From 8143e9c383e36f600680e1a28c833df7bd5bd60b Mon Sep 17 00:00:00 2001
From: nsfisis <nsfisis@gmail.com>
Date: Sun, 26 Jul 2026 00:38:03 +0900
Subject: fix(symfony-string): convert between ASCII and UTF-8 instead of
 panicking

mb_detect_encoding reports "ASCII" for pure ASCII input, so the
conversion paths of toCodePointString/toByteString are reachable: the
formatter's addLineBreaks feeds the detected encoding straight back into
them. Port PHP's mb_convert_encoding calls, which the shim already
handles for ASCII/UTF-8.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---
 .../src/symfony/string/byte_string.rs              | 27 ++++++++++++++++------
 .../src/symfony/string/code_point_string.rs        | 15 +++++++-----
 2 files changed, 29 insertions(+), 13 deletions(-)

diff --git a/crates/shirabe-external-packages/src/symfony/string/byte_string.rs b/crates/shirabe-external-packages/src/symfony/string/byte_string.rs
index cd8eae86..8ad183e3 100644
--- a/crates/shirabe-external-packages/src/symfony/string/byte_string.rs
+++ b/crates/shirabe-external-packages/src/symfony/string/byte_string.rs
@@ -14,19 +14,32 @@ impl ByteString {
         }
     }
 
+    /// `from_encoding` is `""` for PHP's `null`.
     pub fn to_code_point_string(&self, from_encoding: &str) -> CodePointString {
-        // The source `string` is always valid UTF-8 (Rust `String`/`&str` guarantee). PHP takes the
-        // early-return branch whenever `preg_match('//u', ...)` holds for a UTF-8/null encoding, so
-        // the result mirrors the input bytes verbatim. The `mb_detect_encoding`/`iconv` conversion
-        // path only applies to genuinely non-UTF-8 byte strings, which cannot occur here.
+        // The source `string` is always valid UTF-8 (Rust `String`/`&str` guarantee), so
+        // `preg_match('//u', ...)` always holds.
         if matches!(from_encoding, "" | "utf8" | "utf-8" | "UTF8" | "UTF-8") {
             return CodePointString {
                 string: self.string.clone(),
             };
         }
 
-        // TODO(phase-d): non-UTF-8 source encodings would require mb_convert_encoding/iconv-style
-        // decoding, which is unreachable for the UTF-8-only inputs Shirabe currently produces.
-        todo!()
+        let valid_encoding = shirabe_php_shim::mb_detect_encoding(
+            &self.string,
+            Some(vec![from_encoding.to_string()]),
+            true,
+        )
+        .is_some();
+        // PHP throws InvalidArgumentException. Callers detect `from_encoding` from this very
+        // string, so a mismatch is a programming error.
+        assert!(valid_encoding, "Invalid \"{}\" string.", from_encoding);
+
+        CodePointString {
+            string: shirabe_php_shim::mb_convert_encoding(
+                self.string.clone().into_bytes(),
+                "UTF-8",
+                from_encoding,
+            ),
+        }
     }
 }
diff --git a/crates/shirabe-external-packages/src/symfony/string/code_point_string.rs b/crates/shirabe-external-packages/src/symfony/string/code_point_string.rs
index 15a6cfcb..97ee2508 100644
--- a/crates/shirabe-external-packages/src/symfony/string/code_point_string.rs
+++ b/crates/shirabe-external-packages/src/symfony/string/code_point_string.rs
@@ -75,17 +75,20 @@ impl CodePointString {
         Self { string }
     }
 
+    /// `to_encoding` is `""` for PHP's `null`.
     pub fn to_byte_string(&self, to_encoding: &str) -> String {
-        // The source is always valid UTF-8, so PHP's `toByteString` returns the string verbatim
-        // whenever the target is null/UTF-8 (the only encodings reached here). The
-        // mb_convert_encoding/iconv path applies only to non-UTF-8 targets, which do not occur.
+        // A CodePointString is an AbstractUnicodeString, so PHP's `$fromEncoding` is always
+        // 'UTF-8' and the string is returned verbatim for a null/UTF-8 target.
         if matches!(to_encoding, "" | "utf8" | "utf-8" | "UTF8" | "UTF-8") {
             return self.string.clone();
         }
 
-        // TODO(phase-d): converting to a non-UTF-8 target encoding needs mb_convert_encoding/iconv,
-        // unreachable for Shirabe's UTF-8-only output.
-        todo!()
+        // PHP falls back to iconv() only when mb_convert_encoding() rejects the target encoding.
+        shirabe_php_shim::mb_convert_encoding(
+            self.string.clone().into_bytes(),
+            to_encoding,
+            "UTF-8",
+        )
     }
 }
 
-- 
cgit v1.3.1