url: fast path ascii domains, do not run ToASCII

zimbabao · zimbabao · commit 861604f2b135 · 2017-05-16T08:26:41.000-07:00
To match browser behavior fast path ascii only domains and do not run ToASCII on them. Fixes: nodejs#12965 Refs: nodejs#12966 Refs: whatwg/url#309
diff --git a/node.gyp b/node.gyp
@@ -192,7 +192,6 @@
         'src/spawn_sync.cc',
         'src/string_bytes.cc',
         'src/string_search.cc',
-        'src/string_utils.cc',
         'src/stream_base.cc',
         'src/stream_wrap.cc',
         'src/tcp_wrap.cc',
@@ -629,7 +628,6 @@
         '<(OBJ_PATH)<(OBJ_SEPARATOR)util.<(OBJ_SUFFIX)',
         '<(OBJ_PATH)<(OBJ_SEPARATOR)string_bytes.<(OBJ_SUFFIX)',
         '<(OBJ_PATH)<(OBJ_SEPARATOR)string_search.<(OBJ_SUFFIX)',
-        '<(OBJ_PATH)<(OBJ_SEPARATOR)string_utils.<(OBJ_SUFFIX)',
         '<(OBJ_PATH)<(OBJ_SEPARATOR)stream_base.<(OBJ_SUFFIX)',
         '<(OBJ_PATH)<(OBJ_SEPARATOR)node_constants.<(OBJ_SUFFIX)',
         '<(OBJ_PATH)<(OBJ_SEPARATOR)node_revert.<(OBJ_SUFFIX)',
diff --git a/src/node_url.cc b/src/node_url.cc
@@ -131,10 +131,6 @@ enum url_error_cb_args {
     return str.length() >= 2 && name(str[0], str[1]);                         \
   }
 
-CHAR_TEST(8, IsLowerCaseASCII, (ch >='a' && ch <= 'z'))
-
-CHAR_TEST(8, IsLowerCaseASCII, (ch >='a' && ch <= 'z'))
-
 // https://infra.spec.whatwg.org/#ascii-tab-or-newline
 CHAR_TEST(8, IsASCIITabOrNewline, (ch == '\t' || ch == '\n' || ch == '\r'))
 
@@ -865,9 +861,7 @@ static url_host_type ParseHost(url_host* host,
   if (!stringutils::ContainsNonAscii(buf, strlen(buf))) {
     // Lowercase ASCII domains
     for (size_t n = 0; n < decoded.size(); n++) {
-      if (!IsLowerCaseASCII(decoded[n])) {
-        decoded[n] = ASCIILowercase(decoded[n]);
-      }
+      decoded[n] = ASCIILowercase(decoded[n]);
     }
   } else {
     // Then we have to Unicode IDNA toASCII
diff --git a/src/string_utils.cc b/src/string_utils.cc
diff --git a/src/string_utils.h b/src/string_utils.h
@@ -2,13 +2,59 @@
 #ifndef SRC_STRING_UTILS_H_
 #define SRC_STRING_UTILS_H_
 
-#include "env.h"
-#include "env-inl.h"
-#include "util.h"
+#include <cstddef>
+#include <cstdint>
 
 namespace node {
 namespace stringutils {
-  bool ContainsNonAscii(const char* src, size_t len);
+  inline static bool contains_non_ascii_slow(const char* buf, size_t len) {
+    for (size_t i = 0; i < len; ++i) {
+      if (buf[i] & 0x80)
+        return true;
+      }
+      return false;
+  }
+
+  inline bool ContainsNonAscii(const char* src, size_t len) {
+    if (len < 16) {
+        return contains_non_ascii_slow(src, len);
+    }
+
+    const unsigned bytes_per_word = sizeof(uintptr_t);
+    const unsigned align_mask = bytes_per_word - 1;
+    const unsigned unaligned = reinterpret_cast<uintptr_t>(src) & align_mask;
+
+    if (unaligned > 0) {
+        const unsigned n = bytes_per_word - unaligned;
+        if (contains_non_ascii_slow(src, n))
+        return true;
+        src += n;
+        len -= n;
+    }
+
+
+    #if defined(_WIN64) || defined(_LP64)
+    const uintptr_t mask = 0x8080808080808080ll;
+    #else
+    const uintptr_t mask = 0x80808080l;
+    #endif
+
+    const uintptr_t* srcw = reinterpret_cast<const uintptr_t*>(src);
+
+    for (size_t i = 0, n = len / bytes_per_word; i < n; ++i) {
+        if (srcw[i] & mask)
+        return true;
+    }
+
+    const unsigned remainder = len & align_mask;
+    if (remainder > 0) {
+        const size_t offset = len - remainder;
+        if (contains_non_ascii_slow(src + offset, remainder))
+        return true;
+    }
+
+    return false;
+  }
 }  // namespace stringutils
 }  // namespace node
 
