[PGPRO-8894] Correction of expression calculation for case recheckPhrase=true

kovdb75 · kovdb75 · commit 8c490862ca0a · 2023-10-04T01:27:11.000+03:00
diff --git a/Makefile b/Makefile
@@ -30,7 +30,7 @@ REGRESS = security rum rum_validate rum_hash ruminv timestamp orderby orderby_ha
 	int2 int4 int8 float4 float8 money oid \
     time timetz date interval \
     macaddr inet cidr text varchar char bytea bit varbit \
-	numeric rum_weight
+	numeric rum_weight expr
 
 TAP_TESTS = 1
 
diff --git a/expected/expr.out b/expected/expr.out
@@ -0,0 +1,26 @@
+CREATE TABLE documents (
+    en text not null,
+    score float not null,
+    textsearch_index_en_col tsvector
+);
+INSERT INTO documents VALUES ('the pet cat is in the shed', 56, to_tsvector('english', 'the pet cat is in the shed'));
+CREATE INDEX textsearch_index_en ON documents
+    USING rum (textsearch_index_en_col rum_tsvector_addon_ops, score)
+    WITH (attach = 'score', to = 'textsearch_index_en_col');
+SET enable_seqscan=off;
+-- should be 1 row
+SELECT * FROM documents WHERE textsearch_index_en_col @@ ('pet'::tsquery <-> ('dog'::tsquery || 'cat'::tsquery));
+             en             | score | textsearch_index_en_col  
+----------------------------+-------+--------------------------
+ the pet cat is in the shed |    56 | 'cat':3 'pet':2 'shed':7
+(1 row)
+
+SET enable_seqscan=on;
+-- 1 row
+SELECT * FROM documents WHERE textsearch_index_en_col @@ ('pet'::tsquery <-> ('dog'::tsquery || 'cat'::tsquery));
+             en             | score | textsearch_index_en_col  
+----------------------------+-------+--------------------------
+ the pet cat is in the shed |    56 | 'cat':3 'pet':2 'shed':7
+(1 row)
+
+DROP TABLE documents;
diff --git a/sql/expr.sql b/sql/expr.sql
@@ -0,0 +1,21 @@
+CREATE TABLE documents (
+    en text not null,
+    score float not null,
+    textsearch_index_en_col tsvector
+);
+
+INSERT INTO documents VALUES ('the pet cat is in the shed', 56, to_tsvector('english', 'the pet cat is in the shed'));
+
+CREATE INDEX textsearch_index_en ON documents
+    USING rum (textsearch_index_en_col rum_tsvector_addon_ops, score)
+    WITH (attach = 'score', to = 'textsearch_index_en_col');
+
+SET enable_seqscan=off;
+-- should be 1 row
+SELECT * FROM documents WHERE textsearch_index_en_col @@ ('pet'::tsquery <-> ('dog'::tsquery || 'cat'::tsquery));
+
+SET enable_seqscan=on;
+-- 1 row
+SELECT * FROM documents WHERE textsearch_index_en_col @@ ('pet'::tsquery <-> ('dog'::tsquery || 'cat'::tsquery));
+
+DROP TABLE documents;
diff --git a/src/rum_ts_utils.c b/src/rum_ts_utils.c
@@ -284,7 +284,14 @@ checkcondition_rum(void *checkval, QueryOperand *val, ExecPhraseData *data)
 		 * addInfo
 		 */
 		if (gcv->recheckPhrase)
-			return ((val->weight) ? TS_MAYBE : TS_YES);
+		{
+			/*
+			 * We cannot return TS_YES here (if "val->weight > 0"), because
+			 * data->npos = 0 and we have incorrect porocessing of this result
+			 * at the upper levels. So return TS_MAYBE.
+			 */
+			return TS_MAYBE;
+		}
 
 		positions = DatumGetByteaP(gcv->addInfo[j]);
 		ptrt = (char *) VARDATA_ANY(positions);
