Problem with aborting entire transactions on error

From: Zbigniew <zbigniew2011(at)gmail(dot)com>
To: pgsql-general(at)postgresql(dot)org
Subject: Problem with aborting entire transactions on error
Date: 2012-12-10 03:20:35
Message-ID: CALT7RM95O2y9eUCoBPYF3x8zybmLUrwmMp8z4BCLhL2SjxL1aw@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

Hello,

As I read while googling the web, many people complained about this
before. Couldn't it be made optional (can be even with "default ON")?
I understand, that there are situations, when it is a must - for
example, when the rest of queries rely on the result of first ones -
but there are numerous situations, when just skipping a faulty query
is all we need.

A simple - but very common - example: I wanted to perform really large
number of inserts - using transaction, to make it faster - while being
sure the duplicate entries will be skipped. Of course, this job will
be done best by server itself, which is keeping an eye on "primary
key" of the table. Unfortunately: not during a transaction! Any "dupe"
will trash thousands other (proper) entries immediately.

Why is this? My guess is, there is kind of logic in the code, like this:

if { no error during query } {
do it
} else {
withdraw this one
rollback entire transaction
}

Therefore my request - and, as I saw, of many others - would be just
to introduce a little change:

if { no error during query } {
do it
} else {
withdraw this one
if { ROLLBACK_ON_ERROR } {
rollback entire transaction
}
}

(if there's no ROLLBACK_ON_ERROR - it should carry on with the
remaining queries)

Is it really so problematic to introduce such code change, allowing
the users to control this behaviour? Yes, I read about using
"savepoints" - but I think we agree, it's just cumbersome workaround -
and not real solution, like my proposal. All we need is either a
variable to set, or a command, that will allow to modify the present
functionality in the way described above.
--
regards,
Zbigniew

Responses

Browse pgsql-general by date

  From Date Subject
Next Message David Johnston 2012-12-10 04:04:05 Re: Problem with aborting entire transactions on error
Previous Message Jeff Janes 2012-12-10 00:16:31 Re: query performance, though it was timestamps,maybe just table size?