From b7cbb5d55f51ba3ee35cf1e13afadf2dd90d7431 Mon Sep 17 00:00:00 2001
From: Tom Russell <tomalrussell@gmail.com>
Date: Tue, 22 Jan 2019 22:38:00 +0000
Subject: [PATCH] Draft search box db table

---
 migrations/008.search.up.sql | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)
 create mode 100644 migrations/008.search.up.sql

diff --git a/migrations/008.search.up.sql b/migrations/008.search.up.sql
new file mode 100644
index 00000000..c780c6d0
--- /dev/null
+++ b/migrations/008.search.up.sql
@@ -0,0 +1,31 @@
+--
+-- Set up search table for text search over locations
+--
+-- uses extension: CREATE EXTENSION pg_trgm;
+--
+
+CREATE TABLE IF NOT EXISTS search_locations (
+    -- internal unique id
+    search_id serial PRIMARY KEY,
+    -- string to match against in search (E1 0AB / Hackney / Commercial Road...)
+    search_str text,
+    -- search class for hint (postcode / borough / road name...)
+    search_class text,
+    -- geometry as EPSG:4326 (lat-long) to be used in front end directly
+    center geometry(POINT, 4326),
+    -- zoom level
+    zoom int
+);
+
+-- Index for fuzzy match
+-- see https://www.postgresql.org/docs/current/pgtrgm.html
+-- and https://www.postgresql.org/docs/current/textsearch-indexes.html
+--
+-- Should support (I)LIKE or similarity searches:
+-- - SELECT * FROM search_locations WHERE search_str ILIKE 'e1%'
+-- - SELECT *, search_str <-> 'searchterm' AS dist FROM search_locations ORDER BY dist LIMIT 5;
+-- - SELECT *, similarity(search_str, 'searchterm') AS dist FROM search_locations
+--       WHERE t % 'searchterm' ORDER BY dist DESC, t LIMIT 5;
+-- Docs suggest the second query will perform better than the third, when only a small number
+-- of closest matches are desired, if combined with a GIST index (not GIN as below),
+CREATE INDEX trgm_idx_search_str ON search_locations USING GIN (search_str gin_trgm_ops);