From 056ae1f9263fd78d8a087a3744d868b52d957651 Mon Sep 17 00:00:00 2001
From: FletcherDares <52580867+FletcherDares@users.noreply.github.com>
Date: Tue, 2 Jun 2026 00:45:55 -0400
Subject: [PATCH] Centralize `ISASCII` early exit for `search_nonascii`
 (#17166)

---
 benchmark/string_coderange_scan.yml | 10 ++++
 string.c                            | 74 +++++++++++++----------------
 2 files changed, 42 insertions(+), 42 deletions(-)
 create mode 100644 benchmark/string_coderange_scan.yml

diff --git a/benchmark/string_coderange_scan.yml b/benchmark/string_coderange_scan.yml
new file mode 100644
index 00000000000..d47bbd2b303
--- /dev/null
+++ b/benchmark/string_coderange_scan.yml
@@ -0,0 +1,10 @@
+prelude: |
+  def unknown(s) = s.b.force_encoding("UTF-8")
+  multibyte   = unknown("\u{00e9}" * 16384)   # best case: every byte non-ASCII
+  alternating = unknown("\u{00e9}a" * 10922)  # worst case: non-ASCII then ASCII
+  ascii       = unknown("a" * 32768)          # baseline
+
+benchmark:
+  coderange_multibyte:   multibyte.dup.valid_encoding?
+  coderange_alternating: alternating.dup.valid_encoding?
+  coderange_ascii:       ascii.dup.valid_encoding?
diff --git a/string.c b/string.c
index 61d4f16679d..134e1254318 100644
--- a/string.c
+++ b/string.c
@@ -713,6 +713,10 @@ search_nonascii(const char *p, const char *e)
 {
     const char *s, *t;
 
+    if (p < e && !ISASCII(*p)) {
+        return p;
+    }
+
 #if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)
 # if SIZEOF_UINTPTR_T == 8
 #  define NONASCII_MASK UINT64_C(0x8080808080808080)
@@ -2297,26 +2301,22 @@ enc_strlen(const char *p, const char *e, rb_encoding *enc, int cr)
         c = 0;
         if (ENC_CODERANGE_CLEAN_P(cr)) {
             while (p < e) {
-                if (ISASCII(*p)) {
-                    q = search_nonascii(p, e);
-                    if (!q)
-                        return c + (e - p);
-                    c += q - p;
-                    p = q;
-                }
+                q = search_nonascii(p, e);
+                if (!q)
+                    return c + (e - p);
+                c += q - p;
+                p = q;
                 p += rb_enc_fast_mbclen(p, e, enc);
                 c++;
             }
         }
         else {
             while (p < e) {
-                if (ISASCII(*p)) {
-                    q = search_nonascii(p, e);
-                    if (!q)
-                        return c + (e - p);
-                    c += q - p;
-                    p = q;
-                }
+                q = search_nonascii(p, e);
+                if (!q)
+                    return c + (e - p);
+                c += q - p;
+                p = q;
                 p += rb_enc_mbclen(p, e, enc);
                 c++;
             }
@@ -2354,15 +2354,13 @@ rb_enc_strlen_cr(const char *p, const char *e, rb_encoding *enc, int *cr)
     else if (rb_enc_asciicompat(enc)) {
         c = 0;
         while (p < e) {
-            if (ISASCII(*p)) {
-                q = search_nonascii(p, e);
-                if (!q) {
-                    if (!*cr) *cr = ENC_CODERANGE_7BIT;
-                    return c + (e - p);
-                }
-                c += q - p;
-                p = q;
+            q = search_nonascii(p, e);
+            if (!q) {
+                if (!*cr) *cr = ENC_CODERANGE_7BIT;
+                return c + (e - p);
             }
+            c += q - p;
+            p = q;
             ret = rb_enc_precise_mbclen(p, e, enc);
             if (MBCLEN_CHARFOUND_P(ret)) {
                 *cr |= ENC_CODERANGE_VALID;
@@ -3020,16 +3018,14 @@ str_nth_len(const char *p, const char *e, long *nthp, rb_encoding *enc)
                 *nthp = nth;
                 return (char *)e;
             }
-            if (ISASCII(*p)) {
-                p2 = search_nonascii(p, e2);
-                if (!p2) {
-                    nth -= e2 - p;
-                    *nthp = nth;
-                    return (char *)e2;
-                }
-                nth -= p2 - p;
-                p = p2;
+            p2 = search_nonascii(p, e2);
+            if (!p2) {
+                nth -= e2 - p;
+                *nthp = nth;
+                return (char *)e2;
             }
+            nth -= p2 - p;
+            p = p2;
             n = rb_enc_mbclen(p, e, enc);
             p += n;
             nth--;
@@ -11850,17 +11846,11 @@ enc_str_scrub(rb_encoding *enc, VALUE str, VALUE repl, int cr)
             else if (MBCLEN_CHARFOUND_P(ret)) {
                 cr = ENC_CODERANGE_VALID;
                 p += MBCLEN_CHARFOUND_LEN(ret);
-                /*
-                 * After a valid multibyte character, skip the following ASCII run.
-                 * If the next byte is already non-ASCII, search_nonascii would only
-                 * rediscover p after its word-at-a-time setup.
-                 */
-                if (p < e && ISASCII(*p)) {
-                    p = search_nonascii(p, e);
-                    if (!p) {
-                        p = e;
-                        break;
-                    }
+                /* After a multibyte character, fast-skip the following ASCII run. */
+                p = search_nonascii(p, e);
+                if (!p) {
+                    p = e;
+                    break;
                 }
             }
             else if (MBCLEN_INVALID_P(ret)) {