Bug in predicate indexes? - Mailing list pgsql-hackers

From Joshua D. Drake
Subject Bug in predicate indexes?
Date
Msg-id 437D5137.6020903@commandprompt.com
Whole thread Raw
Responses Re: Bug in predicate indexes?  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
Hello,

It appears that predicate indexes won't uses indexes on int8 columns 
unless they are casted:

test=# select count(*) from test_key ;count
--------120617
(1 row)


test=# alter table test_key ALTER column id set statistics 25;
ALTER TABLE
test=# analyze test_key ;
test=# create index foo on test_key (id) where id >= 50000;
CREATE INDEX
test=# analyze;
ANALYZE
test=# explain analyze select * from test_key where id > 50000;                                                  QUERY
PLAN
-----------------------------------------------------------------------------------------------------------------Seq
Scanon test_key  (cost=0.00..2276.71 rows=69768 width=16) (actual 
 
time=23.186..289.406 rows=70617 loops=1)  Filter: (id > 50000)Total runtime: 516.917 ms
(3 rows)

test=# create index foo on test_key (id) where id >= 50000::bigint;
CREATE INDEX
test=# analyze;
ANALYZE
test=# explain analyze select * from test_key where id > 50000::bigint;
      QUERY PLAN
 

----------------------------------------------------------------------------------------------------------------------------Index
Scanusing foo on test_key  (cost=0.00..1602.88 rows=70806 
 
width=16) (actual time=0.080..264.001 rows=70617 loops=1)  Index Cond: (id > 50000::bigint)Total runtime: 496.160 ms
(3 rows)

test=# explain analyze select * from test_key where id > 50000;
QUERYPLAN
 

----------------------------------------------------------------------------------------------------------------------------Index
Scanusing foo on test_key  (cost=0.00..1602.88 rows=70806 
 
width=16) (actual time=0.116..262.952 rows=70617 loops=1)  Index Cond: (id > 50000)Total runtime: 488.943 ms
(3 rows)



pgsql-hackers by date:

Previous
From: Greg Stark
Date:
Subject: Re: Improving count(*)
Next
From: Christopher Kings-Lynne
Date:
Subject: Re: Loading 7.4 dump to 8.1 with user-custom search_path