gen_bench_expectations to generate bench expectations files.

bench/bench_util.py

 '''
 Created on May 19, 2011
 
 @author: bungeman
 '''
 
+import os
 import re
 import math
 
 # bench representation algorithm constant names
 ALGORITHM_AVERAGE = 'avg'
 ALGORITHM_MEDIAN = 'med'
 ALGORITHM_MINIMUM = 'min'
 ALGORITHM_25TH_PERCENTILE = '25th'
 
 # Regular expressions used throughout.
@@ skipping 123 matching lines @@
         for new_time in TIME_RE_COMPILED.finditer(times):
             current_time_type = new_time.group(1)
             iters = [float(i) for i in
                      new_time.group(2).strip().split(',')]
             value_dic.setdefault(bench, {}).setdefault(
                 current_config, {}).setdefault(current_time_type, []).append(
                     iters)
             layout_dic.setdefault(bench, {}).setdefault(
                 current_config, {}).setdefault(current_time_type, tile_layout)
 
+def parse_skp_bench_data(directory, revision, rep, default_settings={}):

[borenet, 2014/03/17 17:43:05]

Using an instance as a default is bad news.  Instead, set it to None and do:

if not default_settings:
  default_settings = {}

[benchen, 2014/03/17 19:02:40]

Done.

+    """Parses all the skp bench data in the given directory.
+
+    Args:
+      directory: string of path to input data directory.
+      revision: git hash revision that matches the data to process.
+      rep: bench representation algorithm, see bench_util.py.
+      default_settings: dictionary of other run settings.
+
+    Returns:
+      A list of BenchDataPoint objects.
+    """
+    revision_data_points = []
+    file_list = os.listdir(directory)
+    file_list.sort()
+    for bench_file in file_list:
+        scalar_type = None
+        # Scalar type, if any, is in the bench filename after 'scalar_'.
+        if (bench_file.startswith('bench_' + revision + '_data_')):
+            if bench_file.find('scalar_') > 0:
+                components = bench_file.split('_')
+                scalar_type = components[components.index('scalar') + 1]
+        else:  # Skips non skp bench files.
+            continue
+
+        file_handle = open(directory + '/' + bench_file, 'r')

[borenet, 2014/03/17 17:43:05]

Please use the "with" syntax:

with open('/'.join(directory, bench_file), 'r') as file_handle:
  ...

[benchen, 2014/03/17 19:02:40]

ah just copied it over. Done.
On 2014/03/17 17:43:05, borenet wrote:

+
+        default_settings['scalar'] = scalar_type

[borenet, 2014/03/17 17:43:05]

Should we be modifying something that the caller has passed in?  If so, we need
to document that this is happening.  Otherwise, let's make a copy or avoid
modifying it.

[benchen, 2014/03/17 19:02:40]

Good suggestion. Done.
On 2014/03/17 17:43:05, borenet wrote:

+        revision_data_points.extend(
+                        parse(default_settings, file_handle, rep))
+        file_handle.close()
+
+    return revision_data_points
+
 # TODO(bensong): switch to reading JSON output when available. This way we don't
 # need the RE complexities.
 def parse(settings, lines, representation=None):
     """Parses bench output into a useful data structure.
 
     ({str:str}, __iter__ -> str) -> [BenchDataPoint]
     representation is one of the ALGORITHM_XXX types."""
 
     benches = []
     current_bench = None
@@ skipping 153 matching lines @@
     """
     return '<a href="http://code.google.com/p/skia/source/detail?r=%s">%s</a>'%(
         revision_number, revision_number)
 
 def main():
     foo = [[0.0, 0.0], [0.0, 1.0], [0.0, 2.0], [0.0, 3.0]]
     LinearRegression(foo)
 
 if __name__ == "__main__":
     main()