assymetry updating a boolean (=FALSE faster than =TRUE)

On Tue, May 23, 2006 at 02:55:03PM -0700, George Pavlov wrote:

Here is something that seems anomalous to me: when I set a boolean field
to FALSE performance is much better than when I set it to TRUE. Any
reason for FALSE to be favored over TRUE?

Some details:

vacuum analyze my_table;
update my_table set is_foo=FALSE where some_id = 47;
--142 rows affected, 8047 ms execution time.
vacuum analyze my_table;
update my_table set is_foo=TRUE where some_id = 47;
--142 rows affected, 48609 ms execution time.

I have run these kinds of queries repeatedly and the timing above is
representative--the setting to FALSE case is about 6 times more
performant. The table my_table has about 105K rows and has many other
columns of various types. Thre is a trigger on the table, but it does
not do anything special based on this column's value. The some_id column
is indexed. This is on PG 8.1.3 on Linux.

Are there any indexes containing is_foo?
--
Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com
Pervasive Software http://pervasive.com work: 512-231-6117
vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461

"George Pavlov" <gpavlov@mynewplace.com> writes:

Here is something that seems anomalous to me: when I set a boolean field
to FALSE performance is much better than when I set it to TRUE. Any
reason for FALSE to be favored over TRUE?

It isn't. You're measuring something else than you think you're
measuring ... maybe a partial index? Varying effects of where the
free space is in the table? In a test case I see no significant
difference:

regression=# create table foo(f1 int, f2 bool);
CREATE TABLE
regression=# insert into foo select x, true from generate_series(1,100000) x;
INSERT 0 100000
regression=# \timing
Timing is on.
regression=# vacuum foo;
VACUUM
Time: 318.153 ms
regression=# update foo set f2 = true;
UPDATE 100000
Time: 2553.014 ms
regression=# vacuum foo;
VACUUM
Time: 410.356 ms
regression=# update foo set f2 = false;
UPDATE 100000
Time: 2453.620 ms
regression=# vacuum foo;
VACUUM
Time: 405.955 ms
regression=# update foo set f2 = true;
UPDATE 100000
Time: 2497.413 ms
regression=# vacuum foo;
VACUUM
Time: 408.403 ms
regression=# update foo set f2 = false;
UPDATE 100000
Time: 2458.824 ms
regression=#

regards, tom lane

no partial (or other) index affecting these results. just a trigger the
outcome of which was affected by the value of that particular boolean.
sorry, should have looked more carefully...