use trigram tokenizer instead of filter to maintain ordering

Author: nicpottier

Diff for: indexer.go

@@ -308,20 +308,12 @@ const indexSettings = `
 			"routing_partition_size": 3
 		},
 		"analysis": {
-            "filter": {
-                "trigrams_filter": {
-                    "type":     "ngram",
-                    "min_gram": 3,
-                    "max_gram": 3
-				}
-            },
             "analyzer": {
                 "trigrams": {
                     "type":      "custom",
-                    "tokenizer": "standard",
+                    "tokenizer": "trigram",
                     "filter": [
-                        "lowercase",
-                        "trigrams_filter"
+                        "lowercase"
                     ]
 				},
 				"locations": {
@@ -344,6 +336,11 @@ const indexSettings = `
 				  "type": "pattern",
 				  "pattern": "(.* > )?([^>]+)",
 				  "group": 2
+				},
+				"trigram": {
+					"type" : "ngram",
+					"min_gram" : 3,
+					"max_gram" : 3
 				}
 			},
 			"normalizer": {

Diff for: indexer_test.go

@@ -88,21 +88,25 @@ func TestIndexing(t *testing.T) {
 	// urn substring query
 	query = elastic.NewNestedQuery("urns", elastic.NewBoolQuery().Must(
 		elastic.NewMatchQuery("urns.scheme", "tel"),
-		elastic.NewMatchQuery("urns.path", "779")))
-	assertQuery(t, client, physicalName, query, []int64{1, 3, 4, 5, 6, 7})
+		elastic.NewMatchPhraseQuery("urns.path", "779")))
+	assertQuery(t, client, physicalName, query, []int64{1, 3, 4, 7})
 
 	// urn substring query with more characters (77911)
 	query = elastic.NewNestedQuery("urns", elastic.NewBoolQuery().Must(
 		elastic.NewMatchQuery("urns.scheme", "tel"),
-		elastic.NewMatchQuery("urns.path", "779"),
-		elastic.NewMatchQuery("urns.path", "791"),
-		elastic.NewMatchQuery("urns.path", "911")))
+		elastic.NewMatchPhraseQuery("urns.path", "77911")))
 	assertQuery(t, client, physicalName, query, []int64{1})
 
+	// urn substring query with more characters (600055)
+	query = elastic.NewNestedQuery("urns", elastic.NewBoolQuery().Must(
+		elastic.NewMatchQuery("urns.scheme", "tel"),
+		elastic.NewMatchPhraseQuery("urns.path", "600055")))
+	assertQuery(t, client, physicalName, query, []int64{6})
+
 	// match a contact with multiple tel urns
 	query = elastic.NewNestedQuery("urns", elastic.NewBoolQuery().Must(
 		elastic.NewMatchQuery("urns.scheme", "tel"),
-		elastic.NewMatchQuery("urns.path", "222")))
+		elastic.NewMatchPhraseQuery("urns.path", "222")))
 	assertQuery(t, client, physicalName, query, []int64{1})
 
 	// text query

Diff for: testdb.sql

@@ -115,8 +115,8 @@ INSERT INTO contacts_contacturn(id, contact_id, scheme, org_id, priority, path,
 (3, 2, 'tel', 1, 50, '+12067793333', NULL, 'tel:+12067793333'),
 (4, 3, 'tel', 1, 50, '+12067794444', NULL, 'tel:+12067794444'),
 (5, 4, 'tel', 1, 50, '+12067795555', NULL, 'tel:+12067795555'),
-(6, 5, 'tel', 1, 50, '+12067796666', NULL, 'tel:+12067796666'),
-(7, 6, 'tel', 2, 50, '+12067797777', NULL, 'tel:+12067797777'),
+(6, 5, 'tel', 1, 50, '+12060000556', NULL, 'tel:+12067796666'),
+(7, 6, 'tel', 2, 50, '+12060005577', NULL, 'tel:+12067797777'),
 (8, 7, 'tel', 2, 50, '+12067798888', NULL, 'tel:+12067798888'),
 (9, 8, 'viber', 2, 90, 'viberpath==', NULL, 'viber:viberpath=='),
 (10, 9, 'facebook', 2, 90, 1000001, 'funguy', 'facebook:1000001'),