captures keep the name's diacritics

1 files changed, 35 insertions, 1 deletions

diff --git a/internal/norm/fuzz_test.go b/internal/norm/fuzz_test.go
index fde62a0..27deadc 100644
--- a/internal/norm/fuzz_test.go
+++ b/internal/norm/fuzz_test.go
@@ -2,7 +2,11 @@
 
 package norm
 
-import "testing"
+import (
+	"strings"
+	"testing"
+	"unicode/utf8"
+)
 
 // FuzzTextIdempotent: normalising text twice changes nothing more than
 // normalising it once, under every case and fold setting, and so for names.
@@ -28,3 +32,33 @@ func FuzzTextIdempotent(f *testing.F) {
 		}
 	})
 }
+
+// FuzzFoldMapped: FoldMapped's text is always Fold's, and any part of it
+// leads back to original text that folds to something holding that part.
+func FuzzFoldMapped(f *testing.F) {
+	for _, s := range []string{"Łódź-faktura.pdf", "straße", "áb", "\xff\xfeé", "́a", "Æsir"} {
+		f.Add(s, 0, 2)
+	}
+	f.Fuzz(func(t *testing.T, s string, a, b int) {
+		m := FoldMapped(s)
+		if m.Text != Fold(s) {
+			t.Fatalf("FoldMapped(%q).Text = %q, Fold = %q", s, m.Text, Fold(s))
+		}
+		n := len(m.Text) + 1
+		a, b = (a%n+n)%n, (b%n+n)%n
+		if a > b {
+			a, b = b, a
+		}
+		// A regex match starts and ends on whole characters.
+		for a < len(m.Text) && !utf8.RuneStart(m.Text[a]) {
+			a--
+		}
+		for b < len(m.Text) && !utf8.RuneStart(m.Text[b]) {
+			b--
+		}
+		src := m.Source(a, b)
+		if !strings.Contains(Fold(src), m.Text[a:b]) {
+			t.Fatalf("FoldMapped(%q).Source(%d, %d) = %q, which folds to %q, not holding %q", s, a, b, src, Fold(src), m.Text[a:b])
+		}
+	})
+}