Commitdb98918

committed

Add comments on avoid reuse of parse-time snapshot

Apparently, reusing the parse-time query snapshot for later steps(execution) is a frequently considered optimization ... but it doesn'twork, for reasons discovered in thread [1]. Adding some comments aboutwhy it doesn't really work can relieve some future hackers from wastingtime reimplementing it again.[1]https://postgr.es/m/flat/5075D8DF.6050500@fuzzy.czAuthor: Michail NikolaevDiscussion:https://postgr.es/m/CANtu0ogp6cTvMJObXP8n=k+JtqxY1iT9UV5MbGCpjjPa5crCiw@mail.gmail.com

1 parent1933ae6 commitdb98918Copy full SHA for db98918

File tree

+10

-1

lines changed

+10

-1

lines changed

Lines changed: 10 additions & 1 deletion

Original file line number	Diff line number	Diff line change
`@@ -1155,7 +1155,16 @@ exec_simple_query(const char *query_string)`
`1155`	`1155`	`plantree_list=pg_plan_queries(querytree_list,`
`1156`	`1156`	`CURSOR_OPT_PARALLEL_OK,NULL);`
`1157`	`1157`
`1158`		`-/* Done with the snapshot used for parsing/planning */`
	`1158`	`+/*`
	`1159`	`+ * Done with the snapshot used for parsing/planning.`
	`1160`	`+ *`
	`1161`	`+ * While it looks promising to reuse the same snapshot for query`
	`1162`	`+ * execution (at least for simple protocol), unfortunately it causes`
	`1163`	`+ * execution to use a snapshot that has been acquired before locking`
	`1164`	`+ * any of the tables mentioned in the query. This creates user-`
	`1165`	`+ * visible anomalies, so refrain. Refer to`
	`1166`	`+ * https://postgr.es/m/flat/5075D8DF.6050500@fuzzy.cz for details.`
	`1167`	`+ */`
`1159`	`1168`	`if (snapshot_set)`
`1160`	`1169`	`PopActiveSnapshot();`
`1161`	`1170`

Comments

(0)