gm/rebaseline_server/results.py

#!/usr/bin/python

"""
Copyright 2013 Google Inc.

Use of this source code is governed by a BSD-style license that can be
found in the LICENSE file.

Repackage expected/actual GM results as needed by our HTML rebaseline viewer.
"""

# System-level imports
import argparse
import fnmatch
import json
import logging
import os
import re
import sys
import time

# Imports from within Skia
#
# We need to add the 'gm' directory, so that we can import gm_json.py within
# that directory.  That script allows us to parse the actual-results.json file
# written out by the GM tool.
# Make sure that the 'gm' dir is in the PYTHONPATH, but add it at the *end*
# so any dirs that are already in the PYTHONPATH will be preferred.
PARENT_DIRECTORY = os.path.dirname(os.path.realpath(__file__))
GM_DIRECTORY = os.path.dirname(PARENT_DIRECTORY)
if GM_DIRECTORY not in sys.path:
  sys.path.append(GM_DIRECTORY)
import gm_json
import imagediffdb

IMAGE_FILENAME_RE = re.compile(gm_json.IMAGE_FILENAME_PATTERN)
IMAGE_FILENAME_FORMATTER = '%s_%s.png'  # pass in (testname, config)

FIELDS_PASSED_THRU_VERBATIM = [
    gm_json.JSONKEY_EXPECTEDRESULTS_BUGS,
    gm_json.JSONKEY_EXPECTEDRESULTS_IGNOREFAILURE,
    gm_json.JSONKEY_EXPECTEDRESULTS_REVIEWED,
]
CATEGORIES_TO_SUMMARIZE = [
    'builder', 'test', 'config', 'resultType',
    gm_json.JSONKEY_EXPECTEDRESULTS_IGNOREFAILURE,
    gm_json.JSONKEY_EXPECTEDRESULTS_REVIEWED,
]

RESULTS_ALL = 'all'
RESULTS_FAILURES = 'failures'

class Results(object):
  """ Loads actual and expected results from all builders, supplying combined
  reports as requested.

  Once this object has been constructed, the results (in self._results[])
  are immutable.  If you want to update the results based on updated JSON
  file contents, you will need to create a new Results object."""

  def __init__(self, actuals_root, expected_root, generated_images_root):
    """
    Args:
      actuals_root: root directory containing all actual-results.json files
      expected_root: root directory containing all expected-results.json files
      generated_images_root: directory within which to create all pixel diffs;
          if this directory does not yet exist, it will be created
    """
    time_start = int(time.time())
    self._image_diff_db = imagediffdb.ImageDiffDB(generated_images_root)
    self._actuals_root = actuals_root
    self._expected_root = expected_root
    self._load_actual_and_expected()
    self._timestamp = int(time.time())
    logging.info('Results complete; took %d seconds.' %
                 (self._timestamp - time_start))

  def get_timestamp(self):
    """Return the time at which this object was created, in seconds past epoch
    (UTC).
    """
    return self._timestamp

  def edit_expectations(self, modifications):
    """Edit the expectations stored within this object and write them back
    to disk.

    Note that this will NOT update the results stored in self._results[] ;
    in order to see those updates, you must instantiate a new Results object
    based on the (now updated) files on disk.

    Args:
      modifications: a list of dictionaries, one for each expectation to update:

         [
           {
             'builder': 'Test-Mac10.6-MacMini4.1-GeForce320M-x86-Debug',
             'test': 'bigmatrix',
             'config': '8888',
             'expectedHashType': 'bitmap-64bitMD5',
             'expectedHashDigest': '10894408024079689926',
             'bugs': [123, 456],
             'ignore-failure': false,
             'reviewed-by-human': true,
           },
           ...
         ]

    """
    expected_builder_dicts = Results._read_dicts_from_root(self._expected_root)
    for mod in modifications:
      image_name = IMAGE_FILENAME_FORMATTER % (mod['test'], mod['config'])
      # TODO(epoger): assumes a single allowed digest per test
      allowed_digests = [[mod['expectedHashType'],
                          int(mod['expectedHashDigest'])]]
      new_expectations = {
          gm_json.JSONKEY_EXPECTEDRESULTS_ALLOWEDDIGESTS: allowed_digests,
      }
      for field in FIELDS_PASSED_THRU_VERBATIM:
        value = mod.get(field)
        if value is not None:
          new_expectations[field] = value
      builder_dict = expected_builder_dicts[mod['builder']]
      builder_expectations = builder_dict.get(gm_json.JSONKEY_EXPECTEDRESULTS)
      if not builder_expectations:
        builder_expectations = {}
        builder_dict[gm_json.JSONKEY_EXPECTEDRESULTS] = builder_expectations
      builder_expectations[image_name] = new_expectations
    Results._write_dicts_to_root(expected_builder_dicts, self._expected_root)

  def get_results_of_type(self, type):
    """Return results of some/all tests (depending on 'type' parameter).

    Args:
      type: string describing which types of results to include; must be one
            of the RESULTS_* constants

    Results are returned as a dictionary in this form:

       {
         'categories': # dictionary of categories listed in
                       # CATEGORIES_TO_SUMMARIZE, with the number of times
                       # each value appears within its category
         {
           'resultType': # category name
           {
             'failed': 29, # category value and total number found of that value
             'failure-ignored': 948,
             'no-comparison': 4502,
             'succeeded': 38609,
           },
           'builder':
           {
             'Test-Mac10.6-MacMini4.1-GeForce320M-x86-Debug': 1286,
             'Test-Mac10.6-MacMini4.1-GeForce320M-x86-Release': 1134,
             ...
           },
           ... # other categories from CATEGORIES_TO_SUMMARIZE
         }, # end of 'categories' dictionary

         'testData': # list of test results, with a dictionary for each
         [
           {
             'resultType': 'failed',
             'builder': 'Test-Mac10.6-MacMini4.1-GeForce320M-x86-Debug',
             'test': 'bigmatrix',
             'config': '8888',
             'expectedHashType': 'bitmap-64bitMD5',
             'expectedHashDigest': '10894408024079689926',
             'actualHashType': 'bitmap-64bitMD5',
             'actualHashDigest': '2409857384569',
             'bugs': [123, 456],
             'ignore-failure': false,
             'reviewed-by-human': true,
           },
           ...
         ], # end of 'testData' list
       }
    """
    return self._results[type]

  @staticmethod
  def _ignore_builder(builder):
    """Returns True if we should ignore expectations and actuals for a builder.

    This allows us to ignore builders for which we don't maintain expectations
    (trybots, Valgrind, ASAN, TSAN), and avoid problems like
    https://code.google.com/p/skia/issues/detail?id=2036 ('rebaseline_server
    produces error when trying to add baselines for ASAN/TSAN builders')

    Args:
      builder: name of this builder, as a string

    Returns:
      True if we should ignore expectations and actuals for this builder.
    """
    return (builder.endswith('-Trybot') or
            ('Valgrind' in builder) or
            ('TSAN' in builder) or
            ('ASAN' in builder))

  @staticmethod
  def _read_dicts_from_root(root, pattern='*.json'):
    """Read all JSON dictionaries within a directory tree.

    Args:
      root: path to root of directory tree
      pattern: which files to read within root (fnmatch-style pattern)

    Returns:
      A meta-dictionary containing all the JSON dictionaries found within
      the directory tree, keyed by the builder name of each dictionary.

    Raises:
      IOError if root does not refer to an existing directory
    """
    if not os.path.isdir(root):
      raise IOError('no directory found at path %s' % root)
    meta_dict = {}
    for dirpath, dirnames, filenames in os.walk(root):
      for matching_filename in fnmatch.filter(filenames, pattern):
        builder = os.path.basename(dirpath)
        if Results._ignore_builder(builder):
          continue
        fullpath = os.path.join(dirpath, matching_filename)
        meta_dict[builder] = gm_json.LoadFromFile(fullpath)
    return meta_dict

  @staticmethod
  def _write_dicts_to_root(meta_dict, root, pattern='*.json'):
    """Write all per-builder dictionaries within meta_dict to files under
    the root path.

    Security note: this will only write to files that already exist within
    the root path (as found by os.walk() within root), so we don't need to
    worry about malformed content writing to disk outside of root.
    However, the data written to those files is not double-checked, so it
    could contain poisonous data.

    Args:
      meta_dict: a builder-keyed meta-dictionary containing all the JSON
                 dictionaries we want to write out
      root: path to root of directory tree within which to write files
      pattern: which files to write within root (fnmatch-style pattern)

    Raises:
      IOError if root does not refer to an existing directory
      KeyError if the set of per-builder dictionaries written out was
               different than expected
    """
    if not os.path.isdir(root):
      raise IOError('no directory found at path %s' % root)
    actual_builders_written = []
    for dirpath, dirnames, filenames in os.walk(root):
      for matching_filename in fnmatch.filter(filenames, pattern):
        builder = os.path.basename(dirpath)
        if Results._ignore_builder(builder):
          continue
        per_builder_dict = meta_dict.get(builder)
        if per_builder_dict is not None:
          fullpath = os.path.join(dirpath, matching_filename)
          gm_json.WriteToFile(per_builder_dict, fullpath)
          actual_builders_written.append(builder)

    # Check: did we write out the set of per-builder dictionaries we
    # expected to?
    expected_builders_written = sorted(meta_dict.keys())
    actual_builders_written.sort()
    if expected_builders_written != actual_builders_written:
      raise KeyError(
          'expected to write dicts for builders %s, but actually wrote them '
          'for builders %s' % (
              expected_builders_written, actual_builders_written))

  def _generate_pixel_diffs_if_needed(self, test, expected_image, actual_image):
    """If expected_image and actual_image both exist but are different,
    add the image pair to self._image_diff_db and generate pixel diffs.

    Args:
      test: string; name of test
      expected_image: (hashType, hashDigest) tuple describing the expected image
      actual_image: (hashType, hashDigest) tuple describing the actual image
    """
    if expected_image == actual_image:
      return

    (expected_hashtype, expected_hashdigest) = expected_image
    (actual_hashtype, actual_hashdigest) = actual_image
    if None in [expected_hashtype, expected_hashdigest,
                actual_hashtype, actual_hashdigest]:
      return

    expected_url = gm_json.CreateGmActualUrl(
        test_name=test, hash_type=expected_hashtype,
        hash_digest=expected_hashdigest)
    actual_url = gm_json.CreateGmActualUrl(
        test_name=test, hash_type=actual_hashtype,
        hash_digest=actual_hashdigest)
    self._image_diff_db.add_image_pair(
        expected_image_locator=expected_hashdigest,
        expected_image_url=expected_url,
        actual_image_locator=actual_hashdigest,
        actual_image_url=actual_url)

  def _load_actual_and_expected(self):
    """Loads the results of all tests, across all builders (based on the
    files within self._actuals_root and self._expected_root),
    and stores them in self._results.
    """
    logging.info('Reading actual-results JSON files from %s...' %
                 self._actuals_root)
    actual_builder_dicts = Results._read_dicts_from_root(self._actuals_root)
    logging.info('Reading expected-results JSON files from %s...' %
                 self._expected_root)
    expected_builder_dicts = Results._read_dicts_from_root(self._expected_root)

    categories_all = {}
    categories_failures = {}

    Results._ensure_included_in_category_dict(categories_all,
                                              'resultType', [
        gm_json.JSONKEY_ACTUALRESULTS_FAILED,
        gm_json.JSONKEY_ACTUALRESULTS_FAILUREIGNORED,
        gm_json.JSONKEY_ACTUALRESULTS_NOCOMPARISON,
        gm_json.JSONKEY_ACTUALRESULTS_SUCCEEDED,
        ])
    Results._ensure_included_in_category_dict(categories_failures,
                                              'resultType', [
        gm_json.JSONKEY_ACTUALRESULTS_FAILED,
        gm_json.JSONKEY_ACTUALRESULTS_FAILUREIGNORED,
        gm_json.JSONKEY_ACTUALRESULTS_NOCOMPARISON,
        ])

    data_all = []
    data_failures = []
    builders = sorted(actual_builder_dicts.keys())
    num_builders = len(builders)
    builder_num = 0
    for builder in builders:
      builder_num += 1
      logging.info('Generating pixel diffs for builder #%d of %d, "%s"...' %
                   (builder_num, num_builders, builder))
      actual_results_for_this_builder = (
          actual_builder_dicts[builder][gm_json.JSONKEY_ACTUALRESULTS])
      for result_type in sorted(actual_results_for_this_builder.keys()):
        results_of_this_type = actual_results_for_this_builder[result_type]
        if not results_of_this_type:
          continue
        for image_name in sorted(results_of_this_type.keys()):
          actual_image = results_of_this_type[image_name]

          # Default empty expectations; overwrite these if we find any real ones
          expectations_per_test = None
          expected_image = [None, None]
          try:
            expectations_per_test = (
                expected_builder_dicts
                [builder][gm_json.JSONKEY_EXPECTEDRESULTS][image_name])
            # TODO(epoger): assumes a single allowed digest per test
            expected_image = (
                expectations_per_test
                [gm_json.JSONKEY_EXPECTEDRESULTS_ALLOWEDDIGESTS][0])
          except (KeyError, TypeError):
            # There are several cases in which we would expect to find
            # no expectations for a given test:
            #
            # 1. result_type == NOCOMPARISON
            #   There are no expectations for this test yet!
            #
            # 2. alternate rendering mode failures (e.g. serialized)
            #   In cases like
            #   https://code.google.com/p/skia/issues/detail?id=1684
            #   ('tileimagefilter GM test failing in serialized render mode'),
            #   the gm-actuals will list a failure for the alternate
            #   rendering mode even though we don't have explicit expectations
            #   for the test (the implicit expectation is that it must
            #   render the same in all rendering modes).
            #
            # Don't log type 1, because it is common.
            # Log other types, because they are rare and we should know about
            # them, but don't throw an exception, because we need to keep our
            # tools working in the meanwhile!
            if result_type != gm_json.JSONKEY_ACTUALRESULTS_NOCOMPARISON:
              logging.warning('No expectations found for test: %s' % {
                  'builder': builder,
                  'image_name': image_name,
                  'result_type': result_type,
                  })

          # If this test was recently rebaselined, it will remain in
          # the 'failed' set of actuals until all the bots have
          # cycled (although the expectations have indeed been set
          # from the most recent actuals).  Treat these as successes
          # instead of failures.
          #
          # TODO(epoger): Do we need to do something similar in
          # other cases, such as when we have recently marked a test
          # as ignoreFailure but it still shows up in the 'failed'
          # category?  Maybe we should not rely on the result_type
          # categories recorded within the gm_actuals AT ALL, and
          # instead evaluate the result_type ourselves based on what
          # we see in expectations vs actual checksum?
          if expected_image == actual_image:
            updated_result_type = gm_json.JSONKEY_ACTUALRESULTS_SUCCEEDED
          else:
            updated_result_type = result_type

          (test, config) = IMAGE_FILENAME_RE.match(image_name).groups()
          self._generate_pixel_diffs_if_needed(
              test=test, expected_image=expected_image,
              actual_image=actual_image)
          results_for_this_test = {
              'resultType': updated_result_type,
              'builder': builder,
              'test': test,
              'config': config,
              'actualHashType': actual_image[0],
              'actualHashDigest': str(actual_image[1]),
              'expectedHashType': expected_image[0],
              'expectedHashDigest': str(expected_image[1]),

              # FIELDS_PASSED_THRU_VERBATIM that may be overwritten below...
              gm_json.JSONKEY_EXPECTEDRESULTS_IGNOREFAILURE: False,
          }
          if expectations_per_test:
            for field in FIELDS_PASSED_THRU_VERBATIM:
              results_for_this_test[field] = expectations_per_test.get(field)

          if updated_result_type == gm_json.JSONKEY_ACTUALRESULTS_NOCOMPARISON:
            pass # no diff record to calculate at all
          elif updated_result_type == gm_json.JSONKEY_ACTUALRESULTS_SUCCEEDED:
            results_for_this_test['numDifferingPixels'] = 0
            results_for_this_test['percentDifferingPixels'] = 0
            results_for_this_test['weightedDiffMeasure'] = 0
            results_for_this_test['perceptualDifference'] = 0
            results_for_this_test['maxDiffPerChannel'] = 0
          else:
            try:
              diff_record = self._image_diff_db.get_diff_record(
                  expected_image_locator=expected_image[1],
                  actual_image_locator=actual_image[1])
              results_for_this_test['numDifferingPixels'] = (
                  diff_record.get_num_pixels_differing())
              results_for_this_test['percentDifferingPixels'] = (
                  diff_record.get_percent_pixels_differing())
              results_for_this_test['weightedDiffMeasure'] = (
                  diff_record.get_weighted_diff_measure())
              results_for_this_test['perceptualDifference'] = (
                  diff_record.get_perceptual_difference())
              results_for_this_test['maxDiffPerChannel'] = (
                  diff_record.get_max_diff_per_channel())
            except KeyError:
              logging.warning('unable to find diff_record for ("%s", "%s")' %
                              (expected_image[1], actual_image[1]))
              pass

          Results._add_to_category_dict(categories_all, results_for_this_test)
          data_all.append(results_for_this_test)

          # TODO(epoger): In effect, we have a list of resultTypes that we
          # include in the different result lists (data_all and data_failures).
          # This same list should be used by the calls to
          # Results._ensure_included_in_category_dict() earlier on.
          if updated_result_type != gm_json.JSONKEY_ACTUALRESULTS_SUCCEEDED:
            Results._add_to_category_dict(categories_failures,
                                          results_for_this_test)
            data_failures.append(results_for_this_test)

    self._results = {
      RESULTS_ALL:
        {'categories': categories_all, 'testData': data_all},
      RESULTS_FAILURES:
        {'categories': categories_failures, 'testData': data_failures},
    }

  @staticmethod
  def _add_to_category_dict(category_dict, test_results):
    """Add test_results to the category dictionary we are building.
    (See documentation of self.get_results_of_type() for the format of this
    dictionary.)

    Args:
      category_dict: category dict-of-dicts to add to; modify this in-place
      test_results: test data with which to update category_list, in a dict:
         {
           'category_name': 'category_value',
           'category_name': 'category_value',
           ...
         }
    """
    for category in CATEGORIES_TO_SUMMARIZE:
      category_value = test_results.get(category)
      if not category_dict.get(category):
        category_dict[category] = {}
      if not category_dict[category].get(category_value):
        category_dict[category][category_value] = 0
      category_dict[category][category_value] += 1

  @staticmethod
  def _ensure_included_in_category_dict(category_dict,
                                        category_name, category_values):
    """Ensure that the category name/value pairs are included in category_dict,
    even if there aren't any results with that name/value pair.
    (See documentation of self.get_results_of_type() for the format of this
    dictionary.)

    Args:
      category_dict: category dict-of-dicts to modify
      category_name: category name, as a string
      category_values: list of values we want to make sure are represented
                       for this category
    """
    if not category_dict.get(category_name):
      category_dict[category_name] = {}
    for category_value in category_values:
      if not category_dict[category_name].get(category_value):
        category_dict[category_name][category_value] = 0


def main():
  logging.basicConfig(format='%(asctime)s %(levelname)s %(message)s',
                      datefmt='%m/%d/%Y %H:%M:%S',
                      level=logging.INFO)
  parser = argparse.ArgumentParser()
  parser.add_argument(
      '--actuals', required=True,
      help='Directory containing all actual-result JSON files')
  parser.add_argument(
      '--expectations', required=True,
      help='Directory containing all expected-result JSON files')
  parser.add_argument(
      '--outfile', required=True,
      help='File to write result summary into, in JSON format')
  parser.add_argument(
      '--workdir', default='.workdir',
      help='Directory within which to download images and generate diffs')
  args = parser.parse_args()
  results = Results(actuals_root=args.actuals,
                    expected_root=args.expectations,
                    generated_images_root=args.workdir)
  gm_json.WriteToFile(results.get_results_of_type(RESULTS_ALL), args.outfile)


if __name__ == '__main__':
  main()