rebaseline_server: create ImagePairSet-- holds a number of ImagePairs to examine

gm/rebaseline_server/imagepairset.py

Index: gm/rebaseline_server/imagepairset.py
diff --git a/gm/rebaseline_server/imagepairset.py b/gm/rebaseline_server/imagepairset.py
new file mode 100644
index 0000000000000000000000000000000000000000..d09530bfca2c59e0fd0694299a22725d65fc134e
--- /dev/null
+++ b/gm/rebaseline_server/imagepairset.py
@@ -0,0 +1,125 @@
+#!/usr/bin/python
+
+"""
+Copyright 2014 Google Inc.
+
+Use of this source code is governed by a BSD-style license that can be
+found in the LICENSE file.
+
+ImagePairSet class; see its docstring below.
+"""
+
+import column
+
+# Keys used within dictionary representation of ImagePairSet.
+KEY__COLUMNHEADERS = 'columnHeaders'
+KEY__IMAGEPAIRS = 'imagePairs'
+KEY__IMAGESETS = 'imageSets'
+KEY__IMAGESETS__BASE_URL = 'baseUrl'
+KEY__IMAGESETS__DESCRIPTION = 'description'
+
+
+class ImagePairSet(object):
+  """ A collection of ImagePairs, representing two arbitrary sets of images we

[rmistry, 2014/02/12 19:23:56]

Nit: Single line description please.

[epoger, 2014/02/12 19:59:16]

Done.

+  want to compare.
+
+  These could be:
+  - images generated before and after a code patch
+  - expected and actual images for some tests
+  - or any other pairwise set of images.
+  """
+
+  def __init__(self, descriptions=None):
+    """
+    Args:
+      descriptions: a (string, string) tuple describing the two image sets

[rmistry, 2014/02/12 19:23:56]

Should we mention here what is used if descriptions is not provided?

[epoger, 2014/02/12 19:59:16]

Done.

[epoger, 2014/02/12 19:59:16]

Done.

+    """
+    self._column_header_factories = {}
+    self._descriptions = descriptions or ('setA', 'setB')
+    self._extra_column_tallies = {}  # maps column_id -> values
+                                     #                -> instances_per_value
+    self._image_pair_dicts = []
+
+  def add_image_pair(self, image_pair):
+    """Add an ImagePair; this may be repeated any number of times."""
+    # Special handling when we add the first ImagePair...
+    if not self._image_pair_dicts:
+      self._base_url = image_pair.base_url
+
+    if image_pair.base_url != self._base_url:
+      raise Exception('added ImagePair with base_url "%s" instead of "%s"' % (
+          image_pair.base_url, self._base_url))
+    self._image_pair_dicts.append(image_pair.as_dict())
+    extra_columns_dict = image_pair.extra_columns_dict
+    if extra_columns_dict:
+      for column_id, value in extra_columns_dict.iteritems():
+        self._add_extra_column_entry(column_id, value)
+
+  def set_column_header_factory(self, column_id, column_header_factory):
+    """Override the default settings for one of the extraColumn headers.
+
+    Args:
+      column_id: string; unique ID of this column (must match a key within
+          an ImagePair's extra_columns dictionary)
+      column_header_factory: a ColumnHeaderFactory object
+    """
+    self._column_header_factories[column_id] = column_header_factory
+
+  def get_column_header_factory(self, column_id):
+    """Returns the appropriate ColumnHeaderFactory object for assembling a
+    header for a particular extraColumn.

[rmistry, 2014/02/12 19:23:56]

Nit: Single line please

[epoger, 2014/02/12 19:59:16]

Done.

+
+    Args:
+      column_id: string; unique ID of this column (must match a key within
+          an ImagePair's extra_columns dictionary)
+    """
+    try:
+      return self._column_header_factories[column_id]

[rmistry, 2014/02/12 19:23:56]

How about something like this instead:

if not self._column_header_factories.has_key(column_id):
  self._column_header_factories[column_id] =
column.ColumnHeaderFactory(header_text=column_id)
return self._column_header_factories[column_id]

[epoger, 2014/02/12 19:59:16]

I can do that if you like, but I figured the existing pattern might be more
resilient against race conditions.  (For example, if another thread removed this
column_id from the dictionary AFTER the has_key() check but BEFORE the return, a
KeyError would be raised.)

But I'm sure there are plenty of other parts of this code that are subject to
race conditions anyway.

Do you think there is a significant performance penalty associated with the
exception handling?  If so, I could do it this way:

column_header_factory = self._column_header_factories.get(column_id, None)
if not column_header_factory:
  column_header_factory = column.ColumnHeaderFactory(header_text=column_id)
  self._column_header_factories[column_id] = column_header_factory
return column_header_factory


Actually, that's one line shorter than what I have now.  I'll do that, unless
you disapprove.

[rmistry, 2014/02/13 12:10:02]

No I do not think there is significant performance penalty associated with the
exception handling. I recommended the other block just because it was one line
shorter and your new block is one line shorter so that LGTM.

+    except KeyError:
+      self._column_header_factories[column_id] = column.ColumnHeaderFactory(
+          header_text=column_id)
+      return self._column_header_factories[column_id]
+
+  def _add_extra_column_entry(self, column_id, value):
+    """Record one column_id/value extraColumns pair found within an ImagePair.
+    We use this information to generate tallies within the column header

[rmistry, 2014/02/12 19:23:56]

Nit: newline between the single line description and the larger one please.

[epoger, 2014/02/12 19:59:16]

Done.

+    (how many instances we saw of a particular value, within a particular
+    extraColumn).
+    """
+    try:
+      known_values_for_column = self._extra_column_tallies[column_id]
+    except KeyError:
+      known_values_for_column = {}
+      self._extra_column_tallies[column_id] = known_values_for_column
+    instances_of_this_value = known_values_for_column.get(value, 0)
+    instances_of_this_value += 1
+    known_values_for_column[value] = instances_of_this_value
+
+  def _column_headers_as_dict(self):
+    """
+    Return all column headers as a dictionary, suitable for returning
+    within the ImagePairSet dictionary representation.
+    """
+    asdict = {}
+    for column_id, values_for_column in self._extra_column_tallies.iteritems():
+      column_header_factory = self.get_column_header_factory(column_id)
+      asdict[column_id] = column_header_factory.create_as_dict(
+          values_for_column)
+    return asdict
+
+  def as_dict(self):
+    """
+    Return a dictionary describing this package of ImagePairs, as needed when
+    constructing the JSON representation.  Uses the KEY__* constants as keys.
+    """
+    return {
+        KEY__COLUMNHEADERS: self._column_headers_as_dict(),
+        KEY__IMAGEPAIRS: self._image_pair_dicts,
+        KEY__IMAGESETS: [{
+            KEY__IMAGESETS__BASE_URL: self._base_url,
+            KEY__IMAGESETS__DESCRIPTION: self._descriptions[0],
+        }, {
+            KEY__IMAGESETS__BASE_URL: self._base_url,
+            KEY__IMAGESETS__DESCRIPTION: self._descriptions[1],
+        }],
+    }