reference_resolver.py revision a3f6a49ab37290eeeb8db0f41ec0f1cb74a68be7
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 8import string 9 10from file_system import FileNotFoundError 11 12 13def _ClassifySchemaNode(node_name, api): 14 """Attempt to classify |node_name| in an API, determining whether |node_name| 15 refers to a type, function, event, or property in |api|. 16 """ 17 if '.' in node_name: 18 node_name, rest = node_name.split('.', 1) 19 else: 20 rest = None 21 for key, group in [('types', 'type'), 22 ('functions', 'method'), 23 ('events', 'event'), 24 ('properties', 'property')]: 25 for item in api.get(key, []): 26 if item['name'] == node_name: 27 if rest is not None: 28 ret = _ClassifySchemaNode(rest, item) 29 if ret is not None: 30 return ret 31 else: 32 return group, node_name 33 return None 34 35 36def _MakeKey(namespace, ref): 37 key = '%s/%s' % (namespace, ref) 38 # AppEngine doesn't like keys > 500, but there will be some other stuff 39 # that goes into this key, so truncate it earlier. This shoudn't be 40 # happening anyway unless there's a bug, such as http://crbug.com/314102. 41 max_size = 256 42 if len(key) > max_size: 43 logging.error('Key was >%s characters: %s' % (max_size, key)) 44 key = key[:max_size] 45 return key 46 47 48class ReferenceResolver(object): 49 """Resolves references to $ref's by searching through the APIs to find the 50 correct node. 51 52 $ref's have two forms: 53 54 $ref:api.node - Replaces the $ref with a link to node on the API page. The 55 title is set to the name of the node. 56 57 $ref:[api.node The Title] - Same as the previous form but title is set to 58 "The Title". 59 """ 60 61 # Matches after a $ref: that doesn't have []s. 62 _bare_ref = re.compile('\w+(\.\w+)*') 63 64 class Factory(object): 65 def __init__(self, 66 api_data_source_factory, 67 api_models, 68 object_store_creator): 69 self._api_data_source_factory = api_data_source_factory 70 self._api_models = api_models 71 self._object_store_creator = object_store_creator 72 73 def Create(self): 74 return ReferenceResolver( 75 self._api_data_source_factory.Create(None), 76 self._api_models, 77 self._object_store_creator.Create(ReferenceResolver)) 78 79 def __init__(self, api_data_source, api_models, object_store): 80 self._api_data_source = api_data_source 81 self._api_models = api_models 82 self._object_store = object_store 83 84 def _GetRefLink(self, ref, api_list, namespace): 85 # Check nodes within each API the ref might refer to. 86 parts = ref.split('.') 87 for i, part in enumerate(parts): 88 api_name = '.'.join(parts[:i]) 89 if api_name not in api_list: 90 continue 91 try: 92 api = self._api_data_source.get(api_name, disable_refs=True) 93 except FileNotFoundError: 94 continue 95 name = '.'.join(parts[i:]) 96 # Attempt to find |name| in the API. 97 node_info = _ClassifySchemaNode(name, api) 98 if node_info is None: 99 # Check to see if this ref is a property. If it is, we want the ref to 100 # the underlying type the property is referencing. 101 for prop in api.get('properties', []): 102 # If the name of this property is in the ref text, replace the 103 # property with its type, and attempt to classify it. 104 if prop['name'] in name and 'link' in prop: 105 name_as_prop_type = name.replace(prop['name'], prop['link']['name']) 106 node_info = _ClassifySchemaNode(name_as_prop_type, api) 107 if node_info is not None: 108 name = name_as_prop_type 109 text = ref.replace(prop['name'], prop['link']['name']) 110 break 111 if node_info is None: 112 continue 113 else: 114 text = ref 115 category, node_name = node_info 116 if namespace is not None and text.startswith('%s.' % namespace): 117 text = text[len('%s.' % namespace):] 118 return { 119 'href': '%s.html#%s-%s' % (api_name, category, name.replace('.', '-')), 120 'text': text, 121 'name': node_name 122 } 123 124 # If it's not a reference to an API node it might just be a reference to an 125 # API. Check this last so that links within APIs take precedence over links 126 # to other APIs. 127 if ref in api_list: 128 return { 129 'href': '%s.html' % ref, 130 'text': ref, 131 'name': ref 132 } 133 134 return None 135 136 def GetLink(self, ref, namespace=None, title=None): 137 """Resolve $ref |ref| in namespace |namespace| if not None, returning None 138 if it cannot be resolved. 139 """ 140 db_key = _MakeKey(namespace, ref) 141 link = self._object_store.Get(db_key).Get() 142 if link is None: 143 api_list = self._api_models.GetNames() 144 link = self._GetRefLink(ref, api_list, namespace) 145 if link is None and namespace is not None: 146 # Try to resolve the ref in the current namespace if there is one. 147 link = self._GetRefLink('%s.%s' % (namespace, ref), api_list, namespace) 148 if link is None: 149 return None 150 self._object_store.Set(db_key, link) 151 else: 152 link = deepcopy(link) 153 if title is not None: 154 link['text'] = title 155 return link 156 157 def SafeGetLink(self, ref, namespace=None, title=None): 158 """Resolve $ref |ref| in namespace |namespace|, or globally if None. If it 159 cannot be resolved, pretend like it is a link to a type. 160 """ 161 ref_data = self.GetLink(ref, namespace=namespace, title=title) 162 if ref_data is not None: 163 return ref_data 164 logging.error('$ref %s could not be resolved in namespace %s.' % 165 (ref, namespace)) 166 type_name = ref.rsplit('.', 1)[-1] 167 return { 168 'href': '#type-%s' % type_name, 169 'text': title or ref, 170 'name': ref 171 } 172 173 def ResolveAllLinks(self, text, relative_to='', namespace=None): 174 """This method will resolve all $ref links in |text| using namespace 175 |namespace| if not None. Any links that cannot be resolved will be replaced 176 using the default link format that |SafeGetLink| uses. 177 The links will be generated relative to |relative_to|. 178 """ 179 if text is None or '$ref:' not in text: 180 return text 181 182 # requestPath should be of the form (apps|extensions)/...../page.html. 183 # link_prefix should that the target will point to 184 # (apps|extensions)/target.html. Note multiplying a string by a negative 185 # number gives the empty string. 186 link_prefix = '../' * (relative_to.count('/') - 1) 187 split_text = text.split('$ref:') 188 # |split_text| is an array of text chunks that all start with the 189 # argument to '$ref:'. 190 formatted_text = [split_text[0]] 191 for ref_and_rest in split_text[1:]: 192 title = None 193 if ref_and_rest.startswith('[') and ']' in ref_and_rest: 194 # Text was '$ref:[foo.bar maybe title] other stuff'. 195 ref_with_title, rest = ref_and_rest[1:].split(']', 1) 196 ref_with_title = ref_with_title.split(None, 1) 197 if len(ref_with_title) == 1: 198 # Text was '$ref:[foo.bar] other stuff'. 199 ref = ref_with_title[0] 200 else: 201 # Text was '$ref:[foo.bar title] other stuff'. 202 ref, title = ref_with_title 203 else: 204 # Text was '$ref:foo.bar other stuff'. 205 match = self._bare_ref.match(ref_and_rest) 206 if match is None: 207 ref = '' 208 rest = ref_and_rest 209 else: 210 ref = match.group() 211 rest = ref_and_rest[match.end():] 212 213 ref_dict = self.SafeGetLink(ref, namespace=namespace, title=title) 214 formatted_text.append('<a href="%s%s">%s</a>%s' % 215 (link_prefix, ref_dict['href'], ref_dict['text'], rest)) 216 return ''.join(formatted_text) 217