Support for specifying an output directory (issue #106)

lib/src/directive_parser.dart

Index: lib/src/directive_parser.dart
diff --git a/lib/src/directive_parser.dart b/lib/src/directive_parser.dart
new file mode 100644
index 0000000000000000000000000000000000000000..f5e3b96120e3a77380191a84ed6cbd18c75b536e
--- /dev/null
+++ b/lib/src/directive_parser.dart
@@ -0,0 +1,367 @@
+// Copyright (c) 2012, 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.
+
+/**
+ * A mini parser that extracts top-level directives (library, imports, exports,
+ * and parts) from a dart source file.
+ */
+library directive_parser;
+
+import 'info.dart' show DartCodeInfo, DartDirectiveInfo;
+import 'messages.dart' show Messages;
+
+/** Parse and extract top-level directives from [code]. */
+DartCodeInfo parseDartCode(String code, Messages messages) {
+  return new _DirectiveParser(messages).parse(code);
+}
+
+/** A parser that extracts top-level directives. */
+// TODO(sigmund): add source-span to error messages
+class _DirectiveParser {
+  /** Tokenizer used to parse the input until the end of the directives. */
+  _DirectiveTokenizer tokenizer;
+
+  /** Extracted library identifier, if any. */
+  String libraryName;
+
+  /** Extracted part-of identifier, if any. */
+  String partName;
+
+  /** Extracted imports, exports, and parts, if any. */
+  List<DartDirectiveInfo> directives = <DartDirectiveInfo>[];
+
+  /** Helper for reporting error messages. */
+  Messages messages;
+
+  /** Last token read by the parser. */
+  Token token;
+
+  _DirectiveParser(this.messages);
+
+  /** Parse and extract directives from [code]. */
+  DartCodeInfo parse(String code) {
+    tokenizer = new _DirectiveTokenizer(code);
+    parseTopLevel();
+    return new DartCodeInfo(libraryName, partName, directives,
+        code.substring(token.start));
+  }
+
+  /**
+   * Parse top-level directives and comments, but unlike normal Dart code, stop
+   * as soon as we find actual code.
+   */
+  void parseTopLevel() {
+    token = tokenizer.next();
+    while (token.kind != Token.EOF) {
+      if (token.kind == Token.IDENTIFIER) {
+        if (token.value == 'library') {
+          parseLibrary();
+        } else if (token.value == 'part') {
+          parsePart();
+        } else if (token.value == 'import') {
+          parseImport();
+        } else if (token.value == 'export') {
+          parseExport();
+        } else {
+          break;
+        }
+      } else if (token.kind != Token.COMMENT) {
+        break;
+      }
+      token = tokenizer.next();
+    }
+  }
+
+  /** Parse library declarations: 'library foo.bar;' */
+  parseLibrary() {
+    libraryName = parseQualifiedName();

[Jennifer Messerly, 2012/10/25 04:26:25]

fyi, this is not allowed by Dart language spec when I checked last week... it
only had an identifier and VM didn't like foo.bar :(

[Siggi Cherem (dart-lang), 2012/10/25 05:17:29]

Good question - it's not perfectly clear. The spec grammar says only identifier,
the vm crashes, but the text under the gramar (here
http://www.dartlang.org/docs/spec/latest/dart-language-specification.html#id....)

says:
A named library begins with the word library (possibly prefaced by any
applicable metadata annotations), followed by a **qualified identifier** that
gives the name of the library.

dart2js/dart_analyer accept it too, so I'll keep it in.

+    expectToken(Token.SEMICOLON);
+  }
+
+  /**
+   * Parse either a part declaration or part inclusions. For instance,
+   *     part of foo.bar;
+   * or
+   *     part "foo";
+   */
+  parsePart() {
+    token = tokenizer.next();
+    if (token.kind == Token.IDENTIFIER && token.value == 'of') {
+      partName = parseQualifiedName();
+    } else if (token.kind == Token.STRING) {
+      directives.add(new DartDirectiveInfo('part', token.value));
+      token = tokenizer.next();
+    } else {
+      messages.error('unexpected token: ${token}', null);
+    }
+    expectToken(Token.SEMICOLON);
+  }
+
+  /** Parse a qualified name, such as `one.two.three`. */
+  parseQualifiedName() {
+    List<String> segments = [];
+    while (true) {
+      token = tokenizer.next();
+      if (token.kind != Token.IDENTIFIER) {
+        messages.error('invalid qualified name: $token', null);
+        return null;
+      }
+      segments.add(token.value);
+      token = tokenizer.next();
+      if (token.kind == Token.SEMICOLON) break;
+      if (token.kind != Token.DOT) {
+        messages.error('invalid qualified name: $token', null);
+        return null;
+      }
+    }
+    return Strings.join(segments, '.');
+  }
+
+  /** Parse an import, with optional prefix and show/hide combinators. */
+  parseImport() {
+    token = tokenizer.next();
+    if (token.kind != Token.STRING) {
+      messages.error('expected an import url, but found ${token}', null);

[Jennifer Messerly, 2012/10/25 04:26:25]

would be good to include the file name

[Siggi Cherem (dart-lang), 2012/10/25 05:17:29]

Done.

+      return;
+    }
+    var uri = token.value;

[Jennifer Messerly, 2012/10/25 04:26:25]

does the spec allow sequential string concat? i.e. to break up a long line

[Siggi Cherem (dart-lang), 2012/10/25 05:17:29]

fixed. the spec allows it, but the vm crashes on it.

+
+    // Parse the optional prefix.
+    token = tokenizer.next();
+    var prefix;
+    if (token.kind == Token.IDENTIFIER && token.value == 'as') {
+      token = tokenizer.next();
+      if (token.kind != Token.IDENTIFIER) {
+        messages.error('expected an identifier as prefix, but found ${token}',
+            null);
+        return;
+      }
+      prefix = token.value;
+      token = tokenizer.next();
+    }
+
+    // Parse the optional show/hide combinators.
+    var hide;
+    var show;
+    while (token.kind == Token.IDENTIFIER) {
+      if (token.value == 'hide') {
+        if (hide == null) hide = [];
+        hide.addAll(parseIdentifierList());
+      } else if (token.value == 'show') {
+        if (show == null) show = [];
+        show.addAll(parseIdentifierList());
+      }
+    }
+
+    expectToken(Token.SEMICOLON);
+    directives.add(new DartDirectiveInfo('import', uri, prefix, hide, show));
+  }
+
+  /** Parse an export, with optional show/hide combinators. */
+  parseExport() {
+    token = tokenizer.next();
+    if (token.kind != Token.STRING) {
+      messages.error('expected an export url, but found ${token}', null);
+      return;
+    }
+    var uri = token.value;
+
+    // Parse the optional show/hide combinators.
+    token = tokenizer.next();
+    var hide;
+    var show;
+    while (token.kind == Token.IDENTIFIER) {
+      if (token.value == 'hide') {
+        if (hide == null) hide = [];
+        hide.addAll(parseIdentifierList());
+      } else if (token.value == 'show') {
+        if (show == null) show = [];
+        show.addAll(parseIdentifierList());
+      }
+    }
+
+    expectToken(Token.SEMICOLON);
+    directives.add(new DartDirectiveInfo('export', uri, null, hide, show));
+  }
+
+  /** Parse a list of identifiers of the form `id1, id2, id3` */
+  List<String> parseIdentifierList() {
+    var list = [];
+    do {
+      token = tokenizer.next();
+      if (!expectToken(Token.IDENTIFIER)) return list;
+      list.add(token.value);
+      token = tokenizer.next();
+    } while (token.kind == Token.COMMA);
+    return list;
+  }
+
+  /** Report an error if the last token is not of the expected kind. */
+  bool expectToken(int kind) {
+    if (token.kind != kind) {
+      messages.error(
+          'expected <${Token.KIND_NAMES[kind]}>, but got ${token}', null);
+      return false;
+    }
+    return true;
+  }
+}
+
+/** Set of tokens that we parse out of the dart code. */
+class Token {
+  /** Kind of token, one of the constants below. */
+  final int kind;
+
+  /** Value in the token (filled only for identifiers and strings). */
+  final String value;
+
+  /** Start location for the token in the input string. */
+  final int start;
+
+  /** End location for the token in the input string. */
+  final int end;
+
+  const Token(this.kind, this.start, this.end, [this.value]);
+
+  toString() => '<#Token ${KIND_NAMES[kind]}, $value>';
+
+  static const int COMMENT = 0;
+  static const int STRING = 1;
+  static const int IDENTIFIER = 2;
+  static const int SEMICOLON = 3;
+  static const int DOT = 4;
+  static const int COMMA = 5;
+  static const int EOF = 6;
+  static const List<String> KIND_NAMES =
+      const ['comment', 'string', 'id', 'semicolon', 'dot', 'comma', 'eof'];
+
+}
+
+/**
+ * A simple tokenizer that understands comments, identifiers, strings,
+ * separators, and practically nothing else.
+ */
+class _DirectiveTokenizer {
+  int pos = 0;
+  String _data;
+
+  _DirectiveTokenizer(this._data);
+
+  /** Return the next token. */
+  Token next() {
+    while (true) {
+      if (pos >= _data.length) return new Token(Token.EOF, pos, pos);
+      if (!isWhiteSpace(peek())) break;
+      nextChar();
+    }
+
+    var c = peek();
+    switch (c) {
+      case _SLASH:
+        if (peek(1) == _SLASH) return lineComment();
+        if (peek(1) == _STAR) return blockComment();
+        break;
+      case _SINGLE_QUOTE:
+      case _DOUBLE_QUOTE:
+        return string();
+      case _SEMICOLON:
+        pos++;
+        return new Token(Token.SEMICOLON, pos - 1, pos);
+      case _DOT:
+        pos++;
+        return new Token(Token.DOT, pos - 1, pos);
+      case _COMMA:
+        pos++;
+        return new Token(Token.COMMA, pos - 1, pos);
+      default:
+        if (isIdentifierStart(c)) return identifier();
+        break;
+    }
+    return new Token(Token.EOF, pos, pos);
+  }
+
+  int nextChar() => _data.charCodeAt(pos++);
+  int peek([int skip = 0]) => _data.charCodeAt(pos + skip);
+
+  /** Advance parsing until the end of a string (no tripple quotes allowed). */
+  Token string() {

[Jennifer Messerly, 2012/10/25 04:26:25]

does spec allow raw or triple quoted strings? interpolation (gosh I hope not)?

I can't see any value in implementing those as they would be useless in imports,
but would be worth a note...

[Siggi Cherem (dart-lang), 2012/10/25 05:17:29]

ok added a note - seems that yes, multiline strings are allowed. Interpolation
not, the strings must be consts.

+    int start = pos;
+    int startQuote = nextChar();
+    bool escape = false;
+    while (true) {
+      if (pos >= _data.length) return new Token(Token.EOF, start, pos);
+      int c = nextChar();
+      if (c == startQuote && !escape) break;
+      escape = !escape && c == _BACKSLASH;
+    }
+    return new Token(Token.STRING, start, pos,
+        _data.substring(start + 1, pos - 1));
+  }
+
+  /** Advance parsing until the end of an identifier. */
+  Token identifier() {
+    int start = pos;
+    while (pos < _data.length && isIdentifierChar(peek())) pos++;
+    return new Token(Token.IDENTIFIER, start, pos, _data.substring(start, pos));
+  }
+
+  /** Advance parsing until the end of a line comment. */
+  Token lineComment() {
+    int start = pos;
+    while (pos < _data.length && peek() != _LF) pos++;
+    return new Token(Token.COMMENT, start, pos);
+  }
+
+  /** Advance parsing until the end of a block comment (nesting is allowed). */
+  Token blockComment() {
+    var start = pos;
+    var commentNesting = 0;
+    pos += 2;
+    while (pos < _data.length) {
+      if (peek() == _STAR && peek(1) == _SLASH) {
+        pos += 2;
+        if (commentNesting == 0) break;
+        commentNesting--;
+      } else if (peek() == _SLASH && peek(1) == _STAR) {
+        pos += 2;
+        commentNesting++;
+      } else {
+        pos++;
+      }
+    }
+    return new Token(Token.COMMENT, start, pos);
+  }
+
+  bool isWhiteSpace(int c) => c == _LF || c == _SPACE || c == _CR || c == _TAB;
+  bool isIdentifierStart(int c) => c == _UNDERSCORE || isLetter(c);
+  bool isIdentifierChar(int c) => isIdentifierStart(c) || isNumber(c);
+  bool isNumber(int c) => c >= _ZERO && c <= _NINE;
+  bool isLetter(int c) =>
+      (c >= _LOWER_A && c <= _LOWER_Z) ||
+      (c >= _UPPER_A && c <= _UPPER_Z);
+
+
+  // The following constant character values are used for tokenizing.
+
+  static const int _TAB = 9;
+  static const int _LF = 10;
+  static const int _CR = 13;
+  static const int _SPACE = 32;
+  static const int _DOUBLE_QUOTE = 34; // "
+  static const int _DOLLAR = 36;       // $
+  static const int _SINGLE_QUOTE = 39; // '
+  static const int _STAR = 42;         // *
+  static const int _COMMA = 44;        // ,
+  static const int _DOT = 46;          // .
+  static const int _SLASH = 47;        // /
+  static const int _ZERO = 48;         // 0
+  static const int _NINE = 57;         // 9
+  static const int _SEMICOLON = 59;    // ;
+  static const int _UPPER_A = 65;      // A
+  static const int _UPPER_Z = 90;      // Z
+  static const int _BACKSLASH = 92;    // \
+  static const int _UNDERSCORE = 95;   // _
+  static const int _LOWER_A = 97;      // a
+  static const int _LOWER_Z = 122;     // z
+}