@@ -5,17 +5,226 @@ BEGIN;
55CREATE SCHEMA pgosm_flex_faker ;
66
77
8- CREATE FUNCTION pgosm_flex_faker .location_in_place_landuse()
9- RETURNS BOOLEAN
10- LANGUAGE sql
11- SECURITY DEFINER
12- SET search_path TO ' pgosm_flex_faker, pg_temp'
8+ CREATE PROCEDURE pgosm_flex_faker .location_in_place_landuse ()
9+ LANGUAGE plpgsql
1310AS $$
11+ BEGIN
1412
15- -- Do something
13+ -- Define a custom `place_osm_types` table before executing to customize areas
14+ CREATE TEMP TABLE IF NOT EXISTS place_osm_types AS
15+ SELECT ' retail' AS osm_type
16+ UNION
17+ SELECT ' commercial' AS osm_type
18+ ;
1619
17- SELECT True;
1820
21+ -- Basic selection, provide attributes used to rank locations
22+ DROP TABLE IF EXISTS places_for_shops_1;
23+ CREATE TEMP TABLE places_for_shops_1 AS
24+ WITH base AS (
25+ SELECT osm_id, name, osm_type, admin_level, nest_level,
26+ -- Rounding is assuming SRID 3857, or another unit in Meters or Feet.
27+ ROUND(public .ST_Area (geom)::NUMERIC , 0 ) AS geom_area, geom
28+ FROM osm .place_polygon_nested
29+ -- Using innermost places to reduce likelihood over overlap
30+ WHERE innermost
31+ AND name <> ' '
32+ AND admin_level < 99
33+ ), with_space AS (
34+ -- Within each Place, find how many places are "near" (intersects)
35+ -- or contain the types of places (commercial, retail, residential, etc)
36+ -- defined in place_osm_types
37+ SELECT b .osm_id ,
38+ COUNT (lp .osm_id ) AS near_areas,
39+ COALESCE(SUM (public .ST_Area (lp .geom )), 0 ) AS near_space,
40+ COUNT (c .osm_id ) AS contained_areas,
41+ COALESCE(SUM (public .ST_Area (c .geom )), 0 ) AS contained_space
42+ FROM base b
43+ LEFT JOIN osm .landuse_polygon lp
44+ ON public .ST_Intersects (b .geom , lp .geom )
45+ AND lp .osm_type IN (SELECT osm_type FROM place_osm_types)
46+ LEFT JOIN osm .landuse_polygon c
47+ ON public .ST_Contains (b .geom , c .geom )
48+ AND c .osm_type IN (SELECT osm_type FROM place_osm_types)
49+ GROUP BY b .osm_id
50+ )
51+ SELECT b.* , ws .contained_areas , ws .contained_space ,
52+ ws .near_areas , ws .near_space
53+ FROM base b
54+ INNER JOIN with_space ws ON b .osm_id = ws .osm_id
55+ ;
56+
57+
58+ DROP TABLE IF EXISTS places_for_shops;
59+ CREATE TEMP TABLE places_for_shops AS
60+ SELECT osm_id, name, osm_type, admin_level, contained_areas, contained_space,
61+ near_areas, near_space, geom_area,
62+ contained_space / geom_area AS space_contained_ratio_higher_is_better,
63+ near_space / geom_area AS space_near_ratio_higher_is_better,
64+ geom
65+ FROM places_for_shops_1
66+ ORDER BY space_contained_ratio_higher_is_better DESC ,
67+ space_near_ratio_higher_is_better DESC
68+ ;
69+
70+
71+ /*
72+ * The following scoring logic creates scores for each place depending
73+ * on how it's contained and nearby landuse data compare to the area's
74+ * percentile values.
75+ */
76+ DROP TABLE IF EXISTS place_scores;
77+ CREATE TEMP TABLE place_scores AS
78+ WITH breakpoints AS (
79+ -- Calculate percentiles of space available across all available place inputs
80+ -- This should let each region adjust for the input data
81+ SELECT percentile_cont(0 .25 )
82+ within group (order by contained_space asc )
83+ as contained_space_25_perc,
84+ percentile_cont(0 .50 )
85+ within group (order by contained_space asc )
86+ as contained_space_50_perc,
87+ percentile_cont(0 .90 )
88+ within group (order by near_space asc )
89+ as near_space_90_perc
90+ FROM places_for_shops
91+ WHERE near_areas > 0
92+ )
93+ SELECT p .osm_id ,
94+ -- Actual ranking is arbitrary, they key is understanding that scores
95+ -- under a random value in the next step (where random between 0.0 and 1.0)
96+ -- so increasing the max score here results in some areas almost always
97+ -- being picked
98+ CASE WHEN b .contained_space_50_perc < p .contained_space
99+ THEN .55
100+ WHEN b .contained_space_25_perc < p .contained_space
101+ THEN .35
102+ ELSE .01
103+ END AS contained_space_score,
104+ CASE WHEN b .near_space_90_perc < p .near_space
105+ THEN .1
106+ ELSE .01
107+ END AS near_space_score
108+ FROM places_for_shops p
109+ INNER JOIN breakpoints b ON True
110+ -- Excludes places that aren't even nearby (intersects) an appropriate
111+ -- place type
112+ WHERE p .near_areas > 0
113+ ;
114+
115+
116+ DROP TABLE IF EXISTS selected;
117+ CREATE TEMP TABLE selected AS
118+ WITH a AS (
119+ SELECT p .osm_id ,
120+ s .contained_space_score + s .near_space_score
121+ AS total_score,
122+ random() as rnd
123+ FROM places_for_shops p
124+ INNER JOIN place_scores s
125+ ON p .osm_id = s .osm_id
126+ )
127+ SELECT a .osm_id
128+ FROM a
129+ WHERE a .total_score > a .rnd
130+ ;
131+
132+
133+ -- Selected areas to put points into.
134+ DROP TABLE IF EXISTS faker_place_polygon;
135+ CREATE TEMP TABLE faker_place_polygon AS
136+ SELECT p.*
137+ FROM selected s
138+ INNER JOIN places_for_shops p ON s .osm_id = p .osm_id
139+ ORDER BY p .name
140+ ;
141+
142+ CREATE INDEX gix_faker_place_polygon
143+ ON faker_place_polygon USING GIST (geom)
144+ ;
145+
146+
147+ /*
148+ Ranking roads by osm_type with goal of scoring roads with lower speed
149+ limits higher. Uses helper table loaded by PgOSM Flex.
150+
151+ Uses window function for rank steps, then normalize to 0-1 range.
152+ Finally, squishes range into 0.05 - 0.90 to prevent guarantees of
153+ never or always included.
154+ */
155+ DROP TABLE IF EXISTS road_osm_type_rank;
156+ CREATE TEMP TABLE road_osm_type_rank AS
157+ WITH rank_lower_speed_better AS (
158+ SELECT osm_type, maxspeed_mph,
159+ RANK() OVER (ORDER BY maxspeed_mph desc ) AS rnk_raw
160+ FROM pgosm .road
161+ WHERE route_motor
162+ AND osm_type NOT LIKE ' %link'
163+ ), aggs_for_normalization AS (
164+ SELECT MIN (rnk_raw) AS min_rnk, MAX (rnk_raw) AS max_rnk
165+ FROM rank_lower_speed_better
166+ ), normal_rnk AS (
167+ SELECT r .osm_type , r .maxspeed_mph ,
168+ (rnk_raw * 1 .0 - min_rnk) / (max_rnk - min_rnk)
169+ AS normalized_rnk
170+ FROM rank_lower_speed_better r
171+ JOIN aggs_for_normalization ON True
172+ )
173+ SELECT osm_type, maxspeed_mph,
174+ CASE WHEN normalized_rnk < 0 .05 THEN 0 .05
175+ WHEN normalized_rnk > 0 .9 THEN .9
176+ ELSE normalized_rnk
177+ END AS normalized_rnk
178+ FROM normal_rnk
179+ ;
180+
181+
182+ -- ---------------------------------------
183+ -- Identify roads where a building could be
184+ -- Not using actual buildings / addresses because:
185+ -- -- a) privacy
186+ -- -- b) coverage
187+ DROP TABLE IF EXISTS selected_roads ;
188+ CREATE TEMP TABLE selected_roads AS
189+ WITH road_ranks AS (
190+ SELECT p .osm_id AS place_osm_id, p .name AS place_name,
191+ rr .normalized_rnk AS road_type_score,
192+ r .osm_id AS road_osm_id
193+ FROM faker_place_polygon p
194+ INNER JOIN osm .landuse_polygon c
195+ ON public .ST_Contains (p .geom , c .geom )
196+ AND c .osm_type IN (SELECT osm_type FROM place_osm_types)
197+ INNER JOIN osm .road_line r
198+ ON c .geom && r .geom
199+ AND r .route_motor
200+ AND r .osm_type NOT IN (' service'
201+ AND r .osm_type NOT LIKE ' %link'
202+ INNER JOIN road_osm_type_rank rr
203+ ON r .osm_type = rr .osm_type
204+ ), ranked AS (
205+ SELECT * ,
206+ ROW_NUMBER() OVER (
207+ PARTITION BY place_osm_id
208+ ORDER BY road_type_score DESC , random()) AS rnk
209+ FROM road_ranks
210+ )
211+ SELECT *
212+ FROM ranked
213+ WHERE rnk = 1
214+ ;
215+
216+
217+ DROP TABLE IF EXISTS faker_store_location;
218+ CREATE TEMP TABLE faker_store_location AS
219+ SELECT a .place_osm_id , a .place_name , a .road_osm_id ,
220+ r .osm_type , r .name , r .ref ,
221+ public .ST_LineInterpolatePoint (public .ST_LineMerge (r .geom ), random()) AS geom
222+ FROM selected_roads a
223+ INNER JOIN osm .road_line r ON a .road_osm_id = r .osm_id
224+ ;
225+
226+
227+ END
19228$$
20229;
21230
0 commit comments