }
 }
 
+impl ContentHash for compact_str::CompactString {
+    fn hash(&self, state: &mut impl digest::Update) {
+        self.as_bytes().hash(state);
+    }
+}
+
 impl<T: ContentHash> ContentHash for Option<T> {
     fn hash(&self, state: &mut impl digest::Update) {
         match self {

lib/src/repo_path.rs

@@ -17,6 +17,7 @@
 use std::fmt::{Debug, Error, Formatter};
 use std::path::{Component, Path, PathBuf};
 
+use compact_str::CompactString;
 use itertools::Itertools;
 use thiserror::Error;
 
@@ -25,7 +26,7 @@ use crate::file_util;
 content_hash! {
     #[derive(Clone, PartialEq, Eq, PartialOrd, Ord, Debug, Hash)]
     pub struct RepoPathComponent {
-        value: String,
+        value: CompactString,
     }
 }
 
@@ -48,7 +49,9 @@ impl From<&str> for RepoPathComponent {
 impl From<String> for RepoPathComponent {
     fn from(value: String) -> Self {
         assert!(!value.contains('/'));
-        RepoPathComponent { value }
+        RepoPathComponent {
+            value: value.into(),
+        }
     }
 }
 
@@ -76,7 +79,7 @@ impl RepoPath {
             let components = value
                 .split('/')
                 .map(|value| RepoPathComponent {
-                    value: value.to_string(),
+                    value: value.into(),
                 })
                 .collect();
             RepoPath { components }
@@ -141,7 +144,7 @@ impl RepoPath {
         let repo_path_len: usize = self.components.iter().map(|x| x.as_str().len() + 1).sum();
         let mut result = PathBuf::with_capacity(base.as_os_str().len() + repo_path_len);
         result.push(base);
-        result.extend(self.components.iter().map(|dir| &dir.value));
+        result.extend(self.components.iter().map(|dir| dir.value.as_str()));
         result
     }