perf-o-matic should incrementally update JSON responses

https://bugs.webkit.org/show_bug.cgi?id=79898

Reviewed by Eric Seidel.

Now that Runs object store test_runs and averages separately,
we can update JSON data incrementally without degrading values.

Also fixed the order of arguments passed to DashboardImage.needs_update
in schedule_runs_update. This bug had caused all chart images to be
updated on every new report.

* Websites/webkit-perf.appspot.com/controller.py:
(schedule_runs_update):
* Websites/webkit-perf.appspot.com/models.py:
(Runs.update_or_insert):
(Runs.update_incrementally):
(Runs):
(Runs.get_by_objects):
* Websites/webkit-perf.appspot.com/models_unittest.py:
(RunsTest._create_results):
(RunsTest.test_generate_runs):
(RunsTest.test_update_or_insert):
(RunsTest.test_update_incrementally):
(RunsTest.test_to_json_with_results):
* Websites/webkit-perf.appspot.com/report_process_handler.py:
(ReportProcessHandler.post):

git-svn-id: https://svn.webkit.org/repository/webkit/trunk@110210 268f45cc-cd09-0410-ab3c-d52691b4dbfc

diff --git a/ChangeLog b/ChangeLog
index c3bdb60..3df5af6 100644 (file)
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,33 @@
+2012-03-08  Ryosuke Niwa  <rniwa@webkit.org>
+
+        perf-o-matic should incrementally update JSON responses
+        https://bugs.webkit.org/show_bug.cgi?id=79898
+
+        Reviewed by Eric Seidel.
+
+        Now that Runs object store test_runs and averages separately,
+        we can update JSON data incrementally without degrading values.
+
+        Also fixed the order of arguments passed to DashboardImage.needs_update
+        in schedule_runs_update. This bug had caused all chart images to be
+        updated on every new report.
+
+        * Websites/webkit-perf.appspot.com/controller.py:
+        (schedule_runs_update):
+        * Websites/webkit-perf.appspot.com/models.py:
+        (Runs.update_or_insert):
+        (Runs.update_incrementally):
+        (Runs):
+        (Runs.get_by_objects):
+        * Websites/webkit-perf.appspot.com/models_unittest.py:
+        (RunsTest._create_results):
+        (RunsTest.test_generate_runs):
+        (RunsTest.test_update_or_insert):
+        (RunsTest.test_update_incrementally):
+        (RunsTest.test_to_json_with_results):
+        * Websites/webkit-perf.appspot.com/report_process_handler.py:
+        (ReportProcessHandler.post):
+

 2012-03-08  Max Vujovic  <mvujovic@adobe.com>
 
         Add a method to window.internals to enable testing of inspector highlight rects
 2012-03-08  Max Vujovic  <mvujovic@adobe.com>
 
         Add a method to window.internals to enable testing of inspector highlight rects

diff --git a/Websites/webkit-perf.appspot.com/controller.py b/Websites/webkit-perf.appspot.com/controller.py
index 71d450b..9623d93 100644 (file)
--- a/Websites/webkit-perf.appspot.com/controller.py
+++ b/Websites/webkit-perf.appspot.com/controller.py
@@ -85,10 +85,11 @@ class CachedDashboardHandler(webapp2.RequestHandler):
             schedule_dashboard_update()
 
 
             schedule_dashboard_update()
 
 

-def schedule_runs_update(test_id, branch_id, platform_id):
-    taskqueue.add(url='/api/test/runs/update', params={'id': test_id, 'branchid': branch_id, 'platformid': platform_id})
+def schedule_runs_update(test_id, branch_id, platform_id, regenerate_runs=True):
+    if regenerate_runs:
+        taskqueue.add(url='/api/test/runs/update', params={'id': test_id, 'branchid': branch_id, 'platformid': platform_id})

     for display_days in [7, 30, 90, 365]:
     for display_days in [7, 30, 90, 365]:

-        if DashboardImage.needs_update(branch_id, test_id, platform_id, display_days):
+        if DashboardImage.needs_update(branch_id, platform_id, test_id, display_days):

             taskqueue.add(url='/api/test/runs/chart', params={'id': test_id, 'branchid': branch_id, 'platformid': platform_id,
                 'displayDays': display_days})
 
             taskqueue.add(url='/api/test/runs/chart', params={'id': test_id, 'branchid': branch_id, 'platformid': platform_id,
                 'displayDays': display_days})
 

diff --git a/Websites/webkit-perf.appspot.com/models.py b/Websites/webkit-perf.appspot.com/models.py
index 51f029f..df7ab7e 100644 (file)
--- a/Websites/webkit-perf.appspot.com/models.py
+++ b/Websites/webkit-perf.appspot.com/models.py
@@ -361,26 +361,45 @@ class Runs(db.Model):
 
     @classmethod
     def update_or_insert(cls, branch, platform, test):
 
     @classmethod
     def update_or_insert(cls, branch, platform, test):

-        test_runs = []
-        averages = {}
-        values = []
+        key_name = cls._key_name(branch.id, platform.id, test.id)
+        runs = Runs(key_name=key_name, branch=branch, platform=platform, test=test, json_runs='', json_averages='')

 
         for build, result in cls._generate_runs(branch, platform, test.name):
 
         for build, result in cls._generate_runs(branch, platform, test.name):

-            test_runs.append(cls._entry_from_build_and_result(build, result))
-            # FIXME: Calculate the average. In practice, we wouldn't have more than one value for a given revision.
-            averages[build.revision] = result.value
-            values.append(result.value)
-
-        min_value = min(values) if values else None
-        max_value = max(values) if values else None
+            runs.update_incrementally(build, result, check_duplicates_and_save=False)

 
 

-        key_name = cls._key_name(branch.id, platform.id, test.id)
-        runs = Runs(key_name=key_name, branch=branch, platform=platform, test=test,
-            json_runs=json.dumps(test_runs)[1:-1], json_averages=json.dumps(averages)[1:-1], json_min=min_value, json_max=max_value)

         runs.put()
         memcache.set(key_name, runs.to_json())
         return runs
 
         runs.put()
         memcache.set(key_name, runs.to_json())
         return runs
 

+    def update_incrementally(self, build, result, check_duplicates_and_save=True):
+        new_entry = Runs._entry_from_build_and_result(build, result)
+
+        # Check for duplicate entries
+        if check_duplicates_and_save:
+            revision_is_in_runs = str(build.revision) in json.loads('{' + self.json_averages + '}')
+            if revision_is_in_runs and new_entry[1] in [entry[1] for entry in json.loads('[' + self.json_runs + ']')]:
+                return
+
+        if self.json_runs:
+            self.json_runs += ','
+
+        if self.json_averages:
+            self.json_averages += ','
+
+        self.json_runs += json.dumps(new_entry)
+        # FIXME: Calculate the average. In practice, we wouldn't have more than one value for a given revision.
+        self.json_averages += '"%d": %f' % (build.revision, result.value)
+        self.json_min = min(self.json_min, result.value) if self.json_min != None else result.value
+        self.json_max = max(self.json_max, result.value)
+
+        if check_duplicates_and_save:
+            self.put()
+            memcache.set(self.key().name(), self.to_json())
+
+    @staticmethod
+    def get_by_objects(branch, platform, test):
+        return Runs.get_by_key_name(Runs._key_name(branch.id, platform.id, test.id))
+

     @classmethod
     def json_by_ids(cls, branch_id, platform_id, test_id):
         key_name = cls._key_name(branch_id, platform_id, test_id)
     @classmethod
     def json_by_ids(cls, branch_id, platform_id, test_id):
         key_name = cls._key_name(branch_id, platform_id, test_id)

diff --git a/Websites/webkit-perf.appspot.com/models_unittest.py b/Websites/webkit-perf.appspot.com/models_unittest.py
index 7b5f8f1..fa4856b 100644 (file)
--- a/Websites/webkit-perf.appspot.com/models_unittest.py
+++ b/Websites/webkit-perf.appspot.com/models_unittest.py
@@ -527,16 +527,18 @@ class RunsTest(DataStoreTestsBase):
         super(RunsTest, self).setUp()
         self.testbed.init_memcache_stub()
 
         super(RunsTest, self).setUp()
         self.testbed.init_memcache_stub()
 

-    def _create_results(self, branch, platform, builder, test_name, values, timestamps=None):
+    def _create_results(self, branch, platform, builder, test_name, values, timestamps=None, starting_revision=100):
+        builds = []

         results = []
         for i, value in enumerate(values):
             build = Build(branch=branch, platform=platform, builder=builder,
         results = []
         for i, value in enumerate(values):
             build = Build(branch=branch, platform=platform, builder=builder,

-                buildNumber=i, revision=100 + i, timestamp=timestamps[i] if timestamps else datetime.now())
+                buildNumber=i, revision=starting_revision + i, timestamp=timestamps[i] if timestamps else datetime.now())

             build.put()
             result = TestResult(name=test_name, build=build, value=value)
             result.put()
             build.put()
             result = TestResult(name=test_name, build=build, value=value)
             result.put()

+            builds.append(build)

             results.append(result)
             results.append(result)

-        return results
+        return builds, results

 
     def test_generate_runs(self):
         some_branch = Branch.create_if_possible('some-branch', 'Some Branch')
 
     def test_generate_runs(self):
         some_branch = Branch.create_if_possible('some-branch', 'Some Branch')


@@ -544,7 +546,7 @@ class RunsTest(DataStoreTestsBase):
         some_builder = Builder.get(Builder.create('some-builder', 'Some Builder'))
         some_test = Test.update_or_insert('some-test', some_branch, some_platform)
 
         some_builder = Builder.get(Builder.create('some-builder', 'Some Builder'))
         some_test = Test.update_or_insert('some-test', some_branch, some_platform)
 

-        results = self._create_results(some_branch, some_platform, some_builder, 'some-test', [50.0, 51.0, 52.0, 49.0, 48.0])
+        builds, results = self._create_results(some_branch, some_platform, some_builder, 'some-test', [50.0, 51.0, 52.0, 49.0, 48.0])

         last_i = 0
         for i, (build, result) in enumerate(Runs._generate_runs(some_branch, some_platform, some_test)):
             self.assertEqual(build.buildNumber, i)
         last_i = 0
         for i, (build, result) in enumerate(Runs._generate_runs(some_branch, some_platform, some_test)):
             self.assertEqual(build.buildNumber, i)


@@ -573,15 +575,47 @@ class RunsTest(DataStoreTestsBase):
         runs.delete()
         self.assertThereIsNoInstanceOf(Runs)
 
         runs.delete()
         self.assertThereIsNoInstanceOf(Runs)
 

-        results = self._create_results(some_branch, some_platform, some_builder, 'some-test', [50.0])
+        builds, results = self._create_results(some_branch, some_platform, some_builder, 'some-test', [50.0])

         runs = Runs.update_or_insert(some_branch, some_platform, some_test)
         self.assertOnlyInstance(runs)
         self.assertTrue(runs.json_runs.startswith('[5, [4, 0, 100, null],'))
         runs = Runs.update_or_insert(some_branch, some_platform, some_test)
         self.assertOnlyInstance(runs)
         self.assertTrue(runs.json_runs.startswith('[5, [4, 0, 100, null],'))

-        self.assertEqual(runs.json_averages, '"100": 50.0')
+        self.assertEqual(json.loads('{' + runs.json_averages + '}'), {"100": 50.0})

         self.assertEqual(runs.json_min, 50.0)
         self.assertEqual(runs.json_max, 50.0)
         self.assertNotEqual(memcache.get(Runs._key_name(some_branch.id, some_platform.id, some_test.id)), old_memcache_value)
 
         self.assertEqual(runs.json_min, 50.0)
         self.assertEqual(runs.json_max, 50.0)
         self.assertNotEqual(memcache.get(Runs._key_name(some_branch.id, some_platform.id, some_test.id)), old_memcache_value)
 

+    def test_update_incrementally(self):
+        some_branch = Branch.create_if_possible('some-branch', 'Some Branch')
+        some_platform = Platform.create_if_possible('some-platform', 'Some Platform')
+        some_builder = Builder.get(Builder.create('some-builder', 'Some Builder'))
+        some_test = Test.update_or_insert('some-test', some_branch, some_platform)
+        self.assertThereIsNoInstanceOf(Runs)
+
+        timestamps = [datetime.now(), datetime.now()]
+        builds, results = self._create_results(some_branch, some_platform, some_builder, 'some-test', [50.0, 52.0], timestamps)
+        runs = Runs.update_or_insert(some_branch, some_platform, some_test)
+        self.assertOnlyInstance(runs)
+        self.assertEqual(json.loads('[' + runs.json_runs + ']'),
+            [[5, [4, 0, 100, None], mktime(timestamps[0].timetuple()), 50.0, 0, [], None, None],
+            [7, [6, 1, 101, None], mktime(timestamps[1].timetuple()), 52.0, 0, [], None, None]])
+        self.assertEqual(json.loads('{' + runs.json_averages + '}'), {"100": 50.0, "101": 52.0})
+        self.assertEqual(runs.json_min, 50.0)
+        self.assertEqual(runs.json_max, 52.0)
+
+        timestamps.append(datetime.now())
+        builds, results = self._create_results(some_branch, some_platform, some_builder, 'some-test', [48.0],
+            timestamps[2:], starting_revision=102)
+        runs.update_incrementally(builds[0], results[0])
+
+        self.assertOnlyInstance(runs)
+        self.assertEqual(json.loads('[' + runs.json_runs + ']'),
+            [[5, [4, 0, 100, None], mktime(timestamps[0].timetuple()), 50.0, 0, [], None, None],
+            [7, [6, 1, 101, None], mktime(timestamps[1].timetuple()), 52.0, 0, [], None, None],
+            [9, [8, 0, 102, None], mktime(timestamps[2].timetuple()), 48.0, 0, [], None, None]])
+        self.assertEqual(json.loads('{' + runs.json_averages + '}'), {"100": 50.0, "101": 52.0, "102": 48.0})
+        self.assertEqual(runs.json_min, 48.0)
+        self.assertEqual(runs.json_max, 52.0)
+

     def test_json_by_ids(self):
         some_branch = Branch.create_if_possible('some-branch', 'Some Branch')
         some_platform = Platform.create_if_possible('some-platform', 'Some Platform')
     def test_json_by_ids(self):
         some_branch = Branch.create_if_possible('some-branch', 'Some Branch')
         some_platform = Platform.create_if_possible('some-platform', 'Some Platform')


@@ -622,7 +656,7 @@ class RunsTest(DataStoreTestsBase):
         some_platform = Platform.create_if_possible('some-platform', 'Some Platform')
         some_builder = Builder.get(Builder.create('some-builder', 'Some Builder'))
         some_test = Test.update_or_insert('some-test', some_branch, some_platform)
         some_platform = Platform.create_if_possible('some-platform', 'Some Platform')
         some_builder = Builder.get(Builder.create('some-builder', 'Some Builder'))
         some_test = Test.update_or_insert('some-test', some_branch, some_platform)

-        results = self._create_results(some_branch, some_platform, some_builder, 'some-test', [50.0, 51.0, 52.0, 49.0, 48.0])
+        builds, results = self._create_results(some_branch, some_platform, some_builder, 'some-test', [50.0, 51.0, 52.0, 49.0, 48.0])

 
         value = json.loads(Runs.update_or_insert(some_branch, some_platform, some_test).to_json())
         self.assertEqualUnorderedList(value.keys(), ['test_runs', 'averages', 'min', 'max', 'date_range', 'stat'])
 
         value = json.loads(Runs.update_or_insert(some_branch, some_platform, some_test).to_json())
         self.assertEqualUnorderedList(value.keys(), ['test_runs', 'averages', 'min', 'max', 'date_range', 'stat'])

diff --git a/Websites/webkit-perf.appspot.com/report_process_handler.py b/Websites/webkit-perf.appspot.com/report_process_handler.py
index ecf6e38..7103785 100644 (file)
--- a/Websites/webkit-perf.appspot.com/report_process_handler.py
+++ b/Websites/webkit-perf.appspot.com/report_process_handler.py
@@ -37,6 +37,7 @@ from controller import schedule_dashboard_update
 from controller import schedule_manifest_update
 from models import Build
 from models import ReportLog
 from controller import schedule_manifest_update
 from models import Build
 from models import ReportLog

+from models import Runs

 from models import Test
 from models import TestResult
 
 from models import Test
 from models import TestResult
 


@@ -56,10 +57,15 @@ class ReportProcessHandler(webapp2.RequestHandler):
         platform = log.platform()
         build = Build.get_or_insert_from_log(log)
 
         platform = log.platform()
         build = Build.get_or_insert_from_log(log)
 

-        for test_name, result in log.results().iteritems():
+        for test_name, result_value in log.results().iteritems():

             test = Test.update_or_insert(test_name, branch, platform)
             test = Test.update_or_insert(test_name, branch, platform)

-            TestResult.get_or_insert_from_parsed_json(test_name, build, result)
-            schedule_runs_update(test.id, branch.id, platform.id)
+            result = TestResult.get_or_insert_from_parsed_json(test_name, build, result_value)
+            runs = Runs.get_by_objects(branch, platform, test)
+            regenerate_runs = True
+            if runs:
+                runs.update_incrementally(build, result)
+                regenerate_runs = False
+            schedule_runs_update(test.id, branch.id, platform.id, regenerate_runs)

 
         log = ReportLog.get(log.key())
         log.delete()
 
         log = ReportLog.get(log.key())
         log.delete()