Re: Slow queries salad ;)

On Tue, Apr 25, 2006 at 07:53:15PM +0200, PFC wrote:
What version is this??

> annonces=> EXPLAIN ANALYZE SELECT * FROM test.current WHERE id IN (SELECT
> annonce_id FROM bookmarks WHERE list_id IN ('4'));
> QUERY PLAN
> -------------------------------------------------------------------------------------------------------------------------
> Hash Join (cost=6.58..532.84 rows=140 width=203) (actual
> time=0.747..5.283 rows=150 loops=1)
> Hash Cond: ("outer".id = "inner".annonce_id)
> -> Seq Scan on current (cost=0.00..467.24 rows=11524 width=203)
> (actual time=0.006..3.191 rows=11524 loops=1)
> -> Hash (cost=6.23..6.23 rows=140 width=4) (actual time=0.244..0.244
> rows=150 loops=1)
> -> HashAggregate (cost=4.83..6.23 rows=140 width=4) (actual
> time=0.155..0.184 rows=150 loops=1)
> -> Seq Scan on bookmarks (cost=0.00..4.45 rows=150
> width=4) (actual time=0.008..0.097 rows=150 loops=1)
> Filter: (list_id = 4)
> Total runtime: 5.343 ms
> (8 lignes)
>
> annonces=> set enable_hashjoin TO 0;
> SET
> annonces=> EXPLAIN ANALYZE SELECT * FROM test.current WHERE id IN (SELECT
> annonce_id FROM bookmarks WHERE list_id IN ('4'));
> QUERY PLAN
> --------------------------------------------------------------------------------------------------------------------------------
> Nested Loop (cost=4.83..824.22 rows=140 width=203) (actual
> time=0.219..1.034 rows=150 loops=1)
> -> HashAggregate (cost=4.83..6.23 rows=140 width=4) (actual
> time=0.158..0.199 rows=150 loops=1)
> -> Seq Scan on bookmarks (cost=0.00..4.45 rows=150 width=4)
> (actual time=0.011..0.096 rows=150 loops=1)
> Filter: (list_id = 4)
> -> Index Scan using current_pkey on current (cost=0.00..5.83 rows=1
> width=203) (actual time=0.005..0.005 rows=1 loops=150)
> Index Cond: (current.id = "outer".annonce_id)
> Total runtime: 1.108 ms
> (7 lignes)
>
> Hm, the row estimates on the "bookmarks" table are spot on ; why did
> it choose the hash join ?

Because it thought it would be cheaper; see the estimates. Increasing
effective_cache_size or decreasing random_page_cost would favor the
index scan.

> Now, if I want to access the "all" view which contains the union of
> the "current" and "archive" table :
>
> annonces=> set enable_hashjoin TO 1;
> SET
> annonces=> EXPLAIN ANALYZE SELECT * FROM test.all WHERE id IN (SELECT
> annonce_id FROM bookmarks WHERE list_id IN ('4'));
> QUERY PLAN
> --------------------------------------------------------------------------------------------------------------------------------
> Hash Join (cost=6.58..33484.41 rows=314397 width=36) (actual
> time=8300.484..8311.784 rows=150 loops=1)
> Hash Cond: ("outer"."?column1?" = "inner".annonce_id)
> -> Append (cost=0.00..23596.78 rows=449139 width=219) (actual
> time=6.390..8230.821 rows=448875 loops=1)
> -> Seq Scan on archive (cost=0.00..18638.15 rows=437615
> width=219) (actual time=6.389..8175.491 rows=437351 loops=1)
> -> Seq Scan on current (cost=0.00..467.24 rows=11524 width=203)
> (actual time=0.022..8.985 rows=11524 loops=1)
> -> Hash (cost=6.23..6.23 rows=140 width=4) (actual time=0.255..0.255
> rows=150 loops=1)
> -> HashAggregate (cost=4.83..6.23 rows=140 width=4) (actual
> time=0.168..0.197 rows=150 loops=1)
> -> Seq Scan on bookmarks (cost=0.00..4.45 rows=150
> width=4) (actual time=0.015..0.102 rows=150 loops=1)
> Filter: (list_id = 4)
> Total runtime: 8311.870 ms
> (10 lignes)
>
> annonces=> set enable_hashjoin TO 0;
> SET
> annonces=> EXPLAIN ANALYZE SELECT * FROM test.all WHERE id IN (SELECT
> annonce_id FROM bookmarks WHERE list_id IN ('4'));
> QUERY PLAN
> -------------------------------------------------------------------------------------------------------------------------------------
> Merge Join (cost=79604.61..84994.98 rows=314397 width=36) (actual
> time=6944.229..7109.371 rows=150 loops=1)
> Merge Cond: ("outer".annonce_id = "inner".id)
> -> Sort (cost=11.22..11.57 rows=140 width=4) (actual
> time=0.326..0.355 rows=150 loops=1)
> Sort Key: bookmarks.annonce_id
> -> HashAggregate (cost=4.83..6.23 rows=140 width=4) (actual
> time=0.187..0.218 rows=150 loops=1)
> -> Seq Scan on bookmarks (cost=0.00..4.45 rows=150
> width=4) (actual time=0.028..0.126 rows=150 loops=1)
> Filter: (list_id = 4)
> -> Sort (cost=79593.40..80716.25 rows=449139 width=36) (actual
> time=6789.786..7014.815 rows=448625 loops=1)
> Sort Key: "all".id
> -> Append (cost=0.00..23596.78 rows=449139 width=219) (actual
> time=0.013..391.447 rows=448875 loops=1)
> -> Seq Scan on archive (cost=0.00..18638.15 rows=437615
> width=219) (actual time=0.013..332.353 rows=437351 loops=1)
> -> Seq Scan on current (cost=0.00..467.24 rows=11524
> width=203) (actual time=0.013..8.396 rows=11524 loops=1)
> Total runtime: 37226.846 ms
>
> The IN() is quite small (150 values), but the two large tables are
> seq-scanned... is there a way to avoid this ?

Possibly if you break the view apart...

SELECT ... FROM current WHERE id IN (...)
UNION ALL
SELECT ... FROM archive WHERE id IN (...)

> -------------------------------------------------------------------------------------------------------------------------------------
>
> Another nitpick : let's redo the first query differently.
>
> annonces=> EXPLAIN ANALYZE SELECT * FROM test.current WHERE id IN (SELECT
> annonce_id FROM bookmarks WHERE list_id IN ('4'));
> QUERY PLAN
> -------------------------------------------------------------------------------------------------------------------------
> Hash Join (cost=6.58..532.84 rows=140 width=203) (actual
> time=0.794..5.791 rows=150 loops=1)
> Hash Cond: ("outer".id = "inner".annonce_id)
> -> Seq Scan on current (cost=0.00..467.24 rows=11524 width=203)
> (actual time=0.003..3.554 rows=11524 loops=1)
> -> Hash (cost=6.23..6.23 rows=140 width=4) (actual time=0.265..0.265
> rows=150 loops=1)
> -> HashAggregate (cost=4.83..6.23 rows=140 width=4) (actual
> time=0.179..0.210 rows=150 loops=1)
> -> Seq Scan on bookmarks (cost=0.00..4.45 rows=150
> width=4) (actual time=0.021..0.102 rows=150 loops=1)
> Filter: (list_id = 4)
> Total runtime: 5.853 ms
>
> annonces=> EXPLAIN ANALYZE SELECT a.* FROM test.current a, (SELECT
> DISTINCT annonce_id FROM bookmarks WHERE list_id IN ('4')) AS b WHERE
> a.id=b.annonce_id;
> QUERY PLAN
> -------------------------------------------------------------------------------------------------------------------------------
> Hash Join (cost=12.37..538.63 rows=140 width=203) (actual
> time=0.812..5.362 rows=150 loops=1)
> Hash Cond: ("outer".id = "inner".annonce_id)
> -> Seq Scan on current a (cost=0.00..467.24 rows=11524 width=203)
> (actual time=0.005..3.227 rows=11524 loops=1)
> -> Hash (cost=12.02..12.02 rows=140 width=4) (actual
> time=0.296..0.296 rows=150 loops=1)
> -> Unique (cost=9.87..10.62 rows=140 width=4) (actual
> time=0.215..0.265 rows=150 loops=1)
> -> Sort (cost=9.87..10.25 rows=150 width=4) (actual
> time=0.215..0.226 rows=150 loops=1)
> Sort Key: bookmarks.annonce_id
> -> Seq Scan on bookmarks (cost=0.00..4.45 rows=150
> width=4) (actual time=0.007..0.104 rows=150 loops=1)
> Filter: (list_id = 4)
> Total runtime: 5.429 ms
>
> Hm, it does Sort + Unique + Hash ; the Hash alone would have been
> better.

The hash alone wouldn't have gotten you the same output as a DISTINCT,
though.

> Replacing DISTINCT with GROUP BY removes the sort.
>
> annonces=> EXPLAIN ANALYZE SELECT a.* FROM test.current a, (SELECT
> annonce_id FROM bookmarks WHERE list_id IN ('4') GROUP BY annonce_id) AS b
> WHERE a.id=b.annonce_id;
> QUERY PLAN
> -------------------------------------------------------------------------------------------------------------------------
> Hash Join (cost=7.98..534.24 rows=140 width=203) (actual
> time=0.811..5.557 rows=150 loops=1)
> Hash Cond: ("outer".id = "inner".annonce_id)
> -> Seq Scan on current a (cost=0.00..467.24 rows=11524 width=203)
> (actual time=0.006..3.434 rows=11524 loops=1)
> -> Hash (cost=7.63..7.63 rows=140 width=4) (actual time=0.242..0.242
> rows=150 loops=1)
> -> HashAggregate (cost=4.83..6.23 rows=140 width=4) (actual
> time=0.156..0.186 rows=150 loops=1)
> -> Seq Scan on bookmarks (cost=0.00..4.45 rows=150
> width=4) (actual time=0.008..0.097 rows=150 loops=1)
> Filter: (list_id = 4)
> Total runtime: 5.647 ms

Uhm.. that actually runs slower...
--
Jim C. Nasby, Sr. Engineering Consultant jnasby(at)pervasive(dot)com
Pervasive Software http://pervasive.com work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461