Add comments on avoid reuse of parse-time snapshot

alvherre · alvherre · commit db989184cda7 · 2020-02-28T13:25:26.000-03:00
Apparently, reusing the parse-time query snapshot for later steps (execution) is a frequently considered optimization ... but it doesn't work, for reasons discovered in thread [1]. Adding some comments about why it doesn't really work can relieve some future hackers from wasting time reimplementing it again. [1] https://postgr.es/m/flat/5075D8DF.6050500@fuzzy.cz Author: Michail Nikolaev Discussion: https://postgr.es/m/CANtu0ogp6cTvMJObXP8n=k+JtqxY1iT9UV5MbGCpjjPa5crCiw@mail.gmail.com
diff --git a/src/backend/tcop/postgres.c b/src/backend/tcop/postgres.c
@@ -1155,7 +1155,16 @@ exec_simple_query(const char *query_string)
 		plantree_list = pg_plan_queries(querytree_list,
 										CURSOR_OPT_PARALLEL_OK, NULL);
 
-		/* Done with the snapshot used for parsing/planning */
+		/*
+		 * Done with the snapshot used for parsing/planning.
+		 *
+		 * While it looks promising to reuse the same snapshot for query
+		 * execution (at least for simple protocol), unfortunately it causes
+		 * execution to use a snapshot that has been acquired before locking
+		 * any of the tables mentioned in the query.  This creates user-
+		 * visible anomalies, so refrain.  Refer to
+		 * https://postgr.es/m/flat/5075D8DF.6050500@fuzzy.cz for details.
+		 */
 		if (snapshot_set)
 			PopActiveSnapshot();
 
