Re: unexpected join results - Mailing list pgsql-general

From Dann Corbit
Subject Re: unexpected join results
Date
Msg-id D90A5A6C612A39408103E6ECDD77B829408A97@voyager.corporate.connx.com
Whole thread Raw
In response to unexpected join results  (Matthew Phillips <mphillips@timing.com>)
List pgsql-general
-----Original Message-----
From: Matthew Phillips [mailto:mphillips@timing.com]
Sent: Tuesday, March 18, 2003 10:28 AM
To: pgsql-general@postgresql.org
Subject: [GENERAL] unexpected join results


Here is my problem in the most abstract way I can convey it...

I am doing a simple join on two tables. The query:

select m.source_id, second, begin_time, end_time
from measurements m,
     measurement_list l
where m.source_id = l.source_id
and m.second > l.begin_time
and l.end_time IS NOT NULL;

produces the output:

 source_id |       second        |     begin_time      |      end_time

-----------+---------------------+---------------------+----------------
-----
         2 | 2003-03-18 10:09:45 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:46 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:47 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:48 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:49 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:50 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:51 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:52 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:53 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:54 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:55 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:56 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:57 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:58 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:09:59 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:10:00 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:10:01 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:10:02 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:10:03 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:10:04 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:10:05 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:10:06 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
         2 | 2003-03-18 10:10:07 | 2003-02-18 10:09:06 | 2003-02-18
10:10:03
(23 rows)
source_id is an integer identifier, the rest of the columns are
timestamp without time zone.


This output is expected and perfectly ok... the problem is when I add
one more condition to the query. 'and m.second <= l.end_time'. This
would match rows in which second falls in between begin_time and end
time, something all the above rows do.

but...
select m.source_id, second, begin_time, end_time
from measurements m,
     measurement_list l
where m.source_id = l.source_id
and m.second > l.begin_time
and l.end_time IS NOT NULL
and m.second <= l.end_time;

produces:
 source_id | second | begin_time | end_time
-----------+--------+------------+----------
(0 rows)

But I can look back on the previous query results and see that in every
row the second is less than the end time! I am not exactly an sql guru,
but it seems to me that the output should be identical with the extra
line in or out. Any clues as to what is going on here? Please give me a
quick fix  :-) I have been spinning wheels for half a day on this one. I
can post more schema info if that is needed.

using 7.3.1 on FreeBSD 4.7
-------------------------------------------------------->>
Consider:
 second=2003-03-18 10:09:45  end_time=2003-02-18 10:10:03
Notice that the month is March for second, and February
For end_time.  This is true for all the rows.  The query
Result set is correct.
Any time stamp in march of 2003 will be bigger than one in
February of 2003.
<<--------------------------------------------------------

pgsql-general by date:

Previous
From: Matthew Phillips
Date:
Subject: unexpected join results
Next
From: Matthew Phillips
Date:
Subject: Re: unexpected join results