Improvements to task model

pkg/analyzer/lib/src/task/model.dart

 // Copyright (c) 2015, 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 analyzer.src.task.model;
 
 import 'package:analyzer/src/generated/engine.dart' hide AnalysisTask;
 import 'package:analyzer/src/task/inputs.dart';
 import 'package:analyzer/task/model.dart';
 
 /**
  * A concrete implementation of a [ContributionPoint].
  */
 class ContributionPointImpl<V> extends ResultDescriptorImpl<V> implements
     ContributionPoint<V> {
   /**
    * The results that contribute to this result.
    */
   final List<ResultDescriptor<V>> contributors = <ResultDescriptor<V>>[];
 
   /**
-   * Initialize a newly created contribution point to have the given [name].
+   * Initialize a newly created contribution point to have the given [name] and
+   * [defaultValue].
    */
-  ContributionPointImpl(String name) : super(name);
+  ContributionPointImpl(String name, V defaultValue)

[Paul Berry, 2015/01/26 17:31:06]

Since the "value" associated with a contribution point is always determined by
concatenating together the results of all the tasks that contribute to it, does
it even make sense for a contribution point to have a default value?  Maybe we
should remove the "defaultValue" argument and just pass "null" to super.

[Brian Wilkerson, 2015/01/26 19:47:53]

I think it doesn't. I'll make the change.

+      : super(name, defaultValue);
 
   /**
    * Record that the given analysis [result] contibutes to this result.
    */
   void recordContributor(ResultDescriptor<V> result) {
     contributors.add(result);
   }
 }
 
 /**
  * A concrete implementation of a [ResultDescriptor].
  */
 class ResultDescriptorImpl<V> implements ResultDescriptor<V> {
   /**
    * The name of the result, used for debugging.
    */
   final String name;
 
   /**
-   * Initialize a newly created analysis result to have the given [name]. If a
-   * contribution point is specified, then this result will contribute to it.
+   * Return the default value for results described by this descriptor.
    */
-  ResultDescriptorImpl(this.name, {ContributionPoint contributesTo}) {
+  final V defaultValue;

[Paul Berry, 2015/01/26 17:31:06]

This design forces all results to share the same default value.  If the values
stored in results were immutable data structures, I would be ok with this, but
since we sometimes mutate them (e.g. when annotating an AST with pointers to the
element model), it feels a little dangerous.

Would it be better to instead store a function having type () -> V, so that we
can use it to create a default value on the fly when we need it?

[Brian Wilkerson, 2015/01/26 19:47:53]

Good point! It isn't a problem at the moment because all of the results whose
values can be mutable data structures currently have a default value of 'null',
but we should definitely revisit this, given that plugins will be contributing
result descriptors.

+
+  /**
+   * Initialize a newly created analysis result to have the given [name] and
+   * [defaultValue]. If a contribution point is specified, then this result will
+   * contribute to it.
+   */
+  ResultDescriptorImpl(this.name, this.defaultValue,
+      {ContributionPoint contributesTo}) {
     if (contributesTo is ContributionPointImpl) {
       contributesTo.recordContributor(this);
     }
   }
 
   @override
   TaskInput<V> inputFor(AnalysisTarget target) =>
       new SimpleTaskInput<V>(target, this);
 
   @override
@@ skipping 39 matching lines @@
   AnalysisTask createTask(AnalysisContext context, AnalysisTarget target,
       Map<String, dynamic> inputs) {
     AnalysisTask task = buildTask(context, target);
     task.inputs = inputs;
     return task;
   }
 
   @override
   String toString() => name;
 }