Thread: Conditional INSERT: if not exists

I want to insert a row unless it exists already.  Do I have to write a
stored procedure to do this?

Thanks,
Don

Don Morrison wrote:
> I want to insert a row unless it exists already.  Do I have to write a
> stored procedure to do this?

Hi, Don.

Just do the insert.  If it already exists, it simply isn't inserted.
You will, of course, receive an error that the row exists, but that
isn't a problem.  If you need to do this in a transaction, you can
simply set savepoints and roll back to the savepoint if an insert fails
(and you expected that it could).

Sean

One way would be to create a Primary Index containing all columns in the
table...?

For example - a table called "grps" with 3 colums:
    CREATE TABLE grps
    (
      code varchar(4) NOT NULL,
      description text NOT NULL,
      category varchar(3),
      CONSTRAINT groups_pkey PRIMARY KEY (code,description,category)
    )

An insert on this table will fail if the values for each column in the row
you are attempting to insert already matches and existing row - Example:

 code |              description               | category
------+----------------------------------------+----------
 0001 | GROUP NOT DEFINED                      | OT1
 0009 | NEWLY CREATED PRODUCTS                 | OT1
 0010 | GROUP NOT SET                          | OT1

Then attempting:
    INSERT INTO grps VALUES ('0001', 'GROUP NOT DEFINED', 'OT1');
This will fail as it already matches the first row in the table.

This WILL work (with a different value in the third column):
    INSERT INTO grps VALUES ('0001', 'GROUP NOT DEFINED', 'OT2');

Cheers,
-p


-----Original Message-----
From: pgsql-novice-owner@postgresql.org
[mailto:pgsql-novice-owner@postgresql.org] On Behalf Of Don Morrison
Sent: Tuesday, 22 August 2006 10:43
To: pgsql-novice@postgresql.org
Subject: [NOVICE] Conditional INSERT: if not exists

I want to insert a row unless it exists already.  Do I have to write a
stored procedure to do this?

Thanks,
Don

---------------------------(end of broadcast)---------------------------
TIP 4: Have you searched our list archives?

               http://archives.postgresql.org


*******************Confidentiality and Privilege Notice*******************

The material contained in this message is privileged and confidential to
the addressee.  If you are not the addressee indicated in this message or
responsible for delivery of the message to such person, you may not copy
or deliver this message to anyone, and you should destroy it and kindly
notify the sender by reply email.

Information in this message that does not relate to the official business
of Weatherbeeta must be treated as neither given nor endorsed by Weatherbeeta.
Weatherbeeta, its employees, contractors or associates shall not be liable
for direct, indirect or consequential loss arising from transmission of this
message or any attachments

> > I want to insert a row unless it exists already.  Do I have to write a
> > stored procedure to do this?
>[...]
> isn't a problem.  If you need to do this in a transaction, you can
> simply set savepoints and roll back to the savepoint if an insert fails
> (and you expected that it could).

Yes, that's the problem: I have it in one transaction which I don't
want to fail on error.  I'll look into the savepoint method.
Specifically the transaction I'm writing is a "Z SQL Method" in Zope
(a page templating system and web server).

> > > I want to insert a row unless it exists already.  Do I have to write a
> > > stored procedure to do this?
> >[...]
> > isn't a problem.  If you need to do this in a transaction, you can
> > simply set savepoints and roll back to the savepoint if an insert fails
> > (and you expected that it could).
> Yes, that's the problem: I have it in one transaction which I don't
> want to fail on error.  I'll look into the savepoint method.
> Specifically the transaction I'm writing is a "Z SQL Method" in Zope
> (a page templating system and web server).

I'll have to upgrade to PostgreSQL 8.X. :( 7.X doesn't have savepoint.

Hi,

why not simply put a where condition in you insert :

insert into table values (a,b)
where not exists (select a,b from table)



Don Morrison a écrit :
> I want to insert a row unless it exists already.  Do I have to write a
> stored procedure to do this?
>
> Thanks,
> Don
>
> ---------------------------(end of broadcast)---------------------------
> TIP 4: Have you searched our list archives?
>
>               http://archives.postgresql.org
>

> why not simply put a where condition in you insert :
>
> insert into table values (a,b)
> where not exists (select a,b from table)

The WHERE clause does not fit into the INSERT syntax in that way:

http://www.postgresql.org/docs/8.1/interactive/sql-insert.html

INSERT INTO table [ ( column [, ...] ) ]
    { DEFAULT VALUES | VALUES ( { expression | DEFAULT } [, ...] ) | query }

My problem: if the insert fails because the value already exists, then
this starts a rollback of my entire transaction.  The solution I'm
trying is to create a nested transaction with a savepoint right before
the insert, thus catching the rollback with the nested
transaction...I'm not sure the nested transaction is necessary...maybe
just the savepoint. Example:

...other outer transaction work here...

BEGIN;
SAVEPOINT insert_may_fail;
INSERT INTO table ...;
COMMIT;

...continue next outer transaction work here...

On Wed, Aug 23, 2006 at 12:48:53 -0700,
  Don Morrison <donmorrison@gmail.com> wrote:
>
> My problem: if the insert fails because the value already exists, then
> this starts a rollback of my entire transaction.  The solution I'm

If you know there won't be concurrent inserts or deletes affecting the
row of interest there is a way to do this in the INSERT statement.

> transaction...I'm not sure the nested transaction is necessary...maybe
> just the savepoint. Example:

Using save points creates nested transactions, so you have that overhead by
whatever name it is called.

> If you know there won't be concurrent inserts or deletes affecting the
> row of interest there is a way to do this in the INSERT statement.

I cannot rule out concurrent inserts.

> > transaction...I'm not sure the nested transaction is necessary...maybe
> > just the savepoint. Example:
> Using save points creates nested transactions, so you have that overhead by
> whatever name it is called.

Overhead yes, that was not the question.  The question was is the
extra nesting necessary?

On 8/23/06 5:05 PM, "Don Morrison" <donmorrison@gmail.com> wrote:

>> If you know there won't be concurrent inserts or deletes affecting the
>> row of interest there is a way to do this in the INSERT statement.
>
> I cannot rule out concurrent inserts.
>
>>> transaction...I'm not sure the nested transaction is necessary...maybe
>>> just the savepoint. Example:
>> Using save points creates nested transactions, so you have that overhead by
>> whatever name it is called.
>
> Overhead yes, that was not the question.  The question was is the
> extra nesting necessary?

The answer is yes, I think, if you need your inserts to be done within a
transaction (as opposed to serially).

Sean

> > Overhead yes, that was not the question.  The question was is the
> > extra nesting necessary?
>
> The answer is yes, I think, if you need your inserts to be done within a
> transaction (as opposed to serially).

Thanks for the input everybody. -Don

> isn't a problem.  If you need to do this in a transaction, you can
> simply set savepoints and roll back to the savepoint if an insert fails
> (and you expected that it could).

Unfortunately, for some reason it does a rollback of the entire outer
transaction too, not just the nested one. :(  Anybody know why?

If I use stored procedures will it be easier?

> > isn't a problem.  If you need to do this in a transaction, you can
> > simply set savepoints and roll back to the savepoint if an insert fails
> > (and you expected that it could).
>
> Unfortunately, for some reason it does a rollback of the entire outer
> transaction too, not just the nested one. :(  Anybody know why?
>
> If I use stored procedures will it be easier?
>

When I try to insert a duplicate row, this generates an IntegrityError
on my unique index, which rolls back my entire transaction, when I
really want it to generate a ROLLBACK TO SAVEPOINT for the most recent
savepoint, so only that insert is ignored.  Does anyone know a
configuration trick to get it to do this, or do I have to write a
stored procedure where I explicitly call ROLLBACK TO SAVEPOINT?

Don Morrison wrote:

> My problem: if the insert fails because the value already exists, then
> this starts a rollback of my entire transaction.  The solution I'm
> trying is to create a nested transaction with a savepoint right before
> the insert, thus catching the rollback with the nested
> transaction...I'm not sure the nested transaction is necessary...maybe
> just the savepoint. Example:

You need to:

SAVEPOINT insert_may_fail
try:
    INSERT INTO ...
    RELEASE SAVEPOINT
except:
    ROLLBACK TO SAVEPOINT
    UPDATE ...

You can catch the exception either in Python, or if you are using a ZSQL
method, using <dtml-try>.


--
Stuart Bishop <stuart@stuartbishop.net>
http://www.stuartbishop.net/