Scanner for UTF-8 byte arrays

sdk/lib/_internal/compiler/implementation/source_file.dart

Index: sdk/lib/_internal/compiler/implementation/source_file.dart
diff --git a/sdk/lib/_internal/compiler/implementation/source_file.dart b/sdk/lib/_internal/compiler/implementation/source_file.dart
index 3e7d2b8a37856e63a45226e3e3e9f077f2fff207..25f2afbf708d06b196bb3dabfe3901c778f9cbcd 100644
--- a/sdk/lib/_internal/compiler/implementation/source_file.dart
+++ b/sdk/lib/_internal/compiler/implementation/source_file.dart
@@ -5,42 +5,83 @@
 library source_file;
 
 import 'dart:math';
+import 'dart:convert' show UTF8;
 
 /**
- * Represents a file of source code.
+ * Represents a file of source code. The content can be either a [String] or
+ * a UTF-8 encoded [List<int>] of bytes.
  */
-class SourceFile {
+abstract class SourceFile {
 
   /** The name of the file. */
   final String filename;
 
-  /** The text content of the file. */
-  final String text;
+  SourceFile(this.filename);
+
+  /** The text content of the file represented as a String. */
+  String slowText();
 
-  List<int> _lineStarts;
+  /** The content of the file represented as a UTF-8 encoded [List<int>]. */
+  List<int> slowUtf8Bytes();
 
-  SourceFile(this.filename, this.text);
+  /**
+   * The length of the string representation of this source file, i.e.,
+   * equivalent to [:slowText().length:], but faster.
+   */
+  int get length;
 
+  /**
+   * Sets the string length of this source file. For source files based on UTF-8
+   * byte arrays, the string length is computed and assigned by the scanner.
+   */
+  set length(v);

[ngeoffray, 2013/10/18 10:19:37]

int v?

[lukas, 2013/10/24 16:48:36]

Done.

+
+  /**
+   * A map from line numbers to offsets in the string text representation of
+   * this source file.
+   */
   List<int> get lineStarts {
-    if (_lineStarts == null) {
-      var starts = [0];
-      var index = 0;
-      while (index < text.length) {
-        index = text.indexOf('\n', index) + 1;
-        if (index <= 0) break;
-        starts.add(index);
-      }
-      starts.add(text.length + 1);
-      _lineStarts = starts;
+    if (lineStartsCache == null) {
+      // When reporting errors during scanning, the line numbers are not yet
+      // available and need to be computed using this slow path.
+      lineStartsCache = lineStartsFromString(slowText());
     }
-    return _lineStarts;
+    return lineStartsCache;
   }
 
+  /**
+   * Sets the line numbers map for this source file. This map is computed and
+   * assigned by the scanner, avoiding a separate traversal of the source file.
+   *
+   * The map contains one additional entry at the end of the file, as if the
+   * source file had one more empty line at the end. This simplifies the binary
+   * search in [getLine].
+   */
+  set lineStarts(v) => lineStartsCache = v;

[ngeoffray, 2013/10/18 10:19:37]

v -> List<int> v.

[lukas, 2013/10/24 16:48:36]

Done.

+
+  List<int> lineStartsCache;
+
+  List<int> lineStartsFromString(String text) {
+    var starts = [0];
+    var index = 0;
+    while (index < text.length) {
+      index = text.indexOf('\n', index) + 1;
+      if (index <= 0) break;
+      starts.add(index);
+    }
+    starts.add(text.length + 1); // One additional line start at the end.
+    return starts;
+  }
+
+  /**
+   * Returns the line number for the offset [position] in the string
+   * representation of this source file.
+   */
   int getLine(int position) {
     List<int> starts = lineStarts;
     if (position < 0 || starts.last <= position) {
       throw 'bad position #$position in file $filename with '
-            'length ${text.length}.';
+            'length ${length}.';
     }
     int first = 0;
     int count = starts.length;
@@ -58,10 +99,16 @@ class SourceFile {
     return first;
   }
 
+  /**
+   * Returns the column number for the offset [position] in the string
+   * representation of this source file.
+   */
   int getColumn(int line, int position) {
     return position - lineStarts[line];
   }
 
+  String slowSubstring(int start, int end);
+
   /**
    * Create a pretty string representation from a character position
    * in the file.
@@ -75,12 +122,12 @@ class SourceFile {
         '${filename}:${line + 1}:${column + 1}: $message');
     if (includeText) {
       buf.write('\n');
-      var textLine;
+      String textLine;
       // +1 for 0-indexing, +1 again to avoid the last line of the file
-      if ((line + 2) < _lineStarts.length) {
-        textLine = text.substring(_lineStarts[line], _lineStarts[line+1]);
+      if ((line + 2) < lineStarts.length) {
+        textLine = slowSubstring(lineStarts[line], lineStarts[line+1]);
       } else {
-        textLine = '${text.substring(_lineStarts[line])}\n';
+        textLine = '${slowSubstring(lineStarts[line], length)}\n';
       }
 
       int toColumn = min(column + (end-start), textLine.length);
@@ -101,3 +148,47 @@ class SourceFile {
     return buf.toString();
   }
 }
+
+class Utf8BytesSourceFile extends SourceFile {
+
+  /** The UTF-8 encoded content of the source file. */
+  final List<int> content;
+
+  Utf8BytesSourceFile(String filename, this.content) : super(filename);
+
+  String slowText() => UTF8.decode(content);
+
+  List<int> slowUtf8Bytes() => content;
+
+  String slowSubstring(int start, int end) {
+    // TODO(lry): to make this faster, the scanner could record the UTF-8 slack
+    // for all positions of the source text. We could use [:content.sublist:].
+    return slowText().substring(start, end);
+  }
+
+  int get length {
+    if (lengthCache == -1) {
+      // During scanning the length is not yet assigned, so we use a slow path.
+      length = slowText().length;
+    }
+    return lengthCache;
+  }
+  set length(v) => lengthCache = v;

[ngeoffray, 2013/10/18 10:19:37]

int v

[lukas, 2013/10/24 16:48:36]

Done.

+  int lengthCache = -1;
+}
+
+class StringSourceFile extends SourceFile {
+
+  final String text;
+
+  StringSourceFile(String filename, this.text) : super(filename);
+
+  int get length => text.length;
+  set length(v) { }

[ngeoffray, 2013/10/18 10:19:37]

int v

[lukas, 2013/10/24 16:48:36]

Done.

+
+  String slowText() => text;
+
+  List<int> slowUtf8Bytes() => UTF8.encode(text);
+
+  String slowSubstring(int start, int end) => text.substring(start, end);
+}