[Predator] Add ``Project`` class and ``ClassifySuspect`` method to project and component classifier

appengine/findit/crash/project.py

Index: appengine/findit/crash/project.py
diff --git a/appengine/findit/crash/project.py b/appengine/findit/crash/project.py
new file mode 100644
index 0000000000000000000000000000000000000000..b97a119c67fe0b6cd97f0eca1937b4eff2bc2223
--- /dev/null
+++ b/appengine/findit/crash/project.py
@@ -0,0 +1,116 @@
+# Copyright 2017 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 collections import namedtuple
+import os
+import re
+
+from libs.gitiles.diff import ChangeType
+
+# Some projects like "chromium", it has many dependency projects like
+# "chromium-blink", "chromium-skia", "chromium-pdfium"...etc., for those
+# dep projects, the "chromium" ``Project`` can derive the names from their
+# dep paths.
+# Some other projects like "android_os", "clank", they don't have any dependency
+# projects that are relavent to Predator.
+_PROJECTS_WITH_DEP_PROJECTS = ['chromium']
+
+
+# TODO(http://crbug.com/659346): write the coverage tests.
+class Project(namedtuple('Project',
+    ['name', 'path_regexs', 'function_regexs',
+     'host_directories'])):
+  """A representation of a "project".
+
+  For example: 'android_os', 'clank' or 'chromium'. Notably, a project knows
+  how to identify itself. Hence, given a stack frame, file path or dependency
+  path or whatever, we ask the ``Project``  whether it matches that frame,
+  CL, etc.
+
+  Properties:
+    name (str): The name of the project, like "chromium", "android_os".
+    path_regexs (list of re.RegexObject): Patterns of paths that project has.
+    function_regexs (list of re.RegexObject): Patterns of functions that project
+      has.
+    host_directories (list of str): The root directoris of this project and its
+      dependency projects.
+      N.B. If ``host_directories`` is availabe, this project can match
+      it with the passed-in ``dep_path`` to tell whether a suspect or stack is
+      from this project, else if this information is missing, the project cannot
+      tell that from ``dep_path``, however that doesn't mean the suspect or
+      stack does not belong to this project, we can use other information like
+      ``path_regexs`` or ``function_regexs`` to analyze.
+  """
+  __slots__ = ()
+
+  def __new__(cls, name, path_regexs=None,
+              function_regexs=None, host_directories=None):
+    path_regexs = [re.compile(path_regex) for path_regex in
+                   path_regexs] if path_regexs else []
+    function_regexs = [re.compile(function_regex) for function_regex in
+                       function_regexs] if function_regexs else []
+    host_directories = host_directories or []
+
+    return super(cls, Project).__new__(
+        cls, name, path_regexs, function_regexs, host_directories)
+
+  def MatchesStackFrame(self, frame):
+    """Returns true if this project matches the frame."""
+    for path_regex in self.path_regexs:
+      if (path_regex.match(frame.dep_path + frame.file_path) or
+          path_regex.match(frame.raw_file_path)):
+        return True
+
+    for function_regex in self.function_regexs:
+      if function_regex.match(frame.function):
+        return True
+
+    for host_directory in self.host_directories:
+      if frame.dep_path.startswith(host_directory):
+        return True
+
+    return False
+
+  def MatchesTouchedFile(self, dep_path, touched_file):
+    """Returns true if this project matches the file path."""
+    if touched_file.change_type == ChangeType.DELETE:
+      path = touched_file.old_path
+    else:
+      path = touched_file.new_path
+
+    # if the path matches the path patterns.

[lijeffrey, 2017/01/27 15:42:20]

nit: If

[Sharu Jiang, 2017/01/27 18:39:04]

Done.

+    for path_regex in self.path_regexs:
+      if path_regex.match(os.path.join(dep_path, path)):
+        return True
+
+    # If the dep_path hosted by this project.
+    for host_directory in self.host_directories:
+      if dep_path.startswith(host_directory):
+        return True

[lijeffrey, 2017/01/27 15:42:20]

I think this can be simplified slightly to:

return (
    any(path_regex.match(
        os.path.join(dep_path, path) for path_regex in self.path_regexs)) or
    any(dep_path.startswith(
        host_directory) for host_direcotry in self.host_directories)))

[Sharu Jiang, 2017/01/27 18:39:04]

Hm... I want to keep these two conditions separate to have more readability.

+
+    return False
+
+  def GetName(self, dep_path=None):
+    """Returns the project name based on dep path.
+
+    N.B. (1) If this project doesn't have dep projects, just return the project
+    name. (2) If this project does, return the derived dep project name based on
+    the self.host_directories.
+    """
+    if self.name not in _PROJECTS_WITH_DEP_PROJECTS or dep_path is None:
+      return self.name
+
+    # For chromium project, get the name of the sub project from ``dep_path``.
+    for host_directory in self.host_directories:
+      if dep_path.startswith(host_directory):
+        path = dep_path[len(host_directory):]
+        if not path:
+          return self.name
+
+        return '%s-%s' % (self.name, path.split('/')[0].lower())
+
+    # Unknown path, return the whole path as project name.
+    return '%s-%s' % (self.name,
+                      '_'.join([dep_part for dep_part in dep_path.split('/')
+                                if dep_part]))