12 # after this many changes, a changeset will be closed
15 # this is the scale factor for lat/lon values stored as integers in the database
19 # replace characters which cannot be represented in XML 1.0.
21 str.gsub(/[\x00-\x08\x0b\x0c\x0e-\x1f]/, "?")
25 # changeset class keeps some information about changesets downloaded from the
26 # database - enough to let us know which changesets are closed/open & recently
29 attr_reader :id, :created_at, :closed_at, :num_changes
33 @created_at = Time.parse(row["created_at"])
34 @closed_at = Time.parse(row["closed_at"])
35 @num_changes = row["num_changes"].to_i
39 (@closed_at < t) || (@num_changes >= CHANGES_LIMIT)
46 def activity_between?(t1, t2)
47 ((@closed_at >= t1) && (@closed_at < t2)) || ((@created_at >= t1) && (@created_at < t2))
52 # builds an XML representation of a changeset from the database
53 class ChangesetBuilder
54 def initialize(now, conn)
60 xml = XML::Node.new("changeset")
61 xml["id"] = cs.id.to_s
62 xml["created_at"] = cs.created_at.getutc.xmlschema
63 xml["closed_at"] = cs.closed_at.getutc.xmlschema if cs.closed?(@now)
64 xml["open"] = cs.open?(@now).to_s
65 xml["num_changes"] = cs.num_changes.to_s
67 res = @conn.exec("select u.id, u.display_name, c.min_lat, c.max_lat, c.min_lon, c.max_lon from users u join changesets c on u.id=c.user_id where c.id=#{cs.id}")
68 xml["user"] = xml_sanitize(res[0]["display_name"])
69 xml["uid"] = res[0]["id"]
71 unless res[0]["min_lat"].nil? ||
72 res[0]["max_lat"].nil? ||
73 res[0]["min_lon"].nil? ||
74 res[0]["max_lon"].nil?
75 xml["min_lat"] = (res[0]["min_lat"].to_f / GEO_SCALE).to_s
76 xml["max_lat"] = (res[0]["max_lat"].to_f / GEO_SCALE).to_s
77 xml["min_lon"] = (res[0]["min_lon"].to_f / GEO_SCALE).to_s
78 xml["max_lon"] = (res[0]["max_lon"].to_f / GEO_SCALE).to_s
88 res = @conn.exec("select k, v from changeset_tags where changeset_id=#{cs.id}")
90 tag = XML::Node.new("tag")
91 tag["k"] = xml_sanitize(row["k"])
92 tag["v"] = xml_sanitize(row["v"])
97 def add_comments(xml, cs)
98 # grab the visible changeset comments as well
99 res = @conn.exec("select cc.author_id, u.display_name as author, cc.body, cc.created_at from changeset_comments cc join users u on cc.author_id=u.id where cc.changeset_id=#{cs.id} and cc.visible order by cc.created_at asc")
100 xml["comments_count"] = res.num_tuples.to_s
102 # early return if there aren't any comments
103 return unless res.num_tuples > 0
105 discussion = XML::Node.new("discussion")
107 comment = XML::Node.new("comment")
108 comment["uid"] = row["author_id"]
109 comment["user"] = xml_sanitize(row["author"])
110 comment["date"] = Time.parse(row["created_at"]).getutc.xmlschema
111 text = XML::Node.new("text")
112 text.content = xml_sanitize(row["body"])
114 discussion << comment
121 # sync a file to guarantee it's on disk
129 # sync a directory to guarantee it's on disk. have to recurse to the root
130 # to guarantee sync for newly created directories.
134 io = IO.for_fd(dh.fileno)
142 # state and connections associated with getting changeset data
143 # replicated to a file.
145 def initialize(config)
146 @config = YAML.load(File.read(config))
147 @state = YAML.load(File.read(@config["state_file"]))
148 @conn = PGconn.connect(@config["db"])
149 @now = Time.now.getutc
153 last_run = @state["last_run"]
154 last_run = (@now - 60) if last_run.nil?
155 @state["last_run"] = @now
156 # pretty much all operations on a changeset will modify its closed_at
157 # time (see rails_port's changeset model). so it is probably enough
158 # for us to look at anything that was closed recently, and filter from
161 .exec("select id, created_at, closed_at, num_changes from changesets where closed_at > ((now() at time zone 'utc') - '1 hour'::interval)")
162 .map { |row| Changeset.new(row) }
163 .select { |cs| cs.activity_between?(last_run, @now) }
165 # set for faster presence lookups by ID
166 cs_ids = Set.new(changesets.map(&:id))
168 # but also add any changesets which have new comments
170 .exec("select distinct changeset_id from changeset_comments where created_at >= '#{last_run}' and created_at < '#{@now}' and visible")
171 .map { |row| row["changeset_id"].to_i }
172 .select { |c_id| !cs_ids.include?(c_id) }
176 .exec("select id, created_at, closed_at, num_changes from changesets where id=#{id}")
177 .map { |row| Changeset.new(row) }
178 .each { |cs| changesets << cs }
181 changesets.sort_by(&:id)
184 # creates an XML file containing the changeset information from the
185 # list of changesets output by open_changesets.
186 def changeset_dump(changesets)
187 doc = XML::Document.new
188 doc.root = XML::Node.new("osm")
189 { "version" => "0.6",
190 "generator" => "replicate_changesets.rb",
191 "copyright" => "OpenStreetMap and contributors",
192 "attribution" => "http://www.openstreetmap.org/copyright",
193 "license" => "http://opendatacommons.org/licenses/odbl/1-0/" }
194 .each { |k, v| doc.root[k] = v }
196 builder = ChangesetBuilder.new(@now, @conn)
197 changesets.each do |cs|
198 doc.root << builder.changeset_xml(cs)
204 # saves new state (including the changeset dump xml)
206 File.open(@config["state_file"], "r") do |fl|
207 fl.flock(File::LOCK_EX)
209 sequence = (@state.key?("sequence") ? @state["sequence"] + 1 : 0)
210 data_stem = @config["data_dir"] + format("/%03d/%03d/%03d", sequence / 1000000, (sequence / 1000) % 1000, (sequence % 1000))
211 data_file = data_stem + ".osm.gz"
212 data_state_file = data_stem + ".state.txt"
213 tmp_state = @config["state_file"] + ".tmp"
214 tmp_data = data_file + ".tmp"
215 # try and write the files to tmp locations and then
216 # move them into place later, to avoid in-progress
217 # clashes, or people seeing incomplete files.
219 FileUtils.mkdir_p(File.dirname(data_file))
220 Zlib::GzipWriter.open(tmp_data) do |fh|
221 fh.write(changeset_dump(open_changesets))
223 @state["sequence"] = sequence
224 File.open(tmp_state, "w") do |fh|
225 fh.write(YAML.dump(@state))
228 # fsync the files in their old locations.
232 # sync the directory as well, to ensure that the file is reachable
233 # from the dirent and has been updated to account for any allocations.
234 fdirsync(File.dirname(tmp_data))
235 fdirsync(File.dirname(tmp_state))
237 # sanity check: the files we're moving into place
238 # should be non-empty.
239 raise "Temporary gzip file should exist, but doesn't." unless File.exist?(tmp_data)
240 raise "Temporary state file should exist, but doesn't." unless File.exist?(tmp_state)
241 raise "Temporary gzip file should be non-empty, but isn't." if File.zero?(tmp_data)
242 raise "Temporary state file should be non-empty, but isn't." if File.zero?(tmp_state)
244 FileUtils.mv(tmp_data, data_file)
245 FileUtils.cp(tmp_state, @config["state_file"])
246 FileUtils.mv(tmp_state, data_state_file)
248 # fsync the files in their new locations, in case the inodes have
249 # changed in the move / copy.
251 fsync(@config["state_file"])
252 fsync(data_state_file)
254 # sync the directory as well, to ensure that the file is reachable
255 # from the dirent and has been updated to account for any allocations.
256 fdirsync(File.dirname(data_file))
257 fdirsync(File.dirname(@config["state_file"]))
259 fl.flock(File::LOCK_UN)
262 STDERR.puts("Error! Couldn't update state.")
263 fl.flock(File::LOCK_UN)
271 rep = Replicator.new(ARGV[0])
273 rescue StandardError => e
274 STDERR.puts "ERROR: #{e.message}"