1berry/postgres
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

De-dupicate Memoize cache keys

Browse Source 
It was possible when determining the cache keys for a Memoize path that
if the same expr appeared twice in the parameterized path's ppi_clauses
and/or in the Nested Loop's inner relation's lateral_vars.  If this
happened the Memoize node's cache keys would contain duplicates.  This
isn't a problem for correctness, all it means is that the cache lookups
will be suboptimal due to having redundant work to do on every hash table
lookup and insert.

Here we adjust paraminfo_get_equal_hashops() to look for duplicates and
ignore them when we find them.

Author: David Rowley
Reviewed-by: Richard Guo
Discussion: https://postgr.es/m/422277.1706207562%40sss.pgh.pa.us
This commit is contained in:
David Rowley 2024-01-26 20:51:36 +13:00
parent bd5760df38
commit bc397e5cdb
2 changed files with 21 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
src/backend/optimizer/path/joinpath.c
View File

@ -492,8 +492,16 @@ paraminfo_get_equal_hashops(PlannerInfo *root, ParamPathInfo *param_info,
return false; return false;
} }
*operators = lappend_oid(*operators, hasheqoperator); /*
*param_exprs = lappend(*param_exprs, expr); * 'expr' may already exist as a parameter from a previous item in
* ppi_clauses. No need to include it again, however we'd better
* ensure we do switch into binary mode if required. See below.
*/
if (!list_member(*param_exprs, expr))
{
*operators = lappend_oid(*operators, hasheqoperator);
*param_exprs = lappend(*param_exprs, expr);
}
/* /*
* When the join operator is not hashable then it's possible that * When the join operator is not hashable then it's possible that
@ -536,8 +544,16 @@ paraminfo_get_equal_hashops(PlannerInfo *root, ParamPathInfo *param_info,
return false; return false;
} }
*operators = lappend_oid(*operators, typentry->eq_opr); /*
*param_exprs = lappend(*param_exprs, expr); * 'expr' may already exist as a parameter from the ppi_clauses. No
* need to include it again, however we'd better ensure we do switch
* into binary mode.
*/
if (!list_member(*param_exprs, expr))
{
*operators = lappend_oid(*operators, typentry->eq_opr);
*param_exprs = lappend(*param_exprs, expr);
}
/* /*
* We must go into binary mode as we don't have too much of an idea of * We must go into binary mode as we don't have too much of an idea of

2
src/test/regress/expected/memoize.out
View File

@ -107,7 +107,7 @@ WHERE t1.unique1 < 10;', false);
-> Index Scan using tenk1_unique1 on tenk1 t1 (actual rows=10 loops=N) -> Index Scan using tenk1_unique1 on tenk1 t1 (actual rows=10 loops=N)
Index Cond: (unique1 < 10) Index Cond: (unique1 < 10)
-> Memoize (actual rows=2 loops=N) -> Memoize (actual rows=2 loops=N)
Cache Key: t1.two, t1.two Cache Key: t1.two
Cache Mode: binary Cache Mode: binary
Hits: 8 Misses: 2 Evictions: Zero Overflows: 0 Memory Usage: NkB Hits: 8 Misses: 2 Evictions: Zero Overflows: 0 Memory Usage: NkB
-> Subquery Scan on t2 (actual rows=2 loops=N) -> Subquery Scan on t2 (actual rows=2 loops=N)