Differential graphs on report (#657)

analysis/benchmark_results.py

@@ -17,10 +17,12 @@
 import functools
 
 from analysis import data_utils
+from analysis import coverage_data_utils
 from analysis import stat_tests
 
 
-class BenchmarkResults:  # pylint: disable=too-many-public-methods
+# pylint: disable=too-many-public-methods, too-many-arguments
+class BenchmarkResults:
     """Represents results of various analysis done on benchmark data.
 
     NOTE: Do not create this class manually! Instead, use the |benchmarks|
@@ -31,11 +33,12 @@ class BenchmarkResults:  # pylint: disable=too-many-public-methods
     template, properties are computed on demand and only once.
     """
 
-    def __init__(self, benchmark_name, experiment_df, output_directory,
-                 plotter):
+    def __init__(self, benchmark_name, experiment_df, coverage_dict,
+                 output_directory, plotter):
         self.name = benchmark_name
 
         self._experiment_df = experiment_df
+        self._coverage_dict = coverage_dict
         self._output_directory = output_directory
         self._plotter = plotter
 
@@ -45,6 +48,19 @@ def _prefix_with_benchmark(self, filename):
     def _get_full_path(self, filename):
         return os.path.join(self._output_directory, filename)
 
+    def _get_experiment_filestore_path(self, fuzzer_name):
+        return coverage_data_utils.get_fuzzer_filestore_path(
+            self._benchmark_df, fuzzer_name)
+
+    def get_filestore_name(self, fuzzer_name):
+        """Returns the filestore name of the |fuzzer_name|."""
+        filestore_path = self._get_experiment_filestore_path(fuzzer_name)
+        gcs_prefix = 'gs://'
+        gcs_http_prefix = 'https://storage.googleapis.com/'
+        if filestore_path.startswith(gcs_prefix):
+            filestore_path = filestore_path.replace(gcs_prefix, gcs_http_prefix)
+        return filestore_path
+
     @property
     @functools.lru_cache()
     # TODO(lszekeres): With python3.8+, replace above two decorators with:
@@ -53,16 +69,44 @@ def _benchmark_df(self):
         exp_df = self._experiment_df
         return exp_df[exp_df.benchmark == self.name]
 
+    @property
+    @functools.lru_cache()
+    def fuzzer_names(self):
+        """Names of all fuzzers."""
+        return self._benchmark_df.fuzzer.unique()
+
     @property
     @functools.lru_cache()
     def _benchmark_snapshot_df(self):
         return data_utils.get_benchmark_snapshot(self._benchmark_df)
 
     @property
     @functools.lru_cache()
-    def fuzzers(self):
-        """Fuzzers with valid trials on this benchmark."""
-        return self._benchmark_df.fuzzer.unique()
+    def _benchmark_coverage_dict(self):
+        """Covered regions of each fuzzer on this benchmark."""
+        return coverage_data_utils.get_benchmark_cov_dict(
+            self._coverage_dict, self.name)
+
+    @property
+    @functools.lru_cache()
+    def _benchmark_aggregated_coverage_df(self):
+        """Aggregated covered regions of each fuzzer on this benchmark."""
+        return coverage_data_utils.get_benchmark_aggregated_cov_df(
+            self._benchmark_coverage_dict)
+
+    @property
+    @functools.lru_cache()
+    def _unique_region_dict(self):
+        """Unique regions with the fuzzers that cover it."""
+        return coverage_data_utils.get_unique_region_dict(
+            self._benchmark_coverage_dict)
+
+    @property
+    @functools.lru_cache()
+    def _unique_region_cov_df(self):
+        """Fuzzers with the number of covered unique regions."""
+        return coverage_data_utils.get_unique_region_cov_df(
+            self._unique_region_dict, self.fuzzer_names)
 
     @property
     def fuzzers_with_not_enough_samples(self):
@@ -236,3 +280,30 @@ def better_than_plot(self):
         self._plotter.write_better_than_plot(better_than_table,
                                              self._get_full_path(plot_filename))
         return plot_filename
+
+    @property
+    def unique_coverage_ranking_plot(self):
+        """Ranking plot for unique coverage."""
+        plot_filename = self._prefix_with_benchmark('ranking_unique_region.svg')
+        unique_region_cov_df_combined = self._unique_region_cov_df.merge(
+            self._benchmark_aggregated_coverage_df, on='fuzzer')
+        self._plotter.write_unique_coverage_ranking_plot(
+            unique_region_cov_df_combined, self._get_full_path(plot_filename))
+        return plot_filename
+
+    @property
+    @functools.lru_cache()
+    def pairwise_unique_coverage_table(self):
+        """Pairwise unique coverage table for each pair of fuzzers."""
+        return coverage_data_utils.get_pairwise_unique_coverage_table(
+            self._benchmark_coverage_dict)
+
+    @property
+    def pairwise_unique_coverage_plot(self):
+        """Pairwise unique coverage plot for each pair of fuzzers."""
+        plot_filename = self._prefix_with_benchmark(
+            'pairwise_unique_coverage_plot.svg')
+        self._plotter.write_pairwise_unique_coverage_heatmap_plot(
+            self.pairwise_unique_coverage_table,
+            self._get_full_path(plot_filename))
+        return plot_filename

analysis/coverage_data_utils.py

@@ -0,0 +1,159 @@
+# Copyright 2020 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Utility functions for coverage data calculation."""
+
+import collections
+import json
+import os
+import posixpath
+import tempfile
+import pandas as pd
+
+from common import filestore_utils
+
+
+def get_fuzzer_benchmark_key(fuzzer: str, benchmark: str):
+    """Returns the key in coverage dict for a pair of fuzzer-benchmark."""
+    return fuzzer + ' ' + benchmark
+
+
+def get_fuzzer_filestore_path(benchmark_df, fuzzer):
+    """Gets the filestore_path for |fuzzer| in |benchmark_df|."""
+    fuzzer_df = benchmark_df[benchmark_df.fuzzer == fuzzer]
+    filestore_path = fuzzer_df.experiment_filestore.unique()[0]
+    exp_name = fuzzer_df.experiment.unique()[0]
+    return posixpath.join(filestore_path, exp_name)
+
+
+def get_covered_regions_dict(experiment_df):
+    """Combines json files for different fuzzer-benchmark pair
+    in |experiment_df| and returns a dictionary of the covered regions."""
+    covered_regions_dict = {}
+    benchmarks = experiment_df.benchmark.unique()
+    for benchmark in benchmarks:
+        benchmark_df = experiment_df[experiment_df.benchmark == benchmark]
+        fuzzers = benchmark_df.fuzzer.unique()
+        for fuzzer in fuzzers:
+            fuzzer_covered_regions = get_fuzzer_covered_regions(
+                benchmark_df, benchmark, fuzzer)
+            key = get_fuzzer_benchmark_key(fuzzer, benchmark)
+            covered_regions_dict[key] = fuzzer_covered_regions
+    return covered_regions_dict
+
+
+def get_fuzzer_covered_regions(benchmark_df, benchmark, fuzzer):
+    """Gets the covered regions for |fuzzer| in |benchmark_df| from the json
+    file in the bucket."""
+    with tempfile.TemporaryDirectory() as temp_dir:
+        dst_file = os.path.join(temp_dir, 'tmp.json')
+        src_filestore_path = get_fuzzer_filestore_path(benchmark_df, fuzzer)
+        src_file = posixpath.join(src_filestore_path, 'coverage', 'data',
+                                  benchmark, fuzzer, 'covered_regions.json')
+        filestore_utils.cp(src_file, dst_file)
+        with open(dst_file) as json_file:
+            return json.load(json_file)
+
+
+def get_unique_region_dict(benchmark_coverage_dict):
+    """Returns a dictionary containing the covering fuzzers for each
+    unique region, where the |threshold| defines which regions are unique."""
+    region_dict = collections.defaultdict(list)
+    unique_region_dict = {}
+    threshold_count = 1
+    for fuzzer in benchmark_coverage_dict:
+        for region in benchmark_coverage_dict[fuzzer]:
+            region_dict[region].append(fuzzer)
+    for region, fuzzers in region_dict.items():
+        if len(fuzzers) <= threshold_count:
+            unique_region_dict[region] = fuzzers
+    return unique_region_dict
+
+
+def get_unique_region_cov_df(unique_region_dict, fuzzer_names):
+    """Returns a DataFrame where the two columns are fuzzers and the number
+    of unique regions covered."""
+    fuzzers = collections.defaultdict(int)
+    for region in unique_region_dict:
+        for fuzzer in unique_region_dict[region]:
+            fuzzers[fuzzer] += 1
+    dict_to_transform = {'fuzzer': [], 'unique_regions_covered': []}
+    for fuzzer in fuzzer_names:
+        covered_num = fuzzers[fuzzer]
+        dict_to_transform['fuzzer'].append(fuzzer)
+        dict_to_transform['unique_regions_covered'].append(covered_num)
+    return pd.DataFrame(dict_to_transform)
+
+
+def get_benchmark_cov_dict(coverage_dict, benchmark):
+    """Returns a dictionary to store the covered regions of each fuzzer.
+    Uses a set of tuples to store the covered regions."""
+    benchmark_cov_dict = {}
+    for key_pair, covered_regions in coverage_dict.items():
+        current_fuzzer, current_benchmark = key_pair.split()
+        if current_benchmark == benchmark:
+            covered_regions_in_set = set()
+            for region in covered_regions:
+                covered_regions_in_set.add(tuple(region))
+            benchmark_cov_dict[current_fuzzer] = covered_regions_in_set
+    return benchmark_cov_dict
+
+
+def get_benchmark_aggregated_cov_df(benchmark_coverage_dict):
+    """Returns a dataframe where each row represents a fuzzer and its
+    aggregated coverage number."""
+    dict_to_transform = {'fuzzer': [], 'aggregated_edges_covered': []}
+    for fuzzer in benchmark_coverage_dict:
+        aggregated_edges_covered = len(benchmark_coverage_dict[fuzzer])
+        dict_to_transform['fuzzer'].append(fuzzer)
+        dict_to_transform['aggregated_edges_covered'].append(
+            aggregated_edges_covered)
+    return pd.DataFrame(dict_to_transform)
+
+
+def get_pairwise_unique_coverage_table(benchmark_coverage_dict):
+    """Returns a table that shows the unique coverage between
+    each pair of fuzzers.
+
+    The pairwise unique coverage table is a square matrix where each
+    row and column represents a fuzzer, and each cell contains a number
+    showing the regions covered by the fuzzer of the column but not by
+    the fuzzer of the row."""
+
+    fuzzers = benchmark_coverage_dict.keys()
+
+    pairwise_unique_coverage_values = []
+    for fuzzer_in_row in fuzzers:
+        row = []
+        for fuzzer_in_col in fuzzers:
+            pairwise_unique_coverage_value = get_unique_covered_percentage(
+                benchmark_coverage_dict[fuzzer_in_row],
+                benchmark_coverage_dict[fuzzer_in_col])
+            row.append(pairwise_unique_coverage_value)
+        pairwise_unique_coverage_values.append(row)
+
+    return pd.DataFrame(pairwise_unique_coverage_values,
+                        index=fuzzers,
+                        columns=fuzzers)
+
+
+def get_unique_covered_percentage(fuzzer_row_covered_regions,
+                                  fuzzer_col_covered_regions):
+    """Returns the number of regions covered by the fuzzer of the column
+    but not by the fuzzer of the row."""
+
+    unique_region_count = 0
+    for region in fuzzer_col_covered_regions:
+        if region not in fuzzer_row_covered_regions:
+            unique_region_count += 1
+    return unique_region_count

analysis/data_utils.py

@@ -38,7 +38,8 @@ def validate_data(experiment_df):
 def drop_uninteresting_columns(experiment_df):
     """Returns table with only interesting columns."""
     return experiment_df[[
-        'benchmark', 'fuzzer', 'trial_id', 'time', 'edges_covered'
+        'benchmark', 'fuzzer', 'trial_id', 'time', 'edges_covered',
+        'experiment', 'experiment_filestore'
     ]]
 
 

analysis/experiment_results.py

@@ -21,7 +21,7 @@
 from analysis import stat_tests
 
 
-class ExperimentResults:
+class ExperimentResults:  # pylint: disable=too-many-instance-attributes
     """Provides the main interface for getting various analysis results and
     plots about an experiment, represented by |experiment_df|.
 
@@ -31,11 +31,13 @@ class ExperimentResults:
     template, only the properties needed for the given report will be computed.
     """
 
-    def __init__(self,
-                 experiment_df,
-                 output_directory,
-                 plotter,
-                 experiment_name=None):
+    def __init__(  # pylint: disable=too-many-arguments
+            self,
+            experiment_df,
+            coverage_dict,
+            output_directory,
+            plotter,
+            experiment_name=None):
         if experiment_name:
             self.name = experiment_name
         else:
@@ -63,6 +65,9 @@ def __init__(self,
 
         self._plotter = plotter
 
+        # Dictionary to store the full coverage data.
+        self._coverage_dict = coverage_dict
+
     def _get_full_path(self, filename):
         return os.path.join(self._output_directory, filename)
 
@@ -87,6 +92,7 @@ def benchmarks(self):
         benchmark_names = self._experiment_df.benchmark.unique()
         return [
             benchmark_results.BenchmarkResults(name, self._experiment_df,
+                                               self._coverage_dict,
                                                self._output_directory,
                                                self._plotter)
             for name in sorted(benchmark_names)