Improve basename(). Avoid calling mblen() for ASCII compatible locales.

Author: dstogov

Zend/zend_globals.h

@@ -95,6 +95,10 @@ struct _zend_compiler_globals {
 	bool skip_shebang;
 	bool increment_lineno;
 
+	bool variable_width_locale;   /* UTF-8, Shift-JIS, Big5, ISO 2022, EUC, etc */
+	bool ascii_compatible_locale; /* locale uses ASCII characters as singletons */
+	                              /* and don't use them as lead/trail units     */
+
 	zend_string *doc_comment;
 	uint32_t extra_fn_flags;
 

Zend/zend_operators.c

@@ -30,12 +30,21 @@
 #include "zend_exceptions.h"
 #include "zend_closures.h"
 
+#include <locale.h>
+#ifdef HAVE_LANGINFO_H
+# include <langinfo.h>
+#endif
+
 #ifdef __SSE2__
 #include <emmintrin.h>
 #endif
 
+#if defined(ZEND_WIN32) && !defined(ZTS) && defined(_MSC_VER)
+/* This performance improvement of tolower() on Windows gives 10-18% on bench.php */
+#define ZEND_USE_TOLOWER_L 1
+#endif
+
 #ifdef ZEND_USE_TOLOWER_L
-#include <locale.h>
 static _locale_t current_locale = NULL;
 /* this is true global! may lead to strange effects on ZTS, but so may setlocale() */
 #define zend_tolower(c) _tolower_l(c, current_locale)
@@ -2537,13 +2546,85 @@ ZEND_API bool ZEND_FASTCALL zend_object_is_true(zval *op) /* {{{ */
 }
 /* }}} */
 
-#ifdef ZEND_USE_TOLOWER_L
 ZEND_API void zend_update_current_locale(void) /* {{{ */
 {
+#ifdef ZEND_USE_TOLOWER_L
+# if defined(ZEND_WIN32) && defined(_MSC_VER)
 	current_locale = _get_current_locale();
+# else
+	current_locale = uselocale(0);
+# endif
+#endif
+#if defined(ZEND_WIN32) && defined(_MSC_VER)
+	if (MB_CUR_MAX > 1) {
+		unsigned int cp = ___lc_codepage_func();
+		CG(variable_width_locale) = 1;
+		// TODO: EUC-* are also ASCII compatible ???
+		CG(ascii_compatible_locale) =
+			cp == 65001; /* UTF-8 */
+	} else {
+		CG(variable_width_locale) = 0;
+		CG(ascii_compatible_locale) = 1;
+	}
+#elif defined(MB_CUR_MAX)
+	/* Check if current locale uses variable width encoding */
+	if (MB_CUR_MAX > 1) {
+#if HAVE_NL_LANGINFO
+		const char *charmap = nl_langinfo(CODESET);
+#else
+		char buf[16];
+		const char *charmap = NULL;
+		const char *locale = setlocale(LC_CTYPE, NULL);
+
+		if (locale) {
+			const char *dot = strchr(locale, '.');
+			const char *modifier;
+
+			if (dot) {
+				dot++;
+				modifier = strchr(dot, '@');
+				if (!modifier) {
+					charmap = dot;
+				} else if (modifier - dot < sizeof(buf)) {
+					memcpy(buf, dot, modifier - dot);
+                    buf[modifier - dot] = '\0';
+                    charmap = buf;
+				}
+			}
+		}
+#endif
+		CG(variable_width_locale) = 1;
+		CG(ascii_compatible_locale) = 0;
+
+		if (charmap) {
+			size_t len = strlen(charmap);
+			static const char *ascii_compatible_charmaps[] = {
+				"utf-8",
+				"utf8",
+				// TODO: EUC-* are also ASCII compatible ???
+				NULL
+			};
+			const char **p;
+			/* Check if current locale is ASCII compatible */
+			for (p = ascii_compatible_charmaps; *p; p++) {
+				if (zend_binary_strcasecmp(charmap, len, *p, strlen(*p)) == 0) {
+					CG(ascii_compatible_locale) = 1;
+					break;
+				}
+			}
+		}
+
+	} else {
+		CG(variable_width_locale) = 0;
+		CG(ascii_compatible_locale) = 1;
+	}
+#else
+	/* We can't determine current charset. Assume the worst case */
+	CG(variable_width_locale) = 1;
+	CG(ascii_compatible_locale) = 0;
+#endif
 }
 /* }}} */
-#endif
 
 static zend_always_inline void zend_str_tolower_impl(char *dest, const char *str, size_t length) /* {{{ */ {
 	unsigned char *p = (unsigned char*)str;

Zend/zend_operators.h

@@ -450,16 +450,7 @@ ZEND_API zend_long ZEND_FASTCALL zend_atol(const char *str, size_t str_len);
 #define convert_to_object_ex(zv) convert_to_object(zv)
 #define convert_scalar_to_number_ex(zv) convert_scalar_to_number(zv)
 
-#if defined(ZEND_WIN32) && !defined(ZTS) && defined(_MSC_VER)
-/* This performance improvement of tolower() on Windows gives 10-18% on bench.php */
-#define ZEND_USE_TOLOWER_L 1
-#endif
-
-#ifdef ZEND_USE_TOLOWER_L
 ZEND_API void zend_update_current_locale(void);
-#else
-#define zend_update_current_locale()
-#endif
 
 /* The offset in bytes between the value and type fields of a zval */
 #define ZVAL_OFFSETOF_TYPE	\

ext/standard/string.c

@@ -1465,66 +1465,96 @@ PHP_FUNCTION(strtolower)
 /* {{{ php_basename */
 PHPAPI zend_string *php_basename(const char *s, size_t len, const char *suffix, size_t suffix_len)
 {
-	/* State 0 is directly after a directory separator (or at the start of the string).
-	 * State 1 is everything else. */
-	int state = 0;
-	const char *basename_start = s;
-	const char *basename_end = s;
-	while (len > 0) {
-		int inc_len = (*s == '\0' ? 1 : php_mblen(s, len));
+	const char *basename_start;
+	const char *basename_end;
+
+	if (CG(ascii_compatible_locale)) {
+#ifdef ZEND_WIN32
+		if ((len >= 2) && isalpha((int)((unsigned char *)s)[0]) && (s[1] == ':')) {
+			s += 2;
+			len -= 2;
+		}
+#endif
+
+		basename_end = s + len - 1;
+
+		/* Strip trailing slashes */
+		while (basename_end >= s && IS_SLASH_P(basename_end)) {
+			basename_end--;
+		}
+		if (basename_end < s) {
+			return ZSTR_EMPTY_ALLOC();
+		}
+
+		/* Extract filename */
+		basename_start = basename_end;
+		basename_end++;
+		while (basename_start > s && !IS_SLASH_P(basename_start - 1)) {
+			basename_start--;
+		}
+	} else {
+		/* State 0 is directly after a directory separator (or at the start of the string).
+		 * State 1 is everything else. */
+		int state = 0;
+
+		basename_start = s;
+		basename_end = s;
+		while (len > 0) {
+			int inc_len = (*s == '\0' ? 1 : php_mblen(s, len));
 
-		switch (inc_len) {
-			case 0:
-				goto quit_loop;
-			case 1:
+			switch (inc_len) {
+				case 0:
+					goto quit_loop;
+				case 1:
 #if defined(PHP_WIN32)
-				if (*s == '/' || *s == '\\') {
+					if (*s == '/' || *s == '\\') {
 #else
-				if (*s == '/') {
+					if (*s == '/') {
 #endif
-					if (state == 1) {
-						state = 0;
-						basename_end = s;
-					}
+						if (state == 1) {
+							state = 0;
+							basename_end = s;
+						}
 #if defined(PHP_WIN32)
-				/* Catch relative paths in c:file.txt style. They're not to confuse
-				   with the NTFS streams. This part ensures also, that no drive
-				   letter traversing happens. */
-				} else if ((*s == ':' && (s - basename_start == 1))) {
-					if (state == 0) {
-						basename_start = s;
-						state = 1;
+					/* Catch relative paths in c:file.txt style. They're not to confuse
+					   with the NTFS streams. This part ensures also, that no drive
+					   letter traversing happens. */
+					} else if ((*s == ':' && (s - basename_start == 1))) {
+						if (state == 0) {
+							basename_start = s;
+							state = 1;
+						} else {
+							basename_end = s;
+							state = 0;
+						}
+#endif
 					} else {
-						basename_end = s;
-						state = 0;
+						if (state == 0) {
+							basename_start = s;
+							state = 1;
+						}
+					}
+					break;
+				default:
+					if (inc_len < 0) {
+						/* If character is invalid, treat it like other non-significant characters. */
+						inc_len = 1;
+						php_mb_reset();
 					}
-#endif
-				} else {
 					if (state == 0) {
 						basename_start = s;
 						state = 1;
 					}
-				}
-				break;
-			default:
-				if (inc_len < 0) {
-					/* If character is invalid, treat it like other non-significant characters. */
-					inc_len = 1;
-					php_mb_reset();
-				}
-				if (state == 0) {
-					basename_start = s;
-					state = 1;
-				}
-				break;
+					break;
+			}
+			s += inc_len;
+			len -= inc_len;
 		}
-		s += inc_len;
-		len -= inc_len;
-	}
 
 quit_loop:
-	if (state == 1) {
-		basename_end = s;
+		if (state == 1) {
+			basename_end = s;
+		}
 	}
 
 	if (suffix != NULL && suffix_len < (size_t)(basename_end - basename_start) &&
@@ -4604,7 +4634,6 @@ static zend_string *try_setlocale_str(zend_long cat, zend_string *loc) {
 		retval = setlocale(cat, NULL);
 	}
 # endif
-	zend_update_current_locale();
 	if (!retval) {
 		return NULL;
 	}
@@ -4615,6 +4644,7 @@ static zend_string *try_setlocale_str(zend_long cat, zend_string *loc) {
 
 		BG(locale_changed) = 1;
 		if (cat == LC_CTYPE || cat == LC_ALL) {
+			zend_update_current_locale();
 			if (BG(ctype_string)) {
 				zend_string_release_ex(BG(ctype_string), 0);
 			}