Stop copying tokens in incremental scanner

pkg/analyzer/lib/src/generated/incremental_scanner.dart

Index: pkg/analyzer/lib/src/generated/incremental_scanner.dart
diff --git a/pkg/analyzer/lib/src/generated/incremental_scanner.dart b/pkg/analyzer/lib/src/generated/incremental_scanner.dart
index 75df98d49fa204d02355d6b7f291d19d98178576..04072bcbdfd3a5b68bbf66d79478112e1f4f4b3a 100644
--- a/pkg/analyzer/lib/src/generated/incremental_scanner.dart
+++ b/pkg/analyzer/lib/src/generated/incremental_scanner.dart
@@ -12,14 +12,28 @@ import 'source.dart';
 import 'utilities_collection.dart' show TokenMap;
 
 /**
- * Instances of the class `IncrementalScanner` implement a scanner that scans a subset of a
- * string and inserts the resulting tokens into the middle of an existing token stream.
+ * An `IncrementalScanner` is a scanner that scans a subset of a string and
+ * inserts the resulting tokens into the middle of an existing token stream.
  */
-class IncrementalScanner extends Scanner {
+class IncrementalScanner {
+  /**
+   * The source being scanned.
+   */
+  final Source source;
+
   /**
    * The reader used to access the characters in the source.
    */
-  CharacterReader _reader;
+  final CharacterReader reader;
+
+  /**
+   * The error listener that will be informed of any errors that are found
+   * during the scan.
+   *
+   * TODO(brianwilkerson) Replace this with a list of errors so that we can
+   * update the errors.
+   */
+  final AnalysisErrorListener errorListener;
 
   /**
    * A map from tokens that were copied to the copies of the tokens.
@@ -46,17 +60,11 @@ class IncrementalScanner extends Scanner {
   bool _hasNonWhitespaceChange = false;
 
   /**
-   * Initialize a newly created scanner.
-   *
-   * @param source the source being scanned
-   * @param reader the character reader used to read the characters in the source
-   * @param errorListener the error listener that will be informed of any errors that are found
-   */
-  IncrementalScanner(Source source, CharacterReader reader,
-      AnalysisErrorListener errorListener)
-      : super(source, reader, errorListener) {
-    this._reader = reader;
-  }
+   * Initialize a newly created scanner to scan characters within the given
+   * [source]. The content of the source can be read using the given [reader].
+   * Any errors that are found will be reported to the given [errorListener].
+   */
+  IncrementalScanner(this.source, this.reader, this.errorListener);
 
   /**
    * Return `true` if there were any tokens either added or removed (or both) as a result of
@@ -92,97 +100,87 @@ class IncrementalScanner extends Scanner {
   TokenMap get tokenMap => _tokenMap;
 
   /**
-   * Given the stream of tokens scanned from the original source, the modified source (the result of
-   * replacing one contiguous range of characters with another string of characters), and a
-   * specification of the modification that was made, return a stream of tokens scanned from the
-   * modified source. The original stream of tokens will not be modified.
+   * Given the [stream] of tokens scanned from the original source, the modified
+   * source (the result of replacing one contiguous range of characters with
+   * another string of characters), and a specification of the modification that
+   * was made, update the token stream to reflect the modified source. Return
+   * the first token in the updated token stream.
    *
-   * @param originalStream the stream of tokens scanned from the original source
-   * @param index the index of the first character in both the original and modified source that was
-   *          affected by the modification
-   * @param removedLength the number of characters removed from the original source
-   * @param insertedLength the number of characters added to the modified source
-   */
-  Token rescan(Token originalStream, int index, int removedLength,
-      int insertedLength) {
-    //
-    // Copy all of the tokens in the originalStream whose end is less than the
-    // replacement start. (If the replacement start is equal to the end of an
-    // existing token, then it means that the existing token might have been
-    // modified, so we need to rescan it.)
-    //
-    while (originalStream.type != TokenType.EOF && originalStream.end < index) {
-      originalStream = _copyAndAdvance(originalStream, 0);
+   * The [stream] is expected to be the first non-EOF token in the token stream.
+   *
+   * The modification is specified by the [index] of the first character in both
+   * the original and modified source that was affected by the modification, the
+   * number of characters removed from the original source (the [removedLength])
+   * and the number of characters added to the modified source (the
+   * [insertedLength]).
+   */
+  Token rescan(Token stream, int index, int removedLength, int insertedLength) {
+    Token leftEof = stream.previous;
+    //
+    // Compute the delta between the character index of characters after the
+    // modified region in the original source and the index of the corresponding
+    // character in the modified source.
+    //
+    int delta = insertedLength - removedLength;
+    //
+    // Skip past the tokens whose end is less than the replacement start. (If
+    // the replacement start is equal to the end of an existing token, then it
+    // means that the existing token might have been modified, so we need to
+    // rescan it.)
+    //
+    while (stream.type != TokenType.EOF && stream.end < index) {
+      _tokenMap.put(stream, stream);
+      stream = stream.next;
     }
-    Token oldFirst = originalStream;
-    Token oldLeftToken = originalStream.previous;
-    _leftToken = tail;
+    Token oldFirst = stream;
+    Token oldLeftToken = stream.previous;
+    _leftToken = oldLeftToken;
     //
-    // Skip tokens in the original stream until we find a token whose offset is
-    // greater than the end of the removed region. (If the end of the removed
-    // region is equal to the beginning of an existing token, then it means that
-    // the existing token might have been modified, so we need to rescan it.)
+    // Skip past tokens until we find a token whose offset is greater than the
+    // end of the removed region. (If the end of the removed region is equal to
+    // the beginning of an existing token, then it means that the existing token
+    // might have been modified, so we need to rescan it.)
     //
     int removedEnd = index + (removedLength == 0 ? 0 : removedLength - 1);
-    while (originalStream.type != TokenType.EOF &&
-        originalStream.offset <= removedEnd) {
-      originalStream = originalStream.next;
+    while (stream.type != TokenType.EOF && stream.offset <= removedEnd) {
+      stream = stream.next;
     }
+    //
+    // Figure out which region of characters actually needs to be re-scanned.
+    //
     Token oldLast;
     Token oldRightToken;
-    if (originalStream.type != TokenType.EOF &&
-        removedEnd + 1 == originalStream.offset) {
-      oldLast = originalStream;
-      originalStream = originalStream.next;
-      oldRightToken = originalStream;
+    if (stream.type != TokenType.EOF && removedEnd + 1 == stream.offset) {
+      oldLast = stream;
+      stream = stream.next;
+      oldRightToken = stream;
     } else {
-      oldLast = originalStream.previous;
-      oldRightToken = originalStream;
+      oldLast = stream.previous;
+      oldRightToken = stream;
     }
     //
-    // Compute the delta between the character index of characters after the
-    // modified region in the original source and the index of the corresponding
-    // character in the modified source.
-    //
-    int delta = insertedLength - removedLength;
-    //
     // Compute the range of characters that are known to need to be rescanned.
     // If the index is within an existing token, then we need to start at the
     // beginning of the token.
     //
-    int scanStart = math.min(oldFirst.offset, index);
-    int oldEnd = oldLast.end + delta - 1;
-    int newEnd = index + insertedLength - 1;
-    int scanEnd = math.max(newEnd, oldEnd);
-    //
-    // Starting at the start of the scan region, scan tokens from the
-    // modifiedSource until the end of the just scanned token is greater than or
-    // equal to end of the scan region in the modified source. Include trailing
-    // characters of any token that was split as a result of inserted text,
-    // as in "ab" --> "a.b".
-    //
-    _reader.offset = scanStart - 1;
-    int next = _reader.advance();
-    while (next != -1 && _reader.offset <= scanEnd) {
-      next = bigSwitch(next);
-    }
+    int scanStart = math.max(oldFirst.previous.end, 0);
+    int scanEnd = oldLast.end + delta;
     //
-    // Copy the remaining tokens in the original stream, but apply the delta to
-    // the token's offset.
+    // Starting at the start of the scan region, scan tokens from the modified
+    // source until the end of the just scanned token is greater than or equal
+    // to end of the scan region in the modified source. Include trailing
+    // characters of any token that was split as a result of inserted text, as
+    // in "ab" --> "a.b".
     //
-    if (originalStream.type == TokenType.EOF) {
-      _copyAndAdvance(originalStream, delta);
-      _rightToken = tail;
-      _rightToken.setNextWithoutSettingPrevious(_rightToken);
-    } else {
-      originalStream = _copyAndAdvance(originalStream, delta);
-      _rightToken = tail;
-      while (originalStream.type != TokenType.EOF) {
-        originalStream = _copyAndAdvance(originalStream, delta);
-      }
-      Token eof = _copyAndAdvance(originalStream, delta);
-      eof.setNextWithoutSettingPrevious(eof);
-    }
+    Token replacementStart = _scanRange(scanStart, scanEnd);
+    oldLeftToken.setNext(replacementStart);
+    Token replacementEnd = _findEof(replacementStart).previous;
+    replacementEnd.setNext(stream);
+    //
+    // Apply the delta to the tokens after the last new token.
+    //
+    _updateOffsets(stream, delta);
+    _rightToken = stream;
     //
     // If the index is immediately after an existing token and the inserted
     // characters did not change that original token, then adjust the leftToken
@@ -221,22 +219,7 @@ class IncrementalScanner extends Scanner {
     // parsing.
     // TODO(brianwilkerson) Update the lineInfo.
     //
-    return firstToken;
-  }
-
-  Token _copyAndAdvance(Token originalToken, int delta) {
-    Token copiedToken = originalToken.copy();
-    _tokenMap.put(originalToken, copiedToken);
-    copiedToken.offset += delta;
-    appendToken(copiedToken);
-    Token originalComment = originalToken.precedingComments;
-    Token copiedComment = originalToken.precedingComments;
-    while (originalComment != null) {
-      _tokenMap.put(originalComment, copiedComment);
-      originalComment = originalComment.next;
-      copiedComment = copiedComment.next;
-    }
-    return originalToken.next;
+    return leftEof.next;
   }
 
   /**
@@ -251,4 +234,45 @@ class IncrementalScanner extends Scanner {
       oldToken.type == newToken.type &&
           oldToken.length == newToken.length &&
           oldToken.lexeme == newToken.lexeme;
+
+  /**
+   * Given a [token], return the EOF token that follows the token.
+   */
+  Token _findEof(Token token) {
+    while (token.type != TokenType.EOF) {
+      token = token.next;
+    }
+    return token;
+  }
+
+  /**
+   * Scan the token between the [start] (inclusive) and [end] (exclusive)
+   * offsets.
+   */
+  Token _scanRange(int start, int end) {
+    Scanner scanner = new Scanner(
+        source,
+        new CharacterRangeReader(reader, start, end),
+        errorListener);
+    return scanner.tokenize();
+  }
+
+  /**
+   * Update the offsets of every token from the given [token] to the end of the
+   * stream by adding the given [delta].
+   */
+  void _updateOffsets(Token token, int delta) {
+    while (token.type != TokenType.EOF) {
+      _tokenMap.put(token, token);
+      token.offset += delta;
+      Token comment = token.precedingComments;
+      while (comment != null) {
+        comment.offset += delta;
+        comment = comment.next;
+      }
+      token = token.next;
+    }
+    _tokenMap.put(token, token);
+    token.offset += delta;

[Paul Berry, 2014/11/20 22:41:59]

Does the EOF token ever have precedingComments?  (i.e. because there was a
comment after the last token in the file)

[Brian Wilkerson, 2014/11/20 22:55:02]

Yes, when there are comments at the end of the file. Good catch! I'd like to fix
this in a later CL as well.

+  }
 }