Write custom URI-parser, to avoid using regexp.

sdk/lib/core/uri.dart

Index: sdk/lib/core/uri.dart
diff --git a/sdk/lib/core/uri.dart b/sdk/lib/core/uri.dart
index 41130ad08cea677b05826f3f4574e48842324bf7..38774759593cffcf33fd3eafe50645e4f62bc0ff 100644
--- a/sdk/lib/core/uri.dart
+++ b/sdk/lib/core/uri.dart
@@ -118,17 +118,183 @@ class Uri {
   /**
    * Creates a new URI object by parsing a URI string.
    */
-  static Uri parse(String uri) => new Uri._fromMatch(_splitRe.firstMatch(uri));
-
-  Uri._fromMatch(Match m) :
-    this(scheme: _makeScheme(_emptyIfNull(m[_COMPONENT_SCHEME])),
-         userInfo: _emptyIfNull(m[_COMPONENT_USER_INFO]),
-         host: _eitherOf(
-         m[_COMPONENT_HOST], m[_COMPONENT_HOST_IPV6]),
-         port: _parseIntOrZero(m[_COMPONENT_PORT]),
-         path: _emptyIfNull(m[_COMPONENT_PATH]),
-         query: _emptyIfNull(m[_COMPONENT_QUERY_DATA]),
-         fragment: _emptyIfNull(m[_COMPONENT_FRAGMENT]));
+  static Uri parse(String uri) {

[Søren Gjesse, 2014/02/24 09:12:41]

I think it could be a good idea to include the regexp from the spec in a
documentation comment here.

[Anders Johnsen, 2014/02/24 12:44:33]

Done.

+    // This parsing will not validate percent-encoding, IPv6, etc. When done
+    // it'll call `new Uri(...)` that will perform these validations. This is

[Lasse Reichstein Nielsen, 2014/02/24 13:02:27]

it'll -> it will
that -> which

[Anders Johnsen, 2014/02/24 14:26:49]

Done.

+    // purely splitting up the uri string into components.

[Lasse Reichstein Nielsen, 2014/02/24 13:02:27]

uri -> URI.

[Anders Johnsen, 2014/02/24 14:26:49]

Done.

+    bool isSchemeCharacter(int ch) {

[Lasse Reichstein Nielsen, 2014/02/24 13:02:27]

Consider making this a static private method instead.

[Anders Johnsen, 2014/02/24 14:26:49]

Done.

+      return ch < 128 && ((_schemeTable[ch >> 4] & (1 << (ch & 0x0f))) != 0);
+    }
+
+    bool isRegName(int ch) {
+      return ch < 128 && ((_regNameTable[ch >> 4] & (1 << (ch & 0x0f))) != 0);
+    }
+
+    List<int> codeUnits = uri.codeUnits;
+    int length = codeUnits.length;
+    int index = 0;
+
+    int schemeEndIndex = 0;
+
+    if (length == 0) {
+      return new Uri();
+    }
+
+    if (codeUnits[0] != '/'.codeUnitAt(0)) {

[Søren Gjesse, 2014/02/24 09:12:41]

Use _SLASH (see frequently used character codes at line 1483.

[Anders Johnsen, 2014/02/24 12:44:33]

Done.

[Lasse Reichstein Nielsen, 2014/02/24 13:02:27]

Unless you *know* that this is optimized away by all compilers (and it isn't),
replace it with a constant ("const CHAR_SLASH = 0x..;", which may be local to
function or static and private).

[Anders Johnsen, 2014/02/24 14:26:49]

Any const static members should be compiled away?

+      // Can be scheme.
+      while (index < length) {

[Søren Gjesse, 2014/02/24 09:12:41]

Please provide some more comments, e.g. "Look for valid scheme ending in ':'. If
not found assume schema-less URI."

[Anders Johnsen, 2014/02/24 12:44:33]

Done.

+
+        int codeUnit = codeUnits[index++];
+        if (!isSchemeCharacter(codeUnit)) {
+          if (codeUnit == ':'.codeUnitAt(0)) {

[Søren Gjesse, 2014/02/24 09:12:41]

_COLON, and more below.

[Anders Johnsen, 2014/02/24 12:44:33]

Done.

+            schemeEndIndex = index;
+          } else {
+            // Back up one char, as we meet a special char.

[Søren Gjesse, 2014/02/24 09:12:41]

meet -> met?

[Anders Johnsen, 2014/02/24 12:44:33]

Done.

[Lasse Reichstein Nielsen, 2014/02/24 13:02:27]

as -> since

[Anders Johnsen, 2014/02/24 14:26:49]

Done.

+            index--;
+          }
+          break;
+        }
+      }
+    }
+
+    int userInfoEndIndex = -1;
+    int portIndex = -1;
+    int authorityEndIndex = schemeEndIndex;
+    // If we see '//', it must be a authority.

[Søren Gjesse, 2014/02/24 09:12:41]

it must -> there must
a -> an

[Anders Johnsen, 2014/02/24 12:44:33]

Done.

+    if (authorityEndIndex == index &&
+        authorityEndIndex + 1 < length &&
+        codeUnits[authorityEndIndex] == '/'.codeUnitAt(0) &&
+        codeUnits[authorityEndIndex + 1] == '/'.codeUnitAt(0)) {
+      // Skip '//'.
+      authorityEndIndex += 2;
+      // It can both be host and userInfo.
+      while (authorityEndIndex < length) {
+        int codeUnit = codeUnits[authorityEndIndex++];
+        if (!isRegName(codeUnit)) {
+          if (codeUnit == '['.codeUnitAt(0)) {
+            // IPv6. Skip to '['.

[Søren Gjesse, 2014/02/24 09:12:41]

'[' -> ']'

[Anders Johnsen, 2014/02/24 12:44:33]

Done.

+            authorityEndIndex = codeUnits.indexOf(']'.codeUnitAt(0),
+                                                  authorityEndIndex) + 1;
+            if (authorityEndIndex == 0) {
+              throw new FormatException("Bad end of IPv6 host");
+            }
+          } else if (portIndex == -1 && codeUnit == ':'.codeUnitAt(0)) {
+            // First time ':'.
+            portIndex = authorityEndIndex;
+          } else if (codeUnit == '@'.codeUnitAt(0) ||
+                     codeUnit == ':'.codeUnitAt(0)) {
+            // Second time ':' or first '@'. Must be userInfo.
+            userInfoEndIndex = codeUnits.indexOf('@'.codeUnitAt(0),
+                                                 authorityEndIndex - 1);
+            // Not found. Must be path then.
+            if (userInfoEndIndex == -1) {
+              authorityEndIndex = index;
+              break;
+            }
+            portIndex = -1;
+            authorityEndIndex = userInfoEndIndex + 1;
+            // Now it can only be host:port.
+            while (authorityEndIndex < length) {
+              int codeUnit = codeUnits[authorityEndIndex++];
+              if (!isRegName(codeUnit)) {
+                if (codeUnit == '['.codeUnitAt(0)) {

[Søren Gjesse, 2014/02/24 09:12:41]

Refactor next 5 lines to a local function

authorityEndIndex = ipV6Address(codeUnits, authorityEndIndex);

and use above as well

[Anders Johnsen, 2014/02/24 12:44:33]

Done.

+                  authorityEndIndex = codeUnits.indexOf(']'.codeUnitAt(0),
+                                                        authorityEndIndex) + 1;
+                  if (authorityEndIndex == 0) {
+                    throw new FormatException("Bad end of IPv6 host");
+                  }
+                } else if (codeUnit == ':'.codeUnitAt(0)) {
+                  if (portIndex != -1) {
+                    throw new FormatException("Double port in host");
+                  }
+                  portIndex = authorityEndIndex;
+                } else {
+                  authorityEndIndex--;
+                  break;
+                }
+              }
+            }
+            break;
+          } else {
+            authorityEndIndex--;
+            break;
+          }
+        }
+      }
+    } else {
+      authorityEndIndex = schemeEndIndex;
+    }
+
+    // At path now.
+    int pathEndIndex = authorityEndIndex;
+    while (pathEndIndex < length) {
+      int codeUnit = codeUnits[pathEndIndex++];
+      if (codeUnit == '?'.codeUnitAt(0) ||
+          codeUnit == '#'.codeUnitAt(0)) {
+        pathEndIndex--;
+        break;
+      }
+    }
+
+    // Maybe query.
+    int queryEndIndex = pathEndIndex;
+    if (queryEndIndex < length &&
+        codeUnits[queryEndIndex] == '?'.codeUnitAt(0)) {
+      while (queryEndIndex < length) {
+        int codeUnit = codeUnits[queryEndIndex++];
+        if (codeUnit == '#'.codeUnitAt(0)) {
+          queryEndIndex--;
+          break;
+        }
+      }
+    }
+
+    var scheme = null;
+    if (schemeEndIndex > 0) {
+      scheme = uri.substring(0, schemeEndIndex - 1);
+    }
+
+    var host = "";
+    var userInfo = "";
+    var port = 0;
+    if (schemeEndIndex != authorityEndIndex) {
+      int startIndex = schemeEndIndex + 2;
+      if (userInfoEndIndex > 0) {
+        userInfo = uri.substring(startIndex, userInfoEndIndex);
+        startIndex = userInfoEndIndex + 1;
+      }
+      if (portIndex > 0) {
+        var portStr = uri.substring(portIndex, authorityEndIndex);
+        try {
+          port = int.parse(portStr);
+        } catch (_) {
+          throw new FormatException("Invalid port: '$portStr'");
+        }
+        host = uri.substring(startIndex, portIndex - 1);
+      } else {
+        host = uri.substring(startIndex, authorityEndIndex);
+      }
+    }
+
+    var path = uri.substring(authorityEndIndex, pathEndIndex);
+    var query = "";
+    if (pathEndIndex < queryEndIndex) {
+      query = uri.substring(pathEndIndex + 1, queryEndIndex);
+    }
+    var fragment = "";
+    // If queryEndIndex is not at end (length), there is a fragment.
+    if (queryEndIndex < length) {
+      fragment = uri.substring(queryEndIndex + 1, length);
+    }
+
+    return new Uri(scheme: scheme,
+                   userInfo: userInfo,
+                   host: host,
+                   port: port,
+                   path: path,
+                   query: query,
+                   fragment: fragment);
+  }
 
   /**
    * Creates a new URI from its components.
@@ -719,58 +885,6 @@ class Uri {
     return result.toString();
   }
 
-  static String _emptyIfNull(String val) => val != null ? val : '';
-
-  static int _parseIntOrZero(String val) {
-    if (val != null && val != '') {
-      return int.parse(val);
-    } else {
-      return 0;
-    }
-  }
-
-  static String _eitherOf(String val1, String val2) {
-    if (val1 != null) return val1;
-    if (val2 != null) return val2;
-    return '';
-  }
-
-  // NOTE: This code was ported from: closure-library/closure/goog/uri/utils.js
-  static final RegExp _splitRe = new RegExp(
-      '^'
-      '(?:'
-        '([^:/?#]+)'                    // scheme - ignore special characters
-                                        // used by other URL parts such as :,
-                                        // ?, /, #, and .
-      ':)?'
-      '(?://'
-        '(?:([^/?#]*)@)?'               // userInfo
-        '(?:'
-          r'([\w\d\-\u0100-\uffff.%]*)'
-                                        // host - restrict to letters,
-                                        // digits, dashes, dots, percent
-                                        // escapes, and unicode characters.
-          '|'
-          // TODO(ajohnsen): Only allow a max number of parts?
-          r'\[([A-Fa-f0-9:.]*)\])'
-                                        // IPv6 host - restrict to hex,
-                                        // dot and colon.
-        '(?::([0-9]+))?'                // port
-      ')?'
-      r'([^?#[]+)?'                     // path
-      r'(?:\?([^#]*))?'                 // query
-      '(?:#(.*))?'                      // fragment
-      r'$');
-
-  static const _COMPONENT_SCHEME = 1;
-  static const _COMPONENT_USER_INFO = 2;
-  static const _COMPONENT_HOST = 3;
-  static const _COMPONENT_HOST_IPV6 = 4;
-  static const _COMPONENT_PORT = 5;
-  static const _COMPONENT_PATH = 6;
-  static const _COMPONENT_QUERY_DATA = 7;
-  static const _COMPONENT_FRAGMENT = 8;
-
   /**
    * Returns whether the URI is absolute.
    */
@@ -1625,6 +1739,27 @@ class Uri {
                 //              pqrstuvwxyz   ~
       0x47ff];  // 0x70 - 0x7f  1111111111100010
 
+  // Characters allowed in the reg-name as of RFC 3986.
+  // RFC 3986 Apendix A
+  // reg-name = *( unreserved / pct-encoded / sub-delims )
+  static const _regNameTable = const [
+                //             LSB            MSB
+                //              |              |
+      0x0000,   // 0x00 - 0x0f  0000000000000000
+      0x0000,   // 0x10 - 0x1f  0000000000000000
+                //               !  $%&'()*+,-.
+      0x7ff2,   // 0x20 - 0x2f  0100111111111110
+                //              0123456789 ; =
+      0x2bff,   // 0x30 - 0x3f  1111111111010100
+                //               ABCDEFGHIJKLMNO
+      0xfffe,   // 0x40 - 0x4f  0111111111111111
+                //              PQRSTUVWXYZ    _
+      0x87ff,   // 0x50 - 0x5f  1111111111100001
+                //               abcdefghijklmno
+      0xfffe,   // 0x60 - 0x6f  0111111111111111
+                //              pqrstuvwxyz   ~
+      0x47ff];  // 0x70 - 0x7f  1111111111100010
+
   // Characters allowed in the path as of RFC 3986.
   // RFC 3986 section 3.3.
   // pchar = unreserved / pct-encoded / sub-delims / ":" / "@"