Create a glob package.

pkg/glob/lib/src/parser.dart

+// Copyright (c) 2014, the Dart project authors.  Please see the AUTHORS file
+// for details. All rights reserved. Use of this source code is governed by a
+// BSD-style license that can be found in the LICENSE file.
+
+library glob.single_component;
+
+import 'package:path/path.dart' as p;
+import 'package:string_scanner/string_scanner.dart';
+
+import 'ast.dart';
+import 'utils.dart';
+
+/// A parser for globs.
+class Parser {
+  /// The scanner used to scan the source.
+  final StringScanner _scanner;
+
+  /// The path context for the glob.
+  final p.Context _context;
+
+  Parser(String component, this._context)
+      : _scanner = new StringScanner(component);
+
+  /// Parses an entire glob.
+  SequenceNode parse() => _parseSequence();
+
+  /// Parses a [SequenceNode].
+  ///
+  /// If [inOptions] is true, this is parsing within an [OptionsNode].
+  SequenceNode _parseSequence({bool inOptions: false}) {
+    var nodes = [];
+
+    while (!_scanner.isDone) {
+      if (inOptions && (_scanner.matches(',') || _scanner.matches('}'))) break;
+      nodes.add(_parseNode(inOptions: inOptions));
+    }
+
+    return new SequenceNode(nodes);

[Bob Nystrom, 2014/08/27 00:42:04]

This allows an empty sequence intentionally, right? Document that.

[nweiz, 2014/08/27 01:36:33]

Eh, it should probably error for an empty sequence. Added that and a test.

+  }
+
+  /// Parses an [AstNode].
+  ///
+  /// If [inOptions] is true, this is parsing within an [OptionsNode].
+  AstNode _parseNode({bool inOptions: false}) {
+    var star = _parseStar();
+    if (star != null) return star;
+
+    var anyChar = _parseAnyChar();
+    if (anyChar != null) return anyChar;
+
+    var range = _parseRange();
+    if (range != null) return range;
+
+    var options = _parseOptions();
+    if (options != null) return options;
+
+    return _parseLiteral(inOptions: inOptions);
+  }
+
+  /// Tries to parse a [StarNode] or a [DoubleStarNode].
+  ///
+  /// Returns `null` if there's not one to parse.
+  AstNode _parseStar() {
+    if (!_scanner.scan('*')) return null;
+    return _scanner.scan('*') ? new DoubleStarNode(_context) : new StarNode();
+  }
+
+  /// Tries to parse an [AnyCharNode].
+  ///
+  /// Returns `null` if there's not one to parse.
+  AstNode _parseAnyChar() {
+    if (!_scanner.scan('?')) return null;
+    return new AnyCharNode();
+  }
+
+  /// Tries to parse an [RangeNode].
+  ///
+  /// Returns `null` if there's not one to parse.
+  AstNode _parseRange() {
+    if (!_scanner.scan('[')) return null;
+    if (_scanner.matches(']')) _scanner.error('unexpected "]".');
+    var negated = _scanner.scan('!') || _scanner.scan('^');
+
+    var ranges = [];
+    while (!_scanner.scan(']')) {
+      // Allow a backslash to escape a character.
+      _scanner.scan('\\');
+      var char = _scanner.readChar();
+
+      if (_scanner.scan('-')) {
+        if (_scanner.matches('-') || _scanner.matches(']')) {

[Bob Nystrom, 2014/08/27 00:42:04]

'?' or '^' too?

[nweiz, 2014/08/27 01:36:33]

No, something like [#-?] is fine.

+          _scanner.error('unexpected "${_scanner.lastMatch[0]}".');
+        }
+
+        // Allow a backslash to escape a character.
+        _scanner.scan('\\');
+
+        var end = _scanner.readChar();
+        if (end < char) {
+          _scanner.error("Range out of order.",
+              position: _scanner.position - 3, length: 3);

[Bob Nystrom, 2014/08/27 00:42:04]

This doesn't take into account '\'.

[nweiz, 2014/08/27 01:36:33]

Done.

+        }
+        ranges.add(new Range(char, end));
+      } else {
+        ranges.add(new Range.singleton(char));
+      }
+    }
+
+    return new RangeNode(ranges, negated: negated);
+  }
+
+  /// Tries to parse an [OptionsNode].
+  ///
+  /// Returns `null` if there's not one to parse.
+  AstNode _parseOptions() {
+    if (!_scanner.scan('{')) return null;
+    if (_scanner.matches('}')) _scanner.error('unexpected "}".');
+
+    var options = [_parseSequence(inOptions: true)];
+    while (_scanner.scan(',')) {

[Bob Nystrom, 2014/08/27 00:42:04]

Use a do/while here.

[nweiz, 2014/08/27 01:36:33]

Done.

+      options.add(_parseSequence(inOptions: true));
+    }
+    if (options.length == 1) _scanner.expect(',');

[Bob Nystrom, 2014/08/27 00:42:04]

What's this for?

[nweiz, 2014/08/27 01:36:33]

It disallows single-option groups. Added a comment.

+    _scanner.expect('}');
+
+    return new OptionsNode(options);
+  }
+
+  /// Parses a [LiteralNode].
+  AstNode _parseLiteral({bool inOptions: false}) {
+    var regExp = new RegExp(
+        inOptions ? r'[^*{[?\\}\],()]*' : r'[^*{[?\\}\]()]*');

[Bob Nystrom, 2014/08/27 00:42:04]

Document this.

[nweiz, 2014/08/27 01:36:33]

Done.

+
+    _scanner.scan(regExp);
+    var buffer = new StringBuffer()..write(_scanner.lastMatch[0]);
+
+    while (_scanner.scan('\\')) {
+      buffer.writeCharCode(_scanner.readChar());
+      _scanner.scan(regExp);

[Bob Nystrom, 2014/08/27 00:42:04]

What about multiple sequential escaped characters?

[nweiz, 2014/08/27 01:36:33]

[regExp] just matches zero characters here.

+      buffer.write(_scanner.lastMatch[0]);
+    }
+
+    for (var char in const [']', '(', ')']) {
+      if (_scanner.matches(char)) _scanner.error('unexpected "$char"');
+    }
+    if (!inOptions && _scanner.matches('}')) _scanner.error('unexpected "}"');
+
+    return new LiteralNode(buffer.toString(), _context);
+  }
+}