Commit9caf042

committed

Reorder tests in get_cheapest_path_for_pathkeys().

Checking parallel safety should be even cheaper than cost comparison, sodo that first.Also make some minor, related comment improvements.Richard Guo, reviewed by Aleksander Alekseev, Andy Fan, and me.Discussion:http://postgr.es/m/CAMbWs4-KE2wf4QPj_Sr5mX4QFtBNNKGmxK=+e=KZEGUjdG33=g@mail.gmail.com

1 parentac22a95 commit9caf042Copy full SHA for 9caf042

File tree

-4

lines changed

-4

lines changed

Lines changed: 6 additions & 4 deletions

Original file line number	Diff line number	Diff line change
`@@ -407,7 +407,8 @@ pathkeys_count_contained_in(List keys1, List keys2, int *n_common)`
`407`	`407`	`/*`
`408`	`408`	`* get_cheapest_path_for_pathkeys`
`409`	`409`	`* Find the cheapest path (according to the specified criterion) that`
`410`		`- * satisfies the given pathkeys and parameterization.`
	`410`	`+ * satisfies the given pathkeys and parameterization, and is parallel-safe`
	`411`	`+ * if required.`
`411`	`412`	`* Return NULL if no such path.`
`412`	`413`	`*`
`413`	`414`	`* 'paths' is a list of possible paths that all generate the same relation`
`@@ -429,6 +430,10 @@ get_cheapest_path_for_pathkeys(List paths, List pathkeys,`
`429`	`430`	`{`
`430`	`431`	`Pathpath= (Path)lfirst(l);`
`431`	`432`
	`433`	`+/* If required, reject paths that are not parallel-safe */`
	`434`	`+if (require_parallel_safe&& !path->parallel_safe)`
	`435`	`+continue;`
	`436`	`+`
`432`	`437`	`/*`
`433`	`438`	`* Since cost comparison is a lot cheaper than pathkey comparison, do`
`434`	`439`	`* that first. (XXX is that still true?)`
`@@ -437,9 +442,6 @@ get_cheapest_path_for_pathkeys(List paths, List pathkeys,`
`437`	`442`	`compare_path_costs(matched_path,path,cost_criterion) <=0)`
`438`	`443`	`continue;`
`439`	`444`
`440`		`-if (require_parallel_safe&& !path->parallel_safe)`
`441`		`-continue;`
`442`		`-`
`443`	`445`	`if (pathkeys_contained_in(pathkeys,path->pathkeys)&&`
`444`	`446`	`bms_is_subset(PATH_REQ_OUTER(path),required_outer))`
`445`	`447`	`matched_path=path;`

Comments

(0)