Supersize: Chrome-specific breakdowns, console tweaks

tools/binary_size/libsupersize/precanned_queries.py

Index: tools/binary_size/libsupersize/precanned_queries.py
diff --git a/tools/binary_size/libsupersize/precanned_queries.py b/tools/binary_size/libsupersize/precanned_queries.py
new file mode 100644
index 0000000000000000000000000000000000000000..ddb89698fc38e498214367e991626b421d6a4df8
--- /dev/null
+++ b/tools/binary_size/libsupersize/precanned_queries.py
@@ -0,0 +1,110 @@
+# 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.
+
+"""Contains a set of Chrome-specific size queries."""
+
+import logging
+import models
+
+
+class _Grouper(object):
+  def __init__(self):
+    self.groups = []
+
+  def Add(self, name, group):
+    logging.debug('Computed %s', name)
+    group.name = name
+    self.groups.append(group)
+    return group.Inverted()
+
+  def Finalize(self, remaining):
+    self.groups.sort(key=lambda s:(s.name.startswith('Other'), -s.pss))
+    if remaining:
+      stars = remaining.Filter(lambda s: s.name.startswith('*'))
+      if stars:
+        remaining = stars.Inverted()
+        stars.name = '** Merged Symbols'
+        self.groups.append(stars)
+      remaining.name = 'Other'
+      self.groups.append(remaining)
+
+    logging.debug('Finalized')
+    return models.SymbolGroup(self.groups, is_sorted=True)
+
+
+def CategorizeByChromeComponent(symbols):
+  g = _Grouper()
+
+  # Put things that filter out a lot of symbols at the beginning where possible
+  # to optimize speed.
+  symbols = g.Add('WebRTC', symbols.WhereMatches(r'(?i)webrtc'))
+  symbols = g.Add('Skia', symbols.Filter(lambda s: 'skia/' in s.source_path))
+  symbols = g.Add('V8', symbols.Filter(
+      lambda s: s.source_path.startswith('v8/')))
+
+  # Next, put non-regex queries, since they're a bit faster.
+  symbols = g.Add('ICU', symbols.Filter(lambda s: '/icu/' in s.source_path))
+  symbols = g.Add('Prefetch', symbols.Filter(
+      lambda s: 'resource_prefetch' in s.source_path))
+  symbols = g.Add('Password Manager', symbols.Filter(
+      lambda s: 'password_manager' in s.source_path))
+  symbols = g.Add('Internals Pages', symbols.Filter(
+      lambda s: '_internals' in s.source_path))
+  symbols = g.Add('Autofill', symbols.WhereSourcePathMatches(r'(?i)autofill'))
+  symbols = g.Add('WebGL', symbols.WhereMatches(r'(?i)webgl'))
+  symbols = g.Add('WebBluetooth', symbols.WhereMatches(r'(?i)bluetooth'))
+  symbols = g.Add('WebUSB', symbols.WhereMatches(r'(?i)webusb|(\b|_)usb(\b|_)'))
+  symbols = g.Add('WebVR', symbols.WhereMatches(
+      r'{{_gvr_}}|{{_cwebvr_}}|{{_vr_}}'))
+  symbols = g.Add('FileSystem', symbols.WhereSourcePathMatches(
+      r'content/.*/fileapi|WebKit/.*/filesystem'))
+  symbols = g.Add('WebCrypto', symbols.WhereMatches(r'(?i)webcrypto'))
+  symbols = g.Add('Printing', symbols.WhereMatches(r'printing'))
+  symbols = g.Add('Cast', symbols.WhereSourcePathMatches(
+      r'(?i)(\b|_)cast(\b|_)'))
+  symbols = g.Add('Media Source', symbols.WhereMatches(
+      r'(?i)mediasource|blink::.*TrackDefault|blink::.*SourceBuffer'))
+
+  # XSLT must come before libxml.
+  symbols = g.Add('XSLT', symbols.WhereMatches(r'(?i)xslt'))
+  symbols = g.Add('libxml', symbols.Filter(
+      lambda s: 'libxml' in s.source_path))
+
+  # These have some overlap with above, so need to come afterwards.
+  blink_syms = symbols.WhereSourcePathMatches(r'\b(blink|WebKit)\b')
+  symbols = blink_syms.Inverted()
+  blink_generated = blink_syms.WhereSourceIsGenerated()
+  g.Add('Blink (generated)', blink_generated)
+  g.Add('Blink (non-generated)', blink_generated.Inverted())
+
+  symbols = g.Add('Codecs', symbols.WhereSourcePathMatches(
+      r'^third_party/(libweb[mp]|libpng|libjpeg_turbo|opus|ffmpeg|libvpx)/'))
+  symbols = g.Add('Other Third-Party', symbols.Filter(
+      lambda s: 'third_party' in s.source_path))
+
+  return g.Finalize(symbols)
+
+
+def CategorizeGenerated(symbols):

[estevenson, 2017/05/05 22:24:18]

Could add jni here? Might be super small though.

[agrieve, 2017/05/06 01:18:22]

Done. ~50kb. But the number doesn't capture many of the string literals
involved. Reworking the jni generator so that it assigns all string literals to
variable names would make them show up.

+  g = _Grouper()
+  symbols = symbols.WhereSourceIsGenerated()
+  symbols = g.Add('Protocol Buffers', symbols.Filter(lambda s: (
+      '/protobuf/' in s.object_path or
+      s.object_path.endswith('.pbzero.o') or
+      s.object_path.endswith('.pb.o'))))
+  symbols = g.Add('Mojo', symbols.Filter(lambda s: (
+      '.mojom' in s.source_path or  # Blink uses .mojom-blink.cc
+      s.source_path.startswith('mojo/') or
+      s.name.startswith('mojo::'))))
+  symbols = g.Add('DevTools', symbols.WhereSourcePathMatches(
+      r'\b(?:protocol|devtools)\b'))
+  symbols = g.Add('Blink (bindings)', symbols.Filter(lambda s: (
+      'blink/bindings' in s.source_path or
+      ('WebKit/' in s.object_path and '/bindings/' in s.object_path))))
+  symbols = g.Add('Blink (IDL)', symbols.Filter(lambda s: (
+      'WebKit/Source/core' in s.object_path)))
+  symbols = g.Add('Blink (Other)', symbols.Filter(lambda s: (
+      'WebKit' in s.object_path or 'blink/' in s.object_path)))
+
+  return g.Finalize(symbols)