Keep the first digit of a CSS hex escape

css_test.go

@@ -806,6 +806,26 @@ func TestSelector(t *testing.T) {
 	}
 }
 
+func TestHexEscapeMatchesID(t *testing.T) {
+	s, err := Parse(`#\31 0`)
+	if err != nil {
+		t.Fatal(err)
+	}
+	root, err := html.Parse(strings.NewReader(`<div id="10">x</div><div id="1">y</div><div id="310">z</div>`))
+	if err != nil {
+		t.Fatal(err)
+	}
+	got := s.Select(root)
+	if len(got) != 1 {
+		t.Fatalf("matches=%d, want the element with id 10", len(got))
+	}
+	for _, a := range got[0].Attr {
+		if a.Key == "id" && a.Val != "10" {
+			t.Fatalf("id=%q, want 10", a.Val)
+		}
+	}
+}
+
 func TestBadSelector(t *testing.T) {
 	tests := []struct {
 		sel string

lex.go

@@ -4,6 +4,7 @@ import (
 	"fmt"
 	"strconv"
 	"strings"
+	"unicode"
 	"unicode/utf8"
 )
 
@@ -359,33 +360,37 @@ func (l *lexer) consumeEscape(b *strings.Builder) error {
 		return nil
 	}
 
+	// The digit already consumed is part of the value. At most five more
+	// follow it (six total). Whitespace ends the escape so a later hex
+	// digit stays in the identifier, which is how #\31 0 names id "10".
 	var hexRune strings.Builder
-	n := 0
-	for {
-		r := l.peek()
-		if isHex(r) {
-			l.pop()
-			n++
-			if n > 5 {
-				return l.errorf("too many hex digits consuming escape sequence")
-			}
-			hexRune.WriteRune(r)
-			continue
-		}
+	hexRune.WriteRune(r)
+	for n := 1; n < 6 && isHex(l.peek()); n++ {
+		hexRune.WriteRune(l.pop())
+	}
+	if isHex(l.peek()) {
+		return l.errorf("too many hex digits consuming escape sequence")
+	}
+	if isWhitespace(l.peek()) {
+		l.pop()
+	}
 
-		if isWhitespace(r) {
-			l.pop()
-			continue
-		}
+	s := hexRune.String()
+	val, err := strconv.ParseUint(s, 16, 64)
+	if err != nil {
+		return l.errorf("failed to parse hex escape sequence %s: %v", s, err)
+	}
+	b.WriteRune(escapedCodePoint(val))
+	return nil
+}
 
-		s := hexRune.String()
-		val, err := strconv.ParseUint(s, 16, 64)
-		if err != nil {
-			return l.errorf("failed to parse hex escape sequence %s: %v", s, err)
-		}
-		b.WriteRune(rune(val))
-		return nil
+// escapedCodePoint applies the CSS rule for a hex escape value.
+// Zero, a surrogate, or a value past the Unicode maximum is U+FFFD.
+func escapedCodePoint(val uint64) rune {
+	if val == 0 || val > unicode.MaxRune || (val >= 0xD800 && val <= 0xDFFF) {
+		return '\uFFFD'
 	}
+	return rune(val)
 }
 
 // https://www.w3.org/TR/css-syntax-3/#consume-a-name

lex_test.go

@@ -60,7 +60,7 @@ func TestLexer(t *testing.T) {
 			` "\0a f" `,
 			[]token{
 				tok(tokenWhitespace, " "),
-				tok(tokenString, `"\0a f"`, "¯"),
+				tok(tokenString, `"\0a f"`, "\nf"),
 				tok(tokenWhitespace, " "),
 			},
 		},
@@ -84,6 +84,18 @@ func TestLexer(t *testing.T) {
 				tok(tokenHash, `#\0100`, "#Ā").withFlag(tokenFlagID),
 			},
 		},
+		{
+			`#\31 0`,
+			[]token{
+				tok(tokenHash, `#\31 0`, "#10").withFlag(tokenFlagID),
+			},
+		},
+		{
+			`\26 B`,
+			[]token{
+				tok(tokenIdent, `\26 B`, "&B"),
+			},
+		},
 		{
 			`#foo()`,
 			[]token{