Adding upstream version 124.0.1.upstream/124.0.1

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

34 files changed, 5870 insertions, 0 deletions

diff --git a/third_party/rust/icu_locid/src/databake.rs b/third_party/rust/icu_locid/src/databake.rs
new file mode 100644
index 0000000000..03b7357c5e
--- /dev/null
+++ b/third_party/rust/icu_locid/src/databake.rs
@@ -0,0 +1,23 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use crate::LanguageIdentifier;
+use alloc::string::ToString;
+use databake::*;
+
+impl Bake for LanguageIdentifier {
+    fn bake(&self, env: &CrateEnv) -> TokenStream {
+        env.insert("icu_locid");
+        let repr = self.to_string();
+        if self.variants.len() <= 1 {
+            quote! {
+                icu_locid::langid!(#repr)
+            }
+        } else {
+            quote! {
+                icu_locid::LanguageIdentifier::from_str(#repr).unwrap()
+            }
+        }
+    }
+}
diff --git a/third_party/rust/icu_locid/src/extensions/mod.rs b/third_party/rust/icu_locid/src/extensions/mod.rs
new file mode 100644
index 0000000000..a37bf8b9fc
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/mod.rs
@@ -0,0 +1,324 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+//! Unicode Extensions provide a mechanism to extend the [`LanguageIdentifier`] with
+//! additional bits of information - a combination of a [`LanguageIdentifier`] and [`Extensions`]
+//! is called [`Locale`].
+//!
+//! There are four types of extensions:
+//!
+//!  * [`Unicode Extensions`] - marked as `u`.
+//!  * [`Transform Extensions`] - marked as `t`.
+//!  * [`Private Use Extensions`] - marked as `x`.
+//!  * [`Other Extensions`] - marked as any `a-z` except of `u`, `t` and `x`.
+//!
+//! One can think of extensions as a bag of extra information on top of basic 4 [`subtags`].
+//!
+//! Notice: `Other` extension type is currently not supported.
+//!
+//! # Examples
+//!
+//! ```
+//! use icu::locid::extensions::unicode::{Key, Value};
+//! use icu::locid::Locale;
+//!
+//! let loc: Locale = "en-US-u-ca-buddhist-t-en-us-h0-hybrid-x-foo"
+//!     .parse()
+//!     .expect("Failed to parse.");
+//!
+//! assert_eq!(loc.id.language, "en".parse().unwrap());
+//! assert_eq!(loc.id.script, None);
+//! assert_eq!(loc.id.region, Some("US".parse().unwrap()));
+//! assert_eq!(loc.id.variants.len(), 0);
+//!
+//! let key: Key = "ca".parse().expect("Parsing key failed.");
+//! let value: Value = "buddhist".parse().expect("Parsing value failed.");
+//! assert_eq!(loc.extensions.unicode.keywords.get(&key), Some(&value));
+//! ```
+//!
+//! [`LanguageIdentifier`]: super::LanguageIdentifier
+//! [`Locale`]: super::Locale
+//! [`subtags`]: super::subtags
+//! [`Other Extensions`]: other
+//! [`Private Use Extensions`]: private
+//! [`Transform Extensions`]: transform
+//! [`Unicode Extensions`]: unicode
+pub mod other;
+pub mod private;
+pub mod transform;
+pub mod unicode;
+
+use other::Other;
+use private::Private;
+use transform::Transform;
+use unicode::Unicode;
+
+use alloc::vec::Vec;
+
+use crate::parser::ParserError;
+use crate::parser::SubtagIterator;
+
+/// Defines the type of extension.
+#[derive(Debug, PartialEq, Eq, Clone, Hash, PartialOrd, Ord, Copy)]
+#[non_exhaustive]
+pub enum ExtensionType {
+    /// Transform Extension Type marked as `t`.
+    Transform,
+    /// Unicode Extension Type marked as `u`.
+    Unicode,
+    /// Private Extension Type marked as `x`.
+    Private,
+    /// All other extension types.
+    Other(u8),
+}
+
+impl ExtensionType {
+    pub(crate) const fn try_from_byte(key: u8) -> Result<Self, ParserError> {
+        let key = key.to_ascii_lowercase();
+        match key {
+            b'u' => Ok(Self::Unicode),
+            b't' => Ok(Self::Transform),
+            b'x' => Ok(Self::Private),
+            b'a'..=b'z' => Ok(Self::Other(key)),
+            _ => Err(ParserError::InvalidExtension),
+        }
+    }
+
+    pub(crate) const fn try_from_bytes_manual_slice(
+        bytes: &[u8],
+        start: usize,
+        end: usize,
+    ) -> Result<Self, ParserError> {
+        if end - start != 1 {
+            return Err(ParserError::InvalidExtension);
+        }
+        #[allow(clippy::indexing_slicing)]
+        Self::try_from_byte(bytes[start])
+    }
+}
+
+/// A map of extensions associated with a given [`Locale`](crate::Locale).
+#[derive(Debug, Default, PartialEq, Eq, Clone, Hash)]
+#[non_exhaustive]
+pub struct Extensions {
+    /// A representation of the data for a Unicode extension, when present in the locale identifier.
+    pub unicode: Unicode,
+    /// A representation of the data for a transform extension, when present in the locale identifier.
+    pub transform: Transform,
+    /// A representation of the data for a private-use extension, when present in the locale identifier.
+    pub private: Private,
+    /// A sequence of any other extensions that are present in the locale identifier but are not formally
+    /// [defined](https://unicode.org/reports/tr35/) and represented explicitly as [`Unicode`], [`Transform`],
+    /// and [`Private`] are.
+    pub other: Vec<Other>,
+}
+
+impl Extensions {
+    /// Returns a new empty map of extensions. Same as [`default()`](Default::default()), but is `const`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::Extensions;
+    ///
+    /// assert_eq!(Extensions::new(), Extensions::default());
+    /// ```
+    #[inline]
+    pub const fn new() -> Self {
+        Self {
+            unicode: Unicode::new(),
+            transform: Transform::new(),
+            private: Private::new(),
+            other: Vec::new(),
+        }
+    }
+
+    /// Function to create a new map of extensions containing exactly one unicode extension, callable in `const`
+    /// context.
+    #[inline]
+    pub const fn from_unicode(unicode: Unicode) -> Self {
+        Self {
+            unicode,
+            transform: Transform::new(),
+            private: Private::new(),
+            other: Vec::new(),
+        }
+    }
+
+    /// Returns whether there are no extensions present.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// let loc: Locale = "en-US-u-foo".parse().expect("Parsing failed.");
+    ///
+    /// assert!(!loc.extensions.is_empty());
+    /// ```
+    pub fn is_empty(&self) -> bool {
+        self.unicode.is_empty()
+            && self.transform.is_empty()
+            && self.private.is_empty()
+            && self.other.is_empty()
+    }
+
+    /// Retains the specified extension types, clearing all others.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::ExtensionType;
+    /// use icu::locid::Locale;
+    ///
+    /// let loc: Locale =
+    ///     "und-a-hello-t-mul-u-world-z-zzz-x-extra".parse().unwrap();
+    ///
+    /// let mut only_unicode = loc.clone();
+    /// only_unicode
+    ///     .extensions
+    ///     .retain_by_type(|t| t == ExtensionType::Unicode);
+    /// assert_eq!(only_unicode, "und-u-world".parse().unwrap());
+    ///
+    /// let mut only_t_z = loc.clone();
+    /// only_t_z.extensions.retain_by_type(|t| {
+    ///     t == ExtensionType::Transform || t == ExtensionType::Other(b'z')
+    /// });
+    /// assert_eq!(only_t_z, "und-t-mul-z-zzz".parse().unwrap());
+    /// ```
+    pub fn retain_by_type<F>(&mut self, mut predicate: F)
+    where
+        F: FnMut(ExtensionType) -> bool,
+    {
+        if !predicate(ExtensionType::Unicode) {
+            self.unicode.clear();
+        }
+        if !predicate(ExtensionType::Transform) {
+            self.transform.clear();
+        }
+        if !predicate(ExtensionType::Private) {
+            self.private.clear();
+        }
+        self.other
+            .retain(|o| predicate(ExtensionType::Other(o.get_ext_byte())));
+    }
+
+    pub(crate) fn try_from_iter(iter: &mut SubtagIterator) -> Result<Self, ParserError> {
+        let mut unicode = None;
+        let mut transform = None;
+        let mut private = None;
+        let mut other = Vec::new();
+
+        while let Some(subtag) = iter.next() {
+            if subtag.is_empty() {
+                return Err(ParserError::InvalidExtension);
+            }
+            match subtag.first().map(|b| ExtensionType::try_from_byte(*b)) {
+                Some(Ok(ExtensionType::Unicode)) => {
+                    if unicode.is_some() {
+                        return Err(ParserError::DuplicatedExtension);
+                    }
+                    unicode = Some(Unicode::try_from_iter(iter)?);
+                }
+                Some(Ok(ExtensionType::Transform)) => {
+                    if transform.is_some() {
+                        return Err(ParserError::DuplicatedExtension);
+                    }
+                    transform = Some(Transform::try_from_iter(iter)?);
+                }
+                Some(Ok(ExtensionType::Private)) => {
+                    if private.is_some() {
+                        return Err(ParserError::DuplicatedExtension);
+                    }
+                    private = Some(Private::try_from_iter(iter)?);
+                }
+                Some(Ok(ExtensionType::Other(ext))) => {
+                    if other.iter().any(|o: &Other| o.get_ext_byte() == ext) {
+                        return Err(ParserError::DuplicatedExtension);
+                    }
+                    let parsed = Other::try_from_iter(ext, iter)?;
+                    if let Err(idx) = other.binary_search(&parsed) {
+                        other.insert(idx, parsed);
+                    } else {
+                        return Err(ParserError::InvalidExtension);
+                    }
+                }
+                _ => return Err(ParserError::InvalidExtension),
+            }
+        }
+
+        Ok(Self {
+            unicode: unicode.unwrap_or_default(),
+            transform: transform.unwrap_or_default(),
+            private: private.unwrap_or_default(),
+            other,
+        })
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        let mut wrote_tu = false;
+        // Alphabetic by singleton
+        self.other.iter().try_for_each(|other| {
+            if other.get_ext() > 't' && !wrote_tu {
+                // Since 't' and 'u' are next to each other in alphabetical
+                // order, write both now.
+                self.transform.for_each_subtag_str(f)?;
+                self.unicode.for_each_subtag_str(f)?;
+                wrote_tu = true;
+            }
+            other.for_each_subtag_str(f)?;
+            Ok(())
+        })?;
+
+        if !wrote_tu {
+            self.transform.for_each_subtag_str(f)?;
+            self.unicode.for_each_subtag_str(f)?;
+        }
+
+        // Private must be written last, since it allows single character
+        // keys. Extensions must also be written in alphabetical order,
+        // which would seem to imply that other extensions `y` and `z` are
+        // invalid, but this is not specified.
+        self.private.for_each_subtag_str(f)?;
+        Ok(())
+    }
+}
+
+impl_writeable_for_each_subtag_str_no_test!(Extensions);
+
+#[test]
+fn test_writeable() {
+    use crate::Locale;
+    use writeable::assert_writeable_eq;
+    assert_writeable_eq!(Extensions::new(), "");
+    assert_writeable_eq!(
+        "my-t-my-d0-zawgyi".parse::<Locale>().unwrap().extensions,
+        "t-my-d0-zawgyi",
+    );
+    assert_writeable_eq!(
+        "ar-SA-u-ca-islamic-civil"
+            .parse::<Locale>()
+            .unwrap()
+            .extensions,
+        "u-ca-islamic-civil",
+    );
+    assert_writeable_eq!(
+        "en-001-x-foo-bar".parse::<Locale>().unwrap().extensions,
+        "x-foo-bar",
+    );
+    assert_writeable_eq!(
+        "und-t-m0-true".parse::<Locale>().unwrap().extensions,
+        "t-m0-true",
+    );
+    assert_writeable_eq!(
+        "und-a-foo-t-foo-u-foo-w-foo-z-foo-x-foo"
+            .parse::<Locale>()
+            .unwrap()
+            .extensions,
+        "a-foo-t-foo-u-foo-w-foo-z-foo-x-foo",
+    );
+}
diff --git a/third_party/rust/icu_locid/src/extensions/other/mod.rs b/third_party/rust/icu_locid/src/extensions/other/mod.rs
new file mode 100644
index 0000000000..933128739d
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/other/mod.rs
@@ -0,0 +1,186 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+//! Other Use Extensions is a list of extensions other than unicode,
+//! transform or private.
+//!
+//! Those extensions are treated as a pass-through, and no Unicode related
+//! behavior depends on them.
+//!
+//! The main struct for this extension is [`Other`] which is a list of [`Subtag`]s.
+//!
+//! # Examples
+//!
+//! ```
+//! use icu::locid::extensions::other::Other;
+//! use icu::locid::Locale;
+//!
+//! let mut loc: Locale = "en-US-a-foo-faa".parse().expect("Parsing failed.");
+//! ```
+
+mod subtag;
+
+use crate::helpers::ShortSlice;
+use crate::parser::ParserError;
+use crate::parser::SubtagIterator;
+use alloc::vec::Vec;
+#[doc(inline)]
+pub use subtag::{subtag, Subtag};
+
+/// A list of [`Other Use Extensions`] as defined in [`Unicode Locale
+/// Identifier`] specification.
+///
+/// Those extensions are treated as a pass-through, and no Unicode related
+/// behavior depends on them.
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::extensions::other::{Other, Subtag};
+///
+/// let subtag1: Subtag = "foo".parse().expect("Failed to parse a Subtag.");
+/// let subtag2: Subtag = "bar".parse().expect("Failed to parse a Subtag.");
+///
+/// let other = Other::from_vec_unchecked(b'a', vec![subtag1, subtag2]);
+/// assert_eq!(&other.to_string(), "a-foo-bar");
+/// ```
+///
+/// [`Other Use Extensions`]: https://unicode.org/reports/tr35/#other_extensions
+/// [`Unicode Locale Identifier`]: https://unicode.org/reports/tr35/#Unicode_locale_identifier
+#[derive(Clone, PartialEq, Eq, Debug, Default, Hash, PartialOrd, Ord)]
+pub struct Other {
+    ext: u8,
+    keys: ShortSlice<Subtag>,
+}
+
+impl Other {
+    /// A constructor which takes a pre-sorted list of [`Subtag`].
+    ///
+    /// # Panics
+    ///
+    /// Panics if `ext` is not ASCII alphabetic.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::other::{Other, Subtag};
+    ///
+    /// let subtag1: Subtag = "foo".parse().expect("Failed to parse a Subtag.");
+    /// let subtag2: Subtag = "bar".parse().expect("Failed to parse a Subtag.");
+    ///
+    /// let other = Other::from_vec_unchecked(b'a', vec![subtag1, subtag2]);
+    /// assert_eq!(&other.to_string(), "a-foo-bar");
+    /// ```
+    pub fn from_vec_unchecked(ext: u8, keys: Vec<Subtag>) -> Self {
+        Self::from_short_slice_unchecked(ext, keys.into())
+    }
+
+    pub(crate) fn from_short_slice_unchecked(ext: u8, keys: ShortSlice<Subtag>) -> Self {
+        assert!(ext.is_ascii_alphabetic());
+        Self { ext, keys }
+    }
+
+    pub(crate) fn try_from_iter(ext: u8, iter: &mut SubtagIterator) -> Result<Self, ParserError> {
+        debug_assert!(ext.is_ascii_alphabetic());
+
+        let mut keys = ShortSlice::new();
+        while let Some(subtag) = iter.peek() {
+            if !Subtag::valid_key(subtag) {
+                break;
+            }
+            if let Ok(key) = Subtag::try_from_bytes(subtag) {
+                keys.push(key);
+            }
+            iter.next();
+        }
+
+        Ok(Self::from_short_slice_unchecked(ext, keys))
+    }
+
+    /// Gets the tag character for this extension as a &str.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// let loc: Locale = "und-a-hello-world".parse().unwrap();
+    /// let other_ext = &loc.extensions.other[0];
+    /// assert_eq!(other_ext.get_ext_str(), "a");
+    /// ```
+    pub fn get_ext_str(&self) -> &str {
+        debug_assert!(self.ext.is_ascii_alphabetic());
+        unsafe { core::str::from_utf8_unchecked(core::slice::from_ref(&self.ext)) }
+    }
+
+    /// Gets the tag character for this extension as a char.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// let loc: Locale = "und-a-hello-world".parse().unwrap();
+    /// let other_ext = &loc.extensions.other[0];
+    /// assert_eq!(other_ext.get_ext(), 'a');
+    /// ```
+    pub fn get_ext(&self) -> char {
+        self.ext as char
+    }
+
+    /// Gets the tag character for this extension as a byte.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// let loc: Locale = "und-a-hello-world".parse().unwrap();
+    /// let other_ext = &loc.extensions.other[0];
+    /// assert_eq!(other_ext.get_ext_byte(), b'a');
+    /// ```
+    pub fn get_ext_byte(&self) -> u8 {
+        self.ext
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        f(self.get_ext_str())?;
+        self.keys.iter().map(|t| t.as_str()).try_for_each(f)
+    }
+}
+
+writeable::impl_display_with_writeable!(Other);
+
+impl writeable::Writeable for Other {
+    fn write_to<W: core::fmt::Write + ?Sized>(&self, sink: &mut W) -> core::fmt::Result {
+        sink.write_str(self.get_ext_str())?;
+        for key in self.keys.iter() {
+            sink.write_char('-')?;
+            writeable::Writeable::write_to(key, sink)?;
+        }
+
+        Ok(())
+    }
+
+    fn writeable_length_hint(&self) -> writeable::LengthHint {
+        let mut result = writeable::LengthHint::exact(1);
+        for key in self.keys.iter() {
+            result += writeable::Writeable::writeable_length_hint(key) + 1;
+        }
+        result
+    }
+
+    fn write_to_string(&self) -> alloc::borrow::Cow<str> {
+        if self.keys.is_empty() {
+            return alloc::borrow::Cow::Borrowed(self.get_ext_str());
+        }
+        let mut string =
+            alloc::string::String::with_capacity(self.writeable_length_hint().capacity());
+        let _ = self.write_to(&mut string);
+        alloc::borrow::Cow::Owned(string)
+    }
+}
diff --git a/third_party/rust/icu_locid/src/extensions/other/subtag.rs b/third_party/rust/icu_locid/src/extensions/other/subtag.rs
new file mode 100644
index 0000000000..03be569406
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/other/subtag.rs
@@ -0,0 +1,36 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+impl_tinystr_subtag!(
+    /// A single item used in a list of [`Other`](super::Other) extensions.
+    ///
+    /// The subtag has to be an ASCII alphanumerical string no shorter than
+    /// two characters and no longer than eight.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::other::subtag;
+    ///
+    /// assert_eq!(subtag!("Foo").as_str(), "foo");
+    /// ```
+    Subtag,
+    extensions::other,
+    subtag,
+    extensions_other_subtag,
+    2..=8,
+    s,
+    s.is_ascii_alphanumeric(),
+    s.to_ascii_lowercase(),
+    s.is_ascii_alphanumeric() && s.is_ascii_lowercase(),
+    InvalidExtension,
+    ["foo12"],
+    ["y", "toolooong"],
+);
+
+impl Subtag {
+    pub(crate) const fn valid_key(v: &[u8]) -> bool {
+        2 <= v.len() && v.len() <= 8
+    }
+}
diff --git a/third_party/rust/icu_locid/src/extensions/private/mod.rs b/third_party/rust/icu_locid/src/extensions/private/mod.rs
new file mode 100644
index 0000000000..5b41fdce09
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/private/mod.rs
@@ -0,0 +1,187 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+//! Private Use Extensions is a list of extensions intended for
+//! private use.
+//!
+//! Those extensions are treated as a pass-through, and no Unicode related
+//! behavior depends on them.
+//!
+//! The main struct for this extension is [`Private`] which is a list of [`Subtag`]s.
+//!
+//! # Examples
+//!
+//! ```
+//! use icu::locid::extensions::private::subtag;
+//! use icu::locid::{locale, Locale};
+//!
+//! let mut loc: Locale = "en-US-x-foo-faa".parse().expect("Parsing failed.");
+//!
+//! assert!(loc.extensions.private.contains(&subtag!("foo")));
+//! assert_eq!(loc.extensions.private.iter().next(), Some(&subtag!("foo")));
+//!
+//! loc.extensions.private.clear();
+//!
+//! assert!(loc.extensions.private.is_empty());
+//! assert_eq!(loc, locale!("en-US"));
+//! ```
+
+mod other;
+
+use alloc::vec::Vec;
+use core::ops::Deref;
+
+#[doc(inline)]
+pub use other::{subtag, Subtag};
+
+use crate::helpers::ShortSlice;
+use crate::parser::ParserError;
+use crate::parser::SubtagIterator;
+
+/// A list of [`Private Use Extensions`] as defined in [`Unicode Locale
+/// Identifier`] specification.
+///
+/// Those extensions are treated as a pass-through, and no Unicode related
+/// behavior depends on them.
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::extensions::private::{Private, Subtag};
+///
+/// let subtag1: Subtag = "foo".parse().expect("Failed to parse a Subtag.");
+/// let subtag2: Subtag = "bar".parse().expect("Failed to parse a Subtag.");
+///
+/// let private = Private::from_vec_unchecked(vec![subtag1, subtag2]);
+/// assert_eq!(&private.to_string(), "x-foo-bar");
+/// ```
+///
+/// [`Private Use Extensions`]: https://unicode.org/reports/tr35/#pu_extensions
+/// [`Unicode Locale Identifier`]: https://unicode.org/reports/tr35/#Unicode_locale_identifier
+#[derive(Clone, PartialEq, Eq, Debug, Default, Hash, PartialOrd, Ord)]
+pub struct Private(ShortSlice<Subtag>);
+
+impl Private {
+    /// Returns a new empty list of private-use extensions. Same as [`default()`](Default::default()), but is `const`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::private::Private;
+    ///
+    /// assert_eq!(Private::new(), Private::default());
+    /// ```
+    #[inline]
+    pub const fn new() -> Self {
+        Self(ShortSlice::new())
+    }
+
+    /// A constructor which takes a pre-sorted list of [`Subtag`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::private::{Private, Subtag};
+    ///
+    /// let subtag1: Subtag = "foo".parse().expect("Failed to parse a Subtag.");
+    /// let subtag2: Subtag = "bar".parse().expect("Failed to parse a Subtag.");
+    ///
+    /// let private = Private::from_vec_unchecked(vec![subtag1, subtag2]);
+    /// assert_eq!(&private.to_string(), "x-foo-bar");
+    /// ```
+    pub fn from_vec_unchecked(input: Vec<Subtag>) -> Self {
+        Self(input.into())
+    }
+
+    /// A constructor which takes a single [`Subtag`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::private::{Private, Subtag};
+    ///
+    /// let subtag: Subtag = "foo".parse().expect("Failed to parse a Subtag.");
+    ///
+    /// let private = Private::new_single(subtag);
+    /// assert_eq!(&private.to_string(), "x-foo");
+    /// ```
+    pub const fn new_single(input: Subtag) -> Self {
+        Self(ShortSlice::new_single(input))
+    }
+
+    /// Empties the [`Private`] list.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::private::{Private, Subtag};
+    ///
+    /// let subtag1: Subtag = "foo".parse().expect("Failed to parse a Subtag.");
+    /// let subtag2: Subtag = "bar".parse().expect("Failed to parse a Subtag.");
+    /// let mut private = Private::from_vec_unchecked(vec![subtag1, subtag2]);
+    ///
+    /// assert_eq!(&private.to_string(), "x-foo-bar");
+    ///
+    /// private.clear();
+    ///
+    /// assert_eq!(private, Private::new());
+    /// ```
+    pub fn clear(&mut self) {
+        self.0.clear();
+    }
+
+    pub(crate) fn try_from_iter(iter: &mut SubtagIterator) -> Result<Self, ParserError> {
+        let keys = iter
+            .map(Subtag::try_from_bytes)
+            .collect::<Result<ShortSlice<_>, _>>()?;
+
+        Ok(Self(keys))
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        if self.is_empty() {
+            return Ok(());
+        }
+        f("x")?;
+        self.deref().iter().map(|t| t.as_str()).try_for_each(f)
+    }
+}
+
+writeable::impl_display_with_writeable!(Private);
+
+impl writeable::Writeable for Private {
+    fn write_to<W: core::fmt::Write + ?Sized>(&self, sink: &mut W) -> core::fmt::Result {
+        if self.is_empty() {
+            return Ok(());
+        }
+        sink.write_str("x")?;
+        for key in self.iter() {
+            sink.write_char('-')?;
+            writeable::Writeable::write_to(key, sink)?;
+        }
+        Ok(())
+    }
+
+    fn writeable_length_hint(&self) -> writeable::LengthHint {
+        if self.is_empty() {
+            return writeable::LengthHint::exact(0);
+        }
+        let mut result = writeable::LengthHint::exact(1);
+        for key in self.iter() {
+            result += writeable::Writeable::writeable_length_hint(key) + 1;
+        }
+        result
+    }
+}
+
+impl Deref for Private {
+    type Target = [Subtag];
+
+    fn deref(&self) -> &Self::Target {
+        self.0.deref()
+    }
+}
diff --git a/third_party/rust/icu_locid/src/extensions/private/other.rs b/third_party/rust/icu_locid/src/extensions/private/other.rs
new file mode 100644
index 0000000000..810ffa2f49
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/private/other.rs
@@ -0,0 +1,32 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+impl_tinystr_subtag!(
+    /// A single item used in a list of [`Private`](super::Private) extensions.
+    ///
+    /// The subtag has to be an ASCII alphanumerical string no shorter than
+    /// one character and no longer than eight.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::private::Subtag;
+    ///
+    /// let subtag1: Subtag = "Foo".parse().expect("Failed to parse a Subtag.");
+    ///
+    /// assert_eq!(subtag1.as_str(), "foo");
+    /// ```
+    Subtag,
+    extensions::private,
+    subtag,
+    extensions_private_subtag,
+    1..=8,
+    s,
+    s.is_ascii_alphanumeric(),
+    s.to_ascii_lowercase(),
+    s.is_ascii_alphanumeric() && s.is_ascii_lowercase(),
+    InvalidExtension,
+    ["foo12"],
+    ["toolooong"],
+);
diff --git a/third_party/rust/icu_locid/src/extensions/transform/fields.rs b/third_party/rust/icu_locid/src/extensions/transform/fields.rs
new file mode 100644
index 0000000000..2f12de9d14
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/transform/fields.rs
@@ -0,0 +1,221 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use core::borrow::Borrow;
+use core::iter::FromIterator;
+use litemap::LiteMap;
+
+use super::Key;
+use super::Value;
+
+/// A list of [`Key`]-[`Value`] pairs representing functional information
+/// about content transformations.
+///
+/// Here are examples of fields used in Unicode:
+/// - `s0`, `d0` - Transform source/destination
+/// - `t0` - Machine Translation
+/// - `h0` - Hybrid Locale Identifiers
+///
+/// You can find the full list in [`Unicode BCP 47 T Extension`] section of LDML.
+///
+/// [`Unicode BCP 47 T Extension`]: https://unicode.org/reports/tr35/tr35.html#BCP47_T_Extension
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::extensions::transform::{key, Fields, Key, Value};
+///
+/// let value = "hybrid".parse::<Value>().expect("Failed to parse a Value.");
+/// let fields = [(key!("h0"), value)].into_iter().collect::<Fields>();
+///
+/// assert_eq!(&fields.to_string(), "h0-hybrid");
+/// ```
+#[derive(Clone, PartialEq, Eq, Debug, Default, Hash, PartialOrd, Ord)]
+pub struct Fields(LiteMap<Key, Value>);
+
+impl Fields {
+    /// Returns a new empty list of key-value pairs. Same as [`default()`](Default::default()), but is `const`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::transform::Fields;
+    ///
+    /// assert_eq!(Fields::new(), Fields::default());
+    /// ```
+    #[inline]
+    pub const fn new() -> Self {
+        Self(LiteMap::new())
+    }
+
+    /// Returns `true` if there are no fields.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::transform::Fields;
+    /// use icu::locid::locale;
+    /// use icu::locid::Locale;
+    ///
+    /// let loc1 = Locale::try_from_bytes(b"und-t-h0-hybrid").unwrap();
+    /// let loc2 = locale!("und-u-ca-buddhist");
+    ///
+    /// assert!(!loc1.extensions.transform.fields.is_empty());
+    /// assert!(loc2.extensions.transform.fields.is_empty());
+    /// ```
+    pub fn is_empty(&self) -> bool {
+        self.0.is_empty()
+    }
+
+    /// Empties the [`Fields`] list.
+    ///
+    /// Returns the old list.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::transform::{key, Fields, Value};
+    ///
+    /// let value = "hybrid".parse::<Value>().expect("Failed to parse a Value.");
+    /// let mut fields = [(key!("h0"), value)].into_iter().collect::<Fields>();
+    ///
+    /// assert_eq!(&fields.to_string(), "h0-hybrid");
+    ///
+    /// fields.clear();
+    ///
+    /// assert_eq!(fields, Fields::new());
+    /// ```
+    pub fn clear(&mut self) -> Self {
+        core::mem::take(self)
+    }
+
+    /// Returns `true` if the list contains a [`Value`] for the specified [`Key`].
+    ///
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::transform::{Fields, Key, Value};
+    ///
+    /// let key: Key = "h0".parse().expect("Failed to parse a Key.");
+    /// let value: Value = "hybrid".parse().expect("Failed to parse a Value.");
+    /// let mut fields = [(key, value)].into_iter().collect::<Fields>();
+    ///
+    /// let key: Key = "h0".parse().expect("Failed to parse a Key.");
+    /// assert!(&fields.contains_key(&key));
+    /// ```
+    pub fn contains_key<Q>(&self, key: &Q) -> bool
+    where
+        Key: Borrow<Q>,
+        Q: Ord,
+    {
+        self.0.contains_key(key)
+    }
+
+    /// Returns a reference to the [`Value`] corresponding to the [`Key`].
+    ///
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::transform::{key, Fields, Key, Value};
+    ///
+    /// let value = "hybrid".parse::<Value>().unwrap();
+    /// let fields = [(key!("h0"), value.clone())]
+    ///     .into_iter()
+    ///     .collect::<Fields>();
+    ///
+    /// assert_eq!(fields.get(&key!("h0")), Some(&value));
+    /// ```
+    pub fn get<Q>(&self, key: &Q) -> Option<&Value>
+    where
+        Key: Borrow<Q>,
+        Q: Ord,
+    {
+        self.0.get(key)
+    }
+
+    /// Sets the specified keyword, returning the old value if it already existed.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::transform::{key, Key, Value};
+    /// use icu::locid::Locale;
+    ///
+    /// let lower = "lower".parse::<Value>().expect("valid extension subtag");
+    /// let casefold = "casefold".parse::<Value>().expect("valid extension subtag");
+    ///
+    /// let mut loc: Locale = "en-t-hi-d0-casefold"
+    ///     .parse()
+    ///     .expect("valid BCP-47 identifier");
+    /// let old_value = loc.extensions.transform.fields.set(key!("d0"), lower);
+    ///
+    /// assert_eq!(old_value, Some(casefold));
+    /// assert_eq!(loc, "en-t-hi-d0-lower".parse().unwrap());
+    /// ```
+    pub fn set(&mut self, key: Key, value: Value) -> Option<Value> {
+        self.0.insert(key, value)
+    }
+
+    /// Retains a subset of fields as specified by the predicate function.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::transform::key;
+    /// use icu::locid::Locale;
+    ///
+    /// let mut loc: Locale = "und-t-h0-hybrid-d0-hex-m0-xml".parse().unwrap();
+    ///
+    /// loc.extensions
+    ///     .transform
+    ///     .fields
+    ///     .retain_by_key(|&k| k == key!("h0"));
+    /// assert_eq!(loc, "und-t-h0-hybrid".parse().unwrap());
+    ///
+    /// loc.extensions
+    ///     .transform
+    ///     .fields
+    ///     .retain_by_key(|&k| k == key!("d0"));
+    /// assert_eq!(loc, Locale::UND);
+    /// ```
+    pub fn retain_by_key<F>(&mut self, mut predicate: F)
+    where
+        F: FnMut(&Key) -> bool,
+    {
+        self.0.retain(|k, _| predicate(k))
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        for (k, v) in self.0.iter() {
+            f(k.as_str())?;
+            v.for_each_subtag_str(f)?;
+        }
+        Ok(())
+    }
+
+    /// This needs to be its own method to help with type inference in helpers.rs
+    #[cfg(test)]
+    pub(crate) fn from_tuple_vec(v: Vec<(Key, Value)>) -> Self {
+        v.into_iter().collect()
+    }
+}
+
+impl From<LiteMap<Key, Value>> for Fields {
+    fn from(map: LiteMap<Key, Value>) -> Self {
+        Self(map)
+    }
+}
+
+impl FromIterator<(Key, Value)> for Fields {
+    fn from_iter<I: IntoIterator<Item = (Key, Value)>>(iter: I) -> Self {
+        LiteMap::from_iter(iter).into()
+    }
+}
+
+impl_writeable_for_key_value!(Fields, "h0", "hybrid", "m0", "m0-true");
diff --git a/third_party/rust/icu_locid/src/extensions/transform/key.rs b/third_party/rust/icu_locid/src/extensions/transform/key.rs
new file mode 100644
index 0000000000..afdb31d760
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/transform/key.rs
@@ -0,0 +1,32 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+impl_tinystr_subtag!(
+    /// A key used in a list of [`Fields`](super::Fields).
+    ///
+    /// The key has to be a two ASCII characters long, with the first
+    /// character being alphabetic, and the second being a number.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::transform::Key;
+    ///
+    /// let key1: Key = "k0".parse().expect("Failed to parse a Key.");
+    ///
+    /// assert_eq!(key1.as_str(), "k0");
+    /// ```
+    Key,
+    extensions::transform,
+    key,
+    extensions_transform_key,
+    2..=2,
+    s,
+    s.all_bytes()[0].is_ascii_alphabetic() && s.all_bytes()[1].is_ascii_digit(),
+    s.to_ascii_lowercase(),
+    s.all_bytes()[0].is_ascii_lowercase() && s.all_bytes()[1].is_ascii_digit(),
+    InvalidExtension,
+    ["k0"],
+    ["", "k", "0k", "k12"],
+);
diff --git a/third_party/rust/icu_locid/src/extensions/transform/mod.rs b/third_party/rust/icu_locid/src/extensions/transform/mod.rs
new file mode 100644
index 0000000000..f5bb74e0db
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/transform/mod.rs
@@ -0,0 +1,237 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+//! Transform Extensions provide information on content transformations in a given locale.
+//!
+//! The main struct for this extension is [`Transform`] which contains [`Fields`] and an
+//! optional [`LanguageIdentifier`].
+//!
+//! [`LanguageIdentifier`]: super::super::LanguageIdentifier
+//!
+//! # Examples
+//!
+//! ```
+//! use icu::locid::extensions::transform::{Fields, Key, Transform, Value};
+//! use icu::locid::{LanguageIdentifier, Locale};
+//!
+//! let mut loc: Locale =
+//!     "en-US-t-es-ar-h0-hybrid".parse().expect("Parsing failed.");
+//!
+//! let lang: LanguageIdentifier =
+//!     "es-AR".parse().expect("Parsing LanguageIdentifier failed.");
+//!
+//! let key: Key = "h0".parse().expect("Parsing key failed.");
+//! let value: Value = "hybrid".parse().expect("Parsing value failed.");
+//!
+//! assert_eq!(loc.extensions.transform.lang, Some(lang));
+//! assert!(loc.extensions.transform.fields.contains_key(&key));
+//! assert_eq!(loc.extensions.transform.fields.get(&key), Some(&value));
+//!
+//! assert_eq!(&loc.extensions.transform.to_string(), "t-es-ar-h0-hybrid");
+//! ```
+mod fields;
+mod key;
+mod value;
+
+pub use fields::Fields;
+#[doc(inline)]
+pub use key::{key, Key};
+pub use value::Value;
+
+use crate::helpers::ShortSlice;
+use crate::parser::SubtagIterator;
+use crate::parser::{parse_language_identifier_from_iter, ParserError, ParserMode};
+use crate::subtags::Language;
+use crate::LanguageIdentifier;
+use litemap::LiteMap;
+
+/// A list of [`Unicode BCP47 T Extensions`] as defined in [`Unicode Locale
+/// Identifier`] specification.
+///
+/// Transform extension carries information about source language or script of
+/// transformed content, including content that has been transliterated, transcribed,
+/// or translated, or in some other way influenced by the source (See [`RFC 6497`] for details).
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::extensions::transform::{Key, Value};
+/// use icu::locid::{LanguageIdentifier, Locale};
+///
+/// let mut loc: Locale =
+///     "de-t-en-us-h0-hybrid".parse().expect("Parsing failed.");
+///
+/// let en_us: LanguageIdentifier = "en-US".parse().expect("Parsing failed.");
+///
+/// assert_eq!(loc.extensions.transform.lang, Some(en_us));
+/// let key: Key = "h0".parse().expect("Parsing key failed.");
+/// let value: Value = "hybrid".parse().expect("Parsing value failed.");
+/// assert_eq!(loc.extensions.transform.fields.get(&key), Some(&value));
+/// ```
+/// [`Unicode BCP47 T Extensions`]: https://unicode.org/reports/tr35/#t_Extension
+/// [`RFC 6497`]: https://www.ietf.org/rfc/rfc6497.txt
+/// [`Unicode Locale Identifier`]: https://unicode.org/reports/tr35/#Unicode_locale_identifier
+#[derive(Clone, PartialEq, Eq, Debug, Default, Hash)]
+#[allow(clippy::exhaustive_structs)] // spec-backed stable datastructure
+pub struct Transform {
+    /// The [`LanguageIdentifier`] specified with this locale extension, or `None` if not present.
+    pub lang: Option<LanguageIdentifier>,
+    /// The key-value pairs present in this locale extension, with each extension key subtag
+    /// associated to its provided value subtag.
+    pub fields: Fields,
+}
+
+impl Transform {
+    /// Returns a new empty map of Transform extensions. Same as [`default()`](Default::default()), but is `const`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::transform::Transform;
+    ///
+    /// assert_eq!(Transform::new(), Transform::default());
+    /// ```
+    #[inline]
+    pub const fn new() -> Self {
+        Self {
+            lang: None,
+            fields: Fields::new(),
+        }
+    }
+
+    /// Returns `true` if there are no tfields and no tlang in the `TransformExtensionList`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// let mut loc: Locale = "en-US-t-es-ar".parse().expect("Parsing failed.");
+    ///
+    /// assert!(!loc.extensions.transform.is_empty());
+    /// ```
+    pub fn is_empty(&self) -> bool {
+        self.lang.is_none() && self.fields.is_empty()
+    }
+
+    /// Clears the transform extension, effectively removing it from the locale.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// let mut loc: Locale = "en-US-t-es-ar".parse().unwrap();
+    /// loc.extensions.transform.clear();
+    /// assert_eq!(loc, "en-US".parse().unwrap());
+    /// ```
+    pub fn clear(&mut self) {
+        self.lang = None;
+        self.fields.clear();
+    }
+
+    pub(crate) fn try_from_iter(iter: &mut SubtagIterator) -> Result<Self, ParserError> {
+        let mut tlang = None;
+        let mut tfields = LiteMap::new();
+
+        if let Some(subtag) = iter.peek() {
+            if Language::try_from_bytes(subtag).is_ok() {
+                tlang = Some(parse_language_identifier_from_iter(
+                    iter,
+                    ParserMode::Partial,
+                )?);
+            }
+        }
+
+        let mut current_tkey = None;
+        let mut current_tvalue = ShortSlice::new();
+        let mut has_current_tvalue = false;
+
+        while let Some(subtag) = iter.peek() {
+            if let Some(tkey) = current_tkey {
+                if let Ok(val) = Value::parse_subtag(subtag) {
+                    has_current_tvalue = true;
+                    if let Some(val) = val {
+                        current_tvalue.push(val);
+                    }
+                } else {
+                    if !has_current_tvalue {
+                        return Err(ParserError::InvalidExtension);
+                    }
+                    tfields.try_insert(tkey, Value::from_short_slice_unchecked(current_tvalue));
+                    current_tkey = None;
+                    current_tvalue = ShortSlice::new();
+                    has_current_tvalue = false;
+                    continue;
+                }
+            } else if let Ok(tkey) = Key::try_from_bytes(subtag) {
+                current_tkey = Some(tkey);
+            } else {
+                break;
+            }
+
+            iter.next();
+        }
+
+        if let Some(tkey) = current_tkey {
+            if !has_current_tvalue {
+                return Err(ParserError::InvalidExtension);
+            }
+            tfields.try_insert(tkey, Value::from_short_slice_unchecked(current_tvalue));
+        }
+
+        Ok(Self {
+            lang: tlang,
+            fields: tfields.into(),
+        })
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        if self.is_empty() {
+            return Ok(());
+        }
+        f("t")?;
+        if let Some(lang) = &self.lang {
+            lang.for_each_subtag_str_lowercased(f)?;
+        }
+        self.fields.for_each_subtag_str(f)
+    }
+}
+
+writeable::impl_display_with_writeable!(Transform);
+
+impl writeable::Writeable for Transform {
+    fn write_to<W: core::fmt::Write + ?Sized>(&self, sink: &mut W) -> core::fmt::Result {
+        if self.is_empty() {
+            return Ok(());
+        }
+        sink.write_str("t")?;
+        if let Some(lang) = &self.lang {
+            sink.write_char('-')?;
+            lang.write_lowercased_to(sink)?;
+        }
+        if !self.fields.is_empty() {
+            sink.write_char('-')?;
+            writeable::Writeable::write_to(&self.fields, sink)?;
+        }
+        Ok(())
+    }
+
+    fn writeable_length_hint(&self) -> writeable::LengthHint {
+        if self.is_empty() {
+            return writeable::LengthHint::exact(0);
+        }
+        let mut result = writeable::LengthHint::exact(1);
+        if let Some(lang) = &self.lang {
+            result += writeable::Writeable::writeable_length_hint(lang) + 1;
+        }
+        if !self.fields.is_empty() {
+            result += writeable::Writeable::writeable_length_hint(&self.fields) + 1;
+        }
+        result
+    }
+}
diff --git a/third_party/rust/icu_locid/src/extensions/transform/value.rs b/third_party/rust/icu_locid/src/extensions/transform/value.rs
new file mode 100644
index 0000000000..798e84793d
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/transform/value.rs
@@ -0,0 +1,134 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use crate::helpers::ShortSlice;
+use crate::parser::{ParserError, SubtagIterator};
+use core::ops::RangeInclusive;
+use core::str::FromStr;
+use tinystr::TinyAsciiStr;
+
+/// A value used in a list of [`Fields`](super::Fields).
+///
+/// The value has to be a sequence of one or more alphanumerical strings
+/// separated by `-`.
+/// Each part of the sequence has to be no shorter than three characters and no
+/// longer than 8.
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::extensions::transform::Value;
+///
+/// "hybrid".parse::<Value>().expect("Valid Value.");
+///
+/// "hybrid-foobar".parse::<Value>().expect("Valid Value.");
+///
+/// "no".parse::<Value>().expect_err("Invalid Value.");
+/// ```
+#[derive(Debug, PartialEq, Eq, Clone, Hash, PartialOrd, Ord, Default)]
+pub struct Value(ShortSlice<TinyAsciiStr<{ *TYPE_LENGTH.end() }>>);
+
+const TYPE_LENGTH: RangeInclusive<usize> = 3..=8;
+const TRUE_TVALUE: TinyAsciiStr<8> = tinystr::tinystr!(8, "true");
+
+impl Value {
+    /// A constructor which takes a utf8 slice, parses it and
+    /// produces a well-formed [`Value`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::transform::Value;
+    ///
+    /// let value = Value::try_from_bytes(b"hybrid").expect("Parsing failed.");
+    /// ```
+    pub fn try_from_bytes(input: &[u8]) -> Result<Self, ParserError> {
+        let mut v = ShortSlice::default();
+        let mut has_value = false;
+
+        for subtag in SubtagIterator::new(input) {
+            if !Self::is_type_subtag(subtag) {
+                return Err(ParserError::InvalidExtension);
+            }
+            has_value = true;
+            let val =
+                TinyAsciiStr::from_bytes(subtag).map_err(|_| ParserError::InvalidExtension)?;
+            if val != TRUE_TVALUE {
+                v.push(val);
+            }
+        }
+
+        if !has_value {
+            return Err(ParserError::InvalidExtension);
+        }
+        Ok(Self(v))
+    }
+
+    pub(crate) fn from_short_slice_unchecked(
+        input: ShortSlice<TinyAsciiStr<{ *TYPE_LENGTH.end() }>>,
+    ) -> Self {
+        Self(input)
+    }
+
+    pub(crate) fn is_type_subtag(t: &[u8]) -> bool {
+        TYPE_LENGTH.contains(&t.len()) && t.iter().all(u8::is_ascii_alphanumeric)
+    }
+
+    pub(crate) fn parse_subtag(
+        t: &[u8],
+    ) -> Result<Option<TinyAsciiStr<{ *TYPE_LENGTH.end() }>>, ParserError> {
+        let s = TinyAsciiStr::from_bytes(t).map_err(|_| ParserError::InvalidSubtag)?;
+        if !TYPE_LENGTH.contains(&t.len()) || !s.is_ascii_alphanumeric() {
+            return Err(ParserError::InvalidExtension);
+        }
+
+        let s = s.to_ascii_lowercase();
+
+        if s == TRUE_TVALUE {
+            Ok(None)
+        } else {
+            Ok(Some(s))
+        }
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        if self.0.is_empty() {
+            f("true")?;
+        } else {
+            self.0.iter().map(TinyAsciiStr::as_str).try_for_each(f)?;
+        }
+        Ok(())
+    }
+}
+
+impl FromStr for Value {
+    type Err = ParserError;
+
+    fn from_str(source: &str) -> Result<Self, Self::Err> {
+        Self::try_from_bytes(source.as_bytes())
+    }
+}
+
+impl_writeable_for_each_subtag_str_no_test!(Value, selff, selff.0.is_empty() => alloc::borrow::Cow::Borrowed("true"));
+
+#[test]
+fn test_writeable() {
+    use writeable::assert_writeable_eq;
+
+    let hybrid = "hybrid".parse().unwrap();
+    let foobar = "foobar".parse().unwrap();
+
+    assert_writeable_eq!(Value::default(), "true");
+    assert_writeable_eq!(
+        Value::from_short_slice_unchecked(vec![hybrid].into()),
+        "hybrid"
+    );
+    assert_writeable_eq!(
+        Value::from_short_slice_unchecked(vec![hybrid, foobar].into()),
+        "hybrid-foobar"
+    );
+}
diff --git a/third_party/rust/icu_locid/src/extensions/unicode/attribute.rs b/third_party/rust/icu_locid/src/extensions/unicode/attribute.rs
new file mode 100644
index 0000000000..f6fc53e057
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/unicode/attribute.rs
@@ -0,0 +1,34 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+impl_tinystr_subtag!(
+    /// An attribute used in a set of [`Attributes`](super::Attributes).
+    ///
+    /// An attribute has to be a sequence of alphanumerical characters no
+    /// shorter than three and no longer than eight characters.
+    ///
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::{attribute, Attribute};
+    ///
+    /// let attr: Attribute =
+    ///     "buddhist".parse().expect("Failed to parse an Attribute.");
+    ///
+    /// assert_eq!(attr, attribute!("buddhist"));
+    /// ```
+    Attribute,
+    extensions::unicode,
+    attribute,
+    extensions_unicode_attribute,
+    3..=8,
+    s,
+    s.is_ascii_alphanumeric(),
+    s.to_ascii_lowercase(),
+    s.is_ascii_alphanumeric() && s.is_ascii_lowercase(),
+    InvalidExtension,
+    ["foo12"],
+    ["no", "toolooong"],
+);
diff --git a/third_party/rust/icu_locid/src/extensions/unicode/attributes.rs b/third_party/rust/icu_locid/src/extensions/unicode/attributes.rs
new file mode 100644
index 0000000000..1cdaded306
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/unicode/attributes.rs
@@ -0,0 +1,120 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use super::Attribute;
+
+use crate::helpers::ShortSlice;
+use alloc::vec::Vec;
+use core::ops::Deref;
+
+/// A set of [`Attribute`] elements as defined in [`Unicode Extension Attributes`].
+///
+/// [`Unicode Extension Attributes`]: https://unicode.org/reports/tr35/tr35.html#u_Extension
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::extensions::unicode::{Attribute, Attributes};
+///
+/// let attribute1: Attribute =
+///     "foobar".parse().expect("Failed to parse a variant subtag.");
+///
+/// let attribute2: Attribute = "testing"
+///     .parse()
+///     .expect("Failed to parse a variant subtag.");
+/// let mut v = vec![attribute1, attribute2];
+/// v.sort();
+/// v.dedup();
+///
+/// let attributes: Attributes = Attributes::from_vec_unchecked(v);
+/// assert_eq!(attributes.to_string(), "foobar-testing");
+/// ```
+#[derive(Default, Debug, PartialEq, Eq, Clone, Hash, PartialOrd, Ord)]
+pub struct Attributes(ShortSlice<Attribute>);
+
+impl Attributes {
+    /// Returns a new empty set of attributes. Same as [`default()`](Default::default()), but is `const`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Attributes;
+    ///
+    /// assert_eq!(Attributes::new(), Attributes::default());
+    /// ```
+    #[inline]
+    pub const fn new() -> Self {
+        Self(ShortSlice::new())
+    }
+
+    /// A constructor which takes a pre-sorted list of [`Attribute`] elements.
+    ///
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::{Attribute, Attributes};
+    ///
+    /// let attribute1: Attribute = "foobar".parse().expect("Parsing failed.");
+    /// let attribute2: Attribute = "testing".parse().expect("Parsing failed.");
+    /// let mut v = vec![attribute1, attribute2];
+    /// v.sort();
+    /// v.dedup();
+    ///
+    /// let attributes = Attributes::from_vec_unchecked(v);
+    /// ```
+    ///
+    /// Notice: For performance- and memory-constrained environments, it is recommended
+    /// for the caller to use [`binary_search`](slice::binary_search) instead of [`sort`](slice::sort)
+    /// and [`dedup`](Vec::dedup()).
+    pub fn from_vec_unchecked(input: Vec<Attribute>) -> Self {
+        Self(input.into())
+    }
+
+    pub(crate) fn from_short_slice_unchecked(input: ShortSlice<Attribute>) -> Self {
+        Self(input)
+    }
+
+    /// Empties the [`Attributes`] list.
+    ///
+    /// Returns the old list.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::{attribute, Attribute, Attributes};
+    /// use writeable::assert_writeable_eq;
+    ///
+    /// let mut attributes = Attributes::from_vec_unchecked(vec![
+    ///     attribute!("foobar"),
+    ///     attribute!("testing"),
+    /// ]);
+    ///
+    /// assert_writeable_eq!(attributes, "foobar-testing");
+    ///
+    /// attributes.clear();
+    ///
+    /// assert_writeable_eq!(attributes, "");
+    /// ```
+    pub fn clear(&mut self) -> Self {
+        core::mem::take(self)
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        self.deref().iter().map(|t| t.as_str()).try_for_each(f)
+    }
+}
+
+impl_writeable_for_subtag_list!(Attributes, "foobar", "testing");
+
+impl Deref for Attributes {
+    type Target = [Attribute];
+
+    fn deref(&self) -> &[Attribute] {
+        self.0.deref()
+    }
+}
diff --git a/third_party/rust/icu_locid/src/extensions/unicode/key.rs b/third_party/rust/icu_locid/src/extensions/unicode/key.rs
new file mode 100644
index 0000000000..e008ffd5a8
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/unicode/key.rs
@@ -0,0 +1,32 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+impl_tinystr_subtag!(
+    /// A key used in a list of [`Keywords`](super::Keywords).
+    ///
+    /// The key has to be a two ASCII alphanumerical characters long, with the first
+    /// character being alphanumeric, and the second being alphabetic.
+    ///
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Key;
+    ///
+    /// assert!("ca".parse::<Key>().is_ok());
+    /// ```
+    Key,
+    extensions::unicode,
+    key,
+    extensions_unicode_key,
+    2..=2,
+    s,
+    s.all_bytes()[0].is_ascii_alphanumeric() && s.all_bytes()[1].is_ascii_alphabetic(),
+    s.to_ascii_lowercase(),
+    (s.all_bytes()[0].is_ascii_lowercase() || s.all_bytes()[0].is_ascii_digit())
+        && s.all_bytes()[1].is_ascii_lowercase(),
+    InvalidExtension,
+    ["ca", "8a"],
+    ["a", "a8", "abc"],
+);
diff --git a/third_party/rust/icu_locid/src/extensions/unicode/keywords.rs b/third_party/rust/icu_locid/src/extensions/unicode/keywords.rs
new file mode 100644
index 0000000000..c2839fa44f
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/unicode/keywords.rs
@@ -0,0 +1,393 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use core::borrow::Borrow;
+use core::cmp::Ordering;
+use core::iter::FromIterator;
+use litemap::LiteMap;
+
+use super::Key;
+use super::Value;
+use crate::helpers::ShortSlice;
+use crate::ordering::SubtagOrderingResult;
+
+/// A list of [`Key`]-[`Value`] pairs representing functional information
+/// about locale's internationalization preferences.
+///
+/// Here are examples of fields used in Unicode:
+/// - `hc` - Hour Cycle (`h11`, `h12`, `h23`, `h24`)
+/// - `ca` - Calendar (`buddhist`, `gregory`, ...)
+/// - `fw` - First Day Of the Week (`sun`, `mon`, `sat`, ...)
+///
+/// You can find the full list in [`Unicode BCP 47 U Extension`] section of LDML.
+///
+/// [`Unicode BCP 47 U Extension`]: https://unicode.org/reports/tr35/tr35.html#Key_And_Type_Definitions_
+///
+/// # Examples
+///
+/// Manually build up a [`Keywords`] object:
+///
+/// ```
+/// use icu::locid::{
+///     extensions::unicode::{key, value, Keywords},
+///     locale,
+/// };
+///
+/// let keywords = [(key!("hc"), value!("h23"))]
+///     .into_iter()
+///     .collect::<Keywords>();
+///
+/// assert_eq!(&keywords.to_string(), "hc-h23");
+/// ```
+///
+/// Access a [`Keywords`] object from a [`Locale`]:
+///
+/// ```
+/// use icu::locid::{
+///     extensions::unicode::{key, value},
+///     Locale,
+/// };
+///
+/// let loc: Locale = "und-u-hc-h23-kc-true".parse().expect("Valid BCP-47");
+///
+/// assert_eq!(loc.extensions.unicode.keywords.get(&key!("ca")), None);
+/// assert_eq!(
+///     loc.extensions.unicode.keywords.get(&key!("hc")),
+///     Some(&value!("h23"))
+/// );
+/// assert_eq!(
+///     loc.extensions.unicode.keywords.get(&key!("kc")),
+///     Some(&value!("true"))
+/// );
+///
+/// assert_eq!(loc.extensions.unicode.keywords.to_string(), "hc-h23-kc");
+/// ```
+///
+/// [`Locale`]: crate::Locale
+#[derive(Clone, PartialEq, Eq, Debug, Default, Hash, PartialOrd, Ord)]
+pub struct Keywords(LiteMap<Key, Value, ShortSlice<(Key, Value)>>);
+
+impl Keywords {
+    /// Returns a new empty list of key-value pairs. Same as [`default()`](Default::default()), but is `const`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Keywords;
+    ///
+    /// assert_eq!(Keywords::new(), Keywords::default());
+    /// ```
+    #[inline]
+    pub const fn new() -> Self {
+        Self(LiteMap::new())
+    }
+
+    /// Create a new list of key-value pairs having exactly one pair, callable in a `const` context.
+    #[inline]
+    pub const fn new_single(key: Key, value: Value) -> Self {
+        Self(LiteMap::from_sorted_store_unchecked(
+            ShortSlice::new_single((key, value)),
+        ))
+    }
+
+    /// Returns `true` if there are no keywords.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Keywords;
+    /// use icu::locid::locale;
+    /// use icu::locid::Locale;
+    ///
+    /// let loc1 = Locale::try_from_bytes(b"und-t-h0-hybrid").unwrap();
+    /// let loc2 = locale!("und-u-ca-buddhist");
+    ///
+    /// assert!(loc1.extensions.unicode.keywords.is_empty());
+    /// assert!(!loc2.extensions.unicode.keywords.is_empty());
+    /// ```
+    pub fn is_empty(&self) -> bool {
+        self.0.is_empty()
+    }
+
+    /// Returns `true` if the list contains a [`Value`] for the specified [`Key`].
+    ///
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::{key, value, Keywords};
+    ///
+    /// let keywords = [(key!("ca"), value!("gregory"))]
+    ///     .into_iter()
+    ///     .collect::<Keywords>();
+    ///
+    /// assert!(&keywords.contains_key(&key!("ca")));
+    /// ```
+    pub fn contains_key<Q>(&self, key: &Q) -> bool
+    where
+        Key: Borrow<Q>,
+        Q: Ord,
+    {
+        self.0.contains_key(key)
+    }
+
+    /// Returns a reference to the [`Value`] corresponding to the [`Key`].
+    ///
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::{key, value, Keywords};
+    ///
+    /// let keywords = [(key!("ca"), value!("buddhist"))]
+    ///     .into_iter()
+    ///     .collect::<Keywords>();
+    ///
+    /// assert_eq!(keywords.get(&key!("ca")), Some(&value!("buddhist")));
+    /// ```
+    pub fn get<Q>(&self, key: &Q) -> Option<&Value>
+    where
+        Key: Borrow<Q>,
+        Q: Ord,
+    {
+        self.0.get(key)
+    }
+
+    /// Returns a mutable reference to the [`Value`] corresponding to the [`Key`].
+    ///
+    /// Returns `None` if the key doesn't exist or if the key has no value.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::{key, value, Keywords};
+    ///
+    /// let mut keywords = [(key!("ca"), value!("buddhist"))]
+    ///     .into_iter()
+    ///     .collect::<Keywords>();
+    ///
+    /// if let Some(value) = keywords.get_mut(&key!("ca")) {
+    ///     *value = value!("gregory");
+    /// }
+    /// assert_eq!(keywords.get(&key!("ca")), Some(&value!("gregory")));
+    /// ```
+    pub fn get_mut<Q>(&mut self, key: &Q) -> Option<&mut Value>
+    where
+        Key: Borrow<Q>,
+        Q: Ord,
+    {
+        self.0.get_mut(key)
+    }
+
+    /// Sets the specified keyword, returning the old value if it already existed.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Key;
+    /// use icu::locid::extensions::unicode::Value;
+    /// use icu::locid::extensions::unicode::{key, value};
+    /// use icu::locid::Locale;
+    ///
+    /// let mut loc: Locale = "und-u-hello-ca-buddhist-hc-h12"
+    ///     .parse()
+    ///     .expect("valid BCP-47 identifier");
+    /// let old_value = loc
+    ///     .extensions
+    ///     .unicode
+    ///     .keywords
+    ///     .set(key!("ca"), value!("japanese"));
+    ///
+    /// assert_eq!(old_value, Some(value!("buddhist")));
+    /// assert_eq!(loc, "und-u-hello-ca-japanese-hc-h12".parse().unwrap());
+    /// ```
+    pub fn set(&mut self, key: Key, value: Value) -> Option<Value> {
+        self.0.insert(key, value)
+    }
+
+    /// Removes the specified keyword, returning the old value if it existed.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::{key, Key};
+    /// use icu::locid::Locale;
+    ///
+    /// let mut loc: Locale = "und-u-hello-ca-buddhist-hc-h12"
+    ///     .parse()
+    ///     .expect("valid BCP-47 identifier");
+    /// loc.extensions.unicode.keywords.remove(key!("ca"));
+    /// assert_eq!(loc, "und-u-hello-hc-h12".parse().unwrap());
+    /// ```
+    pub fn remove<Q: Borrow<Key>>(&mut self, key: Q) -> Option<Value> {
+        self.0.remove(key.borrow())
+    }
+
+    /// Clears all Unicode extension keywords, leaving Unicode attributes.
+    ///
+    /// Returns the old Unicode extension keywords.
+    ///
+    /// # Example
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// let mut loc: Locale = "und-u-hello-ca-buddhist-hc-h12".parse().unwrap();
+    /// loc.extensions.unicode.keywords.clear();
+    /// assert_eq!(loc, "und-u-hello".parse().unwrap());
+    /// ```
+    pub fn clear(&mut self) -> Self {
+        core::mem::take(self)
+    }
+
+    /// Retains a subset of keywords as specified by the predicate function.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::key;
+    /// use icu::locid::Locale;
+    ///
+    /// let mut loc: Locale = "und-u-ca-buddhist-hc-h12-ms-metric".parse().unwrap();
+    ///
+    /// loc.extensions
+    ///     .unicode
+    ///     .keywords
+    ///     .retain_by_key(|&k| k == key!("hc"));
+    /// assert_eq!(loc, "und-u-hc-h12".parse().unwrap());
+    ///
+    /// loc.extensions
+    ///     .unicode
+    ///     .keywords
+    ///     .retain_by_key(|&k| k == key!("ms"));
+    /// assert_eq!(loc, Locale::UND);
+    /// ```
+    pub fn retain_by_key<F>(&mut self, mut predicate: F)
+    where
+        F: FnMut(&Key) -> bool,
+    {
+        self.0.retain(|k, _| predicate(k))
+    }
+
+    /// Compare this [`Keywords`] with BCP-47 bytes.
+    ///
+    /// The return value is equivalent to what would happen if you first converted this
+    /// [`Keywords`] to a BCP-47 string and then performed a byte comparison.
+    ///
+    /// This function is case-sensitive and results in a *total order*, so it is appropriate for
+    /// binary search. The only argument producing [`Ordering::Equal`] is `self.to_string()`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Keywords;
+    /// use icu::locid::Locale;
+    /// use std::cmp::Ordering;
+    ///
+    /// let bcp47_strings: &[&str] =
+    ///     &["ca-hebrew", "ca-japanese", "ca-japanese-nu-latn", "nu-latn"];
+    ///
+    /// for ab in bcp47_strings.windows(2) {
+    ///     let a = ab[0];
+    ///     let b = ab[1];
+    ///     assert!(a.cmp(b) == Ordering::Less);
+    ///     let a_kwds = format!("und-u-{}", a)
+    ///         .parse::<Locale>()
+    ///         .unwrap()
+    ///         .extensions
+    ///         .unicode
+    ///         .keywords;
+    ///     assert!(a_kwds.strict_cmp(a.as_bytes()) == Ordering::Equal);
+    ///     assert!(a_kwds.strict_cmp(b.as_bytes()) == Ordering::Less);
+    /// }
+    /// ```
+    pub fn strict_cmp(&self, other: &[u8]) -> Ordering {
+        self.strict_cmp_iter(other.split(|b| *b == b'-')).end()
+    }
+
+    /// Compare this [`Keywords`] with an iterator of BCP-47 subtags.
+    ///
+    /// This function has the same equality semantics as [`Keywords::strict_cmp`]. It is intended as
+    /// a more modular version that allows multiple subtag iterators to be chained together.
+    ///
+    /// For an additional example, see [`SubtagOrderingResult`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Keywords;
+    /// use icu::locid::locale;
+    /// use std::cmp::Ordering;
+    ///
+    /// let subtags: &[&[u8]] = &[b"ca", b"buddhist"];
+    ///
+    /// let kwds = locale!("und-u-ca-buddhist").extensions.unicode.keywords;
+    /// assert_eq!(
+    ///     Ordering::Equal,
+    ///     kwds.strict_cmp_iter(subtags.iter().copied()).end()
+    /// );
+    ///
+    /// let kwds = locale!("und").extensions.unicode.keywords;
+    /// assert_eq!(
+    ///     Ordering::Less,
+    ///     kwds.strict_cmp_iter(subtags.iter().copied()).end()
+    /// );
+    ///
+    /// let kwds = locale!("und-u-nu-latn").extensions.unicode.keywords;
+    /// assert_eq!(
+    ///     Ordering::Greater,
+    ///     kwds.strict_cmp_iter(subtags.iter().copied()).end()
+    /// );
+    /// ```
+    pub fn strict_cmp_iter<'l, I>(&self, mut subtags: I) -> SubtagOrderingResult<I>
+    where
+        I: Iterator<Item = &'l [u8]>,
+    {
+        let r = self.for_each_subtag_str(&mut |subtag| {
+            if let Some(other) = subtags.next() {
+                match subtag.as_bytes().cmp(other) {
+                    Ordering::Equal => Ok(()),
+                    not_equal => Err(not_equal),
+                }
+            } else {
+                Err(Ordering::Greater)
+            }
+        });
+        match r {
+            Ok(_) => SubtagOrderingResult::Subtags(subtags),
+            Err(o) => SubtagOrderingResult::Ordering(o),
+        }
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        for (k, v) in self.0.iter() {
+            f(k.as_str())?;
+            v.for_each_subtag_str(f)?;
+        }
+        Ok(())
+    }
+
+    /// This needs to be its own method to help with type inference in helpers.rs
+    #[cfg(test)]
+    pub(crate) fn from_tuple_vec(v: Vec<(Key, Value)>) -> Self {
+        v.into_iter().collect()
+    }
+}
+
+impl From<LiteMap<Key, Value, ShortSlice<(Key, Value)>>> for Keywords {
+    fn from(map: LiteMap<Key, Value, ShortSlice<(Key, Value)>>) -> Self {
+        Self(map)
+    }
+}
+
+impl FromIterator<(Key, Value)> for Keywords {
+    fn from_iter<I: IntoIterator<Item = (Key, Value)>>(iter: I) -> Self {
+        LiteMap::from_iter(iter).into()
+    }
+}
+
+impl_writeable_for_key_value!(Keywords, "ca", "islamic-civil", "mm", "mm");
diff --git a/third_party/rust/icu_locid/src/extensions/unicode/mod.rs b/third_party/rust/icu_locid/src/extensions/unicode/mod.rs
new file mode 100644
index 0000000000..95f1a2d781
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/unicode/mod.rs
@@ -0,0 +1,237 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+//! Unicode Extensions provide information about user preferences in a given locale.
+//!
+//! The main struct for this extension is [`Unicode`] which contains [`Keywords`] and
+//! [`Attributes`].
+//!
+//!
+//! # Examples
+//!
+//! ```
+//! use icu::locid::extensions::unicode::{attribute, key, value, Unicode};
+//! use icu::locid::Locale;
+//!
+//! let loc: Locale = "en-US-u-foobar-hc-h12".parse().expect("Parsing failed.");
+//!
+//! assert_eq!(
+//!     loc.extensions.unicode.keywords.get(&key!("hc")),
+//!     Some(&value!("h12"))
+//! );
+//! assert!(loc
+//!     .extensions
+//!     .unicode
+//!     .attributes
+//!     .contains(&attribute!("foobar")));
+//! ```
+mod attribute;
+mod attributes;
+mod key;
+mod keywords;
+mod value;
+
+#[doc(inline)]
+pub use attribute::{attribute, Attribute};
+pub use attributes::Attributes;
+#[doc(inline)]
+pub use key::{key, Key};
+pub use keywords::Keywords;
+#[doc(inline)]
+pub use value::{value, Value};
+
+use crate::helpers::ShortSlice;
+use crate::parser::ParserError;
+use crate::parser::SubtagIterator;
+use litemap::LiteMap;
+
+/// Unicode Extensions provide information about user preferences in a given locale.
+///
+/// A list of [`Unicode BCP47 U Extensions`] as defined in [`Unicode Locale
+/// Identifier`] specification.
+///
+/// Unicode extensions provide subtags that specify language and/or locale-based behavior
+/// or refinements to language tags, according to work done by the Unicode Consortium.
+/// (See [`RFC 6067`] for details).
+///
+/// [`Unicode BCP47 U Extensions`]: https://unicode.org/reports/tr35/#u_Extension
+/// [`RFC 6067`]: https://www.ietf.org/rfc/rfc6067.txt
+/// [`Unicode Locale Identifier`]: https://unicode.org/reports/tr35/#Unicode_locale_identifier
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::extensions::unicode::{key, value};
+/// use icu::locid::Locale;
+///
+/// let loc: Locale =
+///     "de-u-hc-h12-ca-buddhist".parse().expect("Parsing failed.");
+///
+/// assert_eq!(
+///     loc.extensions.unicode.keywords.get(&key!("ca")),
+///     Some(&value!("buddhist"))
+/// );
+/// ```
+#[derive(Clone, PartialEq, Eq, Debug, Default, Hash, PartialOrd, Ord)]
+#[allow(clippy::exhaustive_structs)] // spec-backed stable datastructure
+pub struct Unicode {
+    /// The key-value pairs present in this locale extension, with each extension key subtag
+    /// associated to its provided value subtag.
+    pub keywords: Keywords,
+    /// A canonically ordered sequence of single standalone subtags for this locale extension.
+    pub attributes: Attributes,
+}
+
+impl Unicode {
+    /// Returns a new empty map of Unicode extensions. Same as [`default()`](Default::default()), but is `const`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Unicode;
+    ///
+    /// assert_eq!(Unicode::new(), Unicode::default());
+    /// ```
+    #[inline]
+    pub const fn new() -> Self {
+        Self {
+            keywords: Keywords::new(),
+            attributes: Attributes::new(),
+        }
+    }
+
+    /// Returns [`true`] if there list of keywords and attributes is empty.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// let loc: Locale = "en-US-u-foo".parse().expect("Parsing failed.");
+    ///
+    /// assert!(!loc.extensions.unicode.is_empty());
+    /// ```
+    pub fn is_empty(&self) -> bool {
+        self.keywords.is_empty() && self.attributes.is_empty()
+    }
+
+    /// Clears all Unicode extension keywords and attributes, effectively removing
+    /// the Unicode extension.
+    ///
+    /// # Example
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// let mut loc: Locale =
+    ///     "und-t-mul-u-hello-ca-buddhist-hc-h12".parse().unwrap();
+    /// loc.extensions.unicode.clear();
+    /// assert_eq!(loc, "und-t-mul".parse().unwrap());
+    /// ```
+    pub fn clear(&mut self) {
+        self.keywords.clear();
+        self.attributes.clear();
+    }
+
+    pub(crate) fn try_from_iter(iter: &mut SubtagIterator) -> Result<Self, ParserError> {
+        let mut attributes = ShortSlice::new();
+
+        while let Some(subtag) = iter.peek() {
+            if let Ok(attr) = Attribute::try_from_bytes(subtag) {
+                if let Err(idx) = attributes.binary_search(&attr) {
+                    attributes.insert(idx, attr);
+                }
+            } else {
+                break;
+            }
+            iter.next();
+        }
+
+        let mut keywords = LiteMap::new();
+
+        let mut current_keyword = None;
+        let mut current_value = ShortSlice::new();
+
+        while let Some(subtag) = iter.peek() {
+            let slen = subtag.len();
+            if slen == 2 {
+                if let Some(kw) = current_keyword.take() {
+                    keywords.try_insert(kw, Value::from_short_slice_unchecked(current_value));
+                    current_value = ShortSlice::new();
+                }
+                current_keyword = Some(Key::try_from_bytes(subtag)?);
+            } else if current_keyword.is_some() {
+                match Value::parse_subtag(subtag) {
+                    Ok(Some(t)) => current_value.push(t),
+                    Ok(None) => {}
+                    Err(_) => break,
+                }
+            } else {
+                break;
+            }
+            iter.next();
+        }
+
+        if let Some(kw) = current_keyword.take() {
+            keywords.try_insert(kw, Value::from_short_slice_unchecked(current_value));
+        }
+
+        // Ensure we've defined at least one attribute or keyword
+        if attributes.is_empty() && keywords.is_empty() {
+            return Err(ParserError::InvalidExtension);
+        }
+
+        Ok(Self {
+            keywords: keywords.into(),
+            attributes: Attributes::from_short_slice_unchecked(attributes),
+        })
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        if self.is_empty() {
+            return Ok(());
+        }
+        f("u")?;
+        self.attributes.for_each_subtag_str(f)?;
+        self.keywords.for_each_subtag_str(f)?;
+        Ok(())
+    }
+}
+
+writeable::impl_display_with_writeable!(Unicode);
+
+impl writeable::Writeable for Unicode {
+    fn write_to<W: core::fmt::Write + ?Sized>(&self, sink: &mut W) -> core::fmt::Result {
+        if self.is_empty() {
+            return Ok(());
+        }
+        sink.write_str("u")?;
+        if !self.attributes.is_empty() {
+            sink.write_char('-')?;
+            writeable::Writeable::write_to(&self.attributes, sink)?;
+        }
+        if !self.keywords.is_empty() {
+            sink.write_char('-')?;
+            writeable::Writeable::write_to(&self.keywords, sink)?;
+        }
+        Ok(())
+    }
+
+    fn writeable_length_hint(&self) -> writeable::LengthHint {
+        if self.is_empty() {
+            return writeable::LengthHint::exact(0);
+        }
+        let mut result = writeable::LengthHint::exact(1);
+        if !self.attributes.is_empty() {
+            result += writeable::Writeable::writeable_length_hint(&self.attributes) + 1;
+        }
+        if !self.keywords.is_empty() {
+            result += writeable::Writeable::writeable_length_hint(&self.keywords) + 1;
+        }
+        result
+    }
+}
diff --git a/third_party/rust/icu_locid/src/extensions/unicode/value.rs b/third_party/rust/icu_locid/src/extensions/unicode/value.rs
new file mode 100644
index 0000000000..d935656a97
--- /dev/null
+++ b/third_party/rust/icu_locid/src/extensions/unicode/value.rs
@@ -0,0 +1,196 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use crate::helpers::ShortSlice;
+use crate::parser::{ParserError, SubtagIterator};
+use core::ops::RangeInclusive;
+use core::str::FromStr;
+use tinystr::TinyAsciiStr;
+
+/// A value used in a list of [`Keywords`](super::Keywords).
+///
+/// The value has to be a sequence of one or more alphanumerical strings
+/// separated by `-`.
+/// Each part of the sequence has to be no shorter than three characters and no
+/// longer than 8.
+///
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::extensions::unicode::{value, Value};
+/// use writeable::assert_writeable_eq;
+///
+/// assert_writeable_eq!(value!("gregory"), "gregory");
+/// assert_writeable_eq!(
+///     "islamic-civil".parse::<Value>().unwrap(),
+///     "islamic-civil"
+/// );
+///
+/// // The value "true" has the special, empty string representation
+/// assert_eq!(value!("true").to_string(), "");
+/// ```
+#[derive(Debug, PartialEq, Eq, Clone, Hash, PartialOrd, Ord, Default)]
+pub struct Value(ShortSlice<TinyAsciiStr<{ *VALUE_LENGTH.end() }>>);
+
+const VALUE_LENGTH: RangeInclusive<usize> = 3..=8;
+const TRUE_VALUE: TinyAsciiStr<8> = tinystr::tinystr!(8, "true");
+
+impl Value {
+    /// A constructor which takes a utf8 slice, parses it and
+    /// produces a well-formed [`Value`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Value;
+    ///
+    /// Value::try_from_bytes(b"buddhist").expect("Parsing failed.");
+    /// ```
+    pub fn try_from_bytes(input: &[u8]) -> Result<Self, ParserError> {
+        let mut v = ShortSlice::new();
+
+        if !input.is_empty() {
+            for subtag in SubtagIterator::new(input) {
+                let val = Self::subtag_from_bytes(subtag)?;
+                if let Some(val) = val {
+                    v.push(val);
+                }
+            }
+        }
+        Ok(Self(v))
+    }
+
+    /// Const constructor for when the value contains only a single subtag.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Value;
+    ///
+    /// Value::try_from_single_subtag(b"buddhist").expect("valid subtag");
+    /// Value::try_from_single_subtag(b"#####").expect_err("invalid subtag");
+    /// Value::try_from_single_subtag(b"foo-bar").expect_err("not a single subtag");
+    /// ```
+    pub const fn try_from_single_subtag(subtag: &[u8]) -> Result<Self, ParserError> {
+        match Self::subtag_from_bytes(subtag) {
+            Err(_) => Err(ParserError::InvalidExtension),
+            Ok(option) => Ok(Self::from_tinystr(option)),
+        }
+    }
+
+    #[doc(hidden)]
+    pub fn as_tinystr_slice(&self) -> &[TinyAsciiStr<8>] {
+        &self.0
+    }
+
+    #[doc(hidden)]
+    pub const fn as_single_subtag(&self) -> Option<&TinyAsciiStr<8>> {
+        self.0.single()
+    }
+
+    #[doc(hidden)]
+    pub const fn from_tinystr(subtag: Option<TinyAsciiStr<8>>) -> Self {
+        match subtag {
+            None => Self(ShortSlice::new()),
+            Some(val) => {
+                debug_assert!(val.is_ascii_alphanumeric());
+                debug_assert!(!matches!(val, TRUE_VALUE));
+                Self(ShortSlice::new_single(val))
+            }
+        }
+    }
+
+    pub(crate) fn from_short_slice_unchecked(input: ShortSlice<TinyAsciiStr<8>>) -> Self {
+        Self(input)
+    }
+
+    #[doc(hidden)]
+    pub const fn subtag_from_bytes(bytes: &[u8]) -> Result<Option<TinyAsciiStr<8>>, ParserError> {
+        Self::parse_subtag_from_bytes_manual_slice(bytes, 0, bytes.len())
+    }
+
+    pub(crate) fn parse_subtag(t: &[u8]) -> Result<Option<TinyAsciiStr<8>>, ParserError> {
+        Self::parse_subtag_from_bytes_manual_slice(t, 0, t.len())
+    }
+
+    pub(crate) const fn parse_subtag_from_bytes_manual_slice(
+        bytes: &[u8],
+        start: usize,
+        end: usize,
+    ) -> Result<Option<TinyAsciiStr<8>>, ParserError> {
+        let slice_len = end - start;
+        if slice_len > *VALUE_LENGTH.end() || slice_len < *VALUE_LENGTH.start() {
+            return Err(ParserError::InvalidExtension);
+        }
+
+        match TinyAsciiStr::from_bytes_manual_slice(bytes, start, end) {
+            Ok(TRUE_VALUE) => Ok(None),
+            Ok(s) if s.is_ascii_alphanumeric() => Ok(Some(s.to_ascii_lowercase())),
+            Ok(_) => Err(ParserError::InvalidExtension),
+            Err(_) => Err(ParserError::InvalidSubtag),
+        }
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        self.0.iter().map(TinyAsciiStr::as_str).try_for_each(f)
+    }
+}
+
+impl FromStr for Value {
+    type Err = ParserError;
+
+    fn from_str(source: &str) -> Result<Self, Self::Err> {
+        Self::try_from_bytes(source.as_bytes())
+    }
+}
+
+impl_writeable_for_subtag_list!(Value, "islamic", "civil");
+
+/// A macro allowing for compile-time construction of valid Unicode [`Value`] subtag.
+///
+/// The macro only supports single-subtag values.
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::extensions::unicode::{key, value};
+/// use icu::locid::Locale;
+///
+/// let loc: Locale = "de-u-ca-buddhist".parse().unwrap();
+///
+/// assert_eq!(
+///     loc.extensions.unicode.keywords.get(&key!("ca")),
+///     Some(&value!("buddhist"))
+/// );
+/// ```
+///
+/// [`Value`]: crate::extensions::unicode::Value
+#[macro_export]
+#[doc(hidden)]
+macro_rules! extensions_unicode_value {
+    ($value:literal) => {{
+        // What we want:
+        // const R: $crate::extensions::unicode::Value =
+        //     match $crate::extensions::unicode::Value::try_from_single_subtag($value.as_bytes()) {
+        //         Ok(r) => r,
+        //         #[allow(clippy::panic)] // const context
+        //         _ => panic!(concat!("Invalid Unicode extension value: ", $value)),
+        //     };
+        // Workaround until https://github.com/rust-lang/rust/issues/73255 lands:
+        const R: $crate::extensions::unicode::Value =
+            $crate::extensions::unicode::Value::from_tinystr(
+                match $crate::extensions::unicode::Value::subtag_from_bytes($value.as_bytes()) {
+                    Ok(r) => r,
+                    _ => panic!(concat!("Invalid Unicode extension value: ", $value)),
+                },
+            );
+        R
+    }};
+}
+#[doc(inline)]
+pub use extensions_unicode_value as value;
diff --git a/third_party/rust/icu_locid/src/helpers.rs b/third_party/rust/icu_locid/src/helpers.rs
new file mode 100644
index 0000000000..d12435fbf3
--- /dev/null
+++ b/third_party/rust/icu_locid/src/helpers.rs
@@ -0,0 +1,698 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use core::iter::FromIterator;
+
+use alloc::boxed::Box;
+use alloc::vec;
+use alloc::vec::Vec;
+use core::ops::{Deref, DerefMut};
+use litemap::store::*;
+
+/// Internal: A vector that supports no-allocation, constant values if length 0 or 1.
+/// Using ZeroOne(Option<T>) saves 8 bytes in ShortSlice via niche optimization.
+#[derive(Debug, Clone, PartialEq, Eq, Hash, PartialOrd, Ord)]
+pub(crate) enum ShortSlice<T> {
+    ZeroOne(Option<T>),
+    Multi(Box<[T]>),
+}
+
+impl<T> ShortSlice<T> {
+    #[inline]
+    pub const fn new() -> Self {
+        Self::ZeroOne(None)
+    }
+
+    #[inline]
+    pub const fn new_single(item: T) -> Self {
+        Self::ZeroOne(Some(item))
+    }
+
+    pub fn push(&mut self, item: T) {
+        *self = match core::mem::replace(self, Self::ZeroOne(None)) {
+            ShortSlice::ZeroOne(None) => ShortSlice::ZeroOne(Some(item)),
+            ShortSlice::ZeroOne(Some(prev_item)) => {
+                ShortSlice::Multi(vec![prev_item, item].into_boxed_slice())
+            }
+            ShortSlice::Multi(items) => {
+                let mut items = items.into_vec();
+                items.push(item);
+                ShortSlice::Multi(items.into_boxed_slice())
+            }
+        };
+    }
+
+    #[inline]
+    pub const fn single(&self) -> Option<&T> {
+        match self {
+            ShortSlice::ZeroOne(Some(v)) => Some(v),
+            _ => None,
+        }
+    }
+
+    #[inline]
+    pub fn len(&self) -> usize {
+        match self {
+            ShortSlice::ZeroOne(None) => 0,
+            ShortSlice::ZeroOne(_) => 1,
+            ShortSlice::Multi(ref v) => v.len(),
+        }
+    }
+
+    pub fn insert(&mut self, index: usize, elt: T) {
+        assert!(
+            index <= self.len(),
+            "insertion index (is {}) should be <= len (is {})",
+            index,
+            self.len()
+        );
+
+        *self = match core::mem::replace(self, ShortSlice::ZeroOne(None)) {
+            ShortSlice::ZeroOne(None) => ShortSlice::ZeroOne(Some(elt)),
+            ShortSlice::ZeroOne(Some(item)) => {
+                let items = if index == 0 {
+                    vec![elt, item].into_boxed_slice()
+                } else {
+                    vec![item, elt].into_boxed_slice()
+                };
+                ShortSlice::Multi(items)
+            }
+            ShortSlice::Multi(items) => {
+                let mut items = items.into_vec();
+                items.insert(index, elt);
+                ShortSlice::Multi(items.into_boxed_slice())
+            }
+        }
+    }
+
+    pub fn remove(&mut self, index: usize) -> T {
+        assert!(
+            index < self.len(),
+            "removal index (is {}) should be < len (is {})",
+            index,
+            self.len()
+        );
+
+        let (replaced, removed_item) = match core::mem::replace(self, ShortSlice::ZeroOne(None)) {
+            ShortSlice::ZeroOne(None) => unreachable!(),
+            ShortSlice::ZeroOne(Some(v)) => (ShortSlice::ZeroOne(None), v),
+            ShortSlice::Multi(v) => {
+                let mut v = v.into_vec();
+                let removed_item = v.remove(index);
+                match v.len() {
+                    #[allow(clippy::unwrap_used)]
+                    // we know that the vec has exactly one element left
+                    1 => (ShortSlice::ZeroOne(Some(v.pop().unwrap())), removed_item),
+                    // v has at least 2 elements, create a Multi variant
+                    _ => (ShortSlice::Multi(v.into_boxed_slice()), removed_item),
+                }
+            }
+        };
+        *self = replaced;
+        removed_item
+    }
+
+    #[inline]
+    pub fn clear(&mut self) {
+        let _ = core::mem::replace(self, ShortSlice::ZeroOne(None));
+    }
+
+    pub fn retain<F>(&mut self, mut f: F)
+    where
+        F: FnMut(&T) -> bool,
+    {
+        *self = match core::mem::take(self) {
+            Self::ZeroOne(Some(one)) if f(&one) => Self::ZeroOne(Some(one)),
+            Self::ZeroOne(_) => Self::ZeroOne(None),
+            Self::Multi(slice) => {
+                let mut vec = slice.into_vec();
+                vec.retain(f);
+                Self::from(vec)
+            }
+        };
+    }
+}
+
+impl<T> Deref for ShortSlice<T> {
+    type Target = [T];
+
+    fn deref(&self) -> &Self::Target {
+        match self {
+            ShortSlice::ZeroOne(None) => &[],
+            ShortSlice::ZeroOne(Some(v)) => core::slice::from_ref(v),
+            ShortSlice::Multi(v) => v,
+        }
+    }
+}
+
+impl<T> DerefMut for ShortSlice<T> {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        match self {
+            ShortSlice::ZeroOne(None) => &mut [],
+            ShortSlice::ZeroOne(Some(v)) => core::slice::from_mut(v),
+            ShortSlice::Multi(v) => v,
+        }
+    }
+}
+
+impl<T> From<Vec<T>> for ShortSlice<T> {
+    fn from(v: Vec<T>) -> Self {
+        match v.len() {
+            0 => ShortSlice::ZeroOne(None),
+            #[allow(clippy::unwrap_used)] // we know that the vec is not empty
+            1 => ShortSlice::ZeroOne(Some(v.into_iter().next().unwrap())),
+            _ => ShortSlice::Multi(v.into_boxed_slice()),
+        }
+    }
+}
+
+impl<T> Default for ShortSlice<T> {
+    fn default() -> Self {
+        ShortSlice::ZeroOne(None)
+    }
+}
+
+impl<T> FromIterator<T> for ShortSlice<T> {
+    fn from_iter<I: IntoIterator<Item = T>>(iter: I) -> Self {
+        let mut iter = iter.into_iter();
+        match (iter.next(), iter.next()) {
+            (Some(first), Some(second)) => {
+                // Size hint behaviour same as `Vec::extend` + 2
+                let mut vec = Vec::with_capacity(iter.size_hint().0.saturating_add(3));
+                vec.push(first);
+                vec.push(second);
+                vec.extend(iter);
+                Self::Multi(vec.into_boxed_slice())
+            }
+            (first, _) => Self::ZeroOne(first),
+        }
+    }
+}
+
+impl<K, V> StoreConstEmpty<K, V> for ShortSlice<(K, V)> {
+    const EMPTY: ShortSlice<(K, V)> = ShortSlice::ZeroOne(None);
+}
+
+impl<K, V> Store<K, V> for ShortSlice<(K, V)> {
+    #[inline]
+    fn lm_len(&self) -> usize {
+        self.len()
+    }
+
+    #[inline]
+    fn lm_is_empty(&self) -> bool {
+        matches!(self, ShortSlice::ZeroOne(None))
+    }
+
+    #[inline]
+    fn lm_get(&self, index: usize) -> Option<(&K, &V)> {
+        self.get(index).map(|elt| (&elt.0, &elt.1))
+    }
+
+    #[inline]
+    fn lm_last(&self) -> Option<(&K, &V)> {
+        match self {
+            ShortSlice::ZeroOne(v) => v.as_ref(),
+            ShortSlice::Multi(v) => v.last(),
+        }
+        .map(|elt| (&elt.0, &elt.1))
+    }
+
+    #[inline]
+    fn lm_binary_search_by<F>(&self, mut cmp: F) -> Result<usize, usize>
+    where
+        F: FnMut(&K) -> core::cmp::Ordering,
+    {
+        self.binary_search_by(|(k, _)| cmp(k))
+    }
+}
+
+impl<K: Ord, V> StoreFromIterable<K, V> for ShortSlice<(K, V)> {
+    fn lm_sort_from_iter<I: IntoIterator<Item = (K, V)>>(iter: I) -> Self {
+        let v: Vec<(K, V)> = Vec::lm_sort_from_iter(iter);
+        v.into()
+    }
+}
+
+impl<K, V> StoreMut<K, V> for ShortSlice<(K, V)> {
+    fn lm_with_capacity(_capacity: usize) -> Self {
+        ShortSlice::ZeroOne(None)
+    }
+
+    fn lm_reserve(&mut self, _additional: usize) {}
+
+    fn lm_get_mut(&mut self, index: usize) -> Option<(&K, &mut V)> {
+        self.get_mut(index).map(|elt| (&elt.0, &mut elt.1))
+    }
+
+    fn lm_push(&mut self, key: K, value: V) {
+        self.push((key, value))
+    }
+
+    fn lm_insert(&mut self, index: usize, key: K, value: V) {
+        self.insert(index, (key, value))
+    }
+
+    fn lm_remove(&mut self, index: usize) -> (K, V) {
+        self.remove(index)
+    }
+
+    fn lm_clear(&mut self) {
+        self.clear();
+    }
+
+    fn lm_retain<F>(&mut self, mut predicate: F)
+    where
+        F: FnMut(&K, &V) -> bool,
+    {
+        self.retain(|(k, v)| predicate(k, v))
+    }
+}
+
+impl<'a, K: 'a, V: 'a> StoreIterable<'a, K, V> for ShortSlice<(K, V)> {
+    type KeyValueIter =
+        core::iter::Map<core::slice::Iter<'a, (K, V)>, for<'r> fn(&'r (K, V)) -> (&'r K, &'r V)>;
+
+    fn lm_iter(&'a self) -> Self::KeyValueIter {
+        self.iter().map(|elt| (&elt.0, &elt.1))
+    }
+}
+
+impl<K, V> StoreFromIterator<K, V> for ShortSlice<(K, V)> {}
+
+#[test]
+fn test_short_slice_impl() {
+    litemap::testing::check_store::<ShortSlice<(u32, u64)>>();
+}
+
+macro_rules! impl_tinystr_subtag {
+    (
+        $(#[$doc:meta])*
+        $name:ident,
+        $($path:ident)::+,
+        $macro_name:ident,
+        $legacy_macro_name:ident,
+        $len_start:literal..=$len_end:literal,
+        $tinystr_ident:ident,
+        $validate:expr,
+        $normalize:expr,
+        $is_normalized:expr,
+        $error:ident,
+        [$good_example:literal $(,$more_good_examples:literal)*],
+        [$bad_example:literal $(, $more_bad_examples:literal)*],
+    ) => {
+        #[derive(Debug, PartialEq, Eq, Clone, Hash, PartialOrd, Ord, Copy)]
+        #[cfg_attr(feature = "serde", derive(serde::Serialize))]
+        #[repr(transparent)]
+        $(#[$doc])*
+        pub struct $name(tinystr::TinyAsciiStr<$len_end>);
+
+        impl $name {
+            /// A constructor which takes a UTF-8 slice, parses it and
+            #[doc = concat!("produces a well-formed [`", stringify!($name), "`].")]
+            ///
+            /// # Examples
+            ///
+            /// ```
+            #[doc = concat!("use icu_locid::", stringify!($($path::)+), stringify!($name), ";")]
+            ///
+            #[doc = concat!("assert!(", stringify!($name), "::try_from_bytes(b", stringify!($good_example), ").is_ok());")]
+            #[doc = concat!("assert!(", stringify!($name), "::try_from_bytes(b", stringify!($bad_example), ").is_err());")]
+            /// ```
+            pub const fn try_from_bytes(v: &[u8]) -> Result<Self, crate::parser::errors::ParserError> {
+                Self::try_from_bytes_manual_slice(v, 0, v.len())
+            }
+
+            /// Equivalent to [`try_from_bytes(bytes[start..end])`](Self::try_from_bytes),
+            /// but callable in a `const` context (which range indexing is not).
+            pub const fn try_from_bytes_manual_slice(
+                v: &[u8],
+                start: usize,
+                end: usize,
+            ) -> Result<Self, crate::parser::errors::ParserError> {
+                let slen = end - start;
+
+                #[allow(clippy::double_comparisons)] // if len_start == len_end
+                if slen < $len_start || slen > $len_end {
+                    return Err(crate::parser::errors::ParserError::$error);
+                }
+
+                match tinystr::TinyAsciiStr::from_bytes_manual_slice(v, start, end) {
+                    Ok($tinystr_ident) if $validate => Ok(Self($normalize)),
+                    _ => Err(crate::parser::errors::ParserError::$error),
+                }
+            }
+
+            #[doc = concat!("Safely creates a [`", stringify!($name), "`] from its raw format")]
+            /// as returned by [`Self::into_raw`]. Unlike [`Self::try_from_bytes`],
+            /// this constructor only takes normalized values.
+            pub const fn try_from_raw(
+                v: [u8; $len_end],
+            ) -> Result<Self, crate::parser::errors::ParserError> {
+                if let Ok($tinystr_ident) = tinystr::TinyAsciiStr::<$len_end>::try_from_raw(v) {
+                    if $tinystr_ident.len() >= $len_start && $is_normalized {
+                        Ok(Self($tinystr_ident))
+                    } else {
+                        Err(crate::parser::errors::ParserError::$error)
+                    }
+                } else {
+                    Err(crate::parser::errors::ParserError::$error)
+                }
+            }
+
+            #[doc = concat!("Unsafely creates a [`", stringify!($name), "`] from its raw format")]
+            /// as returned by [`Self::into_raw`]. Unlike [`Self::try_from_bytes`],
+            /// this constructor only takes normalized values.
+            ///
+            /// # Safety
+            ///
+            /// This function is safe iff [`Self::try_from_raw`] returns an `Ok`. This is the case
+            /// for inputs that are correctly normalized.
+            pub const unsafe fn from_raw_unchecked(v: [u8; $len_end]) -> Self {
+                Self(tinystr::TinyAsciiStr::from_bytes_unchecked(v))
+            }
+
+            /// Deconstructs into a raw format to be consumed by
+            /// [`from_raw_unchecked`](Self::from_raw_unchecked()) or
+            /// [`try_from_raw`](Self::try_from_raw()).
+            pub const fn into_raw(self) -> [u8; $len_end] {
+                *self.0.all_bytes()
+            }
+
+            #[inline]
+            /// A helper function for displaying as a `&str`.
+            pub const fn as_str(&self) -> &str {
+                self.0.as_str()
+            }
+
+            #[doc(hidden)]
+            pub const fn into_tinystr(&self) -> tinystr::TinyAsciiStr<$len_end> {
+                self.0
+            }
+
+            /// Compare with BCP-47 bytes.
+            ///
+            /// The return value is equivalent to what would happen if you first converted
+            /// `self` to a BCP-47 string and then performed a byte comparison.
+            ///
+            /// This function is case-sensitive and results in a *total order*, so it is appropriate for
+            /// binary search. The only argument producing [`Ordering::Equal`](core::cmp::Ordering::Equal)
+            /// is `self.as_str().as_bytes()`.
+            #[inline]
+            pub fn strict_cmp(self, other: &[u8]) -> core::cmp::Ordering {
+                self.as_str().as_bytes().cmp(other)
+            }
+
+            /// Compare with a potentially unnormalized BCP-47 string.
+            ///
+            /// The return value is equivalent to what would happen if you first parsed the
+            /// BCP-47 string and then performed a structural comparison.
+            ///
+            #[inline]
+            pub fn normalizing_eq(self, other: &str) -> bool {
+                self.as_str().eq_ignore_ascii_case(other)
+            }
+        }
+
+        impl core::str::FromStr for $name {
+            type Err = crate::parser::errors::ParserError;
+
+            fn from_str(source: &str) -> Result<Self, Self::Err> {
+                Self::try_from_bytes(source.as_bytes())
+            }
+        }
+
+        impl<'l> From<&'l $name> for &'l str {
+            fn from(input: &'l $name) -> Self {
+                input.as_str()
+            }
+        }
+
+        impl From<$name> for tinystr::TinyAsciiStr<$len_end> {
+            fn from(input: $name) -> Self {
+                input.into_tinystr()
+            }
+        }
+
+        impl writeable::Writeable for $name {
+            #[inline]
+            fn write_to<W: core::fmt::Write + ?Sized>(&self, sink: &mut W) -> core::fmt::Result {
+                sink.write_str(self.as_str())
+            }
+            #[inline]
+            fn writeable_length_hint(&self) -> writeable::LengthHint {
+                writeable::LengthHint::exact(self.0.len())
+            }
+            #[inline]
+            fn write_to_string(&self) -> alloc::borrow::Cow<str> {
+                alloc::borrow::Cow::Borrowed(self.0.as_str())
+            }
+        }
+
+        writeable::impl_display_with_writeable!($name);
+
+        #[doc = concat!("A macro allowing for compile-time construction of valid [`", stringify!($name), "`] subtags.")]
+        ///
+        /// # Examples
+        ///
+        /// Parsing errors don't have to be handled at runtime:
+        /// ```
+        /// assert_eq!(
+        #[doc = concat!("  icu_locid::", $(stringify!($path), "::",)+ stringify!($macro_name), "!(", stringify!($good_example) ,"),")]
+        #[doc = concat!("  ", stringify!($good_example), ".parse::<icu_locid::", $(stringify!($path), "::",)+ stringify!($name), ">().unwrap()")]
+        /// );
+        /// ```
+        ///
+        /// Invalid input is a compile failure:
+        /// ```compile_fail,E0080
+        #[doc = concat!("icu_locid::", $(stringify!($path), "::",)+ stringify!($macro_name), "!(", stringify!($bad_example) ,");")]
+        /// ```
+        ///
+        #[doc = concat!("[`", stringify!($name), "`]: crate::", $(stringify!($path), "::",)+ stringify!($name))]
+        #[macro_export]
+        #[doc(hidden)]
+        macro_rules! $legacy_macro_name {
+            ($string:literal) => {{
+                use $crate::$($path ::)+ $name;
+                const R: $name =
+                    match $name::try_from_bytes($string.as_bytes()) {
+                        Ok(r) => r,
+                        #[allow(clippy::panic)] // const context
+                        _ => panic!(concat!("Invalid ", $(stringify!($path), "::",)+ stringify!($name), ": ", $string)),
+                    };
+                R
+            }};
+        }
+        #[doc(inline)]
+        pub use $legacy_macro_name as $macro_name;
+
+        #[cfg(feature = "databake")]
+        impl databake::Bake for $name {
+            fn bake(&self, env: &databake::CrateEnv) -> databake::TokenStream {
+                env.insert("icu_locid");
+                let string = self.as_str();
+                databake::quote! { icu_locid::$($path::)+ $macro_name!(#string) }
+            }
+        }
+
+        #[test]
+        fn test_construction() {
+            let maybe = $name::try_from_bytes($good_example.as_bytes());
+            assert!(maybe.is_ok());
+            assert_eq!(maybe, $name::try_from_raw(maybe.unwrap().into_raw()));
+            assert_eq!(maybe.unwrap().as_str(), $good_example);
+            $(
+                let maybe = $name::try_from_bytes($more_good_examples.as_bytes());
+                assert!(maybe.is_ok());
+                assert_eq!(maybe, $name::try_from_raw(maybe.unwrap().into_raw()));
+                assert_eq!(maybe.unwrap().as_str(), $more_good_examples);
+            )*
+            assert!($name::try_from_bytes($bad_example.as_bytes()).is_err());
+            $(
+                assert!($name::try_from_bytes($more_bad_examples.as_bytes()).is_err());
+            )*
+        }
+
+        #[test]
+        fn test_writeable() {
+            writeable::assert_writeable_eq!(&$good_example.parse::<$name>().unwrap(), $good_example);
+            $(
+                writeable::assert_writeable_eq!($more_good_examples.parse::<$name>().unwrap(), $more_good_examples);
+            )*
+        }
+
+        #[cfg(feature = "serde")]
+        impl<'de> serde::Deserialize<'de> for $name {
+            fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
+            where
+                D: serde::de::Deserializer<'de>,
+            {
+                struct Visitor;
+
+                impl<'de> serde::de::Visitor<'de> for Visitor {
+                    type Value = $name;
+
+                    fn expecting(
+                        &self,
+                        formatter: &mut core::fmt::Formatter<'_>,
+                    ) -> core::fmt::Result {
+                        write!(formatter, "a valid BCP-47 {}", stringify!($name))
+                    }
+
+                    fn visit_str<E: serde::de::Error>(self, s: &str) -> Result<Self::Value, E> {
+                        s.parse().map_err(serde::de::Error::custom)
+                    }
+                }
+
+                if deserializer.is_human_readable() {
+                    deserializer.deserialize_string(Visitor)
+                } else {
+                    Self::try_from_raw(serde::de::Deserialize::deserialize(deserializer)?)
+                        .map_err(serde::de::Error::custom)
+                }
+            }
+        }
+
+        // Safety checklist for ULE:
+        //
+        // 1. Must not include any uninitialized or padding bytes (true since transparent over a ULE).
+        // 2. Must have an alignment of 1 byte (true since transparent over a ULE).
+        // 3. ULE::validate_byte_slice() checks that the given byte slice represents a valid slice.
+        // 4. ULE::validate_byte_slice() checks that the given byte slice has a valid length.
+        // 5. All other methods must be left with their default impl.
+        // 6. Byte equality is semantic equality.
+        #[cfg(feature = "zerovec")]
+        unsafe impl zerovec::ule::ULE for $name {
+            fn validate_byte_slice(bytes: &[u8]) -> Result<(), zerovec::ZeroVecError> {
+                let it = bytes.chunks_exact(core::mem::size_of::<Self>());
+                if !it.remainder().is_empty() {
+                    return Err(zerovec::ZeroVecError::length::<Self>(bytes.len()));
+                }
+                for v in it {
+                    // The following can be removed once `array_chunks` is stabilized.
+                    let mut a = [0; core::mem::size_of::<Self>()];
+                    a.copy_from_slice(v);
+                    if Self::try_from_raw(a).is_err() {
+                        return Err(zerovec::ZeroVecError::parse::<Self>());
+                    }
+                }
+                Ok(())
+            }
+        }
+
+        #[cfg(feature = "zerovec")]
+        impl zerovec::ule::AsULE for $name {
+            type ULE = Self;
+            fn to_unaligned(self) -> Self::ULE {
+                self
+            }
+            fn from_unaligned(unaligned: Self::ULE) -> Self {
+                unaligned
+            }
+        }
+
+        #[cfg(feature = "zerovec")]
+        impl<'a> zerovec::maps::ZeroMapKV<'a> for $name {
+            type Container = zerovec::ZeroVec<'a, $name>;
+            type Slice = zerovec::ZeroSlice<$name>;
+            type GetType = $name;
+            type OwnedType = $name;
+        }
+    };
+}
+
+macro_rules! impl_writeable_for_each_subtag_str_no_test {
+    ($type:tt $(, $self:ident, $borrow_cond:expr => $borrow:expr)?) => {
+        impl writeable::Writeable for $type {
+            fn write_to<W: core::fmt::Write + ?Sized>(&self, sink: &mut W) -> core::fmt::Result {
+                let mut initial = true;
+                self.for_each_subtag_str(&mut |subtag| {
+                    if initial {
+                        initial = false;
+                    } else {
+                        sink.write_char('-')?;
+                    }
+                    sink.write_str(subtag)
+                })
+            }
+
+            #[inline]
+            fn writeable_length_hint(&self) -> writeable::LengthHint {
+                let mut result = writeable::LengthHint::exact(0);
+                let mut initial = true;
+                self.for_each_subtag_str::<core::convert::Infallible, _>(&mut |subtag| {
+                    if initial {
+                        initial = false;
+                    } else {
+                        result += 1;
+                    }
+                    result += subtag.len();
+                    Ok(())
+                })
+                .expect("infallible");
+                result
+            }
+
+            $(
+                fn write_to_string(&self) -> alloc::borrow::Cow<str> {
+                    #[allow(clippy::unwrap_used)] // impl_writeable_for_subtag_list's $borrow uses unwrap
+                    let $self = self;
+                    if $borrow_cond {
+                        $borrow
+                    } else {
+                        let mut output = alloc::string::String::with_capacity(self.writeable_length_hint().capacity());
+                        let _ = self.write_to(&mut output);
+                        alloc::borrow::Cow::Owned(output)
+                    }
+                }
+            )?
+        }
+
+        writeable::impl_display_with_writeable!($type);
+    };
+}
+
+macro_rules! impl_writeable_for_subtag_list {
+    ($type:tt, $sample1:literal, $sample2:literal) => {
+        impl_writeable_for_each_subtag_str_no_test!($type, selff, selff.0.len() == 1 => alloc::borrow::Cow::Borrowed(selff.0.get(0).unwrap().as_str()));
+
+        #[test]
+        fn test_writeable() {
+            writeable::assert_writeable_eq!(&$type::default(), "");
+            writeable::assert_writeable_eq!(
+                &$type::from_short_slice_unchecked(alloc::vec![$sample1.parse().unwrap()].into()),
+                $sample1,
+            );
+            writeable::assert_writeable_eq!(
+                &$type::from_short_slice_unchecked(vec![
+                    $sample1.parse().unwrap(),
+                    $sample2.parse().unwrap()
+                ].into()),
+                core::concat!($sample1, "-", $sample2),
+            );
+        }
+    };
+}
+
+macro_rules! impl_writeable_for_key_value {
+    ($type:tt, $key1:literal, $value1:literal, $key2:literal, $expected2:literal) => {
+        impl_writeable_for_each_subtag_str_no_test!($type);
+
+        #[test]
+        fn test_writeable() {
+            writeable::assert_writeable_eq!(&$type::default(), "");
+            writeable::assert_writeable_eq!(
+                &$type::from_tuple_vec(vec![($key1.parse().unwrap(), $value1.parse().unwrap())]),
+                core::concat!($key1, "-", $value1),
+            );
+            writeable::assert_writeable_eq!(
+                &$type::from_tuple_vec(vec![
+                    ($key1.parse().unwrap(), $value1.parse().unwrap()),
+                    ($key2.parse().unwrap(), "true".parse().unwrap())
+                ]),
+                core::concat!($key1, "-", $value1, "-", $expected2),
+            );
+        }
+    };
+}
diff --git a/third_party/rust/icu_locid/src/langid.rs b/third_party/rust/icu_locid/src/langid.rs
new file mode 100644
index 0000000000..eac8c83713
--- /dev/null
+++ b/third_party/rust/icu_locid/src/langid.rs
@@ -0,0 +1,574 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use core::cmp::Ordering;
+use core::str::FromStr;
+
+use crate::ordering::SubtagOrderingResult;
+use crate::parser::{
+    parse_language_identifier, parse_language_identifier_with_single_variant, ParserError,
+    ParserMode, SubtagIterator,
+};
+use crate::subtags;
+use alloc::string::String;
+use writeable::Writeable;
+
+/// A core struct representing a [`Unicode BCP47 Language Identifier`].
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::{
+///     langid,
+///     subtags::{language, region},
+/// };
+///
+/// let li = langid!("en-US");
+///
+/// assert_eq!(li.language, language!("en"));
+/// assert_eq!(li.script, None);
+/// assert_eq!(li.region, Some(region!("US")));
+/// assert_eq!(li.variants.len(), 0);
+/// ```
+///
+/// # Parsing
+///
+/// Unicode recognizes three levels of standard conformance for any language identifier:
+///
+///  * *well-formed* - syntactically correct
+///  * *valid* - well-formed and only uses registered language, region, script and variant subtags...
+///  * *canonical* - valid and no deprecated codes or structure.
+///
+/// At the moment parsing normalizes a well-formed language identifier converting
+/// `_` separators to `-` and adjusting casing to conform to the Unicode standard.
+///
+/// Any bogus subtags will cause the parsing to fail with an error.
+/// No subtag validation is performed.
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::{
+///     langid,
+///     subtags::{language, region, script, variant},
+/// };
+///
+/// let li = langid!("eN_latn_Us-Valencia");
+///
+/// assert_eq!(li.language, language!("en"));
+/// assert_eq!(li.script, Some(script!("Latn")));
+/// assert_eq!(li.region, Some(region!("US")));
+/// assert_eq!(li.variants.get(0), Some(&variant!("valencia")));
+/// ```
+///
+/// [`Unicode BCP47 Language Identifier`]: https://unicode.org/reports/tr35/tr35.html#Unicode_language_identifier
+#[derive(Default, PartialEq, Eq, Clone, Hash)]
+#[allow(clippy::exhaustive_structs)] // This struct is stable (and invoked by a macro)
+pub struct LanguageIdentifier {
+    /// Language subtag of the language identifier.
+    pub language: subtags::Language,
+    /// Script subtag of the language identifier.
+    pub script: Option<subtags::Script>,
+    /// Region subtag of the language identifier.
+    pub region: Option<subtags::Region>,
+    /// Variant subtags of the language identifier.
+    pub variants: subtags::Variants,
+}
+
+impl LanguageIdentifier {
+    /// A constructor which takes a utf8 slice, parses it and
+    /// produces a well-formed [`LanguageIdentifier`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::LanguageIdentifier;
+    ///
+    /// LanguageIdentifier::try_from_bytes(b"en-US").expect("Parsing failed");
+    /// ```
+    pub fn try_from_bytes(v: &[u8]) -> Result<Self, ParserError> {
+        parse_language_identifier(v, ParserMode::LanguageIdentifier)
+    }
+
+    #[doc(hidden)]
+    #[allow(clippy::type_complexity)]
+    // The return type should be `Result<Self, ParserError>` once the `const_precise_live_drops`
+    // is stabilized ([rust-lang#73255](https://github.com/rust-lang/rust/issues/73255)).
+    pub const fn try_from_bytes_with_single_variant(
+        v: &[u8],
+    ) -> Result<
+        (
+            subtags::Language,
+            Option<subtags::Script>,
+            Option<subtags::Region>,
+            Option<subtags::Variant>,
+        ),
+        ParserError,
+    > {
+        parse_language_identifier_with_single_variant(v, ParserMode::LanguageIdentifier)
+    }
+
+    /// A constructor which takes a utf8 slice which may contain extension keys,
+    /// parses it and produces a well-formed [`LanguageIdentifier`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::{langid, LanguageIdentifier};
+    ///
+    /// let li = LanguageIdentifier::try_from_locale_bytes(b"en-US-x-posix")
+    ///     .expect("Parsing failed.");
+    ///
+    /// assert_eq!(li, langid!("en-US"));
+    /// ```
+    ///
+    /// This method should be used for input that may be a locale identifier.
+    /// All extensions will be lost.
+    pub fn try_from_locale_bytes(v: &[u8]) -> Result<Self, ParserError> {
+        parse_language_identifier(v, ParserMode::Locale)
+    }
+
+    /// The default undefined language "und". Same as [`default()`](Default::default()).
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::LanguageIdentifier;
+    ///
+    /// assert_eq!(LanguageIdentifier::default(), LanguageIdentifier::UND);
+    /// ```
+    pub const UND: Self = Self {
+        language: subtags::Language::UND,
+        script: None,
+        region: None,
+        variants: subtags::Variants::new(),
+    };
+
+    /// This is a best-effort operation that performs all available levels of canonicalization.
+    ///
+    /// At the moment the operation will normalize casing and the separator, but in the future
+    /// it may also validate and update from deprecated subtags to canonical ones.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::LanguageIdentifier;
+    ///
+    /// assert_eq!(
+    ///     LanguageIdentifier::canonicalize("pL_latn_pl").as_deref(),
+    ///     Ok("pl-Latn-PL")
+    /// );
+    /// ```
+    pub fn canonicalize<S: AsRef<[u8]>>(input: S) -> Result<String, ParserError> {
+        let lang_id = Self::try_from_bytes(input.as_ref())?;
+        Ok(lang_id.write_to_string().into_owned())
+    }
+
+    /// Compare this [`LanguageIdentifier`] with BCP-47 bytes.
+    ///
+    /// The return value is equivalent to what would happen if you first converted this
+    /// [`LanguageIdentifier`] to a BCP-47 string and then performed a byte comparison.
+    ///
+    /// This function is case-sensitive and results in a *total order*, so it is appropriate for
+    /// binary search. The only argument producing [`Ordering::Equal`] is `self.to_string()`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::LanguageIdentifier;
+    /// use std::cmp::Ordering;
+    ///
+    /// let bcp47_strings: &[&str] = &[
+    ///     "pl-Latn-PL",
+    ///     "und",
+    ///     "und-Adlm",
+    ///     "und-GB",
+    ///     "und-ZA",
+    ///     "und-fonipa",
+    ///     "zh",
+    /// ];
+    ///
+    /// for ab in bcp47_strings.windows(2) {
+    ///     let a = ab[0];
+    ///     let b = ab[1];
+    ///     assert!(a.cmp(b) == Ordering::Less);
+    ///     let a_langid = a.parse::<LanguageIdentifier>().unwrap();
+    ///     assert!(a_langid.strict_cmp(a.as_bytes()) == Ordering::Equal);
+    ///     assert!(a_langid.strict_cmp(b.as_bytes()) == Ordering::Less);
+    /// }
+    /// ```
+    pub fn strict_cmp(&self, other: &[u8]) -> Ordering {
+        self.strict_cmp_iter(other.split(|b| *b == b'-')).end()
+    }
+
+    /// Compare this [`LanguageIdentifier`] with an iterator of BCP-47 subtags.
+    ///
+    /// This function has the same equality semantics as [`LanguageIdentifier::strict_cmp`]. It is intended as
+    /// a more modular version that allows multiple subtag iterators to be chained together.
+    ///
+    /// For an additional example, see [`SubtagOrderingResult`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::LanguageIdentifier;
+    /// use std::cmp::Ordering;
+    ///
+    /// let subtags: &[&[u8]] = &[b"ca", b"ES", b"valencia"];
+    ///
+    /// let loc = "ca-ES-valencia".parse::<LanguageIdentifier>().unwrap();
+    /// assert_eq!(
+    ///     Ordering::Equal,
+    ///     loc.strict_cmp_iter(subtags.iter().copied()).end()
+    /// );
+    ///
+    /// let loc = "ca-ES".parse::<LanguageIdentifier>().unwrap();
+    /// assert_eq!(
+    ///     Ordering::Less,
+    ///     loc.strict_cmp_iter(subtags.iter().copied()).end()
+    /// );
+    ///
+    /// let loc = "ca-ZA".parse::<LanguageIdentifier>().unwrap();
+    /// assert_eq!(
+    ///     Ordering::Greater,
+    ///     loc.strict_cmp_iter(subtags.iter().copied()).end()
+    /// );
+    /// ```
+    pub fn strict_cmp_iter<'l, I>(&self, mut subtags: I) -> SubtagOrderingResult<I>
+    where
+        I: Iterator<Item = &'l [u8]>,
+    {
+        let r = self.for_each_subtag_str(&mut |subtag| {
+            if let Some(other) = subtags.next() {
+                match subtag.as_bytes().cmp(other) {
+                    Ordering::Equal => Ok(()),
+                    not_equal => Err(not_equal),
+                }
+            } else {
+                Err(Ordering::Greater)
+            }
+        });
+        match r {
+            Ok(_) => SubtagOrderingResult::Subtags(subtags),
+            Err(o) => SubtagOrderingResult::Ordering(o),
+        }
+    }
+
+    /// Compare this `LanguageIdentifier` with a potentially unnormalized BCP-47 string.
+    ///
+    /// The return value is equivalent to what would happen if you first parsed the
+    /// BCP-47 string to a `LanguageIdentifier` and then performed a structural comparison.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::LanguageIdentifier;
+    /// use std::cmp::Ordering;
+    ///
+    /// let bcp47_strings: &[&str] = &[
+    ///     "pl-LaTn-pL",
+    ///     "uNd",
+    ///     "UnD-adlm",
+    ///     "uNd-GB",
+    ///     "UND-FONIPA",
+    ///     "ZH",
+    /// ];
+    ///
+    /// for a in bcp47_strings {
+    ///     assert!(a.parse::<LanguageIdentifier>().unwrap().normalizing_eq(a));
+    /// }
+    /// ```
+    pub fn normalizing_eq(&self, other: &str) -> bool {
+        macro_rules! subtag_matches {
+            ($T:ty, $iter:ident, $expected:expr) => {
+                $iter
+                    .next()
+                    .map(|b| <$T>::try_from_bytes(b) == Ok($expected))
+                    .unwrap_or(false)
+            };
+        }
+
+        let mut iter = SubtagIterator::new(other.as_bytes());
+        if !subtag_matches!(subtags::Language, iter, self.language) {
+            return false;
+        }
+        if let Some(ref script) = self.script {
+            if !subtag_matches!(subtags::Script, iter, *script) {
+                return false;
+            }
+        }
+        if let Some(ref region) = self.region {
+            if !subtag_matches!(subtags::Region, iter, *region) {
+                return false;
+            }
+        }
+        for variant in self.variants.iter() {
+            if !subtag_matches!(subtags::Variant, iter, *variant) {
+                return false;
+            }
+        }
+        iter.next().is_none()
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        f(self.language.as_str())?;
+        if let Some(ref script) = self.script {
+            f(script.as_str())?;
+        }
+        if let Some(ref region) = self.region {
+            f(region.as_str())?;
+        }
+        for variant in self.variants.iter() {
+            f(variant.as_str())?;
+        }
+        Ok(())
+    }
+
+    /// Executes `f` on each subtag string of this `LanguageIdentifier`, with every string in
+    /// lowercase ascii form.
+    ///
+    /// The default canonicalization of language identifiers uses titlecase scripts and uppercase
+    /// regions. However, this differs from [RFC6497 (BCP 47 Extension T)], which specifies:
+    ///
+    /// > _The canonical form for all subtags in the extension is lowercase, with the fields
+    /// ordered by the separators, alphabetically._
+    ///
+    /// Hence, this method is used inside [`Transform Extensions`] to be able to get the correct
+    /// canonicalization of the language identifier.
+    ///
+    /// As an example, the canonical form of locale **EN-LATN-CA-T-EN-LATN-CA** is
+    /// **en-Latn-CA-t-en-latn-ca**, with the script and region parts lowercased inside T extensions,
+    /// but titlecased and uppercased outside T extensions respectively.
+    ///
+    /// [RFC6497 (BCP 47 Extension T)]: https://www.ietf.org/rfc/rfc6497.txt
+    /// [`Transform extensions`]: crate::extensions::transform
+    pub(crate) fn for_each_subtag_str_lowercased<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        f(self.language.as_str())?;
+        if let Some(ref script) = self.script {
+            f(script.into_tinystr().to_ascii_lowercase().as_str())?;
+        }
+        if let Some(ref region) = self.region {
+            f(region.into_tinystr().to_ascii_lowercase().as_str())?;
+        }
+        for variant in self.variants.iter() {
+            f(variant.as_str())?;
+        }
+        Ok(())
+    }
+
+    /// Writes this `LanguageIdentifier` to a sink, replacing uppercase ascii chars with
+    /// lowercase ascii chars.
+    ///
+    /// The default canonicalization of language identifiers uses titlecase scripts and uppercase
+    /// regions. However, this differs from [RFC6497 (BCP 47 Extension T)], which specifies:
+    ///
+    /// > _The canonical form for all subtags in the extension is lowercase, with the fields
+    /// ordered by the separators, alphabetically._
+    ///
+    /// Hence, this method is used inside [`Transform Extensions`] to be able to get the correct
+    /// canonicalization of the language identifier.
+    ///
+    /// As an example, the canonical form of locale **EN-LATN-CA-T-EN-LATN-CA** is
+    /// **en-Latn-CA-t-en-latn-ca**, with the script and region parts lowercased inside T extensions,
+    /// but titlecased and uppercased outside T extensions respectively.
+    ///
+    /// [RFC6497 (BCP 47 Extension T)]: https://www.ietf.org/rfc/rfc6497.txt
+    /// [`Transform extensions`]: crate::extensions::transform
+    pub(crate) fn write_lowercased_to<W: core::fmt::Write + ?Sized>(
+        &self,
+        sink: &mut W,
+    ) -> core::fmt::Result {
+        let mut initial = true;
+        self.for_each_subtag_str_lowercased(&mut |subtag| {
+            if initial {
+                initial = false;
+            } else {
+                sink.write_char('-')?;
+            }
+            sink.write_str(subtag)
+        })
+    }
+}
+
+impl AsRef<LanguageIdentifier> for LanguageIdentifier {
+    fn as_ref(&self) -> &Self {
+        self
+    }
+}
+
+impl AsMut<LanguageIdentifier> for LanguageIdentifier {
+    fn as_mut(&mut self) -> &mut Self {
+        self
+    }
+}
+
+impl core::fmt::Debug for LanguageIdentifier {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        core::fmt::Display::fmt(&self, f)
+    }
+}
+
+impl FromStr for LanguageIdentifier {
+    type Err = ParserError;
+
+    fn from_str(source: &str) -> Result<Self, Self::Err> {
+        Self::try_from_bytes(source.as_bytes())
+    }
+}
+
+impl_writeable_for_each_subtag_str_no_test!(LanguageIdentifier, selff, selff.script.is_none() && selff.region.is_none() && selff.variants.is_empty() => selff.language.write_to_string());
+
+#[test]
+fn test_writeable() {
+    use writeable::assert_writeable_eq;
+    assert_writeable_eq!(LanguageIdentifier::UND, "und");
+    assert_writeable_eq!("und-001".parse::<LanguageIdentifier>().unwrap(), "und-001");
+    assert_writeable_eq!(
+        "und-Mymr".parse::<LanguageIdentifier>().unwrap(),
+        "und-Mymr",
+    );
+    assert_writeable_eq!(
+        "my-Mymr-MM".parse::<LanguageIdentifier>().unwrap(),
+        "my-Mymr-MM",
+    );
+    assert_writeable_eq!(
+        "my-Mymr-MM-posix".parse::<LanguageIdentifier>().unwrap(),
+        "my-Mymr-MM-posix",
+    );
+    assert_writeable_eq!(
+        "zh-macos-posix".parse::<LanguageIdentifier>().unwrap(),
+        "zh-macos-posix",
+    );
+}
+
+/// # Examples
+///
+/// ```
+/// use icu::locid::{langid, subtags::language, LanguageIdentifier};
+///
+/// assert_eq!(LanguageIdentifier::from(language!("en")), langid!("en"));
+/// ```
+impl From<subtags::Language> for LanguageIdentifier {
+    fn from(language: subtags::Language) -> Self {
+        Self {
+            language,
+            ..Default::default()
+        }
+    }
+}
+
+/// # Examples
+///
+/// ```
+/// use icu::locid::{langid, subtags::script, LanguageIdentifier};
+///
+/// assert_eq!(
+///     LanguageIdentifier::from(Some(script!("latn"))),
+///     langid!("und-Latn")
+/// );
+/// ```
+impl From<Option<subtags::Script>> for LanguageIdentifier {
+    fn from(script: Option<subtags::Script>) -> Self {
+        Self {
+            script,
+            ..Default::default()
+        }
+    }
+}
+
+/// # Examples
+///
+/// ```
+/// use icu::locid::{langid, subtags::region, LanguageIdentifier};
+///
+/// assert_eq!(
+///     LanguageIdentifier::from(Some(region!("US"))),
+///     langid!("und-US")
+/// );
+/// ```
+impl From<Option<subtags::Region>> for LanguageIdentifier {
+    fn from(region: Option<subtags::Region>) -> Self {
+        Self {
+            region,
+            ..Default::default()
+        }
+    }
+}
+
+/// Convert from an LSR tuple to a [`LanguageIdentifier`].
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::{
+///     langid,
+///     subtags::{language, region, script},
+///     LanguageIdentifier,
+/// };
+///
+/// let lang = language!("en");
+/// let script = script!("Latn");
+/// let region = region!("US");
+/// assert_eq!(
+///     LanguageIdentifier::from((lang, Some(script), Some(region))),
+///     langid!("en-Latn-US")
+/// );
+/// ```
+impl
+    From<(
+        subtags::Language,
+        Option<subtags::Script>,
+        Option<subtags::Region>,
+    )> for LanguageIdentifier
+{
+    fn from(
+        lsr: (
+            subtags::Language,
+            Option<subtags::Script>,
+            Option<subtags::Region>,
+        ),
+    ) -> Self {
+        Self {
+            language: lsr.0,
+            script: lsr.1,
+            region: lsr.2,
+            ..Default::default()
+        }
+    }
+}
+
+/// Convert from a [`LanguageIdentifier`] to an LSR tuple.
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::{
+///     langid,
+///     subtags::{language, region, script},
+/// };
+///
+/// let lid = langid!("en-Latn-US");
+/// let (lang, script, region) = (&lid).into();
+///
+/// assert_eq!(lang, language!("en"));
+/// assert_eq!(script, Some(script!("Latn")));
+/// assert_eq!(region, Some(region!("US")));
+/// ```
+impl From<&LanguageIdentifier>
+    for (
+        subtags::Language,
+        Option<subtags::Script>,
+        Option<subtags::Region>,
+    )
+{
+    fn from(langid: &LanguageIdentifier) -> Self {
+        (langid.language, langid.script, langid.region)
+    }
+}
diff --git a/third_party/rust/icu_locid/src/lib.rs b/third_party/rust/icu_locid/src/lib.rs
new file mode 100644
index 0000000000..9c6c46ca51
--- /dev/null
+++ b/third_party/rust/icu_locid/src/lib.rs
@@ -0,0 +1,93 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+//! Parsing, manipulating, and serializing Unicode Language and Locale Identifiers.
+//!
+//! This module is published as its own crate ([`icu_locid`](https://docs.rs/icu_locid/latest/icu_locid/))
+//! and as part of the [`icu`](https://docs.rs/icu/latest/icu/) crate. See the latter for more details on the ICU4X project.
+//!
+//! The module provides algorithms for parsing a string into a well-formed language or locale identifier
+//! as defined by [`UTS #35: Unicode LDML 3. Unicode Language and Locale Identifiers`].
+//!
+//! [`Locale`] is the most common structure to use for storing information about a language,
+//! script, region, variants and extensions. In almost all cases, this struct should be used as the
+//! base unit for all locale management operations.
+//!
+//! [`LanguageIdentifier`] is a strict subset of [`Locale`] which can be useful in a narrow range of
+//! cases where [`Unicode Extensions`] are not relevant.
+//!
+//! If in doubt, use [`Locale`].
+//!
+//! # Examples
+//!
+//! ```
+//! use icu::locid::Locale;
+//! use icu::locid::{
+//!     locale,
+//!     subtags::{language, region},
+//! };
+//!
+//! let mut loc: Locale = locale!("en-US");
+//!
+//! assert_eq!(loc.id.language, language!("en"));
+//! assert_eq!(loc.id.script, None);
+//! assert_eq!(loc.id.region, Some(region!("US")));
+//! assert_eq!(loc.id.variants.len(), 0);
+//!
+//! loc.id.region = Some(region!("GB"));
+//!
+//! assert_eq!(loc, locale!("en-GB"));
+//! ```
+//!
+//! For more details, see [`Locale`] and [`LanguageIdentifier`].
+//!
+//! [`UTS #35: Unicode LDML 3. Unicode Language and Locale Identifiers`]: https://unicode.org/reports/tr35/tr35.html#Unicode_Language_and_Locale_Identifiers
+//! [`ICU4X`]: ../icu/index.html
+//! [`Unicode Extensions`]: extensions
+
+// https://github.com/unicode-org/icu4x/blob/main/docs/process/boilerplate.md#library-annotations
+#![cfg_attr(not(any(test, feature = "std")), no_std)]
+#![cfg_attr(
+    not(test),
+    deny(
+        clippy::indexing_slicing,
+        clippy::unwrap_used,
+        clippy::expect_used,
+        clippy::panic,
+        clippy::exhaustive_structs,
+        clippy::exhaustive_enums,
+        missing_debug_implementations,
+    )
+)]
+#![warn(missing_docs)]
+
+extern crate alloc;
+
+#[macro_use]
+mod helpers;
+
+mod langid;
+mod locale;
+mod macros;
+mod ordering;
+mod parser;
+
+pub use langid::LanguageIdentifier;
+pub use locale::Locale;
+pub use ordering::SubtagOrderingResult;
+pub use parser::errors::ParserError;
+
+#[doc(no_inline)]
+pub use ParserError as Error;
+
+pub mod extensions;
+#[macro_use]
+pub mod subtags;
+pub mod zerovec;
+
+#[cfg(feature = "serde")]
+mod serde;
+
+#[cfg(feature = "databake")]
+mod databake;
diff --git a/third_party/rust/icu_locid/src/locale.rs b/third_party/rust/icu_locid/src/locale.rs
new file mode 100644
index 0000000000..e87cdf1a20
--- /dev/null
+++ b/third_party/rust/icu_locid/src/locale.rs
@@ -0,0 +1,511 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use crate::ordering::SubtagOrderingResult;
+use crate::parser::{
+    parse_locale, parse_locale_with_single_variant_single_keyword_unicode_keyword_extension,
+    ParserError, ParserMode, SubtagIterator,
+};
+use crate::{extensions, subtags, LanguageIdentifier};
+use alloc::string::String;
+use core::cmp::Ordering;
+use core::str::FromStr;
+use tinystr::TinyAsciiStr;
+use writeable::Writeable;
+
+/// A core struct representing a [`Unicode Locale Identifier`].
+///
+/// A locale is made of two parts:
+///  * Unicode Language Identifier
+///  * A set of Unicode Extensions
+///
+/// [`Locale`] exposes all of the same fields and methods as [`LanguageIdentifier`], and
+/// on top of that is able to parse, manipulate and serialize unicode extension fields.
+///
+///
+/// # Examples
+///
+/// ```
+/// use icu_locid::{
+///     extensions::unicode::{key, value},
+///     locale,
+///     subtags::{language, region},
+/// };
+///
+/// let loc = locale!("en-US-u-ca-buddhist");
+///
+/// assert_eq!(loc.id.language, language!("en"));
+/// assert_eq!(loc.id.script, None);
+/// assert_eq!(loc.id.region, Some(region!("US")));
+/// assert_eq!(loc.id.variants.len(), 0);
+/// assert_eq!(
+///     loc.extensions.unicode.keywords.get(&key!("ca")),
+///     Some(&value!("buddhist"))
+/// );
+/// ```
+///
+/// # Parsing
+///
+/// Unicode recognizes three levels of standard conformance for a locale:
+///
+///  * *well-formed* - syntactically correct
+///  * *valid* - well-formed and only uses registered language subtags, extensions, keywords, types...
+///  * *canonical* - valid and no deprecated codes or structure.
+///
+/// At the moment parsing normalizes a well-formed locale identifier converting
+/// `_` separators to `-` and adjusting casing to conform to the Unicode standard.
+///
+/// Any bogus subtags will cause the parsing to fail with an error.
+///
+/// No subtag validation or alias resolution is performed.
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::{subtags::*, Locale};
+///
+/// let loc: Locale = "eN_latn_Us-Valencia_u-hC-H12"
+///     .parse()
+///     .expect("Failed to parse.");
+///
+/// assert_eq!(loc.id.language, "en".parse::<Language>().unwrap());
+/// assert_eq!(loc.id.script, "Latn".parse::<Script>().ok());
+/// assert_eq!(loc.id.region, "US".parse::<Region>().ok());
+/// assert_eq!(
+///     loc.id.variants.get(0),
+///     "valencia".parse::<Variant>().ok().as_ref()
+/// );
+/// ```
+/// [`Unicode Locale Identifier`]: https://unicode.org/reports/tr35/tr35.html#Unicode_locale_identifier
+#[derive(Default, PartialEq, Eq, Clone, Hash)]
+#[allow(clippy::exhaustive_structs)] // This struct is stable (and invoked by a macro)
+pub struct Locale {
+    /// The basic language/script/region components in the locale identifier along with any variants.
+    pub id: LanguageIdentifier,
+    /// Any extensions present in the locale identifier.
+    pub extensions: extensions::Extensions,
+}
+
+#[test]
+fn test_sizes() {
+    assert_eq!(core::mem::size_of::<subtags::Language>(), 3);
+    assert_eq!(core::mem::size_of::<subtags::Script>(), 4);
+    assert_eq!(core::mem::size_of::<subtags::Region>(), 3);
+    assert_eq!(core::mem::size_of::<subtags::Variant>(), 8);
+    assert_eq!(core::mem::size_of::<subtags::Variants>(), 16);
+    assert_eq!(core::mem::size_of::<LanguageIdentifier>(), 32);
+
+    assert_eq!(core::mem::size_of::<extensions::transform::Transform>(), 56);
+    assert_eq!(core::mem::size_of::<Option<LanguageIdentifier>>(), 32);
+    assert_eq!(core::mem::size_of::<extensions::transform::Fields>(), 24);
+
+    assert_eq!(core::mem::size_of::<extensions::unicode::Attributes>(), 16);
+    assert_eq!(core::mem::size_of::<extensions::unicode::Keywords>(), 24);
+    assert_eq!(core::mem::size_of::<Vec<extensions::other::Other>>(), 24);
+    assert_eq!(core::mem::size_of::<extensions::private::Private>(), 16);
+    assert_eq!(core::mem::size_of::<extensions::Extensions>(), 136);
+
+    assert_eq!(core::mem::size_of::<Locale>(), 168);
+}
+
+impl Locale {
+    /// A constructor which takes a utf8 slice, parses it and
+    /// produces a well-formed [`Locale`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// Locale::try_from_bytes(b"en-US-u-hc-h12").unwrap();
+    /// ```
+    pub fn try_from_bytes(v: &[u8]) -> Result<Self, ParserError> {
+        parse_locale(v)
+    }
+
+    /// The default undefined locale "und". Same as [`default()`](Default::default()).
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// assert_eq!(Locale::default(), Locale::UND);
+    /// ```
+    pub const UND: Self = Self {
+        id: LanguageIdentifier::UND,
+        extensions: extensions::Extensions::new(),
+    };
+
+    /// This is a best-effort operation that performs all available levels of canonicalization.
+    ///
+    /// At the moment the operation will normalize casing and the separator, but in the future
+    /// it may also validate and update from deprecated subtags to canonical ones.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    ///
+    /// assert_eq!(
+    ///     Locale::canonicalize("pL_latn_pl-U-HC-H12").as_deref(),
+    ///     Ok("pl-Latn-PL-u-hc-h12")
+    /// );
+    /// ```
+    pub fn canonicalize<S: AsRef<[u8]>>(input: S) -> Result<String, ParserError> {
+        let locale = Self::try_from_bytes(input.as_ref())?;
+        Ok(locale.write_to_string().into_owned())
+    }
+
+    /// Compare this [`Locale`] with BCP-47 bytes.
+    ///
+    /// The return value is equivalent to what would happen if you first converted this
+    /// [`Locale`] to a BCP-47 string and then performed a byte comparison.
+    ///
+    /// This function is case-sensitive and results in a *total order*, so it is appropriate for
+    /// binary search. The only argument producing [`Ordering::Equal`] is `self.to_string()`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    /// use std::cmp::Ordering;
+    ///
+    /// let bcp47_strings: &[&str] = &[
+    ///     "pl-Latn-PL",
+    ///     "und",
+    ///     "und-fonipa",
+    ///     "und-t-m0-true",
+    ///     "und-u-ca-hebrew",
+    ///     "und-u-ca-japanese",
+    ///     "zh",
+    /// ];
+    ///
+    /// for ab in bcp47_strings.windows(2) {
+    ///     let a = ab[0];
+    ///     let b = ab[1];
+    ///     assert!(a.cmp(b) == Ordering::Less);
+    ///     let a_loc = a.parse::<Locale>().unwrap();
+    ///     assert!(a_loc.strict_cmp(a.as_bytes()) == Ordering::Equal);
+    ///     assert!(a_loc.strict_cmp(b.as_bytes()) == Ordering::Less);
+    /// }
+    /// ```
+    pub fn strict_cmp(&self, other: &[u8]) -> Ordering {
+        self.strict_cmp_iter(other.split(|b| *b == b'-')).end()
+    }
+
+    /// Compare this [`Locale`] with an iterator of BCP-47 subtags.
+    ///
+    /// This function has the same equality semantics as [`Locale::strict_cmp`]. It is intended as
+    /// a more modular version that allows multiple subtag iterators to be chained together.
+    ///
+    /// For an additional example, see [`SubtagOrderingResult`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::locale;
+    /// use std::cmp::Ordering;
+    ///
+    /// let subtags: &[&[u8]] =
+    ///     &[b"ca", b"ES", b"valencia", b"u", b"ca", b"hebrew"];
+    ///
+    /// let loc = locale!("ca-ES-valencia-u-ca-hebrew");
+    /// assert_eq!(
+    ///     Ordering::Equal,
+    ///     loc.strict_cmp_iter(subtags.iter().copied()).end()
+    /// );
+    ///
+    /// let loc = locale!("ca-ES-valencia");
+    /// assert_eq!(
+    ///     Ordering::Less,
+    ///     loc.strict_cmp_iter(subtags.iter().copied()).end()
+    /// );
+    ///
+    /// let loc = locale!("ca-ES-valencia-u-nu-arab");
+    /// assert_eq!(
+    ///     Ordering::Greater,
+    ///     loc.strict_cmp_iter(subtags.iter().copied()).end()
+    /// );
+    /// ```
+    pub fn strict_cmp_iter<'l, I>(&self, mut subtags: I) -> SubtagOrderingResult<I>
+    where
+        I: Iterator<Item = &'l [u8]>,
+    {
+        let r = self.for_each_subtag_str(&mut |subtag| {
+            if let Some(other) = subtags.next() {
+                match subtag.as_bytes().cmp(other) {
+                    Ordering::Equal => Ok(()),
+                    not_equal => Err(not_equal),
+                }
+            } else {
+                Err(Ordering::Greater)
+            }
+        });
+        match r {
+            Ok(_) => SubtagOrderingResult::Subtags(subtags),
+            Err(o) => SubtagOrderingResult::Ordering(o),
+        }
+    }
+
+    /// Compare this `Locale` with a potentially unnormalized BCP-47 string.
+    ///
+    /// The return value is equivalent to what would happen if you first parsed the
+    /// BCP-47 string to a `Locale` and then performed a structural comparison.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    /// use std::cmp::Ordering;
+    ///
+    /// let bcp47_strings: &[&str] = &[
+    ///     "pl-LaTn-pL",
+    ///     "uNd",
+    ///     "UND-FONIPA",
+    ///     "UnD-t-m0-TrUe",
+    ///     "uNd-u-CA-Japanese",
+    ///     "ZH",
+    /// ];
+    ///
+    /// for a in bcp47_strings {
+    ///     assert!(a.parse::<Locale>().unwrap().normalizing_eq(a));
+    /// }
+    /// ```
+    pub fn normalizing_eq(&self, other: &str) -> bool {
+        macro_rules! subtag_matches {
+            ($T:ty, $iter:ident, $expected:expr) => {
+                $iter
+                    .next()
+                    .map(|b| <$T>::try_from_bytes(b) == Ok($expected))
+                    .unwrap_or(false)
+            };
+        }
+
+        let mut iter = SubtagIterator::new(other.as_bytes());
+        if !subtag_matches!(subtags::Language, iter, self.id.language) {
+            return false;
+        }
+        if let Some(ref script) = self.id.script {
+            if !subtag_matches!(subtags::Script, iter, *script) {
+                return false;
+            }
+        }
+        if let Some(ref region) = self.id.region {
+            if !subtag_matches!(subtags::Region, iter, *region) {
+                return false;
+            }
+        }
+        for variant in self.id.variants.iter() {
+            if !subtag_matches!(subtags::Variant, iter, *variant) {
+                return false;
+            }
+        }
+        if !self.extensions.is_empty() {
+            match extensions::Extensions::try_from_iter(&mut iter) {
+                Ok(exts) => {
+                    if self.extensions != exts {
+                        return false;
+                    }
+                }
+                Err(_) => {
+                    return false;
+                }
+            }
+        }
+        iter.next().is_none()
+    }
+
+    #[doc(hidden)]
+    #[allow(clippy::type_complexity)]
+    pub const fn try_from_bytes_with_single_variant_single_keyword_unicode_extension(
+        v: &[u8],
+    ) -> Result<
+        (
+            subtags::Language,
+            Option<subtags::Script>,
+            Option<subtags::Region>,
+            Option<subtags::Variant>,
+            Option<(extensions::unicode::Key, Option<TinyAsciiStr<8>>)>,
+        ),
+        ParserError,
+    > {
+        parse_locale_with_single_variant_single_keyword_unicode_keyword_extension(
+            v,
+            ParserMode::Locale,
+        )
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        self.id.for_each_subtag_str(f)?;
+        self.extensions.for_each_subtag_str(f)?;
+        Ok(())
+    }
+}
+
+impl FromStr for Locale {
+    type Err = ParserError;
+
+    fn from_str(source: &str) -> Result<Self, Self::Err> {
+        Self::try_from_bytes(source.as_bytes())
+    }
+}
+
+impl From<LanguageIdentifier> for Locale {
+    fn from(id: LanguageIdentifier) -> Self {
+        Self {
+            id,
+            extensions: extensions::Extensions::default(),
+        }
+    }
+}
+
+impl From<Locale> for LanguageIdentifier {
+    fn from(loc: Locale) -> Self {
+        loc.id
+    }
+}
+
+impl AsRef<LanguageIdentifier> for Locale {
+    fn as_ref(&self) -> &LanguageIdentifier {
+        &self.id
+    }
+}
+
+impl AsMut<LanguageIdentifier> for Locale {
+    fn as_mut(&mut self) -> &mut LanguageIdentifier {
+        &mut self.id
+    }
+}
+
+impl core::fmt::Debug for Locale {
+    fn fmt(&self, f: &mut core::fmt::Formatter) -> core::fmt::Result {
+        writeable::Writeable::write_to(self, f)
+    }
+}
+
+impl_writeable_for_each_subtag_str_no_test!(Locale, selff, selff.extensions.is_empty() => selff.id.write_to_string());
+
+#[test]
+fn test_writeable() {
+    use writeable::assert_writeable_eq;
+    assert_writeable_eq!(Locale::UND, "und");
+    assert_writeable_eq!("und-001".parse::<Locale>().unwrap(), "und-001");
+    assert_writeable_eq!("und-Mymr".parse::<Locale>().unwrap(), "und-Mymr");
+    assert_writeable_eq!("my-Mymr-MM".parse::<Locale>().unwrap(), "my-Mymr-MM");
+    assert_writeable_eq!(
+        "my-Mymr-MM-posix".parse::<Locale>().unwrap(),
+        "my-Mymr-MM-posix",
+    );
+    assert_writeable_eq!(
+        "zh-macos-posix".parse::<Locale>().unwrap(),
+        "zh-macos-posix",
+    );
+    assert_writeable_eq!(
+        "my-t-my-d0-zawgyi".parse::<Locale>().unwrap(),
+        "my-t-my-d0-zawgyi",
+    );
+    assert_writeable_eq!(
+        "ar-SA-u-ca-islamic-civil".parse::<Locale>().unwrap(),
+        "ar-SA-u-ca-islamic-civil",
+    );
+    assert_writeable_eq!(
+        "en-001-x-foo-bar".parse::<Locale>().unwrap(),
+        "en-001-x-foo-bar",
+    );
+    assert_writeable_eq!("und-t-m0-true".parse::<Locale>().unwrap(), "und-t-m0-true",);
+}
+
+/// # Examples
+///
+/// ```
+/// use icu::locid::Locale;
+/// use icu::locid::{locale, subtags::language};
+///
+/// assert_eq!(Locale::from(language!("en")), locale!("en"));
+/// ```
+impl From<subtags::Language> for Locale {
+    fn from(language: subtags::Language) -> Self {
+        Self {
+            id: language.into(),
+            ..Default::default()
+        }
+    }
+}
+
+/// # Examples
+///
+/// ```
+/// use icu::locid::Locale;
+/// use icu::locid::{locale, subtags::script};
+///
+/// assert_eq!(Locale::from(Some(script!("latn"))), locale!("und-Latn"));
+/// ```
+impl From<Option<subtags::Script>> for Locale {
+    fn from(script: Option<subtags::Script>) -> Self {
+        Self {
+            id: script.into(),
+            ..Default::default()
+        }
+    }
+}
+
+/// # Examples
+///
+/// ```
+/// use icu::locid::Locale;
+/// use icu::locid::{locale, subtags::region};
+///
+/// assert_eq!(Locale::from(Some(region!("US"))), locale!("und-US"));
+/// ```
+impl From<Option<subtags::Region>> for Locale {
+    fn from(region: Option<subtags::Region>) -> Self {
+        Self {
+            id: region.into(),
+            ..Default::default()
+        }
+    }
+}
+
+/// # Examples
+///
+/// ```
+/// use icu::locid::Locale;
+/// use icu::locid::{
+///     locale,
+///     subtags::{language, region, script},
+/// };
+///
+/// assert_eq!(
+///     Locale::from((
+///         language!("en"),
+///         Some(script!("Latn")),
+///         Some(region!("US"))
+///     )),
+///     locale!("en-Latn-US")
+/// );
+/// ```
+impl
+    From<(
+        subtags::Language,
+        Option<subtags::Script>,
+        Option<subtags::Region>,
+    )> for Locale
+{
+    fn from(
+        lsr: (
+            subtags::Language,
+            Option<subtags::Script>,
+            Option<subtags::Region>,
+        ),
+    ) -> Self {
+        Self {
+            id: lsr.into(),
+            ..Default::default()
+        }
+    }
+}
diff --git a/third_party/rust/icu_locid/src/macros.rs b/third_party/rust/icu_locid/src/macros.rs
new file mode 100644
index 0000000000..4537cd4031
--- /dev/null
+++ b/third_party/rust/icu_locid/src/macros.rs
@@ -0,0 +1,191 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+/// A macro allowing for compile-time construction of valid [`LanguageIdentifier`]s.
+///
+/// The macro will perform syntax canonicalization of the tag.
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::{langid, LanguageIdentifier};
+///
+/// const DE_AT: LanguageIdentifier = langid!("de_at");
+///
+/// let de_at: LanguageIdentifier = "de_at".parse().unwrap();
+///
+/// assert_eq!(DE_AT, de_at);
+/// ```
+///
+/// *Note*: The macro cannot produce language identifiers with more than one variants due to const
+/// limitations (see [`Heap Allocations in Constants`]):
+///
+/// ```compile_fail,E0080
+/// icu::locid::langid!("und-variant1-variant2");
+/// ```
+///
+/// Use runtime parsing instead:
+/// ```
+/// "und-variant1-variant2"
+///     .parse::<icu::locid::LanguageIdentifier>()
+///     .unwrap();
+/// ```
+///
+/// [`LanguageIdentifier`]: crate::LanguageIdentifier
+/// [`Heap Allocations in Constants`]: https://github.com/rust-lang/const-eval/issues/20
+#[macro_export]
+macro_rules! langid {
+    ($langid:literal) => {{
+        const R: $crate::LanguageIdentifier =
+            match $crate::LanguageIdentifier::try_from_bytes_with_single_variant($langid.as_bytes()) {
+                Ok((language, script, region, variant)) => $crate::LanguageIdentifier {
+                    language,
+                    script,
+                    region,
+                    variants: match variant {
+                        Some(v) => $crate::subtags::Variants::from_variant(v),
+                        None => $crate::subtags::Variants::new(),
+                    }
+                },
+                #[allow(clippy::panic)] // const context
+                _ => panic!(concat!("Invalid language code: ", $langid, " . Note langid! macro can only support up to a single variant tag. Use runtime parsing instead.")),
+            };
+        R
+    }};
+}
+
+/// A macro allowing for compile-time construction of valid [`Locale`]s.
+///
+/// The macro will perform syntax canonicalization of the tag.
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::{locale, Locale};
+///
+/// const DE_AT: Locale = locale!("de_at");
+///
+/// let de_at: Locale = "de_at".parse().unwrap();
+///
+/// assert_eq!(DE_AT, de_at);
+/// ```
+///
+/// *Note*: The macro cannot produce locales with more than one variant or multiple extensions
+/// (only single keyword unicode extension is supported) due to const
+/// limitations (see [`Heap Allocations in Constants`]):
+///
+/// ```compile_fail,E0080
+/// icu::locid::locale!("sl-IT-rozaj-biske-1994");
+/// ```
+/// Use runtime parsing instead:
+/// ```
+/// "sl-IT-rozaj-biske-1994"
+///     .parse::<icu::locid::Locale>()
+///     .unwrap();
+/// ```
+///
+/// Locales with multiple keys are not supported
+/// ```compile_fail,E0080
+/// icu::locid::locale!("th-TH-u-ca-buddhist-nu-thai");
+/// ```
+/// Use runtime parsing instead:
+/// ```
+/// "th-TH-u-ca-buddhist-nu-thai"
+///     .parse::<icu::locid::Locale>()
+///     .unwrap();
+/// ```
+///
+/// Locales with attributes are not supported
+/// ```compile_fail,E0080
+/// icu::locid::locale!("en-US-u-foobar-ca-buddhist");
+/// ```
+/// Use runtime parsing instead:
+/// ```
+/// "en-US-u-foobar-ca-buddhist"
+///     .parse::<icu::locid::Locale>()
+///     .unwrap();
+/// ```
+///
+/// Locales with single key but multiple types are not supported
+/// ```compile_fail,E0080
+/// icu::locid::locale!("en-US-u-ca-islamic-umalqura");
+/// ```
+/// Use runtime parsing instead:
+/// ```
+/// "en-US-u-ca-islamic-umalqura"
+///     .parse::<icu::locid::Locale>()
+///     .unwrap();
+/// ```
+/// [`Locale`]: crate::Locale
+/// [`Heap Allocations in Constants`]: https://github.com/rust-lang/const-eval/issues/20
+#[macro_export]
+macro_rules! locale {
+    ($locale:literal) => {{
+        const R: $crate::Locale =
+            match $crate::Locale::try_from_bytes_with_single_variant_single_keyword_unicode_extension(
+                $locale.as_bytes(),
+            ) {
+                Ok((language, script, region, variant, keyword)) => $crate::Locale {
+                    id: $crate::LanguageIdentifier {
+                        language,
+                        script,
+                        region,
+                        variants: match variant {
+                            Some(v) => $crate::subtags::Variants::from_variant(v),
+                            None => $crate::subtags::Variants::new(),
+                        },
+                    },
+                    extensions: match keyword {
+                        Some(k) => $crate::extensions::Extensions::from_unicode(
+                            $crate::extensions::unicode::Unicode {
+                                keywords: $crate::extensions::unicode::Keywords::new_single(
+                                    k.0,
+                                    $crate::extensions::unicode::Value::from_tinystr(k.1),
+                                ),
+
+                                attributes: $crate::extensions::unicode::Attributes::new(),
+                            },
+                        ),
+                        None => $crate::extensions::Extensions::new(),
+                    },
+                },
+                #[allow(clippy::panic)] // const context
+                _ => panic!(concat!(
+                    "Invalid language code: ",
+                    $locale,
+                    " . Note the locale! macro only supports up to one variant tag; \
+                                        unicode extensions are not supported. Use \
+                                        runtime parsing instead."
+                )),
+            };
+        R
+    }};
+}
+
+#[cfg(test)]
+mod test {
+    use crate::LanguageIdentifier;
+    use crate::Locale;
+
+    #[test]
+    fn test_langid_macro_can_parse_langid_with_single_variant() {
+        const DE_AT_FOOBAR: LanguageIdentifier = langid!("de_at-foobar");
+        let de_at_foobar: LanguageIdentifier = "de_at-foobar".parse().unwrap();
+        assert_eq!(DE_AT_FOOBAR, de_at_foobar);
+    }
+
+    #[test]
+    fn test_locale_macro_can_parse_locale_with_single_variant() {
+        const DE_AT_FOOBAR: Locale = locale!("de_at-foobar");
+        let de_at_foobar: Locale = "de_at-foobar".parse().unwrap();
+        assert_eq!(DE_AT_FOOBAR, de_at_foobar);
+    }
+
+    #[test]
+    fn test_locale_macro_can_parse_locale_with_single_keyword_unicode_extension() {
+        const DE_AT_U_CA_FOOBAR: Locale = locale!("de_at-u-ca-foobar");
+        let de_at_u_ca_foobar: Locale = "de_at-u-ca-foobar".parse().unwrap();
+        assert_eq!(DE_AT_U_CA_FOOBAR, de_at_u_ca_foobar);
+    }
+}
diff --git a/third_party/rust/icu_locid/src/ordering.rs b/third_party/rust/icu_locid/src/ordering.rs
new file mode 100644
index 0000000000..c877c60c39
--- /dev/null
+++ b/third_party/rust/icu_locid/src/ordering.rs
@@ -0,0 +1,62 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+//! Utilities for performing ordering operations on locales.
+
+use core::cmp::Ordering;
+
+/// The result of a subtag iterator comparison operation.
+///
+/// See [`Locale::strict_cmp_iter`].
+///
+/// # Examples
+///
+/// Check whether a stream of subtags contains two expected locales back-to-back:
+///
+/// ```
+/// use icu::locid::{locale, Locale, SubtagOrderingResult};
+/// use std::cmp::Ordering;
+///
+/// let subtags = b"en-US-it-IT".split(|b| *b == b'-');
+/// let locales = [locale!("en-US"), locale!("it-IT")];
+/// let mut result = SubtagOrderingResult::Subtags(subtags);
+/// for loc in locales.iter() {
+///     match result {
+///         SubtagOrderingResult::Subtags(it) => {
+///             result = loc.strict_cmp_iter(it);
+///         }
+///         SubtagOrderingResult::Ordering(ord) => break,
+///     }
+/// }
+///
+/// assert_eq!(Ordering::Equal, result.end());
+/// ```
+///
+/// [`Locale::strict_cmp_iter`]: crate::Locale::strict_cmp_iter
+#[allow(clippy::exhaustive_enums)] // well-defined exhaustive enum semantics
+#[derive(Debug)]
+pub enum SubtagOrderingResult<I> {
+    /// Potentially remaining subtags after the comparison operation.
+    Subtags(I),
+    /// Resolved ordering between the locale object and the subtags.
+    Ordering(Ordering),
+}
+
+impl<I> SubtagOrderingResult<I>
+where
+    I: Iterator,
+{
+    /// Invoke this function if there are no remaining locale objects to chain in order to get
+    /// a fully resolved [`Ordering`].
+    #[inline]
+    pub fn end(self) -> Ordering {
+        match self {
+            Self::Subtags(mut it) => match it.next() {
+                Some(_) => Ordering::Less,
+                None => Ordering::Equal,
+            },
+            Self::Ordering(o) => o,
+        }
+    }
+}
diff --git a/third_party/rust/icu_locid/src/parser/errors.rs b/third_party/rust/icu_locid/src/parser/errors.rs
new file mode 100644
index 0000000000..b2262460c1
--- /dev/null
+++ b/third_party/rust/icu_locid/src/parser/errors.rs
@@ -0,0 +1,72 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use displaydoc::Display;
+
+/// List of parser errors that can be generated
+/// while parsing [`LanguageIdentifier`](crate::LanguageIdentifier), [`Locale`](crate::Locale),
+/// [`subtags`](crate::subtags) or [`extensions`](crate::extensions).
+///
+/// Re-exported as [`Error`](crate::Error).
+#[derive(Display, Debug, PartialEq, Copy, Clone)]
+#[non_exhaustive]
+pub enum ParserError {
+    /// Invalid language subtag.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Language;
+    /// use icu::locid::ParserError;
+    ///
+    /// assert_eq!("x2".parse::<Language>(), Err(ParserError::InvalidLanguage));
+    /// ```
+    #[displaydoc("The given language subtag is invalid")]
+    InvalidLanguage,
+
+    /// Invalid script, region or variant subtag.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Region;
+    /// use icu::locid::ParserError;
+    ///
+    /// assert_eq!("#@2X".parse::<Region>(), Err(ParserError::InvalidSubtag));
+    /// ```
+    #[displaydoc("Invalid subtag")]
+    InvalidSubtag,
+
+    /// Invalid extension subtag.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::extensions::unicode::Key;
+    /// use icu::locid::ParserError;
+    ///
+    /// assert_eq!("#@2X".parse::<Key>(), Err(ParserError::InvalidExtension));
+    /// ```
+    #[displaydoc("Invalid extension")]
+    InvalidExtension,
+
+    /// Duplicated extension.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::Locale;
+    /// use icu::locid::ParserError;
+    ///
+    /// assert_eq!(
+    ///     "und-u-hc-h12-u-ca-calendar".parse::<Locale>(),
+    ///     Err(ParserError::DuplicatedExtension)
+    /// );
+    /// ```
+    #[displaydoc("Duplicated extension")]
+    DuplicatedExtension,
+}
+
+#[cfg(feature = "std")]
+impl std::error::Error for ParserError {}
diff --git a/third_party/rust/icu_locid/src/parser/langid.rs b/third_party/rust/icu_locid/src/parser/langid.rs
new file mode 100644
index 0000000000..2c6ddeb037
--- /dev/null
+++ b/third_party/rust/icu_locid/src/parser/langid.rs
@@ -0,0 +1,278 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+pub use super::errors::ParserError;
+use crate::extensions::unicode::{Attribute, Key, Value};
+use crate::extensions::ExtensionType;
+use crate::helpers::ShortSlice;
+use crate::parser::SubtagIterator;
+use crate::LanguageIdentifier;
+use crate::{extensions, subtags};
+use tinystr::TinyAsciiStr;
+
+#[derive(PartialEq, Clone, Copy)]
+pub enum ParserMode {
+    LanguageIdentifier,
+    Locale,
+    Partial,
+}
+
+#[derive(PartialEq, Clone, Copy)]
+enum ParserPosition {
+    Script,
+    Region,
+    Variant,
+}
+
+pub fn parse_language_identifier_from_iter(
+    iter: &mut SubtagIterator,
+    mode: ParserMode,
+) -> Result<LanguageIdentifier, ParserError> {
+    let mut script = None;
+    let mut region = None;
+    let mut variants = ShortSlice::new();
+
+    let language = if let Some(subtag) = iter.next() {
+        subtags::Language::try_from_bytes(subtag)?
+    } else {
+        return Err(ParserError::InvalidLanguage);
+    };
+
+    let mut position = ParserPosition::Script;
+
+    while let Some(subtag) = iter.peek() {
+        if mode != ParserMode::LanguageIdentifier && subtag.len() == 1 {
+            break;
+        }
+
+        if position == ParserPosition::Script {
+            if let Ok(s) = subtags::Script::try_from_bytes(subtag) {
+                script = Some(s);
+                position = ParserPosition::Region;
+            } else if let Ok(s) = subtags::Region::try_from_bytes(subtag) {
+                region = Some(s);
+                position = ParserPosition::Variant;
+            } else if let Ok(v) = subtags::Variant::try_from_bytes(subtag) {
+                if let Err(idx) = variants.binary_search(&v) {
+                    variants.insert(idx, v);
+                }
+                position = ParserPosition::Variant;
+            } else if mode == ParserMode::Partial {
+                break;
+            } else {
+                return Err(ParserError::InvalidSubtag);
+            }
+        } else if position == ParserPosition::Region {
+            if let Ok(s) = subtags::Region::try_from_bytes(subtag) {
+                region = Some(s);
+                position = ParserPosition::Variant;
+            } else if let Ok(v) = subtags::Variant::try_from_bytes(subtag) {
+                if let Err(idx) = variants.binary_search(&v) {
+                    variants.insert(idx, v);
+                }
+                position = ParserPosition::Variant;
+            } else if mode == ParserMode::Partial {
+                break;
+            } else {
+                return Err(ParserError::InvalidSubtag);
+            }
+        } else if let Ok(v) = subtags::Variant::try_from_bytes(subtag) {
+            if let Err(idx) = variants.binary_search(&v) {
+                variants.insert(idx, v);
+            } else {
+                return Err(ParserError::InvalidSubtag);
+            }
+        } else if mode == ParserMode::Partial {
+            break;
+        } else {
+            return Err(ParserError::InvalidSubtag);
+        }
+        iter.next();
+    }
+
+    Ok(LanguageIdentifier {
+        language,
+        script,
+        region,
+        variants: subtags::Variants::from_short_slice_unchecked(variants),
+    })
+}
+
+pub fn parse_language_identifier(
+    t: &[u8],
+    mode: ParserMode,
+) -> Result<LanguageIdentifier, ParserError> {
+    let mut iter = SubtagIterator::new(t);
+    parse_language_identifier_from_iter(&mut iter, mode)
+}
+
+#[allow(clippy::type_complexity)]
+pub const fn parse_locale_with_single_variant_single_keyword_unicode_extension_from_iter(
+    mut iter: SubtagIterator,
+    mode: ParserMode,
+) -> Result<
+    (
+        subtags::Language,
+        Option<subtags::Script>,
+        Option<subtags::Region>,
+        Option<subtags::Variant>,
+        Option<(extensions::unicode::Key, Option<TinyAsciiStr<8>>)>,
+    ),
+    ParserError,
+> {
+    let language;
+    let mut script = None;
+    let mut region = None;
+    let mut variant = None;
+    let mut keyword = None;
+
+    if let (i, Some((start, end))) = iter.next_manual() {
+        iter = i;
+        match subtags::Language::try_from_bytes_manual_slice(iter.slice, start, end) {
+            Ok(l) => language = l,
+            Err(e) => return Err(e),
+        }
+    } else {
+        return Err(ParserError::InvalidLanguage);
+    }
+
+    let mut position = ParserPosition::Script;
+
+    while let Some((start, end)) = iter.peek_manual() {
+        if !matches!(mode, ParserMode::LanguageIdentifier) && end - start == 1 {
+            break;
+        }
+
+        if matches!(position, ParserPosition::Script) {
+            if let Ok(s) = subtags::Script::try_from_bytes_manual_slice(iter.slice, start, end) {
+                script = Some(s);
+                position = ParserPosition::Region;
+            } else if let Ok(r) =
+                subtags::Region::try_from_bytes_manual_slice(iter.slice, start, end)
+            {
+                region = Some(r);
+                position = ParserPosition::Variant;
+            } else if let Ok(v) =
+                subtags::Variant::try_from_bytes_manual_slice(iter.slice, start, end)
+            {
+                // We cannot handle multiple variants in a const context
+                debug_assert!(variant.is_none());
+                variant = Some(v);
+                position = ParserPosition::Variant;
+            } else if matches!(mode, ParserMode::Partial) {
+                break;
+            } else {
+                return Err(ParserError::InvalidSubtag);
+            }
+        } else if matches!(position, ParserPosition::Region) {
+            if let Ok(s) = subtags::Region::try_from_bytes_manual_slice(iter.slice, start, end) {
+                region = Some(s);
+                position = ParserPosition::Variant;
+            } else if let Ok(v) =
+                subtags::Variant::try_from_bytes_manual_slice(iter.slice, start, end)
+            {
+                // We cannot handle multiple variants in a const context
+                debug_assert!(variant.is_none());
+                variant = Some(v);
+                position = ParserPosition::Variant;
+            } else if matches!(mode, ParserMode::Partial) {
+                break;
+            } else {
+                return Err(ParserError::InvalidSubtag);
+            }
+        } else if let Ok(v) = subtags::Variant::try_from_bytes_manual_slice(iter.slice, start, end)
+        {
+            debug_assert!(matches!(position, ParserPosition::Variant));
+            if variant.is_some() {
+                // We cannot handle multiple variants in a const context
+                return Err(ParserError::InvalidSubtag);
+            }
+            variant = Some(v);
+        } else if matches!(mode, ParserMode::Partial) {
+            break;
+        } else {
+            return Err(ParserError::InvalidSubtag);
+        }
+
+        iter = iter.next_manual().0;
+    }
+
+    if matches!(mode, ParserMode::Locale) {
+        if let Some((start, end)) = iter.peek_manual() {
+            match ExtensionType::try_from_bytes_manual_slice(iter.slice, start, end) {
+                Ok(ExtensionType::Unicode) => {
+                    iter = iter.next_manual().0;
+                    if let Some((start, end)) = iter.peek_manual() {
+                        if Attribute::try_from_bytes_manual_slice(iter.slice, start, end).is_ok() {
+                            // We cannot handle Attributes in a const context
+                            return Err(ParserError::InvalidSubtag);
+                        }
+                    }
+
+                    let mut key = None;
+                    let mut current_type = None;
+
+                    while let Some((start, end)) = iter.peek_manual() {
+                        let slen = end - start;
+                        if slen == 2 {
+                            if key.is_some() {
+                                // We cannot handle more than one Key in a const context
+                                return Err(ParserError::InvalidSubtag);
+                            }
+                            match Key::try_from_bytes_manual_slice(iter.slice, start, end) {
+                                Ok(k) => key = Some(k),
+                                Err(e) => return Err(e),
+                            };
+                        } else if key.is_some() {
+                            match Value::parse_subtag_from_bytes_manual_slice(
+                                iter.slice, start, end,
+                            ) {
+                                Ok(Some(t)) => {
+                                    if current_type.is_some() {
+                                        // We cannot handle more than one type in a const context
+                                        return Err(ParserError::InvalidSubtag);
+                                    }
+                                    current_type = Some(t);
+                                }
+                                Ok(None) => {}
+                                Err(e) => return Err(e),
+                            }
+                        } else {
+                            break;
+                        }
+                        iter = iter.next_manual().0
+                    }
+                    if let Some(k) = key {
+                        keyword = Some((k, current_type));
+                    }
+                }
+                // We cannot handle Transform, Private, Other extensions in a const context
+                Ok(_) => return Err(ParserError::InvalidSubtag),
+                Err(e) => return Err(e),
+            }
+        }
+    }
+
+    Ok((language, script, region, variant, keyword))
+}
+
+#[allow(clippy::type_complexity)]
+pub const fn parse_language_identifier_with_single_variant(
+    t: &[u8],
+    mode: ParserMode,
+) -> Result<
+    (
+        subtags::Language,
+        Option<subtags::Script>,
+        Option<subtags::Region>,
+        Option<subtags::Variant>,
+    ),
+    ParserError,
+> {
+    let iter = SubtagIterator::new(t);
+    match parse_locale_with_single_variant_single_keyword_unicode_extension_from_iter(iter, mode) {
+        Ok((l, s, r, v, _)) => Ok((l, s, r, v)),
+        Err(e) => Err(e),
+    }
+}
diff --git a/third_party/rust/icu_locid/src/parser/locale.rs b/third_party/rust/icu_locid/src/parser/locale.rs
new file mode 100644
index 0000000000..175fd3a05b
--- /dev/null
+++ b/third_party/rust/icu_locid/src/parser/locale.rs
@@ -0,0 +1,42 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use tinystr::TinyAsciiStr;
+
+use crate::extensions::{self, Extensions};
+use crate::parser::errors::ParserError;
+use crate::parser::{parse_language_identifier_from_iter, ParserMode, SubtagIterator};
+use crate::{subtags, Locale};
+
+use super::parse_locale_with_single_variant_single_keyword_unicode_extension_from_iter;
+
+pub fn parse_locale(t: &[u8]) -> Result<Locale, ParserError> {
+    let mut iter = SubtagIterator::new(t);
+
+    let id = parse_language_identifier_from_iter(&mut iter, ParserMode::Locale)?;
+    let extensions = if iter.peek().is_some() {
+        Extensions::try_from_iter(&mut iter)?
+    } else {
+        Extensions::default()
+    };
+    Ok(Locale { id, extensions })
+}
+
+#[allow(clippy::type_complexity)]
+pub const fn parse_locale_with_single_variant_single_keyword_unicode_keyword_extension(
+    t: &[u8],
+    mode: ParserMode,
+) -> Result<
+    (
+        subtags::Language,
+        Option<subtags::Script>,
+        Option<subtags::Region>,
+        Option<subtags::Variant>,
+        Option<(extensions::unicode::Key, Option<TinyAsciiStr<8>>)>,
+    ),
+    ParserError,
+> {
+    let iter = SubtagIterator::new(t);
+    parse_locale_with_single_variant_single_keyword_unicode_extension_from_iter(iter, mode)
+}
diff --git a/third_party/rust/icu_locid/src/parser/mod.rs b/third_party/rust/icu_locid/src/parser/mod.rs
new file mode 100644
index 0000000000..4b02f71c9a
--- /dev/null
+++ b/third_party/rust/icu_locid/src/parser/mod.rs
@@ -0,0 +1,231 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+pub mod errors;
+mod langid;
+mod locale;
+
+pub use errors::ParserError;
+pub use langid::{
+    parse_language_identifier, parse_language_identifier_from_iter,
+    parse_language_identifier_with_single_variant,
+    parse_locale_with_single_variant_single_keyword_unicode_extension_from_iter, ParserMode,
+};
+
+pub use locale::{
+    parse_locale, parse_locale_with_single_variant_single_keyword_unicode_keyword_extension,
+};
+
+#[inline]
+const fn is_separator(slice: &[u8], idx: usize) -> bool {
+    #[allow(clippy::indexing_slicing)]
+    let b = slice[idx];
+    b == b'-' || b == b'_'
+}
+
+const fn get_current_subtag(slice: &[u8], idx: usize) -> (usize, usize) {
+    debug_assert!(idx < slice.len());
+
+    // This function is called only on the idx == 0 or on a separator.
+    let (start, mut end) = if is_separator(slice, idx) {
+        // If it's a separator, set the start to idx+1 and advance the idx to the next char.
+        (idx + 1, idx + 1)
+    } else {
+        // If it's idx=0, start is 0 and end is set to 1
+        debug_assert!(idx == 0);
+        (0, 1)
+    };
+
+    while end < slice.len() && !is_separator(slice, end) {
+        // Advance until we reach end of slice or a separator.
+        end += 1;
+    }
+    // Notice: this slice may be empty (start == end) for cases like `"en-"` or `"en--US"`
+    (start, end)
+}
+
+// `SubtagIterator` is a helper iterator for [`LanguageIdentifier`] and [`Locale`] parsing.
+//
+// It is quite extraordinary due to focus on performance and Rust limitations for `const`
+// functions.
+//
+// The iterator is eager and fallible allowing it to reject invalid slices such as `"-"`, `"-en"`,
+// `"en-"` etc.
+//
+// The iterator provides methods available for static users - `next_manual` and `peek_manual`,
+// as well as typical `Peekable` iterator APIs - `next` and `peek`.
+//
+// All methods return an `Option` of a `Result`.
+#[derive(Copy, Clone, Debug)]
+pub struct SubtagIterator<'a> {
+    pub slice: &'a [u8],
+    done: bool,
+    // done + subtag is faster than Option<(usize, usize)>
+    // at the time of writing.
+    subtag: (usize, usize),
+}
+
+impl<'a> SubtagIterator<'a> {
+    pub const fn new(slice: &'a [u8]) -> Self {
+        let subtag = if slice.is_empty() || is_separator(slice, 0) {
+            // This returns (0, 0) which returns Some(b"") for slices like `"-en"` or `"-"`
+            (0, 0)
+        } else {
+            get_current_subtag(slice, 0)
+        };
+        Self {
+            slice,
+            done: false,
+            subtag,
+        }
+    }
+
+    pub const fn next_manual(mut self) -> (Self, Option<(usize, usize)>) {
+        if self.done {
+            return (self, None);
+        }
+        let result = self.subtag;
+        if result.1 < self.slice.len() {
+            self.subtag = get_current_subtag(self.slice, result.1);
+        } else {
+            self.done = true;
+        }
+        (self, Some(result))
+    }
+
+    pub const fn peek_manual(&self) -> Option<(usize, usize)> {
+        if self.done {
+            return None;
+        }
+        Some(self.subtag)
+    }
+
+    pub fn peek(&self) -> Option<&'a [u8]> {
+        #[allow(clippy::indexing_slicing)] // peek_manual returns valid indices
+        self.peek_manual().map(|(s, e)| &self.slice[s..e])
+    }
+}
+
+impl<'a> Iterator for SubtagIterator<'a> {
+    type Item = &'a [u8];
+
+    fn next(&mut self) -> Option<Self::Item> {
+        let (s, res) = self.next_manual();
+        *self = s;
+        #[allow(clippy::indexing_slicing)] // next_manual returns valid indices
+        res.map(|(s, e)| &self.slice[s..e])
+    }
+}
+
+#[cfg(test)]
+mod test {
+    use super::*;
+
+    fn slice_to_str(input: &[u8]) -> &str {
+        std::str::from_utf8(input).unwrap()
+    }
+
+    #[test]
+    fn subtag_iterator_peek_test() {
+        let slice = "de_at-u-ca-foobar";
+        let mut si = SubtagIterator::new(slice.as_bytes());
+
+        assert_eq!(si.peek().map(slice_to_str), Some("de"));
+        assert_eq!(si.peek().map(slice_to_str), Some("de"));
+        assert_eq!(si.next().map(slice_to_str), Some("de"));
+
+        assert_eq!(si.peek().map(slice_to_str), Some("at"));
+        assert_eq!(si.peek().map(slice_to_str), Some("at"));
+        assert_eq!(si.next().map(slice_to_str), Some("at"));
+    }
+
+    #[test]
+    fn subtag_iterator_test() {
+        let slice = "";
+        let mut si = SubtagIterator::new(slice.as_bytes());
+        assert_eq!(si.next().map(slice_to_str), Some(""));
+
+        let slice = "-";
+        let mut si = SubtagIterator::new(slice.as_bytes());
+        assert_eq!(si.next().map(slice_to_str), Some(""));
+
+        let slice = "-en";
+        let mut si = SubtagIterator::new(slice.as_bytes());
+        assert_eq!(si.next().map(slice_to_str), Some(""));
+        assert_eq!(si.next().map(slice_to_str), Some("en"));
+        assert_eq!(si.next(), None);
+
+        let slice = "en";
+        let si = SubtagIterator::new(slice.as_bytes());
+        assert_eq!(si.map(slice_to_str).collect::<Vec<_>>(), vec!["en",]);
+
+        let slice = "en-";
+        let si = SubtagIterator::new(slice.as_bytes());
+        assert_eq!(si.map(slice_to_str).collect::<Vec<_>>(), vec!["en", "",]);
+
+        let slice = "--";
+        let mut si = SubtagIterator::new(slice.as_bytes());
+        assert_eq!(si.next().map(slice_to_str), Some(""));
+        assert_eq!(si.next().map(slice_to_str), Some(""));
+        assert_eq!(si.next().map(slice_to_str), Some(""));
+        assert_eq!(si.next(), None);
+
+        let slice = "-en-";
+        let mut si = SubtagIterator::new(slice.as_bytes());
+        assert_eq!(si.next().map(slice_to_str), Some(""));
+        assert_eq!(si.next().map(slice_to_str), Some("en"));
+        assert_eq!(si.next().map(slice_to_str), Some(""));
+        assert_eq!(si.next(), None);
+
+        let slice = "de_at-u-ca-foobar";
+        let si = SubtagIterator::new(slice.as_bytes());
+        assert_eq!(
+            si.map(slice_to_str).collect::<Vec<_>>(),
+            vec!["de", "at", "u", "ca", "foobar",]
+        );
+    }
+
+    #[test]
+    fn get_current_subtag_test() {
+        let slice = "-";
+        let current = get_current_subtag(slice.as_bytes(), 0);
+        assert_eq!(current, (1, 1));
+
+        let slice = "-en";
+        let current = get_current_subtag(slice.as_bytes(), 0);
+        assert_eq!(current, (1, 3));
+
+        let slice = "-en-";
+        let current = get_current_subtag(slice.as_bytes(), 3);
+        assert_eq!(current, (4, 4));
+
+        let slice = "en-";
+        let current = get_current_subtag(slice.as_bytes(), 0);
+        assert_eq!(current, (0, 2));
+
+        let current = get_current_subtag(slice.as_bytes(), 2);
+        assert_eq!(current, (3, 3));
+
+        let slice = "en--US";
+        let current = get_current_subtag(slice.as_bytes(), 0);
+        assert_eq!(current, (0, 2));
+
+        let current = get_current_subtag(slice.as_bytes(), 2);
+        assert_eq!(current, (3, 3));
+
+        let current = get_current_subtag(slice.as_bytes(), 3);
+        assert_eq!(current, (4, 6));
+
+        let slice = "--";
+        let current = get_current_subtag(slice.as_bytes(), 0);
+        assert_eq!(current, (1, 1));
+
+        let current = get_current_subtag(slice.as_bytes(), 1);
+        assert_eq!(current, (2, 2));
+
+        let slice = "-";
+        let current = get_current_subtag(slice.as_bytes(), 0);
+        assert_eq!(current, (1, 1));
+    }
+}
diff --git a/third_party/rust/icu_locid/src/serde.rs b/third_party/rust/icu_locid/src/serde.rs
new file mode 100644
index 0000000000..3bfe303887
--- /dev/null
+++ b/third_party/rust/icu_locid/src/serde.rs
@@ -0,0 +1,135 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use crate::LanguageIdentifier;
+use alloc::string::ToString;
+use serde::{Deserialize, Deserializer, Serialize, Serializer};
+
+impl Serialize for LanguageIdentifier {
+    fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: Serializer,
+    {
+        serializer.serialize_str(&self.to_string())
+    }
+}
+
+impl<'de> Deserialize<'de> for LanguageIdentifier {
+    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        struct LanguageIdentifierVisitor;
+
+        impl<'de> serde::de::Visitor<'de> for LanguageIdentifierVisitor {
+            type Value = LanguageIdentifier;
+
+            fn expecting(&self, formatter: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
+                write!(formatter, "a valid Unicode Language Identifier")
+            }
+
+            fn visit_str<E>(self, s: &str) -> Result<Self::Value, E>
+            where
+                E: serde::de::Error,
+            {
+                s.parse::<LanguageIdentifier>()
+                    .map_err(serde::de::Error::custom)
+            }
+        }
+
+        deserializer.deserialize_string(LanguageIdentifierVisitor)
+    }
+}
+
+#[test]
+fn json() {
+    use crate::langid;
+    use crate::subtags::{Language, Region, Script};
+
+    assert_eq!(
+        serde_json::to_string(&langid!("en-US")).unwrap(),
+        r#""en-US""#
+    );
+    assert_eq!(
+        serde_json::from_str::<LanguageIdentifier>(r#""en-US""#).unwrap(),
+        langid!("en-US")
+    );
+    assert!(serde_json::from_str::<LanguageIdentifier>(r#""2Xs""#).is_err());
+
+    assert_eq!(
+        serde_json::to_string(&"fr".parse::<Language>().unwrap()).unwrap(),
+        r#""fr""#
+    );
+    assert_eq!(
+        serde_json::from_str::<Language>(r#""fr""#).unwrap(),
+        "fr".parse::<Language>().unwrap()
+    );
+    assert!(serde_json::from_str::<Language>(r#""2Xs""#).is_err());
+
+    assert_eq!(
+        serde_json::to_string(&"Latn".parse::<Script>().unwrap()).unwrap(),
+        r#""Latn""#
+    );
+    assert_eq!(
+        serde_json::from_str::<Script>(r#""Latn""#).unwrap(),
+        "Latn".parse::<Script>().unwrap()
+    );
+    assert!(serde_json::from_str::<Script>(r#""2Xs""#).is_err());
+
+    assert_eq!(
+        serde_json::to_string(&"US".parse::<Region>().unwrap()).unwrap(),
+        r#""US""#
+    );
+    assert_eq!(
+        serde_json::from_str::<Region>(r#""US""#).unwrap(),
+        "US".parse::<Region>().unwrap()
+    );
+    assert!(serde_json::from_str::<Region>(r#""2Xs""#).is_err());
+}
+
+#[test]
+fn postcard() {
+    use crate::langid;
+    use crate::subtags::{Language, Region, Script};
+
+    assert_eq!(
+        postcard::to_stdvec(&langid!("en-US")).unwrap(),
+        &[5, b'e', b'n', b'-', b'U', b'S']
+    );
+    assert_eq!(
+        postcard::from_bytes::<LanguageIdentifier>(&[5, b'e', b'n', b'-', b'U', b'S']).unwrap(),
+        langid!("en-US")
+    );
+    assert!(postcard::from_bytes::<LanguageIdentifier>(&[3, b'2', b'X', b's']).is_err());
+
+    assert_eq!(
+        postcard::to_stdvec(&"fr".parse::<Language>().unwrap()).unwrap(),
+        b"fr\0"
+    );
+    assert_eq!(
+        postcard::from_bytes::<Language>(b"fr\0").unwrap(),
+        "fr".parse::<Language>().unwrap()
+    );
+    assert!(postcard::from_bytes::<Language>(b"2Xs").is_err());
+
+    assert_eq!(
+        postcard::to_stdvec(&"Latn".parse::<Script>().unwrap()).unwrap(),
+        b"Latn"
+    );
+    assert_eq!(
+        postcard::from_bytes::<Script>(b"Latn").unwrap(),
+        "Latn".parse::<Script>().unwrap()
+    );
+    assert!(postcard::from_bytes::<Script>(b"2Xss").is_err());
+
+    assert_eq!(
+        postcard::to_stdvec(&"US".parse::<Region>().unwrap()).unwrap(),
+        b"US\0"
+    );
+    assert_eq!(
+        postcard::from_bytes::<Region>(b"US\0").unwrap(),
+        "US".parse::<Region>().unwrap()
+    );
+    assert!(postcard::from_bytes::<Region>(b"2Xs").is_err());
+}
diff --git a/third_party/rust/icu_locid/src/subtags/language.rs b/third_party/rust/icu_locid/src/subtags/language.rs
new file mode 100644
index 0000000000..6fd08a2d5f
--- /dev/null
+++ b/third_party/rust/icu_locid/src/subtags/language.rs
@@ -0,0 +1,107 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+impl_tinystr_subtag!(
+    /// A language subtag (examples: `"en"`, `"csb"`, `"zh"`, `"und"`, etc.)
+    ///
+    /// [`Language`] represents a Unicode base language code conformant to the
+    /// [`unicode_language_id`] field of the Language and Locale Identifier.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Language;
+    ///
+    /// let language: Language =
+    ///     "en".parse().expect("Failed to parse a language subtag.");
+    /// ```
+    ///
+    /// If the [`Language`] has no value assigned, it serializes to a string `"und"`, which
+    /// can be then parsed back to an empty [`Language`] field.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Language;
+    ///
+    /// assert_eq!(Language::default().as_str(), "und");
+    /// ```
+    ///
+    /// `Notice`: ICU4X uses a narrow form of language subtag of 2-3 characters.
+    /// The specification allows language subtag to optionally also be 5-8 characters
+    /// but that form has not been used and ICU4X does not support it right now.
+    ///
+    /// [`unicode_language_id`]: https://unicode.org/reports/tr35/#unicode_language_id
+    Language,
+    subtags,
+    language,
+    subtags_language,
+    2..=3,
+    s,
+    s.is_ascii_alphabetic(),
+    s.to_ascii_lowercase(),
+    s.is_ascii_alphabetic_lowercase(),
+    InvalidLanguage,
+    ["en", "foo"],
+    ["419", "german", "en1"],
+);
+
+impl Language {
+    /// The default undefined language "und". Same as [`default()`](Default::default()).
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Language;
+    ///
+    /// assert_eq!(Language::default(), Language::UND);
+    /// ```
+    pub const UND: Self = unsafe { Self::from_raw_unchecked(*b"und") };
+
+    /// Resets the [`Language`] subtag to an empty one (equal to `"und"`).
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::{language, Language};
+    ///
+    /// let mut lang = language!("csb");
+    ///
+    /// assert_ne!(lang, Language::UND);
+    ///
+    /// lang.clear();
+    ///
+    /// assert_eq!(lang, Language::UND);
+    /// ```
+    #[inline]
+    pub fn clear(&mut self) {
+        *self = Self::UND
+    }
+
+    /// Tests if the [`Language`] subtag is empty (equal to `"und"`).
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Language;
+    ///
+    /// let mut lang = Language::UND;
+    ///
+    /// assert!(lang.is_empty());
+    ///
+    /// lang.clear();
+    ///
+    /// assert!(lang.is_empty());
+    /// ```
+    #[inline]
+    pub fn is_empty(self) -> bool {
+        self == Self::UND
+    }
+}
+
+impl Default for Language {
+    fn default() -> Language {
+        Language::UND
+    }
+}
diff --git a/third_party/rust/icu_locid/src/subtags/mod.rs b/third_party/rust/icu_locid/src/subtags/mod.rs
new file mode 100644
index 0000000000..9cc04dac8c
--- /dev/null
+++ b/third_party/rust/icu_locid/src/subtags/mod.rs
@@ -0,0 +1,62 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+//! Language Identifier and Locale contains a set of subtags
+//! which represent different fields of the structure.
+//!
+//! * [`Language`] is the only mandatory field, which when empty,
+//! takes the value `und`.
+//! * [`Script`] is an optional field representing the written script used by the locale.
+//! * [`Region`] is the region used by the locale.
+//! * [`Variants`] is a list of optional [`Variant`] subtags containing information about the
+//!                variant adjustments used by the locale.
+//!
+//! Subtags can be used in isolation, and all basic operations such as parsing, syntax canonicalization
+//! and serialization are supported on each individual subtag, but most commonly
+//! they are used to construct a [`LanguageIdentifier`] instance.
+//!
+//! [`Variants`] is a special structure which contains a list of [`Variant`] subtags.
+//! It is wrapped around to allow for sorting and deduplication of variants, which
+//! is one of the required steps of language identifier and locale syntax canonicalization.
+//!
+//! # Examples
+//!
+//! ```
+//! use icu::locid::subtags::{Language, Region, Script, Variant};
+//!
+//! let language: Language =
+//!     "en".parse().expect("Failed to parse a language subtag.");
+//! let script: Script =
+//!     "arab".parse().expect("Failed to parse a script subtag.");
+//! let region: Region =
+//!     "cn".parse().expect("Failed to parse a region subtag.");
+//! let variant: Variant =
+//!     "MacOS".parse().expect("Failed to parse a variant subtag.");
+//!
+//! assert_eq!(language.as_str(), "en");
+//! assert_eq!(script.as_str(), "Arab");
+//! assert_eq!(region.as_str(), "CN");
+//! assert_eq!(variant.as_str(), "macos");
+//! ```
+//!
+//! `Notice`: The subtags are canonicalized on parsing. That means
+//! that all operations work on a canonicalized version of the subtag
+//! and serialization is very cheap.
+//!
+//! [`LanguageIdentifier`]: super::LanguageIdentifier
+mod language;
+mod region;
+mod script;
+mod variant;
+mod variants;
+
+#[doc(inline)]
+pub use language::{language, Language};
+#[doc(inline)]
+pub use region::{region, Region};
+#[doc(inline)]
+pub use script::{script, Script};
+#[doc(inline)]
+pub use variant::{variant, Variant};
+pub use variants::Variants;
diff --git a/third_party/rust/icu_locid/src/subtags/region.rs b/third_party/rust/icu_locid/src/subtags/region.rs
new file mode 100644
index 0000000000..4348f15e79
--- /dev/null
+++ b/third_party/rust/icu_locid/src/subtags/region.rs
@@ -0,0 +1,62 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+impl_tinystr_subtag!(
+    /// A region subtag (examples: `"US"`, `"CN"`, `"AR"` etc.)
+    ///
+    /// [`Region`] represents a Unicode base language code conformant to the
+    /// [`unicode_region_id`] field of the Language and Locale Identifier.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Region;
+    ///
+    /// let region: Region =
+    ///     "DE".parse().expect("Failed to parse a region subtag.");
+    /// ```
+    ///
+    /// [`unicode_region_id`]: https://unicode.org/reports/tr35/#unicode_region_id
+    Region,
+    subtags,
+    region,
+    subtags_region,
+    2..=3,
+    s,
+    if s.len() == 2 {
+        s.is_ascii_alphabetic()
+    } else {
+        s.is_ascii_numeric()
+    },
+    if s.len() == 2 {
+        s.to_ascii_uppercase()
+    } else {
+        s
+    },
+    if s.len() == 2 {
+        s.is_ascii_alphabetic_uppercase()
+    } else {
+        s.is_ascii_numeric()
+    },
+    InvalidSubtag,
+    ["FR", "123"],
+    ["12", "FRA", "b2"],
+);
+
+impl Region {
+    /// Returns true if the Region has an alphabetic code.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Region;
+    ///
+    /// let region = Region::try_from_bytes(b"us").expect("Parsing failed.");
+    ///
+    /// assert!(region.is_alphabetic());
+    /// ```
+    pub fn is_alphabetic(&self) -> bool {
+        self.0.len() == 2
+    }
+}
diff --git a/third_party/rust/icu_locid/src/subtags/script.rs b/third_party/rust/icu_locid/src/subtags/script.rs
new file mode 100644
index 0000000000..79ead0390c
--- /dev/null
+++ b/third_party/rust/icu_locid/src/subtags/script.rs
@@ -0,0 +1,33 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+impl_tinystr_subtag!(
+    /// A script subtag (examples: `"Latn"`, `"Arab"`, etc.)
+    ///
+    /// [`Script`] represents a Unicode base language code conformant to the
+    /// [`unicode_script_id`] field of the Language and Locale Identifier.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Script;
+    ///
+    /// let script: Script =
+    ///     "Latn".parse().expect("Failed to parse a script subtag.");
+    /// ```
+    ///
+    /// [`unicode_script_id`]: https://unicode.org/reports/tr35/#unicode_script_id
+    Script,
+    subtags,
+    script,
+    subtags_script,
+    4..=4,
+    s,
+    s.is_ascii_alphabetic(),
+    s.to_ascii_titlecase(),
+    s.is_ascii_alphabetic_titlecase(),
+    InvalidSubtag,
+    ["Latn"],
+    ["Latin"],
+);
diff --git a/third_party/rust/icu_locid/src/subtags/variant.rs b/third_party/rust/icu_locid/src/subtags/variant.rs
new file mode 100644
index 0000000000..c60b138659
--- /dev/null
+++ b/third_party/rust/icu_locid/src/subtags/variant.rs
@@ -0,0 +1,35 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+impl_tinystr_subtag!(
+    /// A variant subtag (examples: `"macos"`, `"posix"`, `"1996"` etc.)
+    ///
+    /// [`Variant`] represents a Unicode base language code conformant to the
+    /// [`unicode_variant_id`] field of the Language and Locale Identifier.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Variant;
+    ///
+    /// let variant: Variant =
+    ///     "macos".parse().expect("Failed to parse a variant subtag.");
+    /// ```
+    ///
+    /// [`unicode_variant_id`]: https://unicode.org/reports/tr35/#unicode_variant_id
+    Variant,
+    subtags,
+    variant,
+    subtags_variant,
+    4..=8,
+    s,
+    s.is_ascii_alphanumeric() && (s.len() != 4 || s.all_bytes()[0].is_ascii_digit()),
+    s.to_ascii_lowercase(),
+    s.is_ascii_lowercase()
+        && s.is_ascii_alphanumeric()
+        && (s.len() != 4 || s.all_bytes()[0].is_ascii_digit()),
+    InvalidSubtag,
+    ["posix", "1996"],
+    ["yes"],
+);
diff --git a/third_party/rust/icu_locid/src/subtags/variants.rs b/third_party/rust/icu_locid/src/subtags/variants.rs
new file mode 100644
index 0000000000..ba5ff1bc1a
--- /dev/null
+++ b/third_party/rust/icu_locid/src/subtags/variants.rs
@@ -0,0 +1,128 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+use super::Variant;
+use crate::helpers::ShortSlice;
+
+use alloc::vec::Vec;
+use core::ops::Deref;
+
+/// A list of variants (examples: `["macos", "posix"]`, etc.)
+///
+/// [`Variants`] stores a list of [`Variant`] subtags in a canonical form
+/// by sorting and deduplicating them.
+///
+/// # Examples
+///
+/// ```
+/// use icu::locid::subtags::{variant, Variants};
+///
+/// let mut v = vec![variant!("posix"), variant!("macos")];
+/// v.sort();
+/// v.dedup();
+///
+/// let variants: Variants = Variants::from_vec_unchecked(v);
+/// assert_eq!(variants.to_string(), "macos-posix");
+/// ```
+#[derive(Default, Debug, PartialEq, Eq, Clone, Hash, PartialOrd, Ord)]
+pub struct Variants(ShortSlice<Variant>);
+
+impl Variants {
+    /// Returns a new empty list of variants. Same as [`default()`](Default::default()), but is `const`.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::Variants;
+    ///
+    /// assert_eq!(Variants::new(), Variants::default());
+    /// ```
+    #[inline]
+    pub const fn new() -> Self {
+        Self(ShortSlice::new())
+    }
+
+    /// Creates a new [`Variants`] set from a single [`Variant`].
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::{variant, Variants};
+    ///
+    /// let variants = Variants::from_variant(variant!("posix"));
+    /// ```
+    #[inline]
+    pub const fn from_variant(variant: Variant) -> Self {
+        Self(ShortSlice::new_single(variant))
+    }
+
+    /// Creates a new [`Variants`] set from a [`Vec`].
+    /// The caller is expected to provide sorted and deduplicated vector as
+    /// an input.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::{variant, Variants};
+    ///
+    /// let mut v = vec![variant!("posix"), variant!("macos")];
+    /// v.sort();
+    /// v.dedup();
+    ///
+    /// let variants = Variants::from_vec_unchecked(v);
+    /// ```
+    ///
+    /// Notice: For performance- and memory-constrained environments, it is recommended
+    /// for the caller to use [`binary_search`](slice::binary_search) instead of [`sort`](slice::sort)
+    /// and [`dedup`](Vec::dedup()).
+    pub fn from_vec_unchecked(input: Vec<Variant>) -> Self {
+        Self(input.into())
+    }
+
+    pub(crate) fn from_short_slice_unchecked(input: ShortSlice<Variant>) -> Self {
+        Self(input)
+    }
+
+    /// Empties the [`Variants`] list.
+    ///
+    /// Returns the old list.
+    ///
+    /// # Examples
+    ///
+    /// ```
+    /// use icu::locid::subtags::{variant, Variants};
+    ///
+    /// let mut v = vec![variant!("posix"), variant!("macos")];
+    /// v.sort();
+    /// v.dedup();
+    ///
+    /// let mut variants: Variants = Variants::from_vec_unchecked(v);
+    ///
+    /// assert_eq!(variants.to_string(), "macos-posix");
+    ///
+    /// variants.clear();
+    ///
+    /// assert_eq!(variants, Variants::default());
+    /// ```
+    pub fn clear(&mut self) -> Self {
+        core::mem::take(self)
+    }
+
+    pub(crate) fn for_each_subtag_str<E, F>(&self, f: &mut F) -> Result<(), E>
+    where
+        F: FnMut(&str) -> Result<(), E>,
+    {
+        self.deref().iter().map(|t| t.as_str()).try_for_each(f)
+    }
+}
+
+impl_writeable_for_subtag_list!(Variants, "macos", "posix");
+
+impl Deref for Variants {
+    type Target = [Variant];
+
+    fn deref(&self) -> &[Variant] {
+        self.0.deref()
+    }
+}
diff --git a/third_party/rust/icu_locid/src/zerovec.rs b/third_party/rust/icu_locid/src/zerovec.rs
new file mode 100644
index 0000000000..ba6a3e85d6
--- /dev/null
+++ b/third_party/rust/icu_locid/src/zerovec.rs
@@ -0,0 +1,132 @@
+// This file is part of ICU4X. For terms of use, please see the file
+// called LICENSE at the top level of the ICU4X source tree
+// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).
+
+//! Documentation on zero-copy deserialization of locale types.
+//!
+//! [`Locale`] and [`LanguageIdentifier`] are highly structured types that cannot be directly
+//! stored in a zero-copy data structure, such as those provided by the [`zerovec`] crate.
+//! This page explains how to indirectly store these types in a [`zerovec`].
+//!
+//! There are two main use cases, which have different solutions:
+//!
+//! 1. **Lookup:** You need to locate a locale in a zero-copy vector, such as when querying a map.
+//! 2. **Obtain:** You have a locale stored in a zero-copy vector, and you need to obtain a proper
+//!    [`Locale`] or [`LanguageIdentifier`] for use elsewhere in your program.
+//!
+//! # Lookup
+//!
+//! To perform lookup, store the stringified locale in a canonical BCP-47 form as a byte array,
+//! and then use [`Locale::strict_cmp()`] to perform an efficient, zero-allocation lookup.
+//!
+//! To produce more human-readable serialized output, you can use [`UnvalidatedStr`].
+//!
+//! ```
+//! use icu_locid::Locale;
+//! use zerovec::ule::UnvalidatedStr;
+//! use zerovec::ZeroMap;
+//!
+//! // ZeroMap from locales to integers
+//! let data: &[(&UnvalidatedStr, u32)] = &[
+//!     ("de-DE-u-hc-h12".into(), 5),
+//!     ("en-US-u-ca-buddhist".into(), 10),
+//!     ("my-MM".into(), 15),
+//!     ("sr-Cyrl-ME".into(), 20),
+//!     ("zh-TW".into(), 25),
+//! ];
+//! let zm: ZeroMap<UnvalidatedStr, u32> = data.iter().copied().collect();
+//!
+//! // Get the value associated with a locale
+//! let loc: Locale = "en-US-u-ca-buddhist".parse().unwrap();
+//! let value = zm.get_copied_by(|uvstr| loc.strict_cmp(uvstr).reverse());
+//! assert_eq!(value, Some(10));
+//! ```
+//!
+//! # Obtain
+//!
+//! Obtaining a [`Locale`] or [`LanguageIdentifier`] is not generally a zero-copy operation, since
+//! both of these types may require memory allocation. If possible, architect your code such that
+//! you do not need to obtain a structured type.
+//!
+//! If you need the structured type, such as if you need to manipulate it in some way, there are two
+//! options: storing subtags, and storing a string for parsing.
+//!
+//! ## Storing Subtags
+//!
+//! If the data being stored only contains a limited number of subtags, you can store them as a
+//! tuple, and then construct the [`LanguageIdentifier`] externally.
+//!
+//! ```
+//! use icu_locid::subtags::{Language, Region, Script};
+//! use icu_locid::LanguageIdentifier;
+//! use icu_locid::{
+//!     langid,
+//!     subtags::{language, region, script},
+//! };
+//! use zerovec::ZeroMap;
+//!
+//! // ZeroMap from integer to LSR (language-script-region)
+//! let zm: ZeroMap<u32, (Language, Option<Script>, Option<Region>)> = [
+//!     (5, (language!("de"), None, Some(region!("DE")))),
+//!     (10, (language!("en"), None, Some(region!("US")))),
+//!     (15, (language!("my"), None, Some(region!("MM")))),
+//!     (
+//!         20,
+//!         (language!("sr"), Some(script!("Cyrl")), Some(region!("ME"))),
+//!     ),
+//!     (25, (language!("zh"), None, Some(region!("TW")))),
+//! ]
+//! .into_iter()
+//! .collect();
+//!
+//! // Construct a LanguageIdentifier from a tuple entry
+//! let lid: LanguageIdentifier =
+//!     zm.get_copied(&25).expect("element is present").into();
+//!
+//! assert_eq!(lid, langid!("zh-TW"));
+//! ```
+//!
+//! ## Storing Strings
+//!
+//! If it is necessary to store and obtain an arbitrary locale, it is currently recommended to
+//! store a BCP-47 string and parse it when needed.
+//!
+//! Since the string is stored in an unparsed state, it is not safe to `unwrap` the result from
+//! `Locale::try_from_bytes()`. See [icu4x#831](https://github.com/unicode-org/icu4x/issues/831)
+//! for a discussion on potential data models that could ensure that the locale is valid during
+//! deserialization.
+//!
+//! As above, to produce more human-readable serialized output, you can use [`UnvalidatedStr`].
+//!
+//! ```
+//! use icu_locid::langid;
+//! use icu_locid::Locale;
+//! use zerovec::ule::UnvalidatedStr;
+//! use zerovec::ZeroMap;
+//!
+//! // ZeroMap from integer to locale string
+//! let data: &[(u32, &UnvalidatedStr)] = &[
+//!     (5, "de-DE-u-hc-h12".into()),
+//!     (10, "en-US-u-ca-buddhist".into()),
+//!     (15, "my-MM".into()),
+//!     (20, "sr-Cyrl-ME".into()),
+//!     (25, "zh-TW".into()),
+//!     (30, "INVALID".into()),
+//! ];
+//! let zm: ZeroMap<u32, UnvalidatedStr> = data.iter().copied().collect();
+//!
+//! // Construct a Locale by parsing the string.
+//! let value = zm.get(&25).expect("element is present");
+//! let loc = Locale::try_from_bytes(value);
+//! assert_eq!(loc, Ok(langid!("zh-TW").into()));
+//!
+//! // Invalid entries are fallible
+//! let err_value = zm.get(&30).expect("element is present");
+//! let err_loc = Locale::try_from_bytes(err_value);
+//! assert!(matches!(err_loc, Err(_)));
+//! ```
+//!
+//! [`Locale`]: crate::Locale
+//! [`Locale::strict_cmp()`]: crate::Locale::strict_cmp()
+//! [`LanguageIdentifier`]: crate::LanguageIdentifier
+//! [`UnvalidatedStr`]: zerovec::ule::UnvalidatedStr