2 # DiffReader reads OSM diffs and applies them to the database.
4 # Uses the streaming LibXML "Reader" interface to cut down on memory
5 # usage, so hopefully we can process fairly large diffs.
7 include ConsistencyValidations
9 # maps each element type to the model class which handles it
13 "relation" => Relation
17 # Construct a diff reader by giving it a bunch of XML +data+ to parse
18 # in OsmChange format. All diffs must be limited to a single changeset
19 # given in +changeset+.
20 def initialize(data, changeset)
21 @reader = XML::Reader.string(data)
22 @changeset = changeset
23 # document that's (re-)used to handle elements expanded out of the
24 # diff processing stream.
25 @doc = XML::Document.new
26 @doc.root = XML::Node.new("osm")
30 # Reads the next element from the XML document. Checks the return value
31 # and throws an exception if an error occurred.
33 # NOTE: XML::Reader#read returns false for EOF and raises an
34 # exception if an error occurs.
36 rescue LibXML::XML::Error => e
37 raise OSM::APIBadXMLError.new("changeset", xml, e.message)
41 # An element-block mapping for using the LibXML reader interface.
43 # Since a lot of LibXML reader usage is boilerplate iteration through
44 # elements, it would be better to DRY and do this in a block. This
45 # could also help with error handling...?
47 # if the start element is empty then don't do any processing, as
48 # there won't be any child elements to process!
49 unless @reader.empty_element?
50 # read the first element
53 while @reader.node_type != 15 # end element
54 # because we read elements in DOM-style to reuse their DOM
55 # parsing code, we don't always read an element on each pass
56 # as the call to @reader.next in the innermost loop will take
57 # care of that for us.
58 if @reader.node_type == 1 # element
62 if @reader.has_attributes?
63 attributes[@reader.name] = @reader.value while @reader.move_to_next_attribute == 1
65 @reader.move_to_element
68 yield name, attributes
78 # An element-block mapping for using the LibXML reader interface.
80 # Since a lot of LibXML reader usage is boilerplate iteration through
81 # elements, it would be better to DRY and do this in a block. This
82 # could also help with error handling...?
84 with_element do |model_name, _model_attributes|
85 model = MODELS[model_name]
87 raise OSM::APIBadUserInput, "Unexpected element type #{model_name}, " \
88 "expected node, way or relation."
90 # new in libxml-ruby >= 2, expand returns an element not associated
91 # with a document. this means that there's no encoding parameter,
92 # which means basically nothing works.
93 expanded = @reader.expand
95 # create a new, empty document to hold this expanded node
96 new_node = @doc.import(expanded)
102 # remove element from doc - it will be garbage collected and the
103 # rest of the document is re-used in the next iteration.
104 @doc.root.child.remove!
109 # Checks a few invariants. Others are checked in the model methods
110 # such as save_ and delete_with_history.
111 def check(model, xml, new)
112 raise OSM::APIBadXMLError.new(model, xml) if new.nil?
113 raise OSM::APIChangesetMismatchError.new(new.changeset_id, @changeset.id) unless new.changeset_id == @changeset.id
117 # Consume the XML diff and try to commit it to the database. This code
118 # is *not* transactional, so code which calls it should ensure that the
119 # appropriate transaction block is in place.
121 # On a failure to meet preconditions (e.g: optimistic locking fails)
122 # an exception subclassing OSM::APIError will be thrown.
124 # data structure used for mapping placeholder IDs to real IDs
125 ids = { :node => {}, :way => {}, :relation => {} }
127 # take the first element and check that it is an osmChange element
129 raise OSM::APIBadUserInput, "Document element should be 'osmChange'." if @reader.name != "osmChange"
131 result = OSM::API.new.get_xml_doc
132 result.root.name = "diffResult"
134 # loop at the top level, within the <osmChange> element
135 with_element do |action_name, action_attributes|
138 # create a new element. this code is agnostic of the element type
139 # because all the elements support the methods that we're using.
140 with_model do |model, xml|
141 new = model.from_xml_node(xml, true)
142 check(model, xml, new)
144 # when this element is saved it will get a new ID, so we save it
145 # to produce the mapping which is sent to other elements.
146 placeholder_id = xml["id"].to_i
147 raise OSM::APIBadXMLError.new(model, xml) if placeholder_id.nil?
149 # check if the placeholder ID has been given before and throw
150 # an exception if it has - we can't create the same element twice.
151 model_sym = model.to_s.downcase.to_sym
152 raise OSM::APIBadUserInput, "Placeholder IDs must be unique for created elements." if ids[model_sym].include? placeholder_id
154 # some elements may have placeholders for other elements in the
155 # diff, so we must fix these before saving the element.
156 new.fix_placeholders!(ids, placeholder_id)
158 # create element given user
159 new.create_with_history(@changeset.user)
161 # save placeholder => allocated ID map
162 ids[model_sym][placeholder_id] = new.id
164 # add the result to the document we're building for return.
165 xml_result = XML::Node.new model.to_s.downcase
166 xml_result["old_id"] = placeholder_id.to_s
167 xml_result["new_id"] = new.id.to_s
168 xml_result["new_version"] = new.version.to_s
169 result.root << xml_result
173 # modify an existing element. again, this code doesn't directly deal
174 # with types, but uses duck typing to handle them transparently.
175 with_model do |model, xml|
176 # get the new element from the XML payload
177 new = model.from_xml_node(xml, false)
178 check(model, xml, new)
180 # if the ID is a placeholder then map it to the real ID
181 model_sym = model.to_s.downcase.to_sym
183 is_placeholder = ids[model_sym].include? client_id
184 id = is_placeholder ? ids[model_sym][client_id] : client_id
186 # and the old one from the database
189 # translate any placeholder IDs to their true IDs.
190 new.fix_placeholders!(ids)
193 old.update_from(new, @changeset.user)
195 xml_result = XML::Node.new model.to_s.downcase
196 xml_result["old_id"] = client_id.to_s
197 xml_result["new_id"] = id.to_s
198 # version is updated in "old" through the update, so we must not
199 # return new.version here but old.version!
200 xml_result["new_version"] = old.version.to_s
201 result.root << xml_result
205 # delete action. this takes a payload in API 0.6, so we need to do
206 # most of the same checks that are done for the modify.
207 with_model do |model, xml|
208 # delete doesn't have to contain a full payload, according to
209 # the wiki docs, so we just extract the things we need.
210 new_id = xml["id"].to_i
211 raise OSM::APIBadXMLError.new(model, xml, "ID attribute is required") if new_id.nil?
213 # if the ID is a placeholder then map it to the real ID
214 model_sym = model.to_s.downcase.to_sym
215 is_placeholder = ids[model_sym].include? new_id
216 id = is_placeholder ? ids[model_sym][new_id] : new_id
218 # build the "new" element by modifying the existing one
220 new.changeset_id = xml["changeset"].to_i
221 new.version = xml["version"].to_i
222 check(model, xml, new)
224 # fetch the matching old element from the DB
227 # can a delete have placeholders under any circumstances?
228 # if a way is modified, then deleted is that a valid diff?
229 new.fix_placeholders!(ids)
231 xml_result = XML::Node.new model.to_s.downcase
232 # oh, the irony... the "new" element actually contains the "old" ID
233 # a better name would have been client/server, but anyway...
234 xml_result["old_id"] = new_id.to_s
236 if action_attributes["if-unused"]
238 old.delete_with_history!(new, @changeset.user)
239 rescue OSM::APIAlreadyDeletedError, OSM::APIPreconditionFailedError
240 xml_result["new_id"] = old.id.to_s
241 xml_result["new_version"] = old.version.to_s
244 old.delete_with_history!(new, @changeset.user)
247 result.root << xml_result
251 # no other actions to choose from, so it must be the users fault!
252 raise OSM::APIChangesetActionInvalid, action_name
256 # return the XML document to be rendered back to the client