Entry point for bisect sample comparison.

tracing/tracing/metrics/compare_samples_unittest.py

Index: tracing/tracing/metrics/compare_samples_unittest.py
diff --git a/tracing/tracing/metrics/compare_samples_unittest.py b/tracing/tracing/metrics/compare_samples_unittest.py
new file mode 100644
index 0000000000000000000000000000000000000000..39caa202b7197e3d03d7263b47d0bd5dd074370c
--- /dev/null
+++ b/tracing/tracing/metrics/compare_samples_unittest.py
@@ -0,0 +1,222 @@
+# Copyright 2016 The Chromium Authors. All rights reserved.
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+import json
+import math
+import os
+import random
+import tempfile
+import unittest
+
+from tracing.metrics import compare_samples
+
+
+REJECT_THE_NULL = 'REJECT_THE_NULL'
+FAIL_TO_REJECT_THE_NULL = 'FAIL_TO_REJECT_THE_NULL'
+NEED_MORE_DATA = 'NEED_MORE_DATA'
+
+
+def Mean(l):
+  if len(l):
+    return float(sum(l))/len(l)
+  return 0
+
+
+class CompareSamplesUnittest(unittest.TestCase):
+  def setUp(self):
+    self._tempfiles = []
+    self._tempdir = tempfile.mkdtemp()
+
+  def tearDown(self):
+    for tf in self._tempfiles:
+      try:
+        os.remove(tf)
+      except OSError:
+        pass
+    try:
+      os.rmdir(self._tempdir)
+    except OSError:
+      pass
+
+  def NewJsonTempfile(self, jsonable_contents):
+    _, new_json_file = tempfile.mkstemp(
+        suffix='.json',
+        dir=self._tempdir,
+        text=True)
+    self._tempfiles.append(new_json_file)
+    with open(new_json_file, 'w') as f:
+      json.dump(jsonable_contents, f)
+    return new_json_file
+
+  def MakeMultipleChartJSONHistograms(self, metric, seed, mu, sigma, n, m):
+    result = []
+    random.seed(seed)
+    for _ in range(m):
+      result.append(self.MakeChartJSONHistogram(metric, mu, sigma, n))
+    return result
+
+  def MakeChartJSONHistogram(self, metric, mu, sigma, n):
+    """Creates a histogram for a normally distributed pseudo-random sample.
+
+    This function creates a deterministic pseudo-random sample and stores it in
+    chartjson histogram format to facilitate the testing of the sample
+    comparison logic.
+
+    For simplicity we use sqrt(n) buckets with equal widths.
+
+    Args:
+      metric (str pair): name of chart, name of the trace.
+      seed (hashable obj): to make the sequences deterministic we seed the RNG.
+      mu (float): desired mean for the sample
+      sigma (float): desired standard deviation for the sample
+      n (int): number of values to generate.
+    """
+    chart_name, trace_name = metric
+    values = [random.gauss(mu, sigma) for _ in range(n)]
+    bucket_count = int(math.ceil(math.sqrt(len(values))))
+    width = (max(values) - min(values))/(bucket_count - 1)
+    prev_bucket = min(values)
+    buckets = []
+    for _ in range(bucket_count):
+      buckets.append({'low': prev_bucket,
+                      'high': prev_bucket + width,
+                      'count': 0})
+      prev_bucket += width
+    for value in values:
+      for bucket in buckets:
+        if value >= bucket['low'] and value < bucket['high']:
+          bucket['count'] += 1
+          break
+    charts = {
+        'charts': {
+            chart_name: {
+                trace_name: {
+                    'type': 'histogram',
+                    'buckets': buckets
+                }
+            }
+        }
+    }
+    return self.NewJsonTempfile(charts)
+
+  def MakeChart(self, metric, seed, mu, sigma, n):
+    """Creates a normally distributed pseudo-random sample. (continuous).
+
+    This function creates a deterministic pseudo-random sample and stores it in
+    chartjson format to facilitate the testing of the sample comparison logic.
+
+    Args:
+      metric (str pair): name of chart, name of the trace.
+      seed (hashable obj): to make the sequences deterministic we seed the RNG.
+      mu (float): desired mean for the sample
+      sigma (float): desired standard deviation for the sample
+      n (int): number of values to generate.
+    """
+    chart_name, trace_name = metric
+    random.seed(seed)
+    values = [random.gauss(mu, sigma) for _ in range(n)]
+    charts = {
+        'charts': {
+            chart_name: {
+                trace_name: {
+                    'type': 'list_of_scalar_values',
+                    'values': values}
+            }
+        }
+    }
+    return self.NewJsonTempfile(charts)
+
+  def testCompareClearRegression(self):
+    metric = ('some_chart', 'some_trace')
+    lower_values = ','.join([self.MakeChart(metric=metric, seed='lower',
+                                            mu=10, sigma=1, n=10)])
+    higher_values = ','.join([self.MakeChart(metric=metric, seed='higher',
+                                             mu=20, sigma=2, n=10)])
+    result = json.loads(compare_samples.CompareSamples(
+        lower_values, higher_values, '/'.join(metric)))
+    self.assertEqual(result['result'], REJECT_THE_NULL)
+
+  def testCompareUnlikelyRegressionWithMultipleRuns(self):
+    metric = ('some_chart', 'some_trace')
+    lower_values = ','.join([
+        self.MakeChart(
+            metric=metric, seed='lower%d' % i, mu=10, sigma=1, n=5)
+        for i in range(4)])
+    higher_values = ','.join([
+        self.MakeChart(
+            metric=metric, seed='higher%d' % i, mu=10.01, sigma=0.95, n=5)
+        for i in range(4)])
+    result = json.loads(compare_samples.CompareSamples(
+        lower_values, higher_values, '/'.join(metric)))
+    self.assertEqual(result['result'], FAIL_TO_REJECT_THE_NULL)
+
+  def testCompareInsufficientData(self):
+    metric = ('some_chart', 'some_trace')
+    lower_values = ','.join([self.MakeChart(metric=metric, seed='lower',
+                                            mu=10, sigma=1, n=5)])
+    higher_values = ','.join([self.MakeChart(metric=metric, seed='higher',
+                                             mu=10.40, sigma=0.95, n=5)])
+    result = json.loads(compare_samples.CompareSamples(
+        lower_values, higher_values, '/'.join(metric)))
+    self.assertEqual(result['result'], NEED_MORE_DATA)
+
+  def testCompareMissingFile(self):
+    metric = ('some_chart', 'some_trace')
+    lower_values = ','.join([self.MakeChart(metric=metric, seed='lower',
+                                            mu=10, sigma=1, n=5)])
+    higher_values = '/path/does/not/exist.json'
+    with self.assertRaises(RuntimeError):
+      compare_samples.CompareSamples(
+          lower_values, higher_values, '/'.join(metric))
+
+  def testCompareMissingMetric(self):
+    metric = ('some_chart', 'some_trace')
+    lower_values = ','.join([self.MakeChart(metric=metric, seed='lower',
+                                            mu=10, sigma=1, n=5)])
+    higher_values = ','.join([self.MakeChart(metric=metric, seed='higher',
+                                             mu=20, sigma=2, n=5)])
+    metric = ('some_chart', 'missing_trace')
+    with self.assertRaises(RuntimeError):
+      compare_samples.CompareSamples(
+          lower_values, higher_values, '/'.join(metric))
+
+  def testCompareBadChart(self):
+    metric = ('some_chart', 'some_trace')
+    lower_values = ','.join([self.MakeChart(metric=metric, seed='lower',
+                                            mu=10, sigma=1, n=5)])
+    higher_values = self.NewJsonTempfile(['obviously', 'not', 'a', 'chart]'])
+    with self.assertRaises(RuntimeError):
+      compare_samples.CompareSamples(
+          lower_values, higher_values, '/'.join(metric))
+
+  def testCompareValuesets(self):
+    vs = os.path.join(os.path.dirname(__file__), 'test-valueset.json')
+    result = compare_samples.CompareSamples(
+        vs, vs, 'timeToFirstContentfulPaint/pcv1-cold/'
+        'http___www.rambler.ru_', data_format='valueset')
+    result = json.loads(result)
+    self.assertEqual(result['result'], NEED_MORE_DATA)
+    self.assertAlmostEqual(Mean(result['sample_a']), 75.3177999958396)
+    self.assertAlmostEqual(Mean(result['sample_b']), 75.3177999958396)
+
+  def testCompareBuildbotOutput(self):
+    bb = os.path.join(os.path.dirname(__file__), 'test-buildbot.txt')
+    result = compare_samples.CompareSamples(
+        bb, bb, 'DrawCallPerf_gl/score',
+        data_format='buildbot')
+    result = json.loads(result)
+    self.assertEqual(result['result'], NEED_MORE_DATA)
+    self.assertEqual(Mean(result['sample_a']), 4123)
+    self.assertEqual(Mean(result['sample_b']), 4123)
+
+  def testCompareChartJsonHistogram(self):
+    metric = ('some_chart', 'some_trace')
+    lower_values = ','.join(self.MakeMultipleChartJSONHistograms(
+        metric=metric, seed='lower', mu=10, sigma=1, n=100, m=10))
+    higher_values = ','.join(self.MakeMultipleChartJSONHistograms(
+        metric=metric, seed='higher', mu=20, sigma=2, n=100, m=10))
+    result = json.loads(compare_samples.CompareSamples(
+        lower_values, higher_values, '/'.join(metric)))
+    self.assertEqual(result['result'], REJECT_THE_NULL)
+