From d74aaf33093011b565d31a193678d881ba17b7ff Mon Sep 17 00:00:00 2001
From: Tim Fennis <fennis.tim@gmail.com>
Date: Tue, 8 Sep 2026 15:51:41 +0200
Subject: [PATCH] =?UTF-8?q?test(bench):=20benchmark=20indexing=20in=20long?=
 =?UTF-8?q?=20strings=20=F0=9F=A7=B5?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 benches/programs/string_index_ascii.ndc   | 15 +++++++++++++++
 benches/programs/string_index_unicode.ndc | 15 +++++++++++++++
 2 files changed, 30 insertions(+)
 create mode 100644 benches/programs/string_index_ascii.ndc
 create mode 100644 benches/programs/string_index_unicode.ndc

diff --git a/benches/programs/string_index_ascii.ndc b/benches/programs/string_index_ascii.ndc
new file mode 100644
index 00000000..ee75638a
--- /dev/null
+++ b/benches/programs/string_index_ascii.ndc
@@ -0,0 +1,15 @@
+// 10,000 reads near the end of a 131,072-codepoint ASCII string.
+// Build by doubling so construction is small compared with O(n) indexing.
+let text = "abcd";
+for _ in 0..15 {
+    text <>= text;
+}
+let last = text.len - 1;
+let matches = 0;
+for i in 0..10_000 {
+    if text[last - i % 4] == "d" {
+        matches += 1;
+    }
+}
+assert_eq(matches, 2_500);
+print(matches);
diff --git a/benches/programs/string_index_unicode.ndc b/benches/programs/string_index_unicode.ndc
new file mode 100644
index 00000000..eb4e4c59
--- /dev/null
+++ b/benches/programs/string_index_unicode.ndc
@@ -0,0 +1,15 @@
+// Same codepoint count and access pattern as string_index_ascii.ndc,
+// with a mix of 1-, 2-, 3-, and 4-byte UTF-8 codepoints.
+let text = "aé中😀";
+for _ in 0..15 {
+    text <>= text;
+}
+let last = text.len - 1;
+let matches = 0;
+for i in 0..10_000 {
+    if text[last - i % 4] == "😀" {
+        matches += 1;
+    }
+}
+assert_eq(matches, 2_500);
+print(matches);