Rolling average work queue diagnostic (#24933).

pkg/analysis_server/lib/src/domain_diagnostic.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 src.domain_diagnostic;
 
+import 'dart:async';
 import 'dart:collection';
 import 'dart:core' hide Resource;
 
 import 'package:analysis_server/plugin/protocol/protocol.dart';
 import 'package:analysis_server/src/analysis_server.dart';
 import 'package:analyzer/file_system/file_system.dart';
 import 'package:analyzer/src/context/cache.dart';
 import 'package:analyzer/src/context/context.dart';
 import 'package:analyzer/src/generated/engine.dart'
     hide AnalysisCache, AnalysisContextImpl;
 import 'package:analyzer/src/generated/source.dart';
 import 'package:analyzer/src/generated/utilities_collection.dart';
 import 'package:analyzer/src/task/driver.dart';
 import 'package:analyzer/task/model.dart';
 
-/// Extract context data from the given [context].
-ContextData extractData(AnalysisContext context) {
-  int explicitFiles = 0;
-  int implicitFiles = 0;
-  int workItems = 0;
-  Set<String> exceptions = new HashSet<String>();
-  if (context is AnalysisContextImpl) {
-    // Work Item count.
-    AnalysisDriver driver = context.driver;
-    List<WorkItem> items = driver.currentWorkOrder?.workItems;
-    workItems ??= items?.length;
-    var cache = context.analysisCache;
-    if (cache is AnalysisCache) {
-      Set<AnalysisTarget> countedTargets = new HashSet<AnalysisTarget>();
-      MapIterator<AnalysisTarget, CacheEntry> iterator = cache.iterator();
-      while (iterator.moveNext()) {
-        AnalysisTarget target = iterator.key;
-        if (countedTargets.add(target)) {
-          CacheEntry cacheEntry = iterator.value;
-          if (target is Source) {
-            if (cacheEntry.explicitlyAdded) {
-              explicitFiles++;
-            } else {
-              implicitFiles++;
-            }
-          }
-          // Caught exceptions.
-          if (cacheEntry.exception != null) {
-            exceptions.add(cacheEntry.exception.toString());
-          }
-        }
-      }
-    }
-  }
-  return new ContextData(context.name, explicitFiles, implicitFiles, workItems,
-      exceptions.toList());
+int _workItemCount(AnalysisContextImpl context) {
+  AnalysisDriver driver = context.driver;
+  List<WorkItem> items = driver.currentWorkOrder?.workItems;
+  return items?.length ?? 0;
 }
 
 /// Instances of the class [DiagnosticDomainHandler] implement a
 /// [RequestHandler] that handles requests in the `diagnostic` domain.
 class DiagnosticDomainHandler implements RequestHandler {
   /// The name of the request used to get diagnostic information.
   static const String DIAGNOSTICS = 'diagnostic.getDiagnostics';
 
   /// The analysis server that is using this handler to process requests.
   final AnalysisServer server;
 
+  /// The sampler tracking rolling work queue length averages.
+  Sampler sampler;
+
   /// Initialize a newly created handler to handle requests for the given
   /// [server].
   DiagnosticDomainHandler(this.server);
 
   /// Answer the `diagnostic.diagnostics` request.
   Response computeDiagnostics(Request request) {
+    // Initialize sampler if needed.
+    if (sampler == null) {
+      sampler = new Sampler(server);
+    }
+
     List<ContextData> infos = <ContextData>[];
     server.folderMap.forEach((Folder folder, AnalysisContext context) {
-      infos.add(extractData(context));
+      infos.add(extractData(folder, context));
     });
 
     return new DiagnosticGetDiagnosticsResult(infos).toResponse(request.id);
   }
 
+  /// Extract context data from the given [context].
+  ContextData extractData(Folder folder, AnalysisContext context) {
+    int explicitFiles = 0;
+    int implicitFiles = 0;
+    int workItems = 0;
+    String workItemAverage = '-1';
+    Set<String> exceptions = new HashSet<String>();
+    if (context is AnalysisContextImpl) {
+      workItems = _workItemCount(context);
+      workItemAverage = sampler.getAverage(folder)?.toString() ?? '-1';
+      var cache = context.analysisCache;
+      if (cache is AnalysisCache) {
+        Set<AnalysisTarget> countedTargets = new HashSet<AnalysisTarget>();
+        MapIterator<AnalysisTarget, CacheEntry> iterator = cache.iterator();
+        while (iterator.moveNext()) {
+          AnalysisTarget target = iterator.key;
+          if (countedTargets.add(target)) {
+            CacheEntry cacheEntry = iterator.value;
+            if (target is Source) {
+              if (cacheEntry.explicitlyAdded) {
+                explicitFiles++;
+              } else {
+                implicitFiles++;
+              }
+            }
+            // Caught exceptions.
+            if (cacheEntry.exception != null) {
+              exceptions.add(cacheEntry.exception.toString());
+            }
+          }
+        }
+      }
+    }
+    return new ContextData(context.name, explicitFiles, implicitFiles,
+        workItems, workItemAverage, exceptions.toList());
+  }
+
   @override
   Response handleRequest(Request request) {
     try {
       String requestName = request.method;
       if (requestName == DIAGNOSTICS) {
         return computeDiagnostics(request);
       }
     } on RequestFailure catch (exception) {
       return exception.response;
     }
     return null;
   }
 }
+
+/// Keeps track of a moving average of work item queue lengths mapped to
+/// contexts.
+///
+/// Sampling terminates after [maxSampleCount], if no one expresses interest
+/// by calling [resetTimerCountdown].
+class Sampler {
+  /// Timer interval.
+  static const Duration duration = const Duration(seconds: 1);
+
+  /// Maximum number of samples taken between calls to [reset].
+  static const int maxSampleCount = 30;
+
+  /// Current sample count.
+  int sampleCount = 0;
+
+  /// The shared timer.
+  Timer timer;
+
+  /// Map of contexts (tracked as folders to avoid leaks) to averages.
+  /// TOOD(pq): consider adding GC to remove mappings for deleted folders
+  Map<Folder, _Average> averages = new HashMap<Folder, _Average>();
+
+  final AnalysisServer server;
+  Sampler(this.server) {
+    start();
+    _sample();
+  }
+
+  /// Get the average for the context associated with the given [folder].
+  int getAverage(Folder folder) {
+    resetTimerCountdown();
+    return averages[folder].value;
+  }
+
+  /// Check if we're currently sampling.
+  bool isSampling() => timer?.isActive ?? false;
+
+  /// Reset counter.
+  void resetTimerCountdown() {
+    sampleCount = 0;
+  }
+
+  /// Start sampling.
+  void start() {
+    // No need to (re)start if already sampling.
+    if (isSampling()) {
+      return;
+    }
+    timer = new Timer.periodic(duration, (Timer timer) {
+      _sample();
+      if (sampleCount++ >= maxSampleCount) {
+        timer.cancel();
+      }
+    });
+  }
+
+  /// Stop sampling.
+  void stop() {
+    timer.cancel();
+  }
+
+  /// Take a sample.
+  void _sample() {
+    try {
+      server.folderMap.forEach((Folder folder, AnalysisContext context) {
+        if (context is AnalysisContextImpl) {
+          _Average average = averages[folder];
+          if (average == null) {
+            average = new _Average();
+            averages[folder] = average;
+          }
+          average.addSample(_workItemCount(context));
+        }
+      });
+    } on Exception {
+      stop();
+    }
+  }
+}
+
+/// Simple rolling average sample counter.
+class _Average {
+  num _val;
+
+  final int sampleCount;
+  _Average([this.sampleCount = 20]);
+
+  num get value => _val ?? 0;
+
+  void addSample(num sample) {
+    if (_val == null) {
+      _val = sample;
+    } else {
+      _val =
+          _val * ((sampleCount - 1) / sampleCount) + sample * (1 / sampleCount);
+    }
+  }
+
+  @override
+  String toString() => 'average: ${value}';
+}