geo_coder 0.1.0
Sign up to get free protection for your applications and to get access to all the features.
- data/Gemfile +12 -0
- data/Gemfile.lock +32 -0
- data/History.txt +6 -0
- data/Makefile +13 -0
- data/Manifest.txt +18 -0
- data/README.rdoc +197 -0
- data/Rakefile +53 -0
- data/TODO.txt +8 -0
- data/VERSION +1 -0
- data/bin/build_indexes +8 -0
- data/bin/rebuild_cluster +22 -0
- data/bin/rebuild_metaphones +23 -0
- data/bin/tiger_import +59 -0
- data/demos/demo/app/ext/geocodewrap.rb +84 -0
- data/demos/demo/app/views/index.builder +13 -0
- data/demos/demo/app/views/index.erb +71 -0
- data/demos/demo/config.ru +12 -0
- data/demos/demo/config/bootstraps.rb +130 -0
- data/demos/demo/config/geoenvironment.rb +25 -0
- data/demos/demo/geocoder_helper.rb +12 -0
- data/demos/demo/geocom_geocode.rb +10 -0
- data/demos/demo/main.rb +3 -0
- data/demos/demo/rakefile.rb +17 -0
- data/demos/demo/tmp/restart.txt +0 -0
- data/demos/simpledemo/views/index.builder +13 -0
- data/demos/simpledemo/views/index.erb +69 -0
- data/demos/simpledemo/ws.rb +83 -0
- data/doc/Makefile +7 -0
- data/doc/html4css1.css +279 -0
- data/doc/lookup.rst +193 -0
- data/doc/parsing.rst +125 -0
- data/doc/voidspace.css +147 -0
- data/geo_coder.gemspec +172 -0
- data/lib/geocoder/us.rb +21 -0
- data/lib/geocoder/us/address.rb +290 -0
- data/lib/geocoder/us/constants.rb +670 -0
- data/lib/geocoder/us/database.rb +745 -0
- data/lib/geocoder/us/import.rb +181 -0
- data/lib/geocoder/us/import/tiger.rb +13 -0
- data/lib/geocoder/us/numbers.rb +58 -0
- data/navteq/README +4 -0
- data/navteq/convert.sql +37 -0
- data/navteq/navteq_import +39 -0
- data/navteq/prepare.sql +92 -0
- data/sql/cluster.sql +16 -0
- data/sql/convert.sql +80 -0
- data/sql/create.sql +37 -0
- data/sql/index.sql +12 -0
- data/sql/place.csv +104944 -0
- data/sql/place.sql +104948 -0
- data/sql/setup.sql +78 -0
- data/src/Makefile +13 -0
- data/src/README +14 -0
- data/src/liblwgeom/Makefile +75 -0
- data/src/liblwgeom/box2d.c +54 -0
- data/src/liblwgeom/lex.yy.c +4799 -0
- data/src/liblwgeom/liblwgeom.h +1405 -0
- data/src/liblwgeom/lwalgorithm.c +946 -0
- data/src/liblwgeom/lwalgorithm.h +52 -0
- data/src/liblwgeom/lwcircstring.c +759 -0
- data/src/liblwgeom/lwcollection.c +541 -0
- data/src/liblwgeom/lwcompound.c +118 -0
- data/src/liblwgeom/lwcurvepoly.c +86 -0
- data/src/liblwgeom/lwgeom.c +886 -0
- data/src/liblwgeom/lwgeom_api.c +2201 -0
- data/src/liblwgeom/lwgparse.c +1219 -0
- data/src/liblwgeom/lwgunparse.c +1054 -0
- data/src/liblwgeom/lwline.c +525 -0
- data/src/liblwgeom/lwmcurve.c +125 -0
- data/src/liblwgeom/lwmline.c +137 -0
- data/src/liblwgeom/lwmpoint.c +138 -0
- data/src/liblwgeom/lwmpoly.c +141 -0
- data/src/liblwgeom/lwmsurface.c +129 -0
- data/src/liblwgeom/lwpoint.c +439 -0
- data/src/liblwgeom/lwpoly.c +579 -0
- data/src/liblwgeom/lwsegmentize.c +1047 -0
- data/src/liblwgeom/lwutil.c +369 -0
- data/src/liblwgeom/measures.c +861 -0
- data/src/liblwgeom/postgis_config.h +93 -0
- data/src/liblwgeom/ptarray.c +847 -0
- data/src/liblwgeom/vsprintf.c +179 -0
- data/src/liblwgeom/wktparse.h +126 -0
- data/src/liblwgeom/wktparse.lex +74 -0
- data/src/liblwgeom/wktparse.tab.c +2353 -0
- data/src/liblwgeom/wktparse.tab.h +145 -0
- data/src/liblwgeom/wktparse.y +385 -0
- data/src/libsqlite3_geocoder/Makefile +22 -0
- data/src/libsqlite3_geocoder/Makefile.nix +15 -0
- data/src/libsqlite3_geocoder/Makefile.redhat +15 -0
- data/src/libsqlite3_geocoder/extension.c +121 -0
- data/src/libsqlite3_geocoder/extension.h +13 -0
- data/src/libsqlite3_geocoder/levenshtein.c +42 -0
- data/src/libsqlite3_geocoder/metaphon.c +278 -0
- data/src/libsqlite3_geocoder/util.c +37 -0
- data/src/libsqlite3_geocoder/wkb_compress.c +54 -0
- data/src/metaphone/Makefile +7 -0
- data/src/metaphone/README +49 -0
- data/src/metaphone/extension.c +37 -0
- data/src/metaphone/metaphon.c +251 -0
- data/src/shp2sqlite/Makefile +37 -0
- data/src/shp2sqlite/Makefile.nix +36 -0
- data/src/shp2sqlite/Makefile.redhat +35 -0
- data/src/shp2sqlite/dbfopen.c +1595 -0
- data/src/shp2sqlite/getopt.c +695 -0
- data/src/shp2sqlite/getopt.h +127 -0
- data/src/shp2sqlite/shapefil.h +500 -0
- data/src/shp2sqlite/shp2sqlite.c +1974 -0
- data/src/shp2sqlite/shpopen.c +1894 -0
- data/tests/address.rb +236 -0
- data/tests/benchmark.rb +20 -0
- data/tests/constants.rb +57 -0
- data/tests/data/address-sample.csv +52 -0
- data/tests/data/db-test.csv +57 -0
- data/tests/data/locations.csv +4 -0
- data/tests/database.rb +137 -0
- data/tests/generate.rb +34 -0
- data/tests/numbers.rb +46 -0
- data/tests/run.rb +11 -0
- metadata +237 -0
@@ -0,0 +1,745 @@
|
|
1
|
+
require 'rubygems'
|
2
|
+
require 'sqlite3'
|
3
|
+
require 'text'
|
4
|
+
|
5
|
+
require 'set'
|
6
|
+
require 'pp'
|
7
|
+
require 'time'
|
8
|
+
require 'thread'
|
9
|
+
|
10
|
+
require 'geocoder/us/address'
|
11
|
+
|
12
|
+
module Geocoder
|
13
|
+
end
|
14
|
+
|
15
|
+
module Geocoder::US
|
16
|
+
# Provides an interface to a Geocoder::US database.
|
17
|
+
class Database
|
18
|
+
Street_Weight = 3.0
|
19
|
+
Number_Weight = 2.0
|
20
|
+
Parity_Weight = 1.25
|
21
|
+
City_Weight = 1.0
|
22
|
+
@@mutex = Mutex.new
|
23
|
+
|
24
|
+
# Takes the path of an SQLite 3 database prepared for Geocoder::US
|
25
|
+
# as the sole mandatory argument. The helper argument points to the
|
26
|
+
# Geocoder::US SQLite plugin; the module looks for this in the same
|
27
|
+
# directory as database.rb by default. The cache_size argument is
|
28
|
+
# measured in kilobytes and is used to set the SQLite cache size; larger
|
29
|
+
# values will trade memory for speed in long-running processes.
|
30
|
+
def initialize (filename, options = {})
|
31
|
+
defaults = {:debug => false, :cache_size => 50000,
|
32
|
+
:helper => "sqlite3.so", :threadsafe => false,
|
33
|
+
:create => false}
|
34
|
+
options = defaults.merge options
|
35
|
+
raise ArgumentError, "can't find database #{filename}" \
|
36
|
+
unless options[:create] or File.exists? filename
|
37
|
+
@db = SQLite3::Database.new( filename )
|
38
|
+
@st = {}
|
39
|
+
@debug = options[:debug]
|
40
|
+
@threadsafe = options[:threadsafe]
|
41
|
+
tune options[:helper], options[:cache_size]
|
42
|
+
end
|
43
|
+
|
44
|
+
def synchronize
|
45
|
+
if not @threadsafe
|
46
|
+
@@mutex.synchronize { yield }
|
47
|
+
else
|
48
|
+
yield
|
49
|
+
end
|
50
|
+
end
|
51
|
+
|
52
|
+
#private
|
53
|
+
|
54
|
+
# Load the SQLite extension and tune the database settings.
|
55
|
+
# q.v. http://web.utk.edu/~jplyon/sqlite/SQLite_optimization_FAQ.html
|
56
|
+
def tune (helper, cache_size)
|
57
|
+
synchronize do
|
58
|
+
@db.create_function("levenshtein", 2) do |func, word1, word2|
|
59
|
+
test1, test2 = [word1, word2].map {|w|
|
60
|
+
w.to_s.gsub(/\W/o, "").downcase
|
61
|
+
}
|
62
|
+
dist = Text::Levenshtein.distance(test1, test2)
|
63
|
+
result = dist.to_f / [test1.length, test2.length].max
|
64
|
+
func.set_result result
|
65
|
+
end
|
66
|
+
@db.create_function("metaphone", 2) do |func, string, len|
|
67
|
+
test = string.to_s.gsub(/\W/o, "")
|
68
|
+
if test =~ /^(\d+)/o
|
69
|
+
mph = $1
|
70
|
+
elsif test =~ /^([wy])$/io
|
71
|
+
mph = $1
|
72
|
+
else
|
73
|
+
mph = Text::Metaphone.metaphone test
|
74
|
+
end
|
75
|
+
func.result = mph[0...len.to_i]
|
76
|
+
end
|
77
|
+
@db.create_function("nondigit_prefix", 1) do |func, string|
|
78
|
+
string.to_s =~ /^(.*\D)?(\d+)$/o
|
79
|
+
func.result = ($1 || "")
|
80
|
+
end
|
81
|
+
@db.create_function("digit_suffix", 1) do |func, string|
|
82
|
+
string.to_s =~ /^(.*\D)?(\d+)$/o
|
83
|
+
func.result = ($2 || "")
|
84
|
+
end
|
85
|
+
#@db.enable_load_extension(1)
|
86
|
+
#@db.load_extension(helper)
|
87
|
+
#@db.enable_load_extension(0)
|
88
|
+
@db.cache_size = cache_size
|
89
|
+
@db.temp_store = "memory"
|
90
|
+
@db.synchronous = "off"
|
91
|
+
end
|
92
|
+
end
|
93
|
+
|
94
|
+
# Return a cached SQLite statement object, preparing it first if
|
95
|
+
# it's not already in the cache.
|
96
|
+
def prepare (sql)
|
97
|
+
$stderr.print "SQL : #{sql}\n" if @debug
|
98
|
+
synchronize do
|
99
|
+
@st[sql] ||= @db.prepare sql
|
100
|
+
end
|
101
|
+
return @st[sql]
|
102
|
+
end
|
103
|
+
|
104
|
+
def flush_statements
|
105
|
+
@st = {}
|
106
|
+
end
|
107
|
+
|
108
|
+
# Generate enough SQL placeholders for a list of objects.
|
109
|
+
def placeholders_for (list)
|
110
|
+
(["?"] * list.length).join(",")
|
111
|
+
end
|
112
|
+
|
113
|
+
# Generate enough SQL placeholders for a list of objects.
|
114
|
+
def metaphone_placeholders_for (list)
|
115
|
+
(["metaphone(?,5)"] * list.length).join(",")
|
116
|
+
end
|
117
|
+
|
118
|
+
# Execute an SQL statement, bind a list of parameters, and
|
119
|
+
# return the result as a list of hashes.
|
120
|
+
def execute (sql, *params)
|
121
|
+
st = prepare(sql)
|
122
|
+
execute_statement st, *params
|
123
|
+
end
|
124
|
+
|
125
|
+
|
126
|
+
# Execute an SQLite statement object, bind the parameters,
|
127
|
+
# map the column names to symbols, and return the rows
|
128
|
+
# as a list of hashes.
|
129
|
+
def execute_statement (st, *params)
|
130
|
+
if @debug
|
131
|
+
start = Time.now
|
132
|
+
$stderr.print "EXEC: #{params.inspect}\n" if !params.empty?
|
133
|
+
end
|
134
|
+
rows = []
|
135
|
+
synchronize do
|
136
|
+
result = st.execute(*params)
|
137
|
+
columns = result.columns.map {|c| c.to_sym}
|
138
|
+
result.each {|row|
|
139
|
+
rows << Hash[*(columns.zip(row).flatten)]}
|
140
|
+
|
141
|
+
end
|
142
|
+
if @debug
|
143
|
+
runtime = format("%.3f", Time.now - start)
|
144
|
+
$stderr.print "ROWS: #{rows.length} (#{runtime}s)\n"
|
145
|
+
end
|
146
|
+
rows.reverse!
|
147
|
+
end
|
148
|
+
|
149
|
+
def places_by_zip (city, zip)
|
150
|
+
execute("SELECT *, levenshtein(?, city) AS city_score
|
151
|
+
FROM place WHERE zip = ? order by priority desc LIMIT 1;", city, zip)
|
152
|
+
end
|
153
|
+
|
154
|
+
# Query the place table for by city, optional state, and zip.
|
155
|
+
# The metaphone index on the place table is used to match
|
156
|
+
# city names.
|
157
|
+
def places_by_city (city, tokens, state)
|
158
|
+
if city.nil?
|
159
|
+
city = ""
|
160
|
+
end
|
161
|
+
if state.nil? or state.empty?
|
162
|
+
and_state = ""
|
163
|
+
args = [city] + tokens.clone
|
164
|
+
else
|
165
|
+
and_state = "AND state = ?"
|
166
|
+
args = [city] + tokens.clone + [state]
|
167
|
+
end
|
168
|
+
metaphones = metaphone_placeholders_for tokens
|
169
|
+
execute("SELECT *, levenshtein(?, city) AS city_score
|
170
|
+
FROM place WHERE city_phone IN (#{metaphones}) #{and_state} order by priority desc LIMIT 1;", *args)
|
171
|
+
end
|
172
|
+
|
173
|
+
# Generate an SQL query and set of parameters against the feature and range
|
174
|
+
# tables for a street name and optional building number. The SQL is
|
175
|
+
# used by candidate_records and more_candidate_records to filter results
|
176
|
+
# by ZIP code.
|
177
|
+
def features_by_street (street, tokens)
|
178
|
+
metaphones = (["metaphone(?,5)"] * tokens.length).join(",")
|
179
|
+
sql = "
|
180
|
+
SELECT feature.*, levenshtein(?, street) AS street_score
|
181
|
+
FROM feature
|
182
|
+
WHERE street_phone IN (#{metaphones})"
|
183
|
+
params = [street] + tokens
|
184
|
+
return [sql, params]
|
185
|
+
end
|
186
|
+
|
187
|
+
# Query the feature and range tables for a set of ranges, given a
|
188
|
+
# building number, street name, and list of candidate ZIP codes.
|
189
|
+
# The metaphone and ZIP code indexes on the feature table are
|
190
|
+
# used to match results.
|
191
|
+
def features_by_street_and_zip (street, tokens, zips)
|
192
|
+
sql, params = features_by_street(street, tokens)
|
193
|
+
in_list = placeholders_for zips
|
194
|
+
sql += " AND feature.zip IN (#{in_list})"
|
195
|
+
params += zips
|
196
|
+
execute sql, *params
|
197
|
+
end
|
198
|
+
|
199
|
+
# Query the feature and range tables for a set of ranges, given a
|
200
|
+
# building number, street name, and list of candidate ZIP codes.
|
201
|
+
# The ZIP codes are reduced to a set of 3-digit prefixes, broadening
|
202
|
+
# the search area.
|
203
|
+
def more_features_by_street_and_zip (street, tokens, zips)
|
204
|
+
sql, params = features_by_street(street, tokens)
|
205
|
+
if !zips.empty? and !zips[0].nil?
|
206
|
+
puts "zip results 2"
|
207
|
+
zip3s = zips.map {|z| z[0..2]+'%'}.to_set.to_a
|
208
|
+
like_list = zip3s.map {|z| "feature.zip LIKE ?"}.join(" OR ")
|
209
|
+
sql += " AND (#{like_list})"
|
210
|
+
params += zip3s
|
211
|
+
end
|
212
|
+
st = @db.prepare sql
|
213
|
+
execute_statement st, *params
|
214
|
+
end
|
215
|
+
|
216
|
+
def ranges_by_feature (fids, number, prenum)
|
217
|
+
in_list = placeholders_for fids
|
218
|
+
limit = 4 * fids.length
|
219
|
+
sql = "
|
220
|
+
SELECT feature_edge.fid AS fid, range.*
|
221
|
+
FROM feature_edge, range
|
222
|
+
WHERE fid IN (#{in_list})
|
223
|
+
AND feature_edge.tlid = range.tlid"
|
224
|
+
params = fids.clone
|
225
|
+
unless prenum.nil?
|
226
|
+
sql += " AND prenum = ?"
|
227
|
+
params += [prenum]
|
228
|
+
end
|
229
|
+
sql += "
|
230
|
+
ORDER BY min(abs(fromhn - ?), abs(tohn - ?))
|
231
|
+
LIMIT #{limit};"
|
232
|
+
params += [number, number]
|
233
|
+
execute sql, *params
|
234
|
+
end
|
235
|
+
|
236
|
+
# Query the edge table for a list of edges matching a list of edge IDs.
|
237
|
+
def edges (edge_ids)
|
238
|
+
in_list = placeholders_for edge_ids
|
239
|
+
sql = "SELECT edge.* FROM edge WHERE edge.tlid IN (#{in_list})"
|
240
|
+
execute sql, *edge_ids
|
241
|
+
end
|
242
|
+
|
243
|
+
# Query the range table for all ranges associated with the given
|
244
|
+
# list of edge IDs.
|
245
|
+
def range_ends (edge_ids)
|
246
|
+
in_list = placeholders_for edge_ids
|
247
|
+
sql = "SELECT tlid, side,
|
248
|
+
min(fromhn) > min(tohn) AS flipped,
|
249
|
+
min(fromhn) AS from0, max(tohn) AS to0,
|
250
|
+
min(tohn) AS from1, max(fromhn) AS to1
|
251
|
+
FROM range WHERE tlid IN (#{in_list})
|
252
|
+
GROUP BY tlid, side;"
|
253
|
+
execute(sql, *edge_ids).map {|r|
|
254
|
+
if r[:flipped] == "0"
|
255
|
+
r[:flipped] = false
|
256
|
+
r[:fromhn], r[:tohn] = r[:from0], r[:to0]
|
257
|
+
else
|
258
|
+
r[:flipped] = true
|
259
|
+
r[:fromhn], r[:tohn] = r[:from1], r[:to1]
|
260
|
+
end
|
261
|
+
[:from0, :to0, :from1, :to1].each {|k| r.delete k}
|
262
|
+
r
|
263
|
+
}
|
264
|
+
end
|
265
|
+
|
266
|
+
def intersections_by_fid (fids)
|
267
|
+
in_list = placeholders_for fids
|
268
|
+
sql = "
|
269
|
+
CREATE TEMPORARY TABLE intersection AS
|
270
|
+
SELECT fid, substr(geometry,1,8) AS point
|
271
|
+
FROM feature_edge, edge
|
272
|
+
WHERE feature_edge.tlid = edge.tlid
|
273
|
+
AND fid IN (#{in_list})
|
274
|
+
UNION
|
275
|
+
SELECT fid, substr(geometry,length(geometry)-7,8) AS point
|
276
|
+
FROM feature_edge, edge
|
277
|
+
WHERE feature_edge.tlid = edge.tlid
|
278
|
+
AND fid IN (#{in_list});
|
279
|
+
CREATE INDEX intersect_pt_idx ON intersection (point);"
|
280
|
+
execute sql, *(fids + fids)
|
281
|
+
# the a.fid < b.fid inequality guarantees consistent ordering of street
|
282
|
+
# names in the output
|
283
|
+
results = execute "
|
284
|
+
SELECT a.fid AS fid1, b.fid AS fid2, a.point
|
285
|
+
FROM intersection a, intersection b, feature f1, feature f2
|
286
|
+
WHERE a.point = b.point AND a.fid < b.fid
|
287
|
+
AND f1.fid = a.fid AND f2.fid = b.fid
|
288
|
+
AND f1.zip = f2.zip
|
289
|
+
AND f1.paflag = 'P' AND f2.paflag = 'P';"
|
290
|
+
execute "DROP TABLE intersection;"
|
291
|
+
flush_statements # the CREATE/DROP TABLE invalidates prepared statements
|
292
|
+
results
|
293
|
+
end
|
294
|
+
|
295
|
+
# Query the place table for notional "primary" place names for each of a
|
296
|
+
# list of ZIP codes. Since the place table shipped with this code is
|
297
|
+
# bespoke, and constructed from a variety of public domain sources,
|
298
|
+
# the primary name for a ZIP is not always the "right" one.
|
299
|
+
def primary_places (zips)
|
300
|
+
in_list = placeholders_for zips
|
301
|
+
sql = "SELECT * FROM place WHERE zip IN (#{in_list}) order by priority desc;"
|
302
|
+
execute sql, *zips
|
303
|
+
end
|
304
|
+
|
305
|
+
# Given a list of rows, find the unique values for a given key.
|
306
|
+
def unique_values (rows, key)
|
307
|
+
rows.map {|r| r[key]}.to_set.to_a
|
308
|
+
end
|
309
|
+
|
310
|
+
# Convert a list of rows into a hash keyed by the given keys.
|
311
|
+
def rows_to_h (rows, *keys)
|
312
|
+
hash = {}
|
313
|
+
rows.each {|row| (hash[row.values_at(*keys)] ||= []) << row; }
|
314
|
+
hash
|
315
|
+
end
|
316
|
+
|
317
|
+
# Merge the values in the list of rows given in src into the
|
318
|
+
# list of rows in dest, matching rows on the given list of keys.
|
319
|
+
# May generate more than one row in dest for each input dest row.
|
320
|
+
def merge_rows! (dest, src, *keys)
|
321
|
+
src = rows_to_h src, *keys
|
322
|
+
dest.map! {|row|
|
323
|
+
vals = row.values_at(*keys)
|
324
|
+
if src.key? vals
|
325
|
+
src[vals].map {|row2| row.merge row2}
|
326
|
+
else
|
327
|
+
[row]
|
328
|
+
end
|
329
|
+
}
|
330
|
+
dest.flatten!
|
331
|
+
end
|
332
|
+
|
333
|
+
def find_candidates (address)
|
334
|
+
places = []
|
335
|
+
candidates = []
|
336
|
+
|
337
|
+
city = address.city.sort {|a,b|a.length <=> b.length}[0]
|
338
|
+
if(!address.zip.empty? && !address.zip.nil?)
|
339
|
+
places = places_by_zip city, address.zip
|
340
|
+
end
|
341
|
+
places = places_by_city city, address.city_parts, address.state if places.empty?
|
342
|
+
return [] if places.empty?
|
343
|
+
|
344
|
+
address.city = unique_values places, :city
|
345
|
+
return places if address.street.empty?
|
346
|
+
|
347
|
+
zips = unique_values places, :zip
|
348
|
+
street = address.street.sort {|a,b|a.length <=> b.length}[0]
|
349
|
+
# puts "street parts = #{address.street_parts.inspect}"
|
350
|
+
candidates = features_by_street_and_zip street, address.street_parts, zips
|
351
|
+
|
352
|
+
if candidates.empty?
|
353
|
+
candidates = more_features_by_street_and_zip street, address.street_parts, zips
|
354
|
+
end
|
355
|
+
|
356
|
+
merge_rows! candidates, places, :zip
|
357
|
+
candidates
|
358
|
+
end
|
359
|
+
|
360
|
+
# Given a query hash and a list of candidates, assign :number
|
361
|
+
# and :precision values to each candidate. If the query building
|
362
|
+
# number is inside the candidate range, set the number on the result
|
363
|
+
# and set the precision to :range; otherwise, find the closest
|
364
|
+
# corner and set precision to :street.
|
365
|
+
def assign_number! (hn, candidates)
|
366
|
+
hn = 0 unless hn
|
367
|
+
for candidate in candidates
|
368
|
+
fromhn, tohn = candidate[:fromhn].to_i, candidate[:tohn].to_i
|
369
|
+
if (hn >= fromhn and hn <= tohn) or (hn <= fromhn and hn >= tohn)
|
370
|
+
candidate[:number] = hn.to_s
|
371
|
+
candidate[:precision] = :range
|
372
|
+
else
|
373
|
+
candidate[:number] = ((hn - fromhn).abs < (hn - tohn).abs ?
|
374
|
+
candidate[:fromhn] : candidate[:tohn]).to_s
|
375
|
+
candidate[:precision] = :street
|
376
|
+
end
|
377
|
+
end
|
378
|
+
end
|
379
|
+
|
380
|
+
def add_ranges! (address, candidates)
|
381
|
+
number = address.number.to_i
|
382
|
+
fids = unique_values candidates, :fid
|
383
|
+
ranges = ranges_by_feature fids, number, address.prenum
|
384
|
+
ranges = ranges_by_feature fids, number, nil unless !ranges.empty?
|
385
|
+
merge_rows! candidates, ranges, :fid
|
386
|
+
assign_number! number, candidates
|
387
|
+
end
|
388
|
+
|
389
|
+
def merge_edges! (candidates)
|
390
|
+
edge_ids = unique_values candidates, :tlid
|
391
|
+
records = edges edge_ids
|
392
|
+
merge_rows! candidates, records, :tlid
|
393
|
+
candidates.reject! {|record| record[:tlid].nil?}
|
394
|
+
edge_ids
|
395
|
+
end
|
396
|
+
|
397
|
+
def extend_ranges! (candidates)
|
398
|
+
edge_ids = merge_edges! candidates
|
399
|
+
full_ranges = range_ends edge_ids
|
400
|
+
merge_rows! candidates, full_ranges, :tlid, :side
|
401
|
+
end
|
402
|
+
|
403
|
+
# Score a list of candidates. For each candidate:
|
404
|
+
# * For each item in the query:
|
405
|
+
# ** if the query item is blank but the candidate is not, score 0.15;
|
406
|
+
# otherwise, if both are blank, score 1.0.
|
407
|
+
# ** If both items are set, compute the scaled Levenshtein-Damerau distance
|
408
|
+
# between them, and add that value (between 0.0 and 1.0) to the score.
|
409
|
+
# * Add 0.5 to the score for each numbered end of the range that matches
|
410
|
+
# the parity of the query number.
|
411
|
+
# * Add 1.0 if the query number is in the candidate range, otherwise
|
412
|
+
# add a fractional value for the notional distance between the
|
413
|
+
# closest candidate corner and the query.
|
414
|
+
# * Finally, divide the score by the total number of comparisons.
|
415
|
+
# The result should be between 0.0 and 1.0, with 1.0 indicating a
|
416
|
+
# perfect match.
|
417
|
+
def score_candidates! (address, candidates)
|
418
|
+
for candidate in candidates
|
419
|
+
candidate[:components] = {}
|
420
|
+
compare = [:prenum, :state, :zip]
|
421
|
+
denominator = compare.length + Street_Weight + City_Weight
|
422
|
+
|
423
|
+
street_score = (1.0 - candidate[:street_score].to_f) * Street_Weight
|
424
|
+
candidate[:components][:street] = street_score
|
425
|
+
city_score = (1.0 - candidate[:city_score].to_f) * City_Weight
|
426
|
+
candidate[:components][:city] = city_score
|
427
|
+
score = street_score + city_score
|
428
|
+
|
429
|
+
compare.each {|key|
|
430
|
+
src = address.send(key); src = src ? src.downcase : ""
|
431
|
+
dest = candidate[key]; dest = dest ? dest.downcase : ""
|
432
|
+
item_score = (src == dest) ? 1 : 0
|
433
|
+
candidate[:components][key] = item_score
|
434
|
+
score += item_score
|
435
|
+
}
|
436
|
+
|
437
|
+
if address.number and !address.number.empty?
|
438
|
+
parity = subscore = 0.0
|
439
|
+
fromhn, tohn, assigned, hn = [
|
440
|
+
candidate[:fromhn],
|
441
|
+
candidate[:tohn],
|
442
|
+
candidate[:number],
|
443
|
+
address.number].map {|s|s.to_i}
|
444
|
+
if candidate[:precision] == :range
|
445
|
+
subscore += Number_Weight
|
446
|
+
elsif assigned > 0
|
447
|
+
# only credit number subscore if assigned
|
448
|
+
subscore += Number_Weight/(assigned - hn).abs.to_f
|
449
|
+
end
|
450
|
+
candidate[:components][:number] = subscore
|
451
|
+
if hn > 0 and assigned > 0
|
452
|
+
# only credit parity if a number was given *and* assigned
|
453
|
+
parity += Parity_Weight/2.0 if fromhn % 2 == hn % 2
|
454
|
+
parity += Parity_Weight/2.0 if tohn % 2 == hn % 2
|
455
|
+
end
|
456
|
+
candidate[:components][:parity] = parity
|
457
|
+
score += subscore + parity
|
458
|
+
denominator += Number_Weight + Parity_Weight
|
459
|
+
end
|
460
|
+
candidate[:components][:total] = score.to_f
|
461
|
+
candidate[:components][:denominator] = denominator
|
462
|
+
candidate[:score] = score.to_f / denominator
|
463
|
+
end
|
464
|
+
end
|
465
|
+
|
466
|
+
# Find the candidates in a list of candidates that are tied for the
|
467
|
+
# top score and prune the remainder from the list.
|
468
|
+
def best_candidates! (candidates)
|
469
|
+
candidates.sort! {|a,b| b[:score] <=> a[:score]}
|
470
|
+
#candidates.reverse_each {|c| print "#{c[:number]} #{c[:state]} #{c[:city]} #{c[:raw_score]} #{c[:number_score]} #{c[:street_score]} #{c[:city_score]}\n" }
|
471
|
+
candidates.delete_if {|record| record[:score] < candidates[0][:score]}
|
472
|
+
end
|
473
|
+
|
474
|
+
# Compute the fractional interpolation distance for a query number along an
|
475
|
+
# edge, given all of the ranges for the same side of that edge.
|
476
|
+
def interpolation_distance (candidate)
|
477
|
+
fromhn, tohn, number = candidate.values_at(:fromhn, :tohn, :number).map{|x| x.to_i}
|
478
|
+
$stderr.print "NUM : #{fromhn} < #{number} < #{tohn} (flipped? #{candidate[:flipped]})\n" if @debug
|
479
|
+
# don't need this anymore since range_ends was improved...
|
480
|
+
fromhn, tohn = tohn, fromhn if fromhn > tohn
|
481
|
+
if fromhn > number
|
482
|
+
0.0
|
483
|
+
elsif tohn < number
|
484
|
+
1.0
|
485
|
+
else
|
486
|
+
(number - fromhn) / (tohn - fromhn).to_f
|
487
|
+
end
|
488
|
+
end
|
489
|
+
|
490
|
+
# Unpack an array of little-endian 4-byte ints, and convert them into
|
491
|
+
# signed floats by dividing by 10^6, inverting the process used by the
|
492
|
+
# compress_wkb_line() function in the SQLite helper extension.
|
493
|
+
def unpack_geometry (geom)
|
494
|
+
points = []
|
495
|
+
if !geom.nil?
|
496
|
+
# Pete - The database format is completely different to the one
|
497
|
+
# expected by the code, so I've done some detective work to
|
498
|
+
# figure out what it should be. It looks like the format is
|
499
|
+
# | 1 byte Type | 4 byte SRID | 4 byte element count| 8 byte double coordinates *
|
500
|
+
# I've added new code to read this, and commented out the old.
|
501
|
+
info = geom.unpack('CVVD*')
|
502
|
+
coords = info.slice(3, info.length)
|
503
|
+
points << [coords.shift, coords.shift] until coords.empty?
|
504
|
+
|
505
|
+
# coords = geom.unpack "V*" # little-endian 4-byte long ints
|
506
|
+
#
|
507
|
+
## now map them into signed floats
|
508
|
+
# coords.map! {|i| ( i > (1 << 31) ? i - (1 << 32) : i ) / 1_000_000.0}
|
509
|
+
# points << [coords.shift, coords.shift] until coords.empty?
|
510
|
+
end
|
511
|
+
points
|
512
|
+
end
|
513
|
+
|
514
|
+
# Calculate the longitude scaling for the average of two latitudes.
|
515
|
+
def scale_lon (lat1,lat2)
|
516
|
+
# an approximation in place of lookup.rst (10e) and (10g)
|
517
|
+
# = scale longitude distances by the cosine of the latitude
|
518
|
+
# (or, actually, the mean of two latitudes)
|
519
|
+
# -- is this even necessary?
|
520
|
+
Math.cos((lat1+lat2) / 2 * Math::PI / 180)
|
521
|
+
end
|
522
|
+
|
523
|
+
# Simple Euclidean distances between two 2-D coordinate pairs, scaled
|
524
|
+
# along the longitudinal axis by scale_lon.
|
525
|
+
def distance (a, b)
|
526
|
+
dx = (b[0] - a[0]) * scale_lon(a[1], b[1])
|
527
|
+
dy = (b[1] - a[1])
|
528
|
+
Math.sqrt(dx ** 2 + dy ** 2)
|
529
|
+
end
|
530
|
+
|
531
|
+
# Find an interpolated point along a list of linestring vertices
|
532
|
+
# proportional to the given fractional distance along the line.
|
533
|
+
def interpolate (points, fraction)
|
534
|
+
$stderr.print "POINTS: #{points.inspect}" if @debug
|
535
|
+
return points[0] if fraction == 0.0
|
536
|
+
return points[-1] if fraction == 1.0
|
537
|
+
total = 0.0
|
538
|
+
(1...points.length).each {|n| total += distance(points[n-1], points[n])}
|
539
|
+
target = total * fraction
|
540
|
+
for n in 1...points.length
|
541
|
+
step = distance(points[n-1], points[n])
|
542
|
+
if step < target
|
543
|
+
target -= step
|
544
|
+
else
|
545
|
+
scale = scale_lon(points[n][1], points[n-1][1])
|
546
|
+
dx = (points[n][0] - points[n-1][0]) * (target/step) * scale
|
547
|
+
dy = (points[n][1] - points[n-1][1]) * (target/step)
|
548
|
+
found = [points[n-1][0]+dx, points[n-1][1]+dy]
|
549
|
+
return found.map {|x| format("%.6f", x).to_f}
|
550
|
+
end
|
551
|
+
end
|
552
|
+
# raise "Can't happen!"
|
553
|
+
end
|
554
|
+
|
555
|
+
# Find and replace the city, state, and county information
|
556
|
+
# in a list of candidates with the primary place information
|
557
|
+
# for the ZIP codes in the candidate list.
|
558
|
+
def canonicalize_places! (candidates)
|
559
|
+
zips_used = unique_values(candidates, :zip)
|
560
|
+
pri_places = rows_to_h primary_places(zips_used), :zip
|
561
|
+
candidates.map! {|record|
|
562
|
+
current_places = pri_places[[record[:zip]]]
|
563
|
+
# FIXME: this should never happen!
|
564
|
+
return [] unless current_places
|
565
|
+
top_priority = current_places.map{|p| p[:priority]}.min
|
566
|
+
current_places.select {|p| p[:priority] == top_priority}.map {|p|
|
567
|
+
record.merge({
|
568
|
+
:city => p[:city],
|
569
|
+
:state => p[:state],
|
570
|
+
:fips_county => p[:fips_county]
|
571
|
+
})
|
572
|
+
}
|
573
|
+
}
|
574
|
+
candidates.flatten!
|
575
|
+
end
|
576
|
+
|
577
|
+
# Clean up a candidate record by formatting the score, replacing nil
|
578
|
+
# values with empty strings, and deleting artifacts from database
|
579
|
+
# queries.
|
580
|
+
def clean_record! (record)
|
581
|
+
record[:score] = format("%.3f", record[:score]).to_f \
|
582
|
+
unless record[:score].nil?
|
583
|
+
record.keys.each {|k| record[k] = "" if record[k].nil? } # clean up nils
|
584
|
+
record.delete :components unless @debug
|
585
|
+
record.delete_if {|k,v| k.is_a? Fixnum or
|
586
|
+
[:geometry, :side, :tlid, :fid, :fid1, :fid2, :street_phone,
|
587
|
+
:city_phone, :fromhn, :tohn, :paflag, :flipped, :street_score,
|
588
|
+
:city_score, :priority, :fips_class, :fips_place, :status].include? k}
|
589
|
+
end
|
590
|
+
|
591
|
+
def best_places (address, places, canonicalize=false)
|
592
|
+
return [] unless !places.empty?
|
593
|
+
score_candidates! address, places
|
594
|
+
best_candidates! places
|
595
|
+
canonicalize_places! places if canonicalize
|
596
|
+
|
597
|
+
# uniqify places
|
598
|
+
by_name = rows_to_h(places, :city, :state)
|
599
|
+
if !by_name.nil?
|
600
|
+
begin
|
601
|
+
by_name.values.each {|v|
|
602
|
+
v.sort! {|a,b|
|
603
|
+
a[:zip] <=> b[:zip]
|
604
|
+
}}
|
605
|
+
rescue
|
606
|
+
|
607
|
+
end
|
608
|
+
places = by_name.map {|k,v| v[0]}
|
609
|
+
|
610
|
+
places.each {|record| clean_record! record}
|
611
|
+
places.each {|record|
|
612
|
+
record[:precision] = (record[:zip] == address.zip ? :zip : :city)
|
613
|
+
}
|
614
|
+
end
|
615
|
+
places
|
616
|
+
end
|
617
|
+
|
618
|
+
# Given an Address object, return a list of possible geocodes by place
|
619
|
+
# name. If canonicalize is true, attempt to return the "primary" postal
|
620
|
+
# place name for the given city, state, or ZIP.
|
621
|
+
def geocode_place (address, canonicalize=false)
|
622
|
+
places = []
|
623
|
+
places = places_by_zip address.text, address.zip if !address.zip.empty? or !address.zip.nil?
|
624
|
+
places = places_by_city address.text, address.city_parts, address.state if places.empty?
|
625
|
+
best_places address, places, canonicalize
|
626
|
+
end
|
627
|
+
|
628
|
+
def geocode_intersection (address, canonical_place=false)
|
629
|
+
candidates = find_candidates address
|
630
|
+
return [] if candidates.empty?
|
631
|
+
return best_places(address, candidates, canonical_place) if candidates[0][:street].nil?
|
632
|
+
|
633
|
+
features = rows_to_h candidates, :fid
|
634
|
+
intersects = intersections_by_fid features.keys.flatten
|
635
|
+
intersects.map! {|record|
|
636
|
+
feat1, feat2 = record.values_at(:fid1, :fid2).map {|k| features[[k]][0]}
|
637
|
+
record.merge! feat1
|
638
|
+
record[:street1] = record.delete(:street)
|
639
|
+
record[:street2] = feat2[:street]
|
640
|
+
record[:lon], record[:lat] = unpack_geometry(record.delete(:point))[0]
|
641
|
+
record[:precision] = :intersection
|
642
|
+
record[:street_score] = (feat1[:street_score].to_f + feat2[:street_score].to_f)/2
|
643
|
+
record
|
644
|
+
}
|
645
|
+
#pp(intersects)
|
646
|
+
|
647
|
+
score_candidates! address, intersects
|
648
|
+
best_candidates! intersects
|
649
|
+
|
650
|
+
by_point = rows_to_h(intersects, :lon, :lat)
|
651
|
+
candidates = by_point.values.map {|records| records[0]}
|
652
|
+
|
653
|
+
canonicalize_places! candidates if canonical_place
|
654
|
+
candidates.each {|record| clean_record! record}
|
655
|
+
candidates
|
656
|
+
end
|
657
|
+
|
658
|
+
# Given an Address object, return a list of possible geocodes by address
|
659
|
+
# range interpolation. If canonicalize is true, attempt to return the
|
660
|
+
# "primary" street and place names, if they are different from the ones
|
661
|
+
# given.
|
662
|
+
def geocode_address (address, canonical_place=false)
|
663
|
+
candidates = find_candidates address
|
664
|
+
return [] if candidates.empty?
|
665
|
+
return best_places(address, candidates, canonical_place) if candidates[0][:street].nil?
|
666
|
+
|
667
|
+
score_candidates! address, candidates
|
668
|
+
best_candidates! candidates
|
669
|
+
|
670
|
+
#candidates.sort {|a,b| b[:score] <=> a[:score]}.each {|candidate|
|
671
|
+
add_ranges! address, candidates
|
672
|
+
score_candidates! address, candidates
|
673
|
+
#pp candidates.sort {|a,b| b[:score] <=> a[:score]}
|
674
|
+
best_candidates! candidates
|
675
|
+
|
676
|
+
# sometimes multiple fids match the same tlid
|
677
|
+
by_tlid = rows_to_h candidates, :tlid
|
678
|
+
candidates = by_tlid.values.map {|records| records[0]}
|
679
|
+
|
680
|
+
# if no number is assigned in the query, only return one
|
681
|
+
# result for each street/zip combo
|
682
|
+
if !address.number.empty?
|
683
|
+
extend_ranges! candidates
|
684
|
+
else
|
685
|
+
by_street = rows_to_h candidates, :street, :zip
|
686
|
+
candidates = by_street.values.map {|records| records[0]}
|
687
|
+
merge_edges! candidates
|
688
|
+
end
|
689
|
+
candidates.map {|record|
|
690
|
+
dist = interpolation_distance record
|
691
|
+
$stderr.print "DIST: #{dist}\n" if @debug
|
692
|
+
points = unpack_geometry record[:geometry]
|
693
|
+
points.reverse! if record[:flipped]
|
694
|
+
record[:lon], record[:lat] = interpolate points, dist
|
695
|
+
}
|
696
|
+
|
697
|
+
canonicalize_places! candidates if canonical_place
|
698
|
+
|
699
|
+
candidates.each {|record| clean_record! record}
|
700
|
+
candidates
|
701
|
+
end
|
702
|
+
|
703
|
+
public
|
704
|
+
|
705
|
+
# Geocode a given address or place name string. The max_penalty and cutoff
|
706
|
+
# arguments are passed to the Address parse functions. If canonicalize is
|
707
|
+
# true, attempt to return the "primary" street and place names, if they are
|
708
|
+
# different from the ones given.
|
709
|
+
#
|
710
|
+
# Returns possible candidate matches as a list of hashes.
|
711
|
+
#
|
712
|
+
# * The :lat and :lon values of each hash store the range-interpolated
|
713
|
+
# address coordinates as latitude and longitude in the WGS84 spheroid.
|
714
|
+
# * The :precision value may be one of :city, :zip, :street, or :range, in
|
715
|
+
# order of increasing precision.
|
716
|
+
# * The :score value will be a float between 0.0 and 1.0 representing
|
717
|
+
# the approximate "goodness" of the candidate match.
|
718
|
+
# * The other values in the hash will represent various structured
|
719
|
+
# components of the address and place name.
|
720
|
+
def geocode (info_to_geocode, canonical_place=false)
|
721
|
+
address = Address.new info_to_geocode
|
722
|
+
$stderr.print "ADDR: #{address.inspect}\n" if @debug
|
723
|
+
return [] if address.city.empty? and address.zip.empty?
|
724
|
+
results = []
|
725
|
+
start_time = Time.now if @debug
|
726
|
+
if address.po_box? and !address.zip.empty?
|
727
|
+
results = geocode_place address, canonical_place
|
728
|
+
end
|
729
|
+
if address.intersection? and !address.street.empty? and address.number.empty?
|
730
|
+
results = geocode_intersection address, canonical_place
|
731
|
+
end
|
732
|
+
if results.empty? and !address.street.empty?
|
733
|
+
results = geocode_address address, canonical_place
|
734
|
+
end
|
735
|
+
if results.empty?
|
736
|
+
results = geocode_place address, canonical_place
|
737
|
+
end
|
738
|
+
if @debug
|
739
|
+
runtime = format("%.3f", Time.now - start_time)
|
740
|
+
$stderr.print "DONE: #{runtime}s\n"
|
741
|
+
end
|
742
|
+
results
|
743
|
+
end
|
744
|
+
end
|
745
|
+
end
|