Avoid reallocation in preg_replace() if nothing was replaced

5 files changed, 58 insertions, 56 deletions

diff --git a/ext/fileinfo/libmagic/funcs.c b/ext/fileinfo/libmagic/funcs.c
index 4df0a2a496..82f4cd38cd 100644
--- a/ext/fileinfo/libmagic/funcs.c
+++ b/ext/fileinfo/libmagic/funcs.c
@@ -457,7 +457,7 @@ file_replace(struct magic_set *ms, const char *pat, const char *rep)
 	zval_ptr_dtor(&patt);
 
 	ZVAL_STRING(&repl, rep);
-	res = php_pcre_replace_impl(pce, ms->o.buf, strlen(ms->o.buf), &repl, 0, -1, &rep_cnt);
+	res = php_pcre_replace_impl(pce, NULL, ms->o.buf, strlen(ms->o.buf), &repl, 0, -1, &rep_cnt);
 
 	zval_ptr_dtor(&repl);
 	if (NULL == res) {
diff --git a/ext/pcre/php_pcre.c b/ext/pcre/php_pcre.c
index c255fd9174..8a6ecb5817 100644
--- a/ext/pcre/php_pcre.c
+++ b/ext/pcre/php_pcre.c
@@ -998,6 +998,7 @@ static zend_string *preg_do_repl_func(zval *function, char *subject, int *offset
 /* {{{ php_pcre_replace
  */
 PHPAPI zend_string *php_pcre_replace(zend_string *regex,
+							  zend_string *subject_str,
 							  char *subject, int subject_len,
 							  zval *replace_val, int is_callable_replace,
 							  int limit, int *replace_count)
@@ -1009,13 +1010,13 @@ PHPAPI zend_string *php_pcre_replace(zend_string *regex,
 		return NULL;
 	}
 
-	return php_pcre_replace_impl(pce, subject, subject_len, replace_val,
+	return php_pcre_replace_impl(pce, subject_str, subject, subject_len, replace_val,
 		is_callable_replace, limit, replace_count);
 }
 /* }}} */
 
 /* {{{ php_pcre_replace_impl() */
-PHPAPI zend_string *php_pcre_replace_impl(pcre_cache_entry *pce, char *subject, int subject_len, zval *replace_val,
+PHPAPI zend_string *php_pcre_replace_impl(pcre_cache_entry *pce, zend_string *subject_str, char *subject, int subject_len, zval *replace_val,
 	int is_callable_replace, int limit, int *replace_count)
 {
 	pcre_extra		*extra = pce->extra;/* Holds results of studying */
@@ -1084,8 +1085,8 @@ PHPAPI zend_string *php_pcre_replace_impl(pcre_cache_entry *pce, char *subject,
 		}
 	}
 
-	alloc_len = 2 * subject_len;
-	result = zend_string_alloc(alloc_len * sizeof(char), 0);
+	alloc_len = 0;
+	result = NULL;
 
 	/* Initialize */
 	match = NULL;
@@ -1148,9 +1149,17 @@ PHPAPI zend_string *php_pcre_replace_impl(pcre_cache_entry *pce, char *subject,
 				}
 			}
 
-			if (new_len > alloc_len) {
-				alloc_len = alloc_len + 2 * new_len;
-				result = zend_string_realloc(result, alloc_len, 0);
+			if (new_len >= alloc_len) {
+				if (alloc_len == 0) {
+					alloc_len = 2 * subject_len;
+					if (new_len >= alloc_len) {
+						alloc_len = alloc_len + 2 * new_len;
+					}
+					result = zend_string_alloc(alloc_len, 0);
+				} else {
+					alloc_len = alloc_len + 2 * new_len;
+					result = zend_string_realloc(result, alloc_len, 0);
+				}
 			}
 			/* copy the part of the string before the match */
 			memcpy(&result->val[result_len], piece, match-piece);
@@ -1205,6 +1214,10 @@ PHPAPI zend_string *php_pcre_replace_impl(pcre_cache_entry *pce, char *subject,
 				memcpy(&result->val[result_len], piece, 1);
 				result_len++;
 			} else {
+				if (!result && subject_str) {
+					result = zend_string_copy(subject_str);
+					break;
+				}
 				new_len = result_len + subject_len - start_offset;
 				if (new_len > alloc_len) {
 					alloc_len = new_len; /* now we know exactly how long it is */
@@ -1214,12 +1227,15 @@ PHPAPI zend_string *php_pcre_replace_impl(pcre_cache_entry *pce, char *subject,
 				memcpy(&result->val[result_len], piece, subject_len - start_offset);
 				result_len += subject_len - start_offset;
 				result->val[result_len] = '\0';
+				result->len = result_len;
 				break;
 			}
 		} else {
 			pcre_handle_exec_error(count);
-			zend_string_free(result);
-			result = NULL;
+			if (result) {
+				zend_string_free(result);
+				result = NULL;
+			}
 			break;
 		}
 
@@ -1233,9 +1249,6 @@ PHPAPI zend_string *php_pcre_replace_impl(pcre_cache_entry *pce, char *subject,
 		start_offset = offsets[1];
 	}
 
-	if (result) {
-		result->len = result_len;
-	}
 	if (size_offsets <= 32) {
 		free_alloca(offsets, use_heap);
 	} else {
@@ -1300,6 +1313,7 @@ static zend_string *php_replace_in_subject(zval *regex, zval *replace, zval *sub
 			/* Do the actual replacement and put the result back into subject_str
 			   for further replacements. */
 			if ((result = php_pcre_replace(regex_str,
+										   subject_str,
 										   subject_str->val,
 										   (int)subject_str->len,
 										   replace_value,
@@ -1320,6 +1334,7 @@ static zend_string *php_replace_in_subject(zval *regex, zval *replace, zval *sub
 		return subject_str;
 	} else {
 		result = php_pcre_replace(Z_STR_P(regex),
+								  subject_str,
 								  subject_str->val,
 								  (int)subject_str->len,
 								  replace,
diff --git a/ext/pcre/php_pcre.h b/ext/pcre/php_pcre.h
index dc923fe62f..88f810493f 100644
--- a/ext/pcre/php_pcre.h
+++ b/ext/pcre/php_pcre.h
@@ -33,7 +33,7 @@
 #include <locale.h>
 #endif
 
-PHPAPI zend_string *php_pcre_replace(zend_string *regex, char *subject, int subject_len, zval *replace_val, int is_callable_replace, int limit, int *replace_count);
+PHPAPI zend_string *php_pcre_replace(zend_string *regex, zend_string *subject_str, char *subject, int subject_len, zval *replace_val, int is_callable_replace, int limit, int *replace_count);
 PHPAPI pcre* pcre_get_compiled_regex(zend_string *regex, pcre_extra **extra, int *options);
 PHPAPI pcre* pcre_get_compiled_regex_ex(zend_string *regex, pcre_extra **extra, int *preg_options, int *coptions);
 
@@ -59,7 +59,7 @@ PHPAPI pcre_cache_entry* pcre_get_compiled_regex_cache(zend_string *regex);
 PHPAPI void  php_pcre_match_impl(  pcre_cache_entry *pce, char *subject, int subject_len, zval *return_value,
 	zval *subpats, int global, int use_flags, zend_long flags, zend_long start_offset);
 
-PHPAPI zend_string *php_pcre_replace_impl(pcre_cache_entry *pce, char *subject, int subject_len, zval *return_value,
+PHPAPI zend_string *php_pcre_replace_impl(pcre_cache_entry *pce, zend_string *subject_str, char *subject, int subject_len, zval *return_value,
 	int is_callable_replace, int limit, int *replace_count);
 
 PHPAPI void  php_pcre_split_impl(  pcre_cache_entry *pce, char *subject, int subject_len, zval *return_value,
diff --git a/ext/spl/spl_iterators.c b/ext/spl/spl_iterators.c
index bb7424a395..21b3c8c905 100644
--- a/ext/spl/spl_iterators.c
+++ b/ext/spl/spl_iterators.c
@@ -2028,10 +2028,9 @@ SPL_METHOD(CallbackFilterIterator, accept)
 SPL_METHOD(RegexIterator, accept)
 {
 	spl_dual_it_object *intern;
-	char *subject;
-	zend_string *result;
-	int subject_len, use_copy, count = 0;
-	zval *subject_ptr, subject_copy, zcount, *replacement, tmp_replacement, rv;
+	zend_string *result, *subject;
+	int count = 0;
+	zval zcount, *replacement, tmp_replacement, rv;
 
 	if (zend_parse_parameters_none() == FAILURE) {
 		return;
@@ -2046,51 +2045,40 @@ SPL_METHOD(RegexIterator, accept)
 	}
 
 	if (intern->u.regex.flags & REGIT_USE_KEY) {
-		subject_ptr = &intern->current.key;
+		subject = zval_get_string(&intern->current.key);
 	} else {
-		subject_ptr = &intern->current.data;
+		subject = zval_get_string(&intern->current.data);
 	}
 
-	ZVAL_UNDEF(&subject_copy);
-	use_copy = zend_make_printable_zval(subject_ptr, &subject_copy);
-	if (use_copy) {
-		subject = Z_STRVAL(subject_copy);
-		subject_len = (int)Z_STRLEN(subject_copy);
-	} else {
-		subject = Z_STRVAL_P(subject_ptr);
-		subject_len = (int)Z_STRLEN_P(subject_ptr);
-	}
-
-	use_copy = 0;
 	switch (intern->u.regex.mode)
 	{
 		case REGIT_MODE_MAX: /* won't happen but makes compiler happy */
 		case REGIT_MODE_MATCH:
-			count = pcre_exec(intern->u.regex.pce->re, intern->u.regex.pce->extra, subject, subject_len, 0, 0, NULL, 0);
+			count = pcre_exec(intern->u.regex.pce->re, intern->u.regex.pce->extra, subject->val, subject->len, 0, 0, NULL, 0);
 			RETVAL_BOOL(count >= 0);
 			break;
 
 		case REGIT_MODE_ALL_MATCHES:
 		case REGIT_MODE_GET_MATCH:
-			if (!use_copy) {
-				subject = estrndup(subject, subject_len);
-				use_copy = 1;
-			}
+//???			if (!use_copy) {
+//???				subject = estrndup(subject, subject_len);
+//???				use_copy = 1;
+//???			}
 			zval_ptr_dtor(&intern->current.data);
 			ZVAL_UNDEF(&intern->current.data);
-			php_pcre_match_impl(intern->u.regex.pce, subject, subject_len, &zcount,
+			php_pcre_match_impl(intern->u.regex.pce, subject->val, subject->len, &zcount,
 				&intern->current.data, intern->u.regex.mode == REGIT_MODE_ALL_MATCHES, intern->u.regex.use_flags, intern->u.regex.preg_flags, 0);
 			RETVAL_BOOL(Z_LVAL(zcount) > 0);
 			break;
 
 		case REGIT_MODE_SPLIT:
-			if (!use_copy) {
-				subject = estrndup(subject, subject_len);
-				use_copy = 1;
-			}
+//???			if (!use_copy) {
+//???				subject = estrndup(subject, subject_len);
+//???				use_copy = 1;
+//???			}
 			zval_ptr_dtor(&intern->current.data);
 			ZVAL_UNDEF(&intern->current.data);
-			php_pcre_split_impl(intern->u.regex.pce, subject, subject_len, &intern->current.data, -1, intern->u.regex.preg_flags);
+			php_pcre_split_impl(intern->u.regex.pce, subject->val, subject->len, &intern->current.data, -1, intern->u.regex.preg_flags);
 			count = zend_hash_num_elements(Z_ARRVAL(intern->current.data));
 			RETVAL_BOOL(count > 1);
 			break;
@@ -2103,7 +2091,7 @@ SPL_METHOD(RegexIterator, accept)
 				convert_to_string(&tmp_replacement);
 				replacement = &tmp_replacement;
 			}
-			result = php_pcre_replace_impl(intern->u.regex.pce, subject, subject_len, replacement, 0, -1, &count);
+			result = php_pcre_replace_impl(intern->u.regex.pce, subject, subject->val, subject->len, replacement, 0, -1, &count);
 
 			if (intern->u.regex.flags & REGIT_USE_KEY) {
 				zval_ptr_dtor(&intern->current.key);
@@ -2122,13 +2110,7 @@ SPL_METHOD(RegexIterator, accept)
 	if (intern->u.regex.flags & REGIT_INVERTED) {
 		RETVAL_BOOL(Z_TYPE_P(return_value) != IS_TRUE);
 	}
-
-	if (use_copy) {
-		efree(subject);
-	}
-	if (!Z_ISUNDEF(subject_copy)) {
-		zval_ptr_dtor(&subject_copy);
-	}
+	zend_string_release(subject);
 } /* }}} */
 
 /* {{{ proto string RegexIterator::getRegex()
diff --git a/win32/sendmail.c b/win32/sendmail.c
index 0254d82ee9..fd7424dda7 100644
--- a/win32/sendmail.c
+++ b/win32/sendmail.c
@@ -168,15 +168,16 @@ static zend_string *php_win32_mail_trim_header(char *header)
 	regex = zend_string_init(PHP_WIN32_MAIL_UNIFY_PATTERN, sizeof(PHP_WIN32_MAIL_UNIFY_PATTERN)-1, 0);
 
 	result = php_pcre_replace(regex,
-				  header, (int)strlen(header),
+				  NULL, header, (int)strlen(header),
 				  &replace,
 				  0,
 				  -1,
 				  NULL);
 
+	zval_ptr_dtor(&replace);
+	zend_string_release(regex);
+
 	if (NULL == result) {
-		zval_ptr_dtor(&replace);
-		zend_string_free(regex);
 		return NULL;
 	}
 
@@ -184,12 +185,16 @@ static zend_string *php_win32_mail_trim_header(char *header)
 	regex = zend_string_init(PHP_WIN32_MAIL_RMVDBL_PATTERN, sizeof(PHP_WIN32_MAIL_RMVDBL_PATTERN)-1, 0);
 
 	result2 = php_pcre_replace(regex,
-				   result->val, (int)result->len,
+				   result, result->val, (int)result->len,
 				   &replace,
 				  0,
 				  -1,
 				  NULL);
-	return result;
+	zval_ptr_dtor(&replace);
+	zend_string_release(regex);
+	zend_string_release(result);
+
+	return result2;
 #else
 	/* In case we don't have PCRE support (for whatever reason...) simply do nothing and return the unmodified header */
 	return estrdup(header);