First incomplete version of lexeme hashing in index

Author: Maksim Milyutin

rum--1.0.sql

@@ -80,12 +80,17 @@ RETURNS bytea
 AS 'MODULE_PATHNAME'
 LANGUAGE C IMMUTABLE STRICT;
 
+CREATE FUNCTION rum_cmp_tslexeme(bytea, bytea)
+RETURNS integer
+AS 'MODULE_PATHNAME'
+LANGUAGE C IMMUTABLE STRICT;
+
 CREATE OPERATOR CLASS rum_tsvector_ops
 FOR TYPE tsvector USING rum
 AS
         OPERATOR        1       @@ (tsvector, tsquery),
         OPERATOR        2       <=> (tsvector, tsquery) FOR ORDER BY pg_catalog.float_ops,
-        FUNCTION        1       gin_cmp_tslexeme(text, text),
+        FUNCTION        1       rum_cmp_tslexeme(bytea, bytea),
         FUNCTION        2       rum_extract_tsvector(tsvector,internal,internal,internal,internal),
         FUNCTION        3       rum_extract_tsquery(tsquery,internal,smallint,internal,internal,internal,internal),
         FUNCTION        4       rum_tsquery_consistent(internal,smallint,tsvector,int,internal,internal,internal,internal),
@@ -94,7 +99,7 @@ AS
         FUNCTION        7       rum_tsquery_pre_consistent(internal,smallint,tsvector,int,internal,internal,internal,internal),
         FUNCTION        8       rum_tsquery_distance(internal,smallint,tsvector,int,internal,internal,internal,internal,internal),
         FUNCTION        10      rum_ts_join_pos(internal, internal),
-        STORAGE         text;
+        STORAGE         bytea;
 -- timestamp ops
 
 CREATE FUNCTION timestamp_distance(timestamp, timestamp)

rum.h

@@ -243,7 +243,7 @@ typedef signed char RumNullCategory;
  */
 #define RumGetDownlink(itup)	RumItemPointerGetBlockNumber(&(itup)->t_tid)
 #define RumSetDownlink(itup,blkno)	ItemPointerSet(&(itup)->t_tid, blkno, InvalidOffsetNumber)
-
+CREATE INDEX rumidx ON test_rum USING rum (a rum_tsvector_ops);
 
 /*
  * Data (posting tree) pages

rum_ts_utils.c

@@ -11,7 +11,9 @@
 
 #include "postgres.h"
 
+#include "access/hash.h"
 #include "access/htup_details.h"
+#include "catalog/pg_collation.h"
 #include "catalog/pg_type.h"
 #include "funcapi.h"
 #include "miscadmin.h"
@@ -25,6 +27,7 @@
 
 #include <math.h>
 
+PG_FUNCTION_INFO_V1(rum_cmp_tslexeme);
 PG_FUNCTION_INFO_V1(rum_extract_tsvector);
 PG_FUNCTION_INFO_V1(rum_extract_tsquery);
 PG_FUNCTION_INFO_V1(rum_tsvector_config);
@@ -503,11 +506,15 @@ rum_extract_tsvector(PG_FUNCTION_ARGS)
 		for (i = 0; i < vector->size; i++)
 		{
 			text	   *txt;
+			bytea	   *hash_value;
 			bytea	   *posData;
 			int			posDataSize;
 
 			txt = cstring_to_text_with_len(STRPTR(vector) + we->pos, we->len);
-			entries[i] = PointerGetDatum(txt);
+			hash_value = (bytea *) palloc(VARHDRSZ + sizeof(int32));
+			SET_VARSIZE(hash_value, VARHDRSZ + sizeof(int32));
+			*VARDATA(hash_value) = DirectFunctionCall1(hashtext, PointerGetDatum(txt));
+			entries[i] = PointerGetDatum(hash_value);
 
 			if (we->haspos)
 			{
@@ -586,10 +593,14 @@ rum_extract_tsquery(PG_FUNCTION_ARGS)
 		for (i = 0; i < (*nentries); i++)
 		{
 			text	   *txt;
+			bytea	   *hash_value;
 
 			txt = cstring_to_text_with_len(GETOPERAND(query) + operands[i]->distance,
 										   operands[i]->length);
-			entries[i] = PointerGetDatum(txt);
+			hash_value = (bytea *) palloc(VARHDRSZ + sizeof(int32));
+			SET_VARSIZE(hash_value, VARHDRSZ + sizeof(int32));
+			*VARDATA(hash_value) = DirectFunctionCall1(hashtext, PointerGetDatum(txt));
+			entries[i] = PointerGetDatum(hash_value);
 			partialmatch[i] = operands[i]->prefix;
 			(*extra_data)[i] = (Pointer) map_item_operand;
 		}
@@ -1389,3 +1400,17 @@ rum_ts_join_pos(PG_FUNCTION_ARGS)
 
 	PG_RETURN_BYTEA_P(result);
 }
+
+Datum
+rum_cmp_tslexeme(PG_FUNCTION_ARGS)
+{
+	bytea	   *arg1 = PG_GETARG_BYTEA_P(0);
+	bytea	   *arg2 = PG_GETARG_BYTEA_P(1);
+	int32		a = *VARDATA(arg1);
+	int32		b = *VARDATA(arg2);
+	int			cmp;
+
+	cmp = (a > b) ? 1 : ((a == b) ? 0 : -1);
+
+	PG_RETURN_INT32(cmp);
+}

sql/rum.sql

@@ -100,17 +100,17 @@ DELETE FROM tst WHERE i = 5;
 VACUUM tst;
 INSERT INTO tst SELECT i%10, to_tsvector('simple', substr(md5(i::text), 1, 1)) FROM generate_series(14001,15000) i;
 
-set enable_bitmapscan=off;
-explain (costs off)
-SELECT a <=> to_tsquery('pg_catalog.english', 'w:*'), *
-	FROM test_rum
-	WHERE a @@ to_tsquery('pg_catalog.english', 'w:*')
-	ORDER BY a <=> to_tsquery('pg_catalog.english', 'w:*');
-SELECT a <=> to_tsquery('pg_catalog.english', 'w:*'), *
-	FROM test_rum
-	WHERE a @@ to_tsquery('pg_catalog.english', 'w:*')
-	ORDER BY a <=> to_tsquery('pg_catalog.english', 'w:*');
-SELECT a <=> to_tsquery('pg_catalog.english', 'b:*'), *
-	FROM test_rum
-	WHERE a @@ to_tsquery('pg_catalog.english', 'b:*')
-	ORDER BY a <=> to_tsquery('pg_catalog.english', 'b:*');
+--  set enable_bitmapscan=off;
+--  explain (costs off)
+--  SELECT a <=> to_tsquery('pg_catalog.english', 'w:*'), *
+	--  FROM test_rum
+	--  WHERE a @@ to_tsquery('pg_catalog.english', 'w:*')
+	--  ORDER BY a <=> to_tsquery('pg_catalog.english', 'w:*');
+--  SELECT a <=> to_tsquery('pg_catalog.english', 'w:*'), *
+	--  FROM test_rum
+	--  WHERE a @@ to_tsquery('pg_catalog.english', 'w:*')
+	--  ORDER BY a <=> to_tsquery('pg_catalog.english', 'w:*');
+--  SELECT a <=> to_tsquery('pg_catalog.english', 'b:*'), *
+	--  FROM test_rum
+	--  WHERE a @@ to_tsquery('pg_catalog.english', 'b:*')
+	--  ORDER BY a <=> to_tsquery('pg_catalog.english', 'b:*');