Use arcstr for package, extra, and group names

Author: charliermarsh

Cargo.lock

@@ -71,6 +71,7 @@ uv-workspace = { path = "crates/uv-workspace" }
 
 anstream = { version = "0.6.15" }
 anyhow = { version = "1.0.89" }
+arcstr = { version = "1.2.0" }
 async-channel = { version = "2.3.1" }
 async-compression = { version = "0.4.12", features = ["bzip2", "gzip", "xz", "zstd"] }
 async-trait = { version = "0.1.82" }

Cargo.toml

@@ -11,6 +11,7 @@ doctest = false
 workspace = true
 
 [dependencies]
+arcstr = { workspace = true }
 rkyv = { workspace = true }
 schemars = { workspace = true, optional = true }
 serde = { workspace = true, features = ["derive"] }

crates/uv-normalize/Cargo.toml

@@ -4,6 +4,7 @@ use std::str::FromStr;
 
 use serde::{Deserialize, Deserializer, Serialize};
 
+use crate::small_string::SmallString;
 use crate::{validate_and_normalize_owned, validate_and_normalize_ref, InvalidNameError};
 
 /// The normalized name of an extra dependency.
@@ -16,7 +17,7 @@ use crate::{validate_and_normalize_owned, validate_and_normalize_ref, InvalidNam
 /// - <https://packaging.python.org/en/latest/specifications/name-normalization/>
 #[derive(Debug, Default, Clone, PartialEq, Eq, Hash, PartialOrd, Ord, Serialize)]
 #[cfg_attr(feature = "schemars", derive(schemars::JsonSchema))]
-pub struct ExtraName(String);
+pub struct ExtraName(SmallString);
 
 impl ExtraName {
     /// Create a validated, normalized extra name.

crates/uv-normalize/src/extra_name.rs

@@ -5,6 +5,7 @@ use std::sync::LazyLock;
 
 use serde::{Deserialize, Deserializer, Serialize, Serializer};
 
+use crate::small_string::SmallString;
 use crate::{validate_and_normalize_owned, validate_and_normalize_ref, InvalidNameError};
 
 /// The normalized name of a dependency group.
@@ -14,7 +15,7 @@ use crate::{validate_and_normalize_owned, validate_and_normalize_ref, InvalidNam
 /// - <https://packaging.python.org/en/latest/specifications/name-normalization/>
 #[derive(Debug, Default, Clone, PartialEq, Eq, Hash, PartialOrd, Ord)]
 #[cfg_attr(feature = "schemars", derive(schemars::JsonSchema))]
-pub struct GroupName(String);
+pub struct GroupName(SmallString);
 
 impl GroupName {
     /// Create a validated, normalized group name.

crates/uv-normalize/src/group_name.rs

@@ -1,6 +1,7 @@
 use std::error::Error;
 use std::fmt::{Display, Formatter};
 
+use crate::small_string::SmallString;
 pub use dist_info_name::DistInfoName;
 pub use extra_name::ExtraName;
 pub use group_name::{GroupName, DEV_DEPENDENCIES};
@@ -10,21 +11,31 @@ mod dist_info_name;
 mod extra_name;
 mod group_name;
 mod package_name;
+mod small_string;
 
 /// Validate and normalize an owned package or extra name.
-pub(crate) fn validate_and_normalize_owned(name: String) -> Result<String, InvalidNameError> {
+pub(crate) fn validate_and_normalize_owned(name: String) -> Result<SmallString, InvalidNameError> {
     if is_normalized(&name)? {
-        Ok(name)
+        Ok(SmallString::from(name))
     } else {
-        validate_and_normalize_ref(name)
+        Ok(SmallString::from(normalize(&name)?))
     }
 }
 
 /// Validate and normalize an unowned package or extra name.
 pub(crate) fn validate_and_normalize_ref(
     name: impl AsRef<str>,
-) -> Result<String, InvalidNameError> {
+) -> Result<SmallString, InvalidNameError> {
     let name = name.as_ref();
+    if is_normalized(name)? {
+        Ok(SmallString::from(name))
+    } else {
+        Ok(SmallString::from(normalize(name)?))
+    }
+}
+
+/// Normalize an unowned package or extra name.
+fn normalize(name: &str) -> Result<String, InvalidNameError> {
     let mut normalized = String::with_capacity(name.len());
 
     let mut last = None;
@@ -136,9 +147,14 @@ mod tests {
             "FrIeNdLy-._.-bArD",
         ];
         for input in inputs {
-            assert_eq!(validate_and_normalize_ref(input).unwrap(), "friendly-bard");
             assert_eq!(
-                validate_and_normalize_owned(input.to_string()).unwrap(),
+                validate_and_normalize_ref(input).unwrap().as_ref(),
+                "friendly-bard"
+            );
+            assert_eq!(
+                validate_and_normalize_owned(input.to_string())
+                    .unwrap()
+                    .as_ref(),
                 "friendly-bard"
             );
         }
@@ -169,9 +185,11 @@ mod tests {
         // Unchanged
         let unchanged = ["friendly-bard", "1okay", "okay2"];
         for input in unchanged {
-            assert_eq!(validate_and_normalize_ref(input).unwrap(), input);
+            assert_eq!(validate_and_normalize_ref(input).unwrap().as_ref(), input);
             assert_eq!(
-                validate_and_normalize_owned(input.to_string()).unwrap(),
+                validate_and_normalize_owned(input.to_string())
+                    .unwrap()
+                    .as_ref(),
                 input
             );
             assert!(is_normalized(input).unwrap());

crates/uv-normalize/src/lib.rs

@@ -1,8 +1,9 @@
+use serde::{Deserialize, Deserializer, Serialize};
 use std::borrow::Cow;
+use std::cmp::PartialEq;
 use std::str::FromStr;
 
-use serde::{Deserialize, Deserializer, Serialize};
-
+use crate::small_string::SmallString;
 use crate::{validate_and_normalize_owned, validate_and_normalize_ref, InvalidNameError};
 
 /// The normalized name of a package.
@@ -27,7 +28,7 @@ use crate::{validate_and_normalize_owned, validate_and_normalize_ref, InvalidNam
 )]
 #[cfg_attr(feature = "schemars", derive(schemars::JsonSchema))]
 #[rkyv(derive(Debug))]
-pub struct PackageName(String);
+pub struct PackageName(SmallString);
 
 impl PackageName {
     /// Create a validated, normalized package name.
@@ -56,7 +57,7 @@ impl PackageName {
 
             Cow::Owned(owned_string)
         } else {
-            Cow::Borrowed(self.0.as_str())
+            Cow::Borrowed(self.0.as_ref())
         }
     }
 

crates/uv-normalize/src/package_name.rs

@@ -0,0 +1,102 @@
+use std::cmp::PartialEq;
+use std::ops::Deref;
+//
+// use rkyv::rancor::{Fallible, Source};
+// use rkyv::ser::{Allocator, Writer};
+// use rkyv::string::{ArchivedString, StringResolver};
+// use rkyv::{Archive, Place};
+
+/// An optimized small string type for short identifiers, like package names.
+///
+/// Represented as an [`arcstr::ArcStr`] internally.
+#[derive(Debug, Default, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub(crate) struct SmallString(arcstr::ArcStr);
+
+impl From<&str> for SmallString {
+    fn from(s: &str) -> Self {
+        Self(s.into())
+    }
+}
+
+impl From<String> for SmallString {
+    fn from(s: String) -> Self {
+        Self(s.into())
+    }
+}
+
+impl AsRef<str> for SmallString {
+    fn as_ref(&self) -> &str {
+        &self.0
+    }
+}
+
+impl Deref for SmallString {
+    type Target = str;
+
+    fn deref(&self) -> &Self::Target {
+        &self.0
+    }
+}
+
+/// A [`serde::Serialize`] implementation for [`SmallString`].
+impl serde::Serialize for SmallString {
+    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: serde::Serializer,
+    {
+        self.0.serialize(serializer)
+    }
+}
+
+/// An [`rkyv`] implementation for [`SmallString`].
+impl rkyv::Archive for SmallString {
+    type Archived = rkyv::string::ArchivedString;
+    type Resolver = rkyv::string::StringResolver;
+
+    #[inline]
+    fn resolve(&self, resolver: Self::Resolver, out: rkyv::Place<Self::Archived>) {
+        rkyv::string::ArchivedString::resolve_from_str(&self.0, resolver, out);
+    }
+}
+
+impl<S> rkyv::Serialize<S> for SmallString
+where
+    S: rkyv::rancor::Fallible + rkyv::ser::Allocator + rkyv::ser::Writer + ?Sized,
+    S::Error: rkyv::rancor::Source,
+{
+    fn serialize(&self, serializer: &mut S) -> Result<Self::Resolver, S::Error> {
+        rkyv::string::ArchivedString::serialize_from_str(&self.0, serializer)
+    }
+}
+
+impl<D: rkyv::rancor::Fallible + ?Sized> rkyv::Deserialize<SmallString, D>
+    for rkyv::string::ArchivedString
+{
+    fn deserialize(&self, _deserializer: &mut D) -> Result<SmallString, D::Error> {
+        Ok(SmallString::from(self.as_str()))
+    }
+}
+
+impl PartialEq<SmallString> for rkyv::string::ArchivedString {
+    fn eq(&self, other: &SmallString) -> bool {
+        **other == **self
+    }
+}
+
+impl PartialOrd<SmallString> for rkyv::string::ArchivedString {
+    fn partial_cmp(&self, other: &SmallString) -> Option<::core::cmp::Ordering> {
+        Some(self.as_str().cmp(other))
+    }
+}
+
+/// An [`schemars::JsonSchema`] implementation for [`SmallString`].
+#[cfg(feature = "schemars")]
+impl schemars::JsonSchema for SmallString {
+    fn schema_name() -> String {
+        String::schema_name()
+    }
+
+    fn json_schema(_gen: &mut schemars::gen::SchemaGenerator) -> schemars::schema::Schema {
+        String::json_schema(_gen)
+    }
+}

crates/uv-normalize/src/small_string.rs

@@ -488,6 +488,7 @@ fn encode_package_group(package: &PackageName, group: &GroupName) -> ExtraName {
 #[cfg(test)]
 mod tests {
     use super::*;
+    use std::str::FromStr;
 
     use uv_pypi_types::ConflictSet;
 
@@ -516,7 +517,7 @@ mod tests {
 
     /// Shortcut for creating a package name.
     fn create_package(name: &str) -> PackageName {
-        PackageName::new(name.to_string()).unwrap()
+        PackageName::from_str(name).unwrap()
     }
 
     /// Shortcut for creating an extra name.