tools: Add a tool to dump the resource_prefetch_predictor database.

tools/resource_prefetch_predictor/prefetch_predictor_tool.py

Index: tools/resource_prefetch_predictor/prefetch_predictor_tool.py
diff --git a/tools/resource_prefetch_predictor/prefetch_predictor_tool.py b/tools/resource_prefetch_predictor/prefetch_predictor_tool.py
new file mode 100755
index 0000000000000000000000000000000000000000..01abed8b42a1e821bfffffdf98cdd03cbd1d1a25
--- /dev/null
+++ b/tools/resource_prefetch_predictor/prefetch_predictor_tool.py
@@ -0,0 +1,82 @@
+#!/usr/bin/python
+# Copyright 2016 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.
+
+"""Inspection of the prefetch predictor database."""
+
+import argparse
+import operator
+import sqlite3
+
+class ResourceType(object):
+  STYLESHEET = 2
+  SCRIPT = 3
+
+
+class Entry(object):
+  def __init__(
+      self, main_page_url, resource_url, resource_type, number_of_hits,
+      number_of_misses, consecutive_misses, average_position):
+    self.main_page_url = main_page_url
+    self.resource_url = resource_url
+    self.resource_type = resource_type
+    self.number_of_hits = int(number_of_hits)
+    self.number_of_misses = int(number_of_misses)
+    self.consecutive_misses = int(consecutive_misses)
+    self.average_position = int(average_position)
+    self.confidence = float(number_of_hits) / (
+        number_of_hits + number_of_misses)
+    self.score = self._Score()
+
+  def _Score(self):

[mattcary, 2016/07/21 14:04:29]

Add a comment about where this score calculation comes from?

[Benoit L, 2016/07/21 14:48:42]

/facepalm
Forgot to upload the latest revision of the patch.
Done.

+    multiplier = 1
+    if self.resource_type in (ResourceType.STYLESHEET, ResourceType.SCRIPT):
+      multiplier = 2
+    return multiplier * 100 - self.average_position
+
+  @classmethod
+  def FromRow(cls, row):
+    return Entry(*row)
+
+  HEADER = (
+      'score,main_page_url,resource_type,number_of_hits,number_of_misses,'
+      'consecutive_misses,average_position,confidence,resource_url')
+
+  def __str__(self):
+    return '%f,%s,%d,%d,%d,%d,%d,%f\t%s' % (
+      self.score, self.main_page_url, self.resource_type,
+      self.number_of_hits, self.number_of_misses, self.consecutive_misses,
+      self.average_position, self.confidence, self.resource_url)
+
+
+def FilterAndSort(entries, domain):
+  result = filter(

[mattcary, 2016/07/21 14:04:29]

Ditto about where these threshold come from.

[Benoit L, 2016/07/21 14:48:42]

Done.

+      lambda x: ((domain is None or x.main_page_url == domain)
+                 and x.confidence > .7
+                 and x.number_of_hits >= 2), entries)
+  return sorted(result, key=operator.attrgetter('score'), reverse=True)
+
+
+def DatabaseStats(filename, domain):
+  connection = sqlite3.connect(filename)
+  c = connection.cursor()
+  query = 'SELECT * FROM resource_prefetch_predictor_host'

[mattcary, 2016/07/21 14:04:29]

It would be safer to name the fields that you're extracting since you depend on
the order in Entry.FromRow.

[Benoit L, 2016/07/21 14:48:42]

Thanks!
Done.

+  entries = [Entry.FromRow(row) for row in c.execute(query)]
+  prefetched = FilterAndSort(entries, domain)
+  print Entry.HEADER
+  for x in prefetched:
+    print x
+
+
+def main():
+  parser = argparse.ArgumentParser()
+  parser.add_argument('-f', dest='database_filename', required=True,
+                      help='Path to the database')
+  parser.add_argument('-d', dest='domain', default=None, help='Domain')
+  args = parser.parse_args()
+  DatabaseStats(args.database_filename, args.domain)
+
+
+if __name__ == '__main__':
+  main()