I've got some server functions that were previously written using
libpq.
They work fine. However because they need to open their own connection to
insert into tables and write into large objects, etc, they aren't
transaction safe. That is to say, if a user starts a transaction,
executes my
Eric Davies [EMAIL PROTECTED] writes:
Now I'm trying to make the functions transaction safe. In theory, all
I need to do is convert my libpq calls to equivalent SPI calls.
However, there don't appear to be any SPI equivalents for the large
object functions like lo_write and lo_read.
What
On Tue, Feb 13, 2001 at 09:52:21AM -0500, Tom Lane wrote:
DaVinci [EMAIL PROTECTED] writes:
procedencia int2 references procedencia,
línea int2 references línea,
empresa int2 references empresa,
Depending on the data values you are working with, it
DaVinci wrote:
On Tue, Feb 13, 2001 at 09:52:21AM -0500, Tom Lane wrote:
DaVinci [EMAIL PROTECTED] writes:
procedenciaint2 references procedencia,
lnea int2 references lnea,
empresaint2 references empresa,
Depending on the data values you are
On Wed, Feb 14, 2001 at 06:34:01AM -0500, Jan Wieck wrote:
DaVinci wrote:
Problem is with referential integrity like you say. But I don't understand
well reason. Interlocks should appear only when values of referenced tables
are changed, added or removed, shouldn't they?. But that is
DaVinci wrote:
On Wed, Feb 14, 2001 at 06:34:01AM -0500, Jan Wieck wrote:
The problem is, that a referential integrity constraint needs
to ensure that from the moment the constraint got checked
until your INSERT got made persistent (COMMIT), nobody else
has a chance
Title: RE: [GENERAL] transaction safety
Hi,
Well, the number is 'locked', because once it's given to you, that's it, it's gone from the 'list of available numbers' (i.e.: the sequence). However, between the insert, and the read of the ID, if another transaction performs an insert, it does
On Tue, Feb 13, 2001 at 09:56:18AM -, Michael Ansley wrote:
Hi,
Well, the number is 'locked', because once it's given to you, that's it,
it's gone from the 'list of available numbers' (i.e.: the sequence).
However, between the insert, and the read of the ID, if another transaction
Title: RE: [GENERAL] transaction safety
OK, someone want to answer this? I have always been under the impression that Postgres would not block under these circumstances, however, this is clearly blocking, for no apparently good reason.
I have just run a test on my own server
Hi all.
I want to create a new tuple of main info and detail tuples (in
different tables) that are joined by a key field. Key field is created by
generator automatically and I need that number in order to assign to detail
tuples. How can I to get that number in a safe way?.
Thanks for
DaVinci wrote:
Hi all.
I want to create a new tuple of main info and detail tuples (in
different tables) that are joined by a key field. Key field is created by
generator automatically and I need that number in order to assign to detail
tuples. How can I to get that number in a
On Mon, Feb 12, 2001 at 11:08:55AM +, Oliver Elphick wrote:
DaVinci wrote:
Hi all.
I want to create a new tuple of main info and detail tuples (in
different tables) that are joined by a key field. Key field is created by
generator automatically and I need that number in
On Mon, Feb 12, 2001 at 01:08:01PM -, Michael Ansley wrote:
The number returned by the sequence for the serial ID is retained within the
session, and so it can be returned by calling currval, e.g.:
...[a detailed example]...
Typically, the insert for a person, and for all the
Title: RE: [GENERAL] transaction safety
There should be no locks on the table for insert. The sequence number is retained by the session, not locked on the database. What this means is that somebody else may start their insert after you, and then you roll back, leaving a gap in the sequences
DaVinci [EMAIL PROTECTED] writes:
On Mon, Feb 12, 2001 at 01:08:01PM -, Michael Ansley wrote:
Typically, the insert for a person, and for all the associated addresses
would be done in the same transaction so that if the insert for one of the
addresses failed, then the whole lot would role
On Mon, Feb 12, 2001 at 10:22:30AM -0500, Tom Lane wrote:
DaVinci [EMAIL PROTECTED] writes:
On Mon, Feb 12, 2001 at 01:08:01PM -, Michael Ansley wrote:
Typically, the insert for a person, and for all the associated addresses
would be done in the same transaction so that if the insert
16 matches
Mail list logo