Better performance on long lists containing comments.

lib/src/source_visitor.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 dart_style.src.source_visitor;
 
 import 'package:analyzer/analyzer.dart';
 import 'package:analyzer/src/generated/scanner.dart';
 import 'package:analyzer/src/generated/source.dart';
 
@@ skipping 1862 matching lines @@
     // Force all of the surrounding collections to split.
     for (var i = 0; i < _collectionSplits.length; i++) {
       _collectionSplits[i] = true;
     }
 
     // Add this collection to the stack.
     _collectionSplits.add(false);
 
     _startLiteralBody(leftBracket);
 
-    // Always use a hard rule to split the elements. The parent chunk of
-    // the collection will handle the unsplit case, so this only comes
-    // into play when the collection is split.
-    var rule = new Rule.hard();
-    builder.startRule(rule);
-
     // If a collection contains a line comment, we assume it's a big complex
     // blob of data with some documented structure. In that case, the user
     // probably broke the elements into lines deliberately, so preserve those.
     var preserveNewlines = _containsLineComments(elements, rightBracket);
 
+    var rule;
+    var lineRule;
+    if (preserveNewlines) {
+      // Newlines are significant, so we'll explicitly write those. Elements
+      // on the same line all share an argument-list-like rule that allows
+      // splitting between zero, one, or all of them. This is faster in long
+      // lists than using individual splits after each element.
+      lineRule = new TypeArgumentRule();
+      builder.startLazyRule(lineRule);
+    } else {
+      // Newlines aren't significant, so use a hard rule to split the elements.
+      // The parent chunk of the collection will handle the unsplit case, so
+      // this only comes into play when the collection is split.
+      rule = new Rule.hard();
+      builder.startRule(rule);
+    }
+
     for (var element in elements) {
       if (element != elements.first) {
         if (preserveNewlines) {
+          // See if the next element is on the next line.
           if (_endLine(element.beginToken.previous) !=
               _startLine(element.beginToken)) {
             oneOrTwoNewlines();
+
+            // Start a new rule for the new line.
+            builder.endRule();
+            lineRule = new TypeArgumentRule();
+            builder.startLazyRule(lineRule);
           } else {
-            soloSplit();
+            lineRule.beforeArgument(split());
           }
         } else {
           builder.split(nest: false, space: true);
         }
       }
 
       builder.nestExpression();
       visit(element);
 
       // The comma after the element.
@@ skipping 485 matching lines @@
   /// Gets the 1-based line number that the beginning of [token] lies on.
   int _startLine(Token token) => _lineInfo.getLocation(token.offset).lineNumber;
 
   /// Gets the 1-based line number that the end of [token] lies on.
   int _endLine(Token token) => _lineInfo.getLocation(token.end).lineNumber;
 
   /// Gets the 1-based column number that the beginning of [token] lies on.
   int _startColumn(Token token) =>
       _lineInfo.getLocation(token.offset).columnNumber;
 }