Docserver: Fix 3 issues relating to finding API references in sample files,

chrome/common/extensions/docs/server2/samples_data_source.py

 # Copyright (c) 2012 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.
 
 import hashlib
 import json
 import logging
 import posixpath
 import re
 import traceback
@@ skipping 38 matching lines @@
     def Create(self, request):
       '''Returns a new SamplesDataSource bound to |request|.
       '''
       return SamplesDataSource(self._extensions_cache,
                                self._apps_cache,
                                self._extension_samples_path,
                                self._base_path,
                                request)
 
     def _GetAPIItems(self, js_file):
-      chrome_regex = '(chrome\.[a-zA-Z0-9\.]+)'
-      calls = set(re.findall(chrome_regex, js_file))
-      # Find APIs that have been assigned into variables.
-      assigned_vars = dict(re.findall('var\s*([^\s]+)\s*=\s*%s;' % chrome_regex,
-                                      js_file))
-      # Replace the variable name with the full API name.
-      for var_name, value in assigned_vars.iteritems():
-        js_file = js_file.replace(var_name, value)
-      return calls.union(re.findall(chrome_regex, js_file))
+      chrome_pattern = r'chrome[\w.]+'
+      # Add API calls that appear normally, like "chrome.runtime.connect".
+      calls = set(re.findall(chrome_pattern, js_file))
+      # Add API calls that have been assigned into variables, like
+      # "var storageArea = chrome.storage.sync; storageArea.get", which should
+      # be expanded like "chrome.storage.sync.get".
+      for match in re.finditer(r'var\s+(\w+)\s*=\s*(%s);' % chrome_pattern,
+                               js_file):
+        var_name, api_prefix = match.groups()
+        for var_match in re.finditer(r'\b%s\.([\w.]+)\b' % re.escape(var_name),

[Jeffrey Yasskin, 2013/11/04 22:50:30]

You could probably speed this up a bit (remove the |V| factor) by replacing the
first %s with '('+'|'.join(map(re.escape), all_var_names)+')', and then looking
up the first group in the dictionary, but what you have is probably still faster
than copying js_file a bunch of times.

[not at google - send to devlin, 2013/11/04 23:39:41]

Yeah. Constructing regex's like this does scare me. I'm not confident in how the
engines are implemented.

Highly likely it wouldn't matter; but still, like you say, this works well
enough.

+                                     js_file):
+          api_suffix, = var_match.groups()

[Jeffrey Yasskin, 2013/11/04 22:50:30]

You could replace the unpacking with var_match.group(1).

[not at google - send to devlin, 2013/11/04 23:39:41]

I kind of like way that it matches line 67 here; and the group-1 thing creeps me
out. Even if the trailing comma looks like a typo.

+          calls.add('%s.%s' % (api_prefix, api_suffix))
+      return calls
 
     def _GetDataFromManifest(self, path, file_system):
       manifest = file_system.ReadSingle(path + '/manifest.json').Get()
       try:
         manifest_json = json.loads(json_comment_eater.Nom(manifest))
       except ValueError as e:
         logging.error('Error parsing manifest.json for %s: %s' % (path, e))
         return None
       l10n_data = {
         'name': manifest_json.get('name', ''),
@@ skipping 45 matching lines @@
         for item in sorted(api_items):
           if len(item.split('.')) < 3:
             continue
           if item.endswith('.removeListener') or item.endswith('.hasListener'):
             continue
           if item.endswith('.addListener'):
             item = item[:-len('.addListener')]
           if item.startswith('chrome.'):
             item = item[len('chrome.'):]
           ref_data = self._ref_resolver.GetLink(item)
+          # TODO(kalman): What about references like chrome.storage.sync.get?
+          # That should link to either chrome.storage.sync or
+          # chrome.storage.StorageArea.get (or probably both).
+          # TODO(kalman): Filter out API-only references? This can happen when
+          # the API namespace is assigned to a variable, but it's very hard to
+          # to disambiguate.
           if ref_data is None:
             continue
           api_calls.append({
             'name': ref_data['text'],
             'link': ref_data['href']
           })
 
         sample_base_path = sample_path.split('/', 1)[1]
         if is_apps:
           url = url_constants.GITHUB_BASE + '/' + sample_base_path
@@ skipping 87 matching lines @@
       else:
         dict_['id'] = self._GetSampleId(name)
         return_list.append(dict_)
     return return_list
 
   def get(self, key):
     return {
       'apps': lambda: self._CreateSamplesDict('apps'),
       'extensions': lambda: self._CreateSamplesDict('extensions')
     }.get(key, lambda: {})()