Ensure islower/isupper handles strs without chars

joshuamegnauth54 · joshuamegnauth54 · commit e52405fa57e9 · 2026-04-12T21:38:40.000-04:00
This fixes a regression mentioned by CodeRabbit. I also figured out how
to check a string's case without allocation using Unicode properties.
Thus, this commit removes `icu_casemap` again. `icu_casemap` and my old
solution is required for a robust case check, but it seems like the
current code is fine for Python.
diff --git a/Cargo.lock b/Cargo.lock
diff --git a/Cargo.toml b/Cargo.toml
@@ -222,16 +222,13 @@ strum = "0.28"
 strum_macros = "0.28"
 syn = "2"
 thiserror = "2.0"
-icu_casemap = "2"
-icu_locale = "2"
 icu_properties = "2"
 icu_normalizer = "2"
 unicode-casing = "0.1.1"
 unic-ucd-age = "0.9.0"
 unicode_names2 = "2.0.0"
 widestring = "1.2.0"
 windows-sys = "0.61.2"
-writeable = "0.6.1"
 wasm-bindgen = "0.2.106"
 
 # Lints
diff --git a/crates/vm/Cargo.toml b/crates/vm/Cargo.toml
@@ -86,10 +86,7 @@ timsort = "0.1.2"
 # TODO: use unic for this; needed for title case:
 # https://github.com/RustPython/RustPython/pull/832#discussion_r275428939
 unicode-casing = { workspace = true }
-icu_casemap = { workspace = true }
-icu_locale = { workspace = true }
 icu_properties = { workspace = true }
-writeable = { workspace = true }
 
 [target.'cfg(unix)'.dependencies]
 rustix = { workspace = true }
diff --git a/crates/vm/src/anystr.rs b/crates/vm/src/anystr.rs
@@ -4,10 +4,11 @@ use crate::{
     convert::TryFromBorrowedObject,
     function::OptionalOption,
 };
-use icu_casemap::CaseMapper;
-use icu_locale::LanguageIdentifier;
+use icu_properties::{
+    CodePointSetData,
+    props::{Alphabetic, ChangesWhenLowercased, ChangesWhenUppercased},
+};
 use num_traits::{cast::ToPrimitive, sign::Signed};
-use writeable::Writeable;
 
 use core::ops::Range;
 
@@ -410,19 +411,17 @@ pub trait AnyStr {
     //  _Py_bytes_islower
     //  unicode_islower_impl
     fn py_islower(&self) -> bool {
+        let case_change = CodePointSetData::new::<ChangesWhenLowercased>();
+        let alphabetic = CodePointSetData::new::<Alphabetic>();
         let mut lower = false;
-        let mut lowercased = String::with_capacity(self.bytes_len());
-        let cm = CaseMapper::new();
         for chunk in self.as_bytes().utf8_chunks().map(|c| c.valid()) {
-            let writer = cm.lowercase(chunk, &LanguageIdentifier::UNKNOWN);
-            lowercased.clear();
-            writer
-                .write_to(&mut lowercased)
-                .expect("Writing to a buffer is infallible");
-            if chunk != lowercased {
+            if chunk.chars().any(|c| case_change.contains(c)) {
                 return false;
             }
-            lower = true;
+
+            if !lower && chunk.chars().any(|c| alphabetic.contains(c)) {
+                lower = true;
+            }
         }
         lower
     }
@@ -431,19 +430,17 @@ pub trait AnyStr {
     //   Py_bytes_isupper
     //  unicode_isupper_impl
     fn py_isupper(&self) -> bool {
+        let case_change = CodePointSetData::new::<ChangesWhenUppercased>();
+        let alphabetic = CodePointSetData::new::<Alphabetic>();
         let mut upper = false;
-        let mut uppercased = String::with_capacity(self.bytes_len());
-        let cm = CaseMapper::new();
         for chunk in self.as_bytes().utf8_chunks().map(|c| c.valid()) {
-            let writer = cm.uppercase(chunk, &LanguageIdentifier::UNKNOWN);
-            uppercased.clear();
-            writer
-                .write_to(&mut uppercased)
-                .expect("Writing to a buffer is infallible");
-            if chunk != uppercased {
+            if chunk.chars().any(|c| case_change.contains(c)) {
                 return false;
             }
-            upper = true;
+
+            if !upper && chunk.chars().any(|c| alphabetic.contains(c)) {
+                upper = true;
+            }
         }
         upper
     }
diff --git a/extra_tests/snippets/builtin_str.py b/extra_tests/snippets/builtin_str.py
@@ -222,6 +222,12 @@
 assert "abc\t12345\txyz".expandtabs() == "abc     12345   xyz"
 assert "-".join(["1", "2", "3"]) == "1-2-3"
 assert "HALLO".isupper()
+assert not "123".isupper()
+assert not "123".islower()
+assert not "\U0001f431".isupper()
+assert not "\U0001f431".islower()
+assert "\U0001f431 CAT".isupper()
+assert "\U0001f431 cat".islower()
 assert "\u0295".islower()
 assert "\u1c89".isupper()
 assert "hello, my name is".partition("my ") == ("hello, ", "my ", "name is")
