Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commita90c950

Browse files
committed
Prevent overly large and NaN row estimates in relations
Given a query with enough joins, it was possible that the query planner,after multiplying the row estimates with the join selectivity that theestimated number of rows would exceed the limits of the double data typeand become infinite.To give an indication on how extreme a case is required to hit this, theparticular example case reported required 379 joins to a table without anystatistics, which resulted in the 1.0/DEFAULT_NUM_DISTINCT being used forthe join selectivity. This eventually caused the row estimates to goinfinite and resulted in an assert failure in initial_cost_mergejoin()where the infinite row estimated was multiplied by an outerstartsel of 0.0resulting in NaN. The failing assert verified that NaN <= Inf, which isfalse.To get around this we use clamp_row_est() to cap row estimates at amaximum of 1e100. This value is thought to be low enough that costsderived from it would remain within the bounds of what the double type canrepresent.Aside from fixing the failing Assert, this also has the added benefit ofmaking it so add_path() will still receive proper numerical values ascosts which will allow it to make more sane choices when determining thecheaper path in extreme cases such as the one described above.Additionally, we also get rid of the isnan() checks in the join costingfunctions. The actual case which originally triggered those checks to beadded in the first place never made it to the mailing lists. It seemslikely that the new code being added to clamp_row_est() will result inthose becoming checks redundant, so just remove them.The fairly harmless assert failure problem does also exist in thebackbranches, however, a more minimalistic fix will be applied there.Reported-by: Onder KalaciReviewed-by: Tom LaneDiscussion:https://postgr.es/m/DM6PR21MB1211FF360183BCA901B27F04D80B0@DM6PR21MB1211.namprd21.prod.outlook.com
1 parentd5a9a66 commita90c950

File tree

1 file changed

+22
-13
lines changed

1 file changed

+22
-13
lines changed

‎src/backend/optimizer/path/costsize.c

Lines changed: 22 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -107,6 +107,13 @@
107107
*/
108108
#defineAPPEND_CPU_COST_MULTIPLIER 0.5
109109

110+
/*
111+
* Maximum value for row estimates. We cap row estimates to this to help
112+
* ensure that costs based on these estimates remain within the range of what
113+
* double can represent. add_path() wouldn't act sanely given infinite or NaN
114+
* cost values.
115+
*/
116+
#defineMAXIMUM_ROWCOUNT 1e100
110117

111118
doubleseq_page_cost=DEFAULT_SEQ_PAGE_COST;
112119
doublerandom_page_cost=DEFAULT_RANDOM_PAGE_COST;
@@ -189,11 +196,14 @@ double
189196
clamp_row_est(doublenrows)
190197
{
191198
/*
192-
* Force estimate to be at least one row, to make explain output look
193-
* better and to avoid possible divide-by-zero when interpolating costs.
194-
* Make it an integer, too.
199+
* Avoid infinite and NaN row estimates. Costs derived from such values
200+
* are going to be useless. Also force the estimate to be at least one
201+
* row, to make explain output look better and to avoid possible
202+
* divide-by-zero when interpolating costs. Make it an integer, too.
195203
*/
196-
if (nrows <=1.0)
204+
if (nrows>MAXIMUM_ROWCOUNT||isnan(nrows))
205+
nrows=MAXIMUM_ROWCOUNT;
206+
elseif (nrows <=1.0)
197207
nrows=1.0;
198208
else
199209
nrows=rint(nrows);
@@ -2737,12 +2747,11 @@ final_cost_nestloop(PlannerInfo *root, NestPath *path,
27372747
QualCostrestrict_qual_cost;
27382748
doublentuples;
27392749

2740-
/* Protect some assumptions below that rowcounts aren't zeroor NaN*/
2741-
if (outer_path_rows <=0||isnan(outer_path_rows))
2750+
/* Protect some assumptions below that rowcounts aren't zero */
2751+
if (outer_path_rows <=0)
27422752
outer_path_rows=1;
2743-
if (inner_path_rows <=0||isnan(inner_path_rows))
2753+
if (inner_path_rows <=0)
27442754
inner_path_rows=1;
2745-
27462755
/* Mark the path with the correct row estimate */
27472756
if (path->path.param_info)
27482757
path->path.rows=path->path.param_info->ppi_rows;
@@ -2952,10 +2961,10 @@ initial_cost_mergejoin(PlannerInfo *root, JoinCostWorkspace *workspace,
29522961
innerendsel;
29532962
Pathsort_path;/* dummy for result of cost_sort */
29542963

2955-
/* Protect some assumptions below that rowcounts aren't zeroor NaN*/
2956-
if (outer_path_rows <=0||isnan(outer_path_rows))
2964+
/* Protect some assumptions below that rowcounts aren't zero */
2965+
if (outer_path_rows <=0)
29572966
outer_path_rows=1;
2958-
if (inner_path_rows <=0||isnan(inner_path_rows))
2967+
if (inner_path_rows <=0)
29592968
inner_path_rows=1;
29602969

29612970
/*
@@ -3185,8 +3194,8 @@ final_cost_mergejoin(PlannerInfo *root, MergePath *path,
31853194
rescannedtuples;
31863195
doublerescanratio;
31873196

3188-
/* Protect some assumptions below that rowcounts aren't zeroor NaN*/
3189-
if (inner_path_rows <=0||isnan(inner_path_rows))
3197+
/* Protect some assumptions below that rowcounts aren't zero */
3198+
if (inner_path_rows <=0)
31903199
inner_path_rows=1;
31913200

31923201
/* Mark the path with the correct row estimate */

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp