reference_resolver.py revision effb81e5f8246d0db0270817048dc992db66e9fb 
1# Copyright (c) 2012 The Chromium Authors. All rights reserved.
2# Use of this source code is governed by a BSD-style license that can be
3# found in the LICENSE file.
4
5from copy import deepcopy
6import logging
7import re
8
9from file_system import FileNotFoundError
10
11
12def _ClassifySchemaNode(node_name, api):
13  """Attempt to classify |node_name| in an API, determining whether |node_name|
14  refers to a type, function, event, or property in |api|.
15  """
16  if '.' in node_name:
17    node_name, rest = node_name.split('.', 1)
18  else:
19    rest = None
20  for key, group in [('types', 'type'),
21                     ('functions', 'method'),
22                     ('events', 'event'),
23                     ('properties', 'property')]:
24    for item in api.get(key, []):
25      if item['name'] == node_name:
26        if rest is not None:
27          ret = _ClassifySchemaNode(rest, item)
28          if ret is not None:
29            return ret
30        else:
31          return group, node_name
32  return None
33
34
35def _MakeKey(namespace, ref):
36  key = '%s/%s' % (namespace, ref)
37  # AppEngine doesn't like keys > 500, but there will be some other stuff
38  # that goes into this key, so truncate it earlier.  This shoudn't be
39  # happening anyway unless there's a bug, such as http://crbug.com/314102.
40  max_size = 256
41  if len(key) > max_size:
42    logging.error('Key was >%s characters: %s' % (max_size, key))
43    key = key[:max_size]
44  return key
45
46
47class ReferenceResolver(object):
48  """Resolves references to $ref's by searching through the APIs to find the
49  correct node.
50
51  $ref's have two forms:
52
53    $ref:api.node - Replaces the $ref with a link to node on the API page. The
54                    title is set to the name of the node.
55
56    $ref:[api.node The Title] - Same as the previous form but title is set to
57                                "The Title".
58  """
59
60  # Matches after a $ref: that doesn't have []s.
61  _bare_ref = re.compile('\w+(\.\w+)*')
62
63  def __init__(self, api_data_source, api_models, object_store):
64    self._api_data_source = api_data_source
65    self._api_models = api_models
66    self._object_store = object_store
67
68  def _GetRefLink(self, ref, api_list, namespace):
69    # Check nodes within each API the ref might refer to.
70    parts = ref.split('.')
71    for i, part in enumerate(parts):
72      api_name = '.'.join(parts[:i])
73      if api_name not in api_list:
74        continue
75      try:
76        api = self._api_data_source.get(api_name, disable_refs=True)
77      except FileNotFoundError:
78        continue
79      name = '.'.join(parts[i:])
80      # Attempt to find |name| in the API.
81      node_info = _ClassifySchemaNode(name, api)
82      if node_info is None:
83        # Check to see if this ref is a property. If it is, we want the ref to
84        # the underlying type the property is referencing.
85        for prop in api.get('properties', []):
86          # If the name of this property is in the ref text, replace the
87          # property with its type, and attempt to classify it.
88          if prop['name'] in name and 'link' in prop:
89            name_as_prop_type = name.replace(prop['name'], prop['link']['name'])
90            node_info = _ClassifySchemaNode(name_as_prop_type, api)
91            if node_info is not None:
92              name = name_as_prop_type
93              text = ref.replace(prop['name'], prop['link']['name'])
94              break
95        if node_info is None:
96          continue
97      else:
98        text = ref
99      category, node_name = node_info
100      if namespace is not None and text.startswith('%s.' % namespace):
101        text = text[len('%s.' % namespace):]
102      api_model = self._api_models.GetModel(api_name).Get()
103      filename = api_model.documentation_options.get('documented_in', api_name)
104      return {
105        'href': '%s#%s-%s' % (filename, category, name.replace('.', '-')),
106        'text': text,
107        'name': node_name
108      }
109
110    # If it's not a reference to an API node it might just be a reference to an
111    # API. Check this last so that links within APIs take precedence over links
112    # to other APIs.
113    if ref in api_list:
114      return {
115        'href': '%s' % ref,
116        'text': ref,
117        'name': ref
118      }
119
120    return None
121
122  def GetLink(self, ref, namespace=None, title=None):
123    """Resolve $ref |ref| in namespace |namespace| if not None, returning None
124    if it cannot be resolved.
125    """
126    db_key = _MakeKey(namespace, ref)
127    link = self._object_store.Get(db_key).Get()
128    if link is None:
129      api_list = self._api_models.GetNames()
130      link = self._GetRefLink(ref, api_list, namespace)
131      if link is None and namespace is not None:
132        # Try to resolve the ref in the current namespace if there is one.
133        link = self._GetRefLink('%s.%s' % (namespace, ref), api_list, namespace)
134      if link is None:
135        return None
136      self._object_store.Set(db_key, link)
137    else:
138      link = deepcopy(link)
139    if title is not None:
140      link['text'] = title
141    return link
142
143  def SafeGetLink(self, ref, namespace=None, title=None):
144    """Resolve $ref |ref| in namespace |namespace|, or globally if None. If it
145    cannot be resolved, pretend like it is a link to a type.
146    """
147    ref_data = self.GetLink(ref, namespace=namespace, title=title)
148    if ref_data is not None:
149      return ref_data
150    logging.error('$ref %s could not be resolved in namespace %s.' %
151        (ref, namespace))
152    type_name = ref.rsplit('.', 1)[-1]
153    return {
154      'href': '#type-%s' % type_name,
155      'text': title or ref,
156      'name': ref
157    }
158
159  # TODO(ahernandez.miralles): This function is no longer needed,
160  # and uses a deprecated style of ref
161  def ResolveAllLinks(self, text, relative_to='', namespace=None):
162    """This method will resolve all $ref links in |text| using namespace
163    |namespace| if not None. Any links that cannot be resolved will be replaced
164    using the default link format that |SafeGetLink| uses.
165    The links will be generated relative to |relative_to|.
166    """
167    if text is None or '$ref:' not in text:
168      return text
169
170    # requestPath should be of the form (apps|extensions)/...../page.html.
171    # link_prefix should  that the target will point to
172    # (apps|extensions)/target.html. Note multiplying a string by a negative
173    # number gives the empty string.
174    link_prefix = '../' * (relative_to.count('/') - 1)
175    split_text = text.split('$ref:')
176    # |split_text| is an array of text chunks that all start with the
177    # argument to '$ref:'.
178    formatted_text = [split_text[0]]
179    for ref_and_rest in split_text[1:]:
180      title = None
181      if ref_and_rest.startswith('[') and ']' in ref_and_rest:
182        # Text was '$ref:[foo.bar maybe title] other stuff'.
183        ref_with_title, rest = ref_and_rest[1:].split(']', 1)
184        ref_with_title = ref_with_title.split(None, 1)
185        if len(ref_with_title) == 1:
186          # Text was '$ref:[foo.bar] other stuff'.
187          ref = ref_with_title[0]
188        else:
189          # Text was '$ref:[foo.bar title] other stuff'.
190          ref, title = ref_with_title
191      else:
192        # Text was '$ref:foo.bar other stuff'.
193        match = self._bare_ref.match(ref_and_rest)
194        if match is None:
195          ref = ''
196          rest = ref_and_rest
197        else:
198          ref = match.group()
199          rest = ref_and_rest[match.end():]
200
201      ref_dict = self.SafeGetLink(ref, namespace=namespace, title=title)
202      formatted_text.append('<a href="%s%s">%s</a>%s' %
203          (link_prefix, ref_dict['href'], ref_dict['text'], rest))
204    return ''.join(formatted_text)
205