easy way to insert same value into 2 columns

For your specific question I don't know the answer.

For this particular case you could use:

create table your_table (
id integer default nextval ('public.zeit_id_seq'::text),
pos_id integer default currval ('public.zeit_id_seq'::text),
...
);

That would work fine as long as you use inserts which don't specify id
if pos_id is not specified (otherwise the currval will throw you an
error cause it cannot be called without nextval being called).

HTH,
Csaba.

Csaba Nagy wrote:

For your specific question I don't know the answer.

For this particular case you could use:

create table your_table (
id integer default nextval ('public.zeit_id_seq'::text),
pos_id integer default currval ('public.zeit_id_seq'::text),
...
);

That would work fine as long as you use inserts which don't specify id
if pos_id is not specified (otherwise the currval will throw you an
error cause it cannot be called without nextval being called).

HTH,
Csaba.

Thanks a lot, but is it save to use?

Do i always get the same value, even if an other insert is changing
the sequence >public.zeit_id_seq< just at the same time?

Regards

Sebastian

Sebastian Boeck wrote:

Thanks a lot, but is it save to use?

Do i always get the same value, even if an other insert is changing the
sequence >public.zeit_id_seq< just at the same time?

You can write a before insert trigger where you retrieve sequence value once and
set for two fields. That way it will ensure that both the fields would get same
value.

HTH

Shridhar

On Thursday 16 October 2003 11:11, Sebastian Boeck wrote:

Csaba Nagy wrote:

For your specific question I don't know the answer.

For this particular case you could use:

create table your_table (
id integer default nextval ('public.zeit_id_seq'::text),
pos_id integer default currval ('public.zeit_id_seq'::text),
...
);

That would work fine as long as you use inserts which don't specify id
if pos_id is not specified (otherwise the currval will throw you an
error cause it cannot be called without nextval being called).

HTH,
Csaba.

Thanks a lot, but is it save to use?

Not really.

Do i always get the same value, even if an other insert is changing
the sequence >public.zeit_id_seq< just at the same time?

Other processes can't interfere - the whole point of sequences is that they
are safe for this sort of thing.

Where you will have problems is that if one of the developers decides it's
more efficient to process fields backwards (zzz...pos_id, id) rather than
forwards (id, pos_id, ...zzz) then it will break.

Use a trigger here. If nothing else so you can stop people like me doing:

INSERT INTO your_table (id,pos_id) VALUES (-1,DEFAULT);

--
Richard Huxton
Archonet Ltd

Thanks a lot, but is it save to use?

Do i always get the same value, even if an other insert is changing
the sequence >public.zeit_id_seq< just at the same time?

Yes. currval() gives you the last id of the sequence in your session.
So if others log in and insert more rows it wont affect your session.

Can I ask why you would wanna do that anyway?

Jacob

On Thu, 2003-10-16 at 12:11, Sebastian Boeck wrote:

Csaba Nagy wrote:

For your specific question I don't know the answer.

For this particular case you could use:

create table your_table (
id integer default nextval ('public.zeit_id_seq'::text),
pos_id integer default currval ('public.zeit_id_seq'::text),
...
);

That would work fine as long as you use inserts which don't specify id
if pos_id is not specified (otherwise the currval will throw you an
error cause it cannot be called without nextval being called).

HTH,
Csaba.

Thanks a lot, but is it save to use?

Do i always get the same value, even if an other insert is changing
the sequence >public.zeit_id_seq< just at the same time?

Regards

Sebastian

---------------------------(end of broadcast)---------------------------
TIP 7: don't forget to increase your free space map settings

--
Venlig hilsen / Best regards,
Jacob Vennervald
System Developer
Proventum Solutions ApS
Tuborg Boulevard 12
2900 Hellerup
Denmark
Phone: +45 36 94 41 66
Mobile: +45 61 68 58 51

Jacob Vennervald wrote:

Thanks a lot, but is it save to use?

Do i always get the same value, even if an other insert is changing
the sequence >public.zeit_id_seq< just at the same time?

Yes. currval() gives you the last id of the sequence in your session.
So if others log in and insert more rows it wont affect your session.

Can I ask why you would wanna do that anyway?

Jacob

Sure!

I have several tables with an inheritance hierachy.
The first one (zeit) has a Serialfield called id.
All other tables inherit from this one, so that i have
an absolutely unique id over all tables. But if i want
to reference the id of a table in the middle of the
hierarchy-chain, it won't work, because it's not the
tables own index. Therefore i need an duplicated value.
On this column i can create another index that can be
used as an foreign key by other tables.

Are there any other ways to accomplish this?

Regards

Sebastian

I can't 100% say it's safe to use, but no, other inserts will not
influence the value, because currval will return the last sequence value
used by the current session (i.e. the current connection). And of course
you only have concurrent inserts using different connection, right ?

HTH,
Csaba.

Csaba Nagy wrote:

I can't 100% say it's safe to use, but no, other inserts will not
influence the value, because currval will return the last sequence value
used by the current session (i.e. the current connection). And of course
you only have concurrent inserts using different connection, right ?

HTH,
Csaba.

Yes, that' right.

Thanks

Sebastian

Richard Huxton wrote:

On Thursday 16 October 2003 11:11, Sebastian Boeck wrote:

Csaba Nagy wrote:

For your specific question I don't know the answer.

For this particular case you could use:

create table your_table (
id integer default nextval ('public.zeit_id_seq'::text),
pos_id integer default currval ('public.zeit_id_seq'::text),
...
);

That would work fine as long as you use inserts which don't specify id
if pos_id is not specified (otherwise the currval will throw you an
error cause it cannot be called without nextval being called).

HTH,
Csaba.

Thanks a lot, but is it save to use?

Not really.

Do i always get the same value, even if an other insert is changing
the sequence >public.zeit_id_seq< just at the same time?

Other processes can't interfere - the whole point of sequences is that they
are safe for this sort of thing.

Where you will have problems is that if one of the developers decides it's
more efficient to process fields backwards (zzz...pos_id, id) rather than
forwards (id, pos_id, ...zzz) then it will break.

Use a trigger here. If nothing else so you can stop people like me doing:

INSERT INTO your_table (id,pos_id) VALUES (-1,DEFAULT);

Is it possible to create the function called by a trigger entirely
in SQL, or do i have to use another language?

Regards

Sebastian

Sebastian Boeck wrote:

Is it possible to create the function called by a trigger entirely
in SQL, or do i have to use another language?

You can write trigger/functions in many languages. For this particular need
pl/pgsql would be more than good enough.

Shridhar

On Thu, Oct 16, 2003 at 13:53:27 +0200,
Sebastian Boeck <sebastianboeck@freenet.de> wrote:

Csaba Nagy wrote:

I can't 100% say it's safe to use, but no, other inserts will not
influence the value, because currval will return the last sequence value
used by the current session (i.e. the current connection). And of course
you only have concurrent inserts using different connection, right ?

HTH,
Csaba.

Yes, that' right.

Thanks

As mentioned earlier, the problem is with concurrent inserts, but rather
with which default expression gets calculated first. Today you are likely
to get predictable behavior out of this, but if you do an alter table
or perhaps upgrade to a new version of postgres, your code may break.