Merge pull request #4027 from Silvertree2010/fix/h1953-xml-id-name-round-trip-turns-a-s

Escape spaces before x<hex> and a separator in XML ids

crates/doc/src/xml_id.rs

@@ -32,9 +32,11 @@ pub fn to_object_name(id: &str) -> String {
 pub fn from_object_name(name: &str) -> String {
     let mut out = String::with_capacity(name.len().saturating_mul(2));
     let last = name.chars().count().saturating_sub(1);
-    for (i, c) in name.chars().enumerate() {
-        if c == ' ' && i == last {
-            // A trailing bare `_` could read back as the end of a uniqueness suffix.
+    let chars: Vec<char> = name.chars().collect();
+    for (i, &c) in chars.iter().enumerate() {
+        if c == ' ' && (i == last || starts_fake_escape(chars.get(i + 1..).unwrap_or_default())) {
+            // A trailing bare `_` could read back as the end of a uniqueness suffix, and a bare
+            // `_` before `x<hex>_` would read back as an `_xHH_` escape.
             out.push_str("_x20_");
         } else if c == ' ' {
             out.push('_');
@@ -68,6 +70,19 @@ pub fn is_valid(id: &str) -> bool {
     chars.all(|c| c.is_alphanumeric() || matches!(c, '_' | '.' | '-' | ':'))
 }
 
+/// Whether the encoded `rest` (after a space) starts with `x`, 1 to 6 hex digits and a character
+/// whose encoding begins with `_`, so that a bare `_` before it would form an escape.
+fn starts_fake_escape(rest: &[char]) -> bool {
+    let Some(('x', digits)) = rest.split_first() else {
+        return false;
+    };
+    let n = digits.iter().take_while(|c| c.is_ascii_hexdigit()).count();
+    if n == 0 || n > 6 {
+        return false;
+    }
+    digits.get(n).is_some_and(|&c| c == ' ' || c == '_' || !(c.is_alphanumeric() || matches!(c, '.' | '-' | ':')))
+}
+
 /// `_xHHHH_` at the start of `s`: the character and the escape's length.
 fn escaped(s: &str) -> Option<(char, usize)> {
     let body = s.strip_prefix("_x")?;
@@ -119,6 +134,15 @@ mod tests {
         assert_eq!(to_object_name("a_x-41_b"), "a x-41 b");
     }
 
+    #[test]
+    fn space_before_x_hex_and_underscore_is_not_an_escape() {
+        for name in ["layer x1 copy", "a xa b", "a xe b", "a xb b", "a x1(b", "a x1_b", "a x123456 c"] {
+            let id = from_object_name(name);
+            assert_eq!(to_object_name(&id), name, "name {name:?} -> id {id:?}");
+        }
+        assert_eq!(from_object_name("layer x1 copy"), "layer_x20_x1_copy");
+    }
+
     #[test]
     fn round_trip_names_that_cannot_start_an_xml_name() {
         for name in ["-abc", ".5 mm", "²x", "Ⅷ", "", "1abc", "Layer 1", " lead", "_a", ":a", "#"] {