discard on constraint violation

Lists: pgsql-general
From: Sim Zacks <sim(at)compulab(dot)co(dot)il>
To: PostgreSQL general <pgsql-general(at)postgresql(dot)org>
Subject: discard on constraint violation
Date: 2011-06-27 12:36:48
Message-ID: 4E087960.2020506@compulab.co.il
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-general

Is there a way to tell a table to discard an insert if it violates a
constraint (unique in my case) instead of giving an error? I don't want
the overhead of a trigger on each row inserted.

My situation is that I'm reading data from an external website with lots
of duplicate data. I am reading the data in a plpythonu function. I
never update, only insert. A trigger on the table that has to check if
each row exists before trying the insert is too much overhead.

I tried in plpython:

create or replace function testme() returns bool as
$$
try:
plpy.execute("insert into
reports.survey_types(typeid,name)values(1,'test')")

return true

except:

plpy.notice('the constraint was violated')
return false
$$language 'plpythonu';

where the insert violates a unique constraint and it never gets to the
except. The function dies and doesn't get to my except with:

WARNING: plpython: in function testme:
DETAIL: <class 'plpy.SPIError'>: Unknown error in PLy_spi_execute_query

ERROR: duplicate key violates unique constraint "survey_types_pkey"
CONTEXT: SQL statement "insert into
reports.survey_types(typeid,name)values(1,'test')"


From: Guillaume Lelarge <guillaume(at)lelarge(dot)info>
To: Sim Zacks <sim(at)compulab(dot)co(dot)il>
Cc: PostgreSQL general <pgsql-general(at)postgresql(dot)org>
Subject: Re: discard on constraint violation
Date: 2011-06-27 12:57:23
Message-ID: 1309179443.14751.15.camel@laptop
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-general

On Mon, 2011-06-27 at 15:36 +0300, Sim Zacks wrote:
> Is there a way to tell a table to discard an insert if it violates a
> constraint (unique in my case) instead of giving an error? I don't want
> the overhead of a trigger on each row inserted.
>

Without a trigger, you can't.

--
Guillaume
http://blog.guillaume.lelarge.info
http://www.dalibo.com


From: "David Johnston" <polobo(at)yahoo(dot)com>
To: "'Sim Zacks'" <sim(at)compulab(dot)co(dot)il>, "'PostgreSQL general'" <pgsql-general(at)postgresql(dot)org>
Subject: Re: discard on constraint violation
Date: 2011-06-27 14:07:36
Message-ID: 005b01cc34d3$91c36fb0$b54a4f10$@yahoo.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Lists: pgsql-general

>>My situation is that I'm reading data from an external website with lots
of duplicate data. I am reading the data in a plpythonu function. I never
update, only insert. A trigger on the table that has to check if each row
exists before trying the insert is too much overhead.

Create a staging table into which you import your external data and then
write a query of the form "INSERT INTO realtable SELECT FROM stagingtable
WHERE stagingtable NOT EXISTS realtable".

David J.