On 15/07/11 07:14, Tom Lane wrote:
> Alvaro Herrera<alvherre@commandprompt.com> writes:
>> Excerpts from Gavin Flower's message of jue jul 14 07:45:00 -0400 2011:
>>> Using pg 9.1beta3, I was found that running a function generated an
>>> error relating to escaped single quotes, yet still produced the answer I
>>> expected!
>> The errors are not spurious. The function doesn't exist now because it
>> didn't get created. What happened is that now your inserts are going
>> into the parent table, not the partitions as your script intended.
> Possibly a more useful answer is "your function appears to be assuming
> that standard_conforming_strings is OFF. As of 9.1 it's ON by default".
>
> regards, tom lane
Thanks Tom (& Alvera),
I checked my postgresql.conf:
standard_conforming_strings = off
I had forgotten I had changed this, and had simply reussed it from earlier!
The silly thing is, that I had been treading up on partitioned tables
and had come across a comment about mistakes could lead to populating
the parent table!
If I could change the sunject of thread, I would change it to start with
'Misleading...'.
My lawyer siuggests defences of 'tiredness due ti it being late at
night' and 'poor eyesight'! :-)
(You can tell I've been reading groklaw.net way too much))
More seriously:
Could a hint be put in after the error message:
'psql:part000.sql:68: ERROR: function measurement_insert_trigger() does
not exist'
when the table has been partitioned, along the lines that 'this might
cause inserts to go into the parent table, is this intended?' or some
such? I am sure many others also get caught. This was purely an
exercise for me, but it would be more serious in production code (yes I
know things 'should' be tested properly first...).
Cheers,
Gavin
