Optimize stripos

Closes GH-7847

Previously stripos would lowercase both the haystack and needle to reuse
strpos. The approach in this PR is similar to strpos. memchr is highly
optimized so we're using it to search for the first character of the
needle in the haystack. If we find it we compare the remaining
characters of the needle manually.

The new implementation seems to perform about half as well as strpos (as
two memchr calls are necessary).

Author: iluuu1994

Zend/zend_operators.h

@@ -919,6 +919,70 @@ static zend_always_inline void zend_unwrap_reference(zval *op) /* {{{ */
 }
 /* }}} */
 
+static zend_always_inline const char *strnichr(const char *ptr, char value, size_t num)
+{
+	char value_lower = zend_tolower_ascii(value);
+	char value_upper = zend_toupper_ascii(value);
+
+	if (value_lower == value_upper) {
+		return (const char *)memchr(ptr, value, num);
+	}
+
+	const char *ptr_lower = (const char *)memchr(ptr, value_lower, num);
+	const char *ptr_upper = (const char *)memchr(ptr, value_upper, num);
+	if (ptr_lower && ptr_upper) {
+		return ptr_lower < ptr_upper ? ptr_lower : ptr_upper;
+	}
+	return ptr_lower ? ptr_lower : ptr_upper;
+}
+
+static zend_always_inline bool strnieq(const char *ptr1, const char *ptr2, size_t num)
+{
+	for (size_t i = 0; i < num; i++) {
+		if (zend_tolower_ascii(*ptr1++) != zend_tolower_ascii(*ptr2++)) {
+			return 0;
+		}
+	}
+	return 1;
+}
+
+static zend_always_inline const char *
+zend_memnistr(const char *haystack, const char *needle, size_t needle_len, const char *end)
+{
+	const char *p = haystack;
+	size_t off_s;
+
+	ZEND_ASSERT(end >= p);
+
+	if (needle_len == 1) {
+		return (const char *)strnichr(p, *needle, (end-p));
+	} else if (UNEXPECTED(needle_len == 0)) {
+		return p;
+	}
+
+	off_s = (size_t)(end - p);
+
+	if (needle_len > off_s) {
+		return NULL;
+	}
+
+	const char ne = zend_tolower_ascii(needle[needle_len-1]);
+	end -= needle_len;
+
+	while (p <= end) {
+		if ((p = (const char *)strnichr(p, *needle, (end-p+1)))) {
+			if (ne == zend_tolower_ascii(p[needle_len-1]) && strnieq(needle+1, p+1, needle_len-2)) {
+				return p;
+			}
+		} else {
+			return NULL;
+		}
+		p++;
+	}
+
+	return NULL;
+}
+
 
 END_EXTERN_C()
 

ext/standard/string.c

@@ -1890,7 +1890,6 @@ PHP_FUNCTION(stripos)
 	const char *found = NULL;
 	zend_string *haystack, *needle;
 	zend_long offset = 0;
-	zend_string *needle_dup = NULL, *haystack_dup;
 
 	ZEND_PARSE_PARAMETERS_START(2, 3)
 		Z_PARAM_STR(haystack)
@@ -1911,19 +1910,14 @@ PHP_FUNCTION(stripos)
 		RETURN_FALSE;
 	}
 
-	haystack_dup = zend_string_tolower(haystack);
-	needle_dup = zend_string_tolower(needle);
-	found = (char*)php_memnstr(ZSTR_VAL(haystack_dup) + offset,
-			ZSTR_VAL(needle_dup), ZSTR_LEN(needle_dup), ZSTR_VAL(haystack_dup) + ZSTR_LEN(haystack));
+	found = (char*)php_memnistr(ZSTR_VAL(haystack) + offset,
+			ZSTR_VAL(needle), ZSTR_LEN(needle), ZSTR_VAL(haystack) + ZSTR_LEN(haystack));
 
 	if (found) {
-		RETVAL_LONG(found - ZSTR_VAL(haystack_dup));
+		RETVAL_LONG(found - ZSTR_VAL(haystack));
 	} else {
 		RETVAL_FALSE;
 	}
-
-	zend_string_release_ex(haystack_dup, 0);
-	zend_string_release_ex(needle_dup, 0);
 }
 /* }}} */
 

main/php.h

@@ -344,6 +344,7 @@ END_EXTERN_C()
 #define phpin zendin
 
 #define php_memnstr zend_memnstr
+#define php_memnistr zend_memnistr
 
 /* functions */
 BEGIN_EXTERN_C()