4 messages in com.mysql.lists.bugsRe: Queries on doubles are not stable| From | Sent On | Attachments |
|---|---|---|
| Alan G. Carter | 30 Nov 2000 14:11 | |
| Sasha Pachev | 01 Dec 2000 16:49 | |
| Thimble Smith | 02 Dec 2000 09:59 | |
| Michael Widenius | 03 Dec 2000 02:35 |
| Subject: | Re: Queries on doubles are not stable |
|---|---|
| From: | Michael Widenius (mon...@mysql.com) |
| Date: | 12/03/2000 02:35:19 AM |
| List: | com.mysql.lists.bugs |
Hi!
"Thimble" == Thimble Smith <ti...@mysql.com> writes:
Thimble> On Fri, Dec 01, 2000 at 05:49:45PM -0700, Sasha Pachev wrote:
On Thursday 30 November 2000 15:11, Alan G. Carter wrote:
/usr/local/mysql/bin/mysql -v -u root --password=$1 << EOF use mysql create table test_doubles (datum double);
insert into test_doubles values (0.5); insert into test_doubles values (1.0); insert into test_doubles values (1.5); insert into test_doubles values (2.0); insert into test_doubles values (2.5);
select * from test_doubles; select * from test_doubles where datum < 1.5; select * from test_doubles where datum > 1.5; select * from test_doubles where datum = 1.5; # # What happens with the bug is: # # 1) 5 rows # 2) 5 rows # 3) 0 rows # 4) 1 rows
The problem with double comparisons is that there is always a margin of error.
Thimble> This would explain the problem if it were row 4 which were Thimble> "incorrect". But it doesn't explain rows 2 and 3 being Thimble> wrong.
Thimble> Alan, I'm not aware of any changes between 3.23.25 and the Thimble> current code that pertain to this - nor have I heard of this Thimble> problem in 3.23.25 from anyone else. So, I'm not sure what Thimble> is making the difference....
Alan, did you compiled MySQL 3.23.25 with a different compiler or on a different machine? I am not aware of any problems with floats that could explain 2) in any MySQL version.
Regards, Monty