[Findit] Plain objects to represent the returned result from running the algorithm,

tools/findit/blame.py

+# Copyright (c) 2014 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.
+
+from threading import Lock, Thread
+
+import crash_utils
+import gitparser
+import svnparser
+import utils
+
+
+class Blame(object):
+  """Represents a blame object.
+
+  The object contains blame information for one line of stack, and this
+  information is shown when there are no CLs that change the crashing files.
+  Attributes:
+    content: The content of the line to find the blame for.
+    component: The component this line is in.
+    stack_frame_index: The stack frame index of this file.
+    file_name: The name of the file.
+    line_number: The line that caused a crash.
+    author: The author of this line on the latest revision.
+    crash_revision: The revision that caused the crash.
+    revision: The latest revision of this line before the crash revision.
+    url: The url of this revision.
+    regression: The regression range of the component, if it exists.
+
+  """
+
+  def __init__(self, content, component, stack_frame_index, file_name,
+               line_number, author, crash_revision, revision, url, regression):
+    # Set all the variables from the arguments.
+    self.content = content
+    self.component = component
+    self.stack_frame_index = stack_frame_index
+    self.file = file_name
+    self.line_number = line_number
+    self.author = author
+    self.revision = revision
+    self.url = url
+    self.regression = regression
+    self.distance = float('inf')
+    revision = int(revision)
+
+    # Calculate the distance, where it measures how far the last revision is
+    # from the regression range.
+    if regression:
+      range1 = int(regression[0])
+      range2 = int(regression[1])
+      self.distance = min(abs(range1 - revision), abs(range2 - revision))
+
+    # If the regression is in SVN but it does not have regression info, check
+    # how far the last revision is from crash revision.
+    elif not utils.IsGitHash(crash_revision):
+      self.distance = abs(int(crash_revision) - revision)
+
+
+class BlameList(object):
+  """Represents a list of blame objects.
+
+  Thread-safe.
+  """
+
+  def __init__(self):
+    self.blame_list = []
+    self.blame_list_lock = Lock()
+
+  def __getitem__(self, index):
+    return self.blame_list[index]
+
+  def AddBlame(self, blame):
+    """Adds blame object to the set."""
+    with self.blame_list_lock:
+      self.blame_list.append(blame)
+
+  def sort(self, key=None):
+    return self.blame_list.sort(key=key)
+
+  def FindBlame(self, callstack, crash_revision_dic, revision_range_dic,
+                url_map, n=10):
+    """Given a stack within a stacktrace, retrieves blame information.
+
+    Only either first 10 or the length of stack, whichever is shorter,

[stgao, 2014/08/09 19:51:02]

10 here is invalid. Same for the in-line comment below.

Whenever you make code change, please ensure comments and documentation are
updated.
This seems to become a common issues in most of your code.

[jeun, 2014/08/12 18:17:33]

Done.

+    results are returned.
+
+    Args:
+      callstack: The list of stack frames.
+      crash_revision_dic: A dictionary that maps component to its crash
+                          revision.
+      revision_range_dic: A dictionary that maps component to its revision
+                          range.
+      url_map: A map from repository type to urls.
+      n: A number of stack frames to show the blame result for.
+    """
+    # Only return blame information for first 10 frames.
+    stackframes = callstack.GetFirstN(n)
+
+    threads = []
+    # Iterate through frames in stack.
+    for stack_frame in stackframes:

[stgao, 2014/08/09 19:51:02]

Naming is not that consistent.
|stackframes| -> |stack_frames|

[jeun, 2014/08/12 18:17:33]

Done.

+      # If the component this line is from does not have a crash revision,
+      # It is not possible to get blame information so ignore this line.
+      component = stack_frame.component
+      if component not in crash_revision_dic:
+        continue
+
+      crash_revision = crash_revision_dic[component]
+      regression = None
+      is_svn = not utils.IsGitHash(crash_revision)
+
+      # If the revision is in SVN, and if revision information is available,
+      # get it. Not for Git because we cannot calculate the distance.
+      if is_svn:
+        repository_parser = svnparser.SVNParser(url_map['svn'])
+        if revision_range_dic and component in revision_range_dic:
+          regression = revision_range_dic[component]
+      else:
+        repository_parser = gitparser.GitParser(url_map['git'])
+
+      # Generate blame entry, one thread for one entry.
+      blame_thread = Thread(
+          target=self.__GenerateBlameEntry,
+          args=[repository_parser, stack_frame, crash_revision, regression])
+      threads.append(blame_thread)
+      blame_thread.start()
+
+    # Join the results before returning.
+    for blame_thread in threads:
+      blame_thread.join()
+
+  def __GenerateBlameEntry(self, repository_parser, stack_frame,
+                           crash_revision, regression):
+    """Generates blame list from the arguments."""
+    stack_frame_index = stack_frame.index
+    component = stack_frame.component
+    file_name = stack_frame.file_name
+    file_path = stack_frame.file_path
+    line = stack_frame.crashed_line_number
+
+    # Parse blame information.
+    parsed_blame_info = repository_parser.ParseBlameInfo(
+        component, file_path, line, crash_revision)
+
+    # If it fails to retrieve information, do not do anything.
+    if not parsed_blame_info:
+      return
+
+    # Create blame object from the parsed info and add it to the list.
+    (content, revision, author, url) = parsed_blame_info
+    blame = Blame(content, component, stack_frame_index, file_name, line,
+                  author, crash_revision, revision, url, regression)
+    self.AddBlame(blame)
+
+  def PrettifyBlame(self):
+    """Returns a string representation of blame results.
+
+    Returns:
+      A string representation of blame_list.
+    """
+    blame_list = self.blame_list
+    return_string = []
+    return_string.append('Below are the blame information of the stacktrace.\n')
+    return_string.append('[\n')
+
+    # If blame info is not available, return the message.
+    if not blame_list:
+      return_string.append('No Blame Information Available.\n')
+      return ''.join(return_string)
+
+    # Sort the blame list by its distance, and its position in stack.
+    blame_list.sort(key=lambda blame: (blame.distance,
+                                       blame.stack_frame_index))
+
+    for blame in blame_list:
+      # If regression information is available, do some filtering.
+      if blame.regression:
+        regressed = blame.regression
+
+        # Discards results that are too far from the regression range.
+        # For example, if regression is 10000:11000, it is very not
+        # likely that a commit from revision 1000 would have caused a crash.
+        if blame.distance > int(regressed[0]) / 4 and blame.distance > int(
+            regressed[1]) / 4:
+          continue
+
+      return_string.append('  {\n')
+      return_string.append(
+          '    suspected_cl: %s\n' % crash_utils.AddHyperlink(
+              blame.revision,
+              blame.url))
+      return_string.append('    component: %s\n' % blame.component)
+      return_string.append('    owner: %s\n' % blame.author)
+      reason = (
+          'The CL changes line %s of file %s from stack %d.' %
+          (blame.line_number, blame.file, blame.stack_frame_index))
+      return_string.append('    reason: %s\n' % reason)
+      if blame.content:
+        return_string.append('    content: %s\n' % blame.content)
+      return_string.append('  }\n')
+
+    return_string.append(']\n')
+    return_string.append('-------------------------------------------\n')
+    return ''.join(return_string)