Thread: problem with new autocommit config parameter and jdbc
Haris,
You can't use jdbc (and probably most other postgres clients) with
autocommit in postgresql.conf turned off.
Hackers,
How should client interfaces handle this new autocommit feature? Is it
best to just issue a set at the beginning of the connection to ensure
that it is always on?
thanks,
--Barry
snpe wrote:
>Hi Dave,
>That is same.Program work with and without quote but row don't deleted.
>Postgresql is 7.3 beta (from cvs) and parameter autocommit in
postgresql.conf
>is off (no auto commit).
>I am tried with db.autocommit(true) after getConnection, but no success
>
>I thin that is bug in JDBC
>PGSql 7.3 beta have new features autocommit on/off and JDBC driver
don't work
>with autocommit off
>
>Thanks
>
>P.S
>I am play ith Oracle JDeveloper 9i and Postgresql and I get error in
prepared
>statement like this error :
>(oracle.jbo.SQLStmtException) JBO-27123: SQL error during call statement
>preparation. Statement: DELETE FROM org_ban WHERE "id"=?
>
>and pgsqlerror is :
>(org.postgresql.util.PSQLException) Malformed stmt [DELETE FROM
org_ban WHERE
>"id"=?] usage : {[? =] call <some_function> ([? [,?]*]) }
>
>I think that JDeveloper call CallableStatement for insert or delete
(select
>and update work fine), but I don't know how.
>
>On Friday 06 September 2002 04:35 pm, Dave Cramer wrote:
>
>
>>Remove the quotes around id, and let me know what happens
>>
>>Dave
>>
>>On Fri, 2002-09-06 at 10:52, snpe wrote:
>>
>>
>>>Hello Dave,
>>> There isn't any error.Program write 'Rows deleted 1', but row hasn't
>>>been deleted
>>>
>>>Thanks
>>>Haris Peco
>>>
>>>On Friday 06 September 2002 04:05 pm, Dave Cramer wrote:
>>>
>>>
>>>>Harris,
>>>>
>>>>What error do you get?
>>>>
>>>>Also you don't need the quotes around id
>>>>
>>>>Dave
>>>>
>>>>On Fri, 2002-09-06 at 10:06, snpe wrote:
>>>>
>>>>
>>>>>Hello,
>>>>> I have simple table with column ID and values '4' in this.
>>>>>I user 7.3 beta1 (from cvs 05.09.2002) and autocommit off in
>>>>>postgresql.conf. Next program don't work .
>>>>>I am tried with compiled postgresql.jar form CVS and with
>>>>>pg73b1jdbc3.jar from 05.09.2002 on jdbc.postgresql.org
>>>>>
>>>>>What is wrong ?
>>>>>
>>>>>regards
>>>>>Haris Peco
>>>>>import java.io.*;
>>>>>import java.sql.*;
>>>>>import java.text.*;
>>>>>
>>>>>public class PrepStatTest
>>>>>{
>>>>> Connection db;
>>>>> String stat="DELETE FROM org_ban WHERE \"id\" = ?";
>>>>> String delid = "4";
>>>>> public PrepStatTest() throws ClassNotFoundException,
>>>>>FileNotFoundException, IOException, SQLException
>>>>> {
>>>>>
Class.forName("org.postgresql.Driver");
>>>>>
db = DriverManager.getConnection("jdbc:postgresql://spnew/snpe",
>>>>>"snpe", "snpe");
>>>>>
PreparedStatement st = db.prepareStatement(stat);
>>>>> st.setString(1, delid);
>>>>> int rowsDeleted = st.executeUpdate();
>>>>>
System.out.println("Rows deleted " + rowsDeleted);
>>>>>
db.commit();
>>>>>
st.close();
>>>>>
db.close();
>>>>> }
>>>>>
>>>>> public static void main(String args[])
>>>>> {
>>>>>
try
>>>>>
{
>>>>>
PrepStatTest test = new PrepStatTest();
>>>>>
}
>>>>>
catch (Exception ex)
>>>>>
{
>>>>>
System.err.println("Exception caught.\n" + ex);
>>>>>
ex.printStackTrace();
>>>>>
}
>>>>> }
>>>>>}
>>>>>
>>>>>
>>>>>---------------------------(end of
>>>>>broadcast)--------------------------- TIP 3: if posting/reading
>>>>>through Usenet, please send an appropriate subscribe-nomail command
>>>>>to majordomo@postgresql.org so that your message can get through to
>>>>>the mailing list cleanly
>>>>>
>>>>>
>>>---------------------------(end of
broadcast)---------------------------
>>>TIP 2: you can get off all lists at once with the unregister command
>>> (send "unregister YourEmailAddressHere" to
majordomo@postgresql.org)
>>>
>>>
>>---------------------------(end of
broadcast)---------------------------
>>TIP 4: Don't 'kill -9' the postmaster
>>
>>
>
>
>---------------------------(end of broadcast)---------------------------
>TIP 3: if posting/reading through Usenet, please send an appropriate
>subscribe-nomail command to majordomo@postgresql.org so that your
>message can get through to the mailing list cleanly
>
>
>
Barry Lind wrote: > Haris, > > You can't use jdbc (and probably most other postgres clients) with > autocommit in postgresql.conf turned off. > > Hackers, > > How should client interfaces handle this new autocommit feature? Is it > best to just issue a set at the beginning of the connection to ensure > that it is always on? Yes, I thought that was the best fix for apps that can't deal with autocommit being off. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
On Saturday 07 September 2002 02:55 am, Bruce Momjian wrote: > Barry Lind wrote: > > Haris, > > > > You can't use jdbc (and probably most other postgres clients) with > > autocommit in postgresql.conf turned off. > > > > Hackers, > > > > How should client interfaces handle this new autocommit feature? Is it > > best to just issue a set at the beginning of the connection to ensure > > that it is always on? > > Yes, I thought that was the best fix for apps that can't deal with > autocommit being off. Can client get information from backend for autocommit (on or off) and that work like psql ?
snpe wrote: > On Saturday 07 September 2002 02:55 am, Bruce Momjian wrote: > > Barry Lind wrote: > > > Haris, > > > > > > You can't use jdbc (and probably most other postgres clients) with > > > autocommit in postgresql.conf turned off. > > > > > > Hackers, > > > > > > How should client interfaces handle this new autocommit feature? Is it > > > best to just issue a set at the beginning of the connection to ensure > > > that it is always on? > > > > Yes, I thought that was the best fix for apps that can't deal with > > autocommit being off. > Can client get information from backend for autocommit (on or off) and that > work like psql ? Sure, you can do SHOW autocommit. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
On Saturday 07 September 2002 04:07 pm, Bruce Momjian wrote: > snpe wrote: > > On Saturday 07 September 2002 02:55 am, Bruce Momjian wrote: > > > Barry Lind wrote: > > > > Haris, > > > > > > > > You can't use jdbc (and probably most other postgres clients) with > > > > autocommit in postgresql.conf turned off. > > > > > > > > Hackers, > > > > > > > > How should client interfaces handle this new autocommit feature? Is > > > > it best to just issue a set at the beginning of the connection to > > > > ensure that it is always on? > > > > > > Yes, I thought that was the best fix for apps that can't deal with > > > autocommit being off. > > > > Can client get information from backend for autocommit (on or off) and > > that work like psql ? > > Sure, you can do SHOW autocommit. I am interesting with JDBC driver. When I make connection in base I want that driver find autocommit mode (from postgresql.conf or call in backend) and set mode true or false thanks
snpe wrote:
>>Sure, you can do SHOW autocommit.
>
> I am interesting with JDBC driver.
> When I make connection in base I want that driver find autocommit mode
> (from postgresql.conf or call in backend) and set mode true or false
>
You could make a call to current_setting in the backend.
I don't know anything about the jdbc driver, but if it's written in C
something like this should work:
text *autocommit = DatumGetTextP(DirectFunctionCall1(current_setting,
CStringGetDatum("autocommit")));
Would this work?
Joe
On Saturday 07 September 2002 09:12 pm, Joe Conway wrote:
> snpe wrote:
> >>Sure, you can do SHOW autocommit.
> >
> > I am interesting with JDBC driver.
> > When I make connection in base I want that driver find autocommit mode
> > (from postgresql.conf or call in backend) and set mode true or false
>
> You could make a call to current_setting in the backend.
>
> I don't know anything about the jdbc driver, but if it's written in C
> something like this should work:
>
> text *autocommit = DatumGetTextP(DirectFunctionCall1(current_setting,
> CStringGetDatum("autocommit")));
>
> Would this work?
Yes.But I don't know like call in JDBC.
Yes it is possible, but according to the jdbc spec, a new connection in jdbc is always initialized to autocommit=true. So jdbc needs to ignore whatever the current server setting is and reset to autocommit=true. --Barry snpe wrote: > On Saturday 07 September 2002 02:55 am, Bruce Momjian wrote: > >>Barry Lind wrote: >> >>>Haris, >>> >>>You can't use jdbc (and probably most other postgres clients) with >>>autocommit in postgresql.conf turned off. >>> >>>Hackers, >>> >>>How should client interfaces handle this new autocommit feature? Is it >>>best to just issue a set at the beginning of the connection to ensure >>>that it is always on? >> >>Yes, I thought that was the best fix for apps that can't deal with >>autocommit being off. > > Can client get information from backend for autocommit (on or off) and that > work like psql ? > > >
Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Barry Lind wrote:
>> How should client interfaces handle this new autocommit feature? Is it
>> best to just issue a set at the beginning of the connection to ensure
>> that it is always on?
> Yes, I thought that was the best fix for apps that can't deal with
> autocommit being off.
If autocommit=off really seriously breaks JDBC then I don't think a
simple SET command at the start of a session is going to do that much
to improve robustness. What if the user issues another SET to turn it
on?
I'd suggest just documenting that it is broken and you can't use it,
until such time as you can get it fixed. Band-aids that only partially
cover the problem don't seem worth the effort to me.
In general I think that autocommit=off is probably going to be very
poorly supported in the 7.3 release. We can document it as being
"work in progress, use at your own risk".
regards, tom lane
On Monday 09 September 2002 08:53 pm, Tom Lane wrote: > Bruce Momjian <pgman@candle.pha.pa.us> writes: > > Barry Lind wrote: > >> How should client interfaces handle this new autocommit feature? Is it > >> best to just issue a set at the beginning of the connection to ensure > >> that it is always on? > > > > Yes, I thought that was the best fix for apps that can't deal with > > autocommit being off. > > If autocommit=off really seriously breaks JDBC then I don't think a > simple SET command at the start of a session is going to do that much > to improve robustness. What if the user issues another SET to turn it > on? > > I'd suggest just documenting that it is broken and you can't use it, > until such time as you can get it fixed. Band-aids that only partially > cover the problem don't seem worth the effort to me. > > In general I think that autocommit=off is probably going to be very > poorly supported in the 7.3 release. We can document it as being > "work in progress, use at your own risk". > I'm use 'autocommit=false' and have problem with psql When any commnad is lost, then next commnad get error for transactions (simple select command).BTW snpe> select * from org_ba; ERROR: relation org_ba does not exists snpe> select * from org_ban; ERROR: current transactions is aborted, queries ignored until end of transaction block snpe> rollback; ROLLBACK snpe> select * from org_ban; this command is ok. regards Haris Peco
On Mon, 2002-09-09 at 17:04, snpe wrote: > I'm use 'autocommit=false' and have problem with psql > When any commnad is lost, then next commnad get error for transactions > (simple select command).BTW > > snpe> select * from org_ba; > ERROR: relation org_ba does not exists > snpe> select * from org_ban; > ERROR: current transactions is aborted, queries ignored until end of > transaction block > snpe> rollback; > ROLLBACK > snpe> select * from org_ban; Maybe I'm missing something, but isn't that the expected behaviour when autocommit is turned off? -- Rod Taylor
On Monday 09 September 2002 11:03 pm, Rod Taylor wrote: > On Mon, 2002-09-09 at 17:04, snpe wrote: > > I'm use 'autocommit=false' and have problem with psql > > When any commnad is lost, then next commnad get error for transactions > > (simple select command).BTW > > > > snpe> select * from org_ba; > > ERROR: relation org_ba does not exists > > snpe> select * from org_ban; > > ERROR: current transactions is aborted, queries ignored until end of > > transaction block > > snpe> rollback; > > ROLLBACK > > snpe> select * from org_ban; > > Maybe I'm missing something, but isn't that the expected behaviour when > autocommit is turned off? I get this every time.When exists command with error next command don't work without explicit rollback and commit (this is not for psql, this error get in with JDeveloper - JDBC driver).When autocommit=ture all is fine haris peco
On Tue, 10 Sep 2002, snpe wrote: > On Monday 09 September 2002 11:03 pm, Rod Taylor wrote: > > On Mon, 2002-09-09 at 17:04, snpe wrote: > > > I'm use 'autocommit=false' and have problem with psql > > > When any commnad is lost, then next commnad get error for transactions > > > (simple select command).BTW > > > > > > snpe> select * from org_ba; > > > ERROR: relation org_ba does not exists > > > snpe> select * from org_ban; > > > ERROR: current transactions is aborted, queries ignored until end of > > > transaction block > > > snpe> rollback; > > > ROLLBACK > > > snpe> select * from org_ban; > > > > Maybe I'm missing something, but isn't that the expected behaviour when > > autocommit is turned off? > I get this every time.When exists command with error next command don't work > without explicit rollback and commit (this is not for psql, this error get in > with JDeveloper - JDBC driver).When autocommit=ture all is fine It starts a transaction, failes the first command and goes into the error has occurred in this transaction state. Seems like reasonable behavior.
On Tuesday 10 September 2002 03:05 am, Stephan Szabo wrote: > On Tue, 10 Sep 2002, snpe wrote: > > On Monday 09 September 2002 11:03 pm, Rod Taylor wrote: > > > On Mon, 2002-09-09 at 17:04, snpe wrote: > > > > I'm use 'autocommit=false' and have problem with psql > > > > When any commnad is lost, then next commnad get error for > > > > transactions (simple select command).BTW > > > > > > > > snpe> select * from org_ba; > > > > ERROR: relation org_ba does not exists > > > > snpe> select * from org_ban; > > > > ERROR: current transactions is aborted, queries ignored until end of > > > > transaction block > > > > snpe> rollback; > > > > ROLLBACK > > > > snpe> select * from org_ban; > > > > > > Maybe I'm missing something, but isn't that the expected behaviour when > > > autocommit is turned off? > > > > I get this every time.When exists command with error next command don't > > work without explicit rollback and commit (this is not for psql, this > > error get in with JDeveloper - JDBC driver).When autocommit=ture all is > > fine > > It starts a transaction, failes the first command and goes into the > error has occurred in this transaction state. Seems like reasonable > behavior. Select command don't start transaction - it is not good Error command don't start transaction - nothing hapen, only typing error regards haris peco
On Tue, 10 Sep 2002, snpe wrote: > On Tuesday 10 September 2002 03:05 am, Stephan Szabo wrote: > > On Tue, 10 Sep 2002, snpe wrote: > > > On Monday 09 September 2002 11:03 pm, Rod Taylor wrote: > > > > On Mon, 2002-09-09 at 17:04, snpe wrote: > > > > > I'm use 'autocommit=false' and have problem with psql > > > > > When any commnad is lost, then next commnad get error for > > > > > transactions (simple select command).BTW > > > > > > > > > > snpe> select * from org_ba; > > > > > ERROR: relation org_ba does not exists > > > > > snpe> select * from org_ban; > > > > > ERROR: current transactions is aborted, queries ignored until end of > > > > > transaction block > > > > > snpe> rollback; > > > > > ROLLBACK > > > > > snpe> select * from org_ban; > > > > > > > > Maybe I'm missing something, but isn't that the expected behaviour when > > > > autocommit is turned off? > > > > > > I get this every time.When exists command with error next command don't > > > work without explicit rollback and commit (this is not for psql, this > > > error get in with JDeveloper - JDBC driver).When autocommit=ture all is > > > fine > > > > It starts a transaction, failes the first command and goes into the > > error has occurred in this transaction state. Seems like reasonable > > behavior. > Select command don't start transaction - it is not good I think you need more justification than "it is not good." If I do a sequence of select statements in autocommit=false, I'd expect the same consistancy as if I'd done begin; select ...; select ...; > Error command don't start transaction - nothing hapen, only typing error If you do an insert that violates a constraint, does that start an transaction or not? I think we have to choose before we start doing the statement not after.
snpe <snpe@snpe.co.yu> writes:
> I'm use 'autocommit=false' and have problem with psql
> When any commnad is lost, then next commnad get error for transactions
> (simple select command).BTW
> snpe> select * from org_ba;
> ERROR: relation org_ba does not exists
> snpe> select * from org_ban;
> ERROR: current transactions is aborted, queries ignored until end of
> transaction block
Um, what's wrong with that?
It seems to me that an application that is using autocommit=off will
expect the first SELECT to start a transaction block. If the first
SELECT fails, then subsequent commands *should* fail until you commit
or rollback. Certainly if you did an explicit BEGIN before the first
SELECT, the above is what you'd get --- why should implicit BEGIN
work differently?
regards, tom lane
On Mon, 9 Sep 2002, Tom Lane wrote:
> snpe <snpe@snpe.co.yu> writes:
>
> > snpe> select * from org_ba;
> > ERROR: relation org_ba does not exists
> > snpe> select * from org_ban;
> > ERROR: current transactions is aborted, queries ignored until end of
> > transaction block
>
> Um, what's wrong with that?
>
> It seems to me that an application that is using autocommit=off will
> expect the first SELECT to start a transaction block.
Yup. In fact, the standard (at least, insofar as I have information
relating to it), specifies that the first SELECT statement above
*must* start a transaction.
From Date's _A Guide to the SQL Standard_ (Fourth Edition):
An SQL-transaction is initiated when the relevant SQL-agent executes
a "transaction-initiating" SQL Statement (see below) and the
SQL-agent does not already have an SQL-transaction in progress.
...
The following SQL statements are _not_ transaction-initiating:
CONNECT
SET CONNECTION
DISCONNECT
SET SESSION AUTHORIZATION
SET CATALOG
SET SCHEMA
SET NAMES
SET TIME ZONE
SET TRANSACTION
SET CONSTRAINTS
COMMIT
ROLLBACK
GET DIAGNOSTICS
Nor, of course, are the nonexecutable statements DECLARE CURSOR,
DECLAR LOCAL TEMPORARY TABLE, BEGIN DECLARE SECTION, SEND DECLARE
SECTIONS, and WHENEVER.
So SELECT ought always to initiate a transaction, if one is not already
in progress. If auto-commit is enabled, of course, that statement may
be committed immediately after execution, if it doesn't fail.
As far as the JDBC driver goes, I'm not too sure of the issues here, but
it should certainly be ensuring that autocommit is enabled, as per the
JDBC specification, when a new connection is created. I see no reason
this couldn't be done with a "SET AUTOCOMMIT TO OFF" or whatever, if
that's necessary to override a possible configuration file setting.
cjs
--
Curt Sampson <cjs@cynic.net> +81 90 7737 2974 http://www.netbsd.org
Don't you know, in this new Dark Age, we're all light. --XTC
On Mon, 9 Sep 2002, Tom Lane wrote: > If autocommit=off really seriously breaks JDBC then I don't think a > simple SET command at the start of a session is going to do that much > to improve robustness. What if the user issues another SET to turn it > on? You mean, to turn it off again? The driver should catch this, in theory. In practice we could probably live with saying, "Don't use SET AUTOCOMMIT; use the methods on the Connection class instead." Probably the driver should be changed for 7.3 just to use the server's SET AUTOCOMMIT functionality.... cjs -- Curt Sampson <cjs@cynic.net> +81 90 7737 2974 http://www.netbsd.org Don't you know, in this new Dark Age, we're all light. --XTC
Curt Sampson <cjs@cynic.net> writes:
> Probably the driver should be changed for 7.3 just to use the server's
> SET AUTOCOMMIT functionality....
That should happen eventually, IMHO, but I am not going to tell the JDBC
developers that they must make it happen for 7.3. They've already got a
pile of much-higher-priority things to fix for 7.3, like schema
compatibility and dropped-column handling.
My feeling about the original complaint is very simple: setting server
autocommit to off is not supported with JDBC (nor is it fully supported
with any other of our frontend clients, right at this instant, though
that may improve somewhat before 7.3 release). If you don't like it,
tough; contribute the required fixes or stop complaining. Someone else
will fix it when they get around to it, but there are bigger problems to
deal with first. Autocommit is only a work-in-progress today, not
something that we promise will do anything useful for anybody.
regards, tom lane
On Tuesday 10 September 2002 04:16 am, Stephan Szabo wrote: > On Tue, 10 Sep 2002, snpe wrote: > > On Tuesday 10 September 2002 03:05 am, Stephan Szabo wrote: > > > On Tue, 10 Sep 2002, snpe wrote: > > > > On Monday 09 September 2002 11:03 pm, Rod Taylor wrote: > > > > > On Mon, 2002-09-09 at 17:04, snpe wrote: > > > > > > I'm use 'autocommit=false' and have problem with psql > > > > > > When any commnad is lost, then next commnad get error for > > > > > > transactions (simple select command).BTW > > > > > > > > > > > > snpe> select * from org_ba; > > > > > > ERROR: relation org_ba does not exists > > > > > > snpe> select * from org_ban; > > > > > > ERROR: current transactions is aborted, queries ignored until end > > > > > > of transaction block > > > > > > snpe> rollback; > > > > > > ROLLBACK > > > > > > snpe> select * from org_ban; > > > > > > > > > > Maybe I'm missing something, but isn't that the expected behaviour > > > > > when autocommit is turned off? > > > > > > > > I get this every time.When exists command with error next command > > > > don't work without explicit rollback and commit (this is not for > > > > psql, this error get in with JDeveloper - JDBC driver).When > > > > autocommit=ture all is fine > > > > > > It starts a transaction, failes the first command and goes into the > > > error has occurred in this transaction state. Seems like reasonable > > > behavior. > > > > Select command don't start transaction - it is not good > > I think you need more justification than "it is not good." If I do a > sequence of select statements in autocommit=false, I'd expect the same > consistancy as if I'd done > begin; > select ...; > select ...; > Ok.You start transaction explicit and this is ok. But simple SELECT don't start transaction. > > Error command don't start transaction - nothing hapen, only typing error > > If you do an insert that violates a constraint, does that start an > transaction or not? I think we have to choose before we start doing the > statement not after. This is typeing error.Nothing happen.That is not transaction. I don't know that is possible, but before start transaction we need parsing command and select or any error don't start transaction This is problem for every client (I know for JDBC) regards Haris Peco
Curt Sampson <cjs@cynic.net> writes:
> From Date's _A Guide to the SQL Standard_ (Fourth Edition):
> ...
> The following SQL statements are _not_ transaction-initiating:
> CONNECT
> SET CONNECTION
> DISCONNECT
> SET SESSION AUTHORIZATION
> SET CATALOG
> SET SCHEMA
> SET NAMES
> SET TIME ZONE
> SET TRANSACTION
> SET CONSTRAINTS
> COMMIT
> ROLLBACK
> GET DIAGNOSTICS
Hm. This brings up a thought I've been turning over for the past
couple days. As of CVS tip, SET commands *do* initiate transactions
if you have autocommit off. By your reading of Date, this is not
spec compliant for certain SET variables: a SET not already within
a transaction should not start a transaction block, at least for the
variables mentioned above. It occurs to me that it'd be reasonable
to make it act that way for all SET variables.
An example of how this would simplify life: consider the problem of
a client that wants to ensure autocommit is on. A simple
SET autocommit TO on;
doesn't work at the moment: if autocommit is off, then you'll need
to issue a COMMIT as well to get out of the implicitly started
transaction. But you don't want to just issue a COMMIT, because
you'll get a nasty ugly WARNING message on stderr if indeed autocommit
was on already. The only warning-free way to issue a SET right now
if you are uncertain about autocommit status is
BEGIN; SET .... ; COMMIT;
Blech. But if SET doesn't start a transaction then you can still
just do SET. This avoids some changes we'll otherwise have to make
in libpq startup, among other places.
Does anyone see any cases where it's important for SET to start
a transaction? (Of course, if you are already *in* a transaction,
the SET will be part of that transaction. The question is whether
we want SET to trigger an implicit BEGIN or not.)
> Nor, of course, are the nonexecutable statements DECLARE CURSOR,
> DECLAR LOCAL TEMPORARY TABLE, BEGIN DECLARE SECTION, SEND DECLARE
> SECTIONS, and WHENEVER.
Hmm. I think the spec's notion of DECLARE must be different from ours.
Our implementation of DECLARE CURSOR both declares and opens the cursor,
and as such it *must* be transaction-initiating; else it's useless.
regards, tom lane
> > > > It starts a transaction, failes the first command and goes into the > > > > error has occurred in this transaction state. Seems like reasonable > > > > behavior. > > > > > > Select command don't start transaction - it is not good > > > > I think you need more justification than "it is not good." If I do a > > sequence of select statements in autocommit=false, I'd expect the same > > consistancy as if I'd done > > begin; > > select ...; > > select ...; > > > Ok.You start transaction explicit and this is ok. > But simple SELECT don't start transaction. Actually someone post a bit from Date's book that implies it does. And, that's still not an justification, it's just a restating of same position. I don't see any reason why the two should be different from a data consistency standpoint, there might be one, but you haven't given any reasons. > > > Error command don't start transaction - nothing hapen, only typing error > > > > If you do an insert that violates a constraint, does that start an > > transaction or not? I think we have to choose before we start doing the > > statement not after. > This is typeing error.Nothing happen.That is not transaction. > I don't know that is possible, but before start transaction we need parsing > command and select or any error don't start transaction Why not? AFAICT it should, the transaction is initiated a statement is run and it fails. Now maybe we shouldn't be going into the wierd disabled statement state, but that's a different argument entirely.
I am waiting for this thread to conclude before deciding exactly what to do for the jdbc driver for 7.3. While using the 'set autocommit true' syntax is nice when talking to a 7.3 server, the jdbc driver also needs to be backwardly compatible with 7.2 and 7.1 servers. So it may just be easier to continue with the current way of doing things, even in the 7.3 case. thanks, --Barry Curt Sampson wrote: > On Mon, 9 Sep 2002, Tom Lane wrote: > > >>If autocommit=off really seriously breaks JDBC then I don't think a >>simple SET command at the start of a session is going to do that much >>to improve robustness. What if the user issues another SET to turn it >>on? > > > You mean, to turn it off again? The driver should catch this, in theory. > > In practice we could probably live with saying, "Don't use SET > AUTOCOMMIT; use the methods on the Connection class instead." > > Probably the driver should be changed for 7.3 just to use the server's > SET AUTOCOMMIT functionality.... > > cjs
Tom Lane wrote: > An example of how this would simplify life: consider the problem of > a client that wants to ensure autocommit is on. A simple > SET autocommit TO on; > doesn't work at the moment: if autocommit is off, then you'll need > to issue a COMMIT as well to get out of the implicitly started > transaction. But you don't want to just issue a COMMIT, because > you'll get a nasty ugly WARNING message on stderr if indeed autocommit > was on already. The only warning-free way to issue a SET right now > if you are uncertain about autocommit status is > BEGIN; SET .... ; COMMIT; > Blech. But if SET doesn't start a transaction then you can still > just do SET. This avoids some changes we'll otherwise have to make > in libpq startup, among other places. > > Does anyone see any cases where it's important for SET to start > a transaction? (Of course, if you are already *in* a transaction, > the SET will be part of that transaction. The question is whether > we want SET to trigger an implicit BEGIN or not.) Uh, well, because we now have SET's rollback in an aborted transaction, there is an issue of whether the SET is part of the transaction or not. Seems it has to be for consistency with our rollback behavior. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
On Tue, 10 Sep 2002, Stephan Szabo wrote: > > > > > It starts a transaction, failes the first command and goes into the > > > > > error has occurred in this transaction state. Seems like reasonable > > > > > behavior. > > > > > > > > Select command don't start transaction - it is not good > > > > > > I think you need more justification than "it is not good." If I do a > > > sequence of select statements in autocommit=false, I'd expect the same > > > consistancy as if I'd done > > > begin; > > > select ...; > > > select ...; > > > > > Ok.You start transaction explicit and this is ok. > > But simple SELECT don't start transaction. > > Actually someone post a bit from Date's book that implies it does. > And, that's still not an justification, it's just a restating of same > position. I don't see any reason why the two should be different from > a data consistency standpoint, there might be one, but you haven't > given any reasons. What if it's a select for update? IF that failed because of a timout on a lock, shouldn't the transaction fail? Or a select into? Either of those should make a transaction fail, and they're just selects.
Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Tom Lane wrote:
>> Does anyone see any cases where it's important for SET to start
>> a transaction? (Of course, if you are already *in* a transaction,
>> the SET will be part of that transaction. The question is whether
>> we want SET to trigger an implicit BEGIN or not.)
> Uh, well, because we now have SET's rollback in an aborted transaction,
> there is an issue of whether the SET is part of the transaction or not.
> Seems it has to be for consistency with our rollback behavior.
Yeah, it must be part of the transaction unless we want to reopen the
SET-rollback can of worms (which I surely don't want to).
However, a SET issued outside any pre-existing transaction block could
form a self-contained transaction without any logical difficulty, even
in autocommit-off mode. The question is whether that's more or less
convenient, or standards-conforming, than what we have.
An alternative that I'd really rather not consider is making SET's
behavior dependent on exactly which variable is being set ...
regards, tom lane
Tom Lane wrote:
> Bruce Momjian <pgman@candle.pha.pa.us> writes:
> > Tom Lane wrote:
> >> Does anyone see any cases where it's important for SET to start
> >> a transaction? (Of course, if you are already *in* a transaction,
> >> the SET will be part of that transaction. The question is whether
> >> we want SET to trigger an implicit BEGIN or not.)
>
> > Uh, well, because we now have SET's rollback in an aborted transaction,
> > there is an issue of whether the SET is part of the transaction or not.
> > Seems it has to be for consistency with our rollback behavior.
>
> Yeah, it must be part of the transaction unless we want to reopen the
> SET-rollback can of worms (which I surely don't want to).
>
> However, a SET issued outside any pre-existing transaction block could
> form a self-contained transaction without any logical difficulty, even
> in autocommit-off mode. The question is whether that's more or less
> convenient, or standards-conforming, than what we have.
That seems messy. What you are saying is that if autocommit is off,
then in:
SET x=1;
UPDATE ...
SET y=2;
ROLLBACK;
that the x=1 doesn't get rolled back bu the y=2 does? I can't see any
good logic for that.
> An alternative that I'd really rather not consider is making SET's
> behavior dependent on exactly which variable is being set ...
Agreed. We discussed that in the SET rollback case and found it was
more trouble that it was worth.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073
On Tue, 10 Sep 2002, scott.marlowe wrote: > On Tue, 10 Sep 2002, Stephan Szabo wrote: > > > > > > > It starts a transaction, failes the first command and goes into the > > > > > > error has occurred in this transaction state. Seems like reasonable > > > > > > behavior. > > > > > > > > > > Select command don't start transaction - it is not good > > > > > > > > I think you need more justification than "it is not good." If I do a > > > > sequence of select statements in autocommit=false, I'd expect the same > > > > consistancy as if I'd done > > > > begin; > > > > select ...; > > > > select ...; > > > > > > > Ok.You start transaction explicit and this is ok. > > > But simple SELECT don't start transaction. > > > > Actually someone post a bit from Date's book that implies it does. > > And, that's still not an justification, it's just a restating of same > > position. I don't see any reason why the two should be different from > > a data consistency standpoint, there might be one, but you haven't > > given any reasons. > > What if it's a select for update? IF that failed because of a timout on a > lock, shouldn't the transaction fail? Or a select into? Either of those > should make a transaction fail, and they're just selects. Yes, but I think it should still work the same as if it had failed in an explicit transaction if autocommit is false (or was that directed at someone else).
Bruce Momjian <pgman@candle.pha.pa.us> writes:
> That seems messy. What you are saying is that if autocommit is off,
> then in:
> SET x=1;
> UPDATE ...
> SET y=2;
> ROLLBACK;
> that the x=1 doesn't get rolled back bu the y=2 does?
Yes, if you weren't in a transaction at the start.
> I can't see any good logic for that.
How about "the SQL spec requires it"? Date seems to think it does,
at least for some variables (of course we have lots of variables
that are not in the spec).
I can't find anything very clear in the SQL92 or SQL99 documents,
and I'm not at home at the moment to look at my copy of Date, but
if Curt's reading is correct then we have spec precedent for acting
this way.
regards, tom lane
On Tue, 10 Sep 2002, Stephan Szabo wrote: > On Tue, 10 Sep 2002, scott.marlowe wrote: > > > On Tue, 10 Sep 2002, Stephan Szabo wrote: > > > > > > > > > It starts a transaction, failes the first command and goes into the > > > > > > > error has occurred in this transaction state. Seems like reasonable > > > > > > > behavior. > > > > > > > > > > > > Select command don't start transaction - it is not good > > > > > > > > > > I think you need more justification than "it is not good." If I do a > > > > > sequence of select statements in autocommit=false, I'd expect the same > > > > > consistancy as if I'd done > > > > > begin; > > > > > select ...; > > > > > select ...; > > > > > > > > > Ok.You start transaction explicit and this is ok. > > > > But simple SELECT don't start transaction. > > > > > > Actually someone post a bit from Date's book that implies it does. > > > And, that's still not an justification, it's just a restating of same > > > position. I don't see any reason why the two should be different from > > > a data consistency standpoint, there might be one, but you haven't > > > given any reasons. > > > > What if it's a select for update? IF that failed because of a timout on a > > lock, shouldn't the transaction fail? Or a select into? Either of those > > should make a transaction fail, and they're just selects. > > Yes, but I think it should still work the same as if it had failed in an > explicit transaction if autocommit is false (or was that directed at > someone else). Sorry, I was agreeing with you, and disagreeing with the guy who was saying that selects shouldn't start a transaction. Should have mentioned that. :-)
Tom Lane wrote: > Bruce Momjian <pgman@candle.pha.pa.us> writes: > > That seems messy. What you are saying is that if autocommit is off, > > then in: > > > SET x=1; > > UPDATE ... > > SET y=2; > > ROLLBACK; > > > that the x=1 doesn't get rolled back bu the y=2 does? > > Yes, if you weren't in a transaction at the start. > > > I can't see any good logic for that. > > How about "the SQL spec requires it"? Date seems to think it does, > at least for some variables (of course we have lots of variables > that are not in the spec). > > I can't find anything very clear in the SQL92 or SQL99 documents, > and I'm not at home at the moment to look at my copy of Date, but > if Curt's reading is correct then we have spec precedent for acting > this way. Spec or not, it looks pretty weird so I would question following the spec on this one. Do we want to say "With autocommit off, SET will be in it's own transaction if it appears before any non-SET command", and "SETs are rolled back except if autocommit off and they appear before any non-SET"? I sure don't. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
On Tuesday 10 September 2002 09:55 pm, Tom Lane wrote: > Bruce Momjian <pgman@candle.pha.pa.us> writes: > > That seems messy. What you are saying is that if autocommit is off, > > then in: > > > > SET x=1; > > UPDATE ... > > SET y=2; > > ROLLBACK; > > > > that the x=1 doesn't get rolled back bu the y=2 does? > > Yes, if you weren't in a transaction at the start. > > > I can't see any good logic for that. > > How about "the SQL spec requires it"? Date seems to think it does, > at least for some variables (of course we have lots of variables > that are not in the spec). > > I can't find anything very clear in the SQL92 or SQL99 documents, > and I'm not at home at the moment to look at my copy of Date, but > if Curt's reading is correct then we have spec precedent for acting > this way. I know what Oracle do (default mode autocommit off except JDBC) : only DML and DDL command start transaction and DDL command end transaction. There is another problem: if select start transaction why error - I will continue transaction. Why invalid command start transaction ? regards haris peco
On Tuesday 10 September 2002 07:46 pm, scott.marlowe wrote: > On Tue, 10 Sep 2002, Stephan Szabo wrote: > > > > > > It starts a transaction, failes the first command and goes into > > > > > > the error has occurred in this transaction state. Seems like > > > > > > reasonable behavior. > > > > > > > > > > Select command don't start transaction - it is not good > > > > > > > > I think you need more justification than "it is not good." If I do a > > > > sequence of select statements in autocommit=false, I'd expect the > > > > same consistancy as if I'd done > > > > begin; > > > > select ...; > > > > select ...; > > > > > > Ok.You start transaction explicit and this is ok. > > > But simple SELECT don't start transaction. > > > > Actually someone post a bit from Date's book that implies it does. > > And, that's still not an justification, it's just a restating of same > > position. I don't see any reason why the two should be different from > > a data consistency standpoint, there might be one, but you haven't > > given any reasons. > > What if it's a select for update? IF that failed because of a timout on a > lock, shouldn't the transaction fail? Or a select into? Either of those > should make a transaction fail, and they're just selects. Ok.Any lock or update,delete, insert (and all ddl command) start transaction (select for update, too), but simple select no.Select don't change data and no transaction - this process cannot lost consistency (any command with error too). And if transaction start, so what ... I will (maybe) continue transaction (I don't end transaction), but I get error. and I must end transaction I think that we must parse command, choose if 'start transaction' and start transaction or no. regards Haris Peco
On Tue, 10 Sep 2002, snpe wrote: > On Tuesday 10 September 2002 07:46 pm, scott.marlowe wrote: > > What if it's a select for update? IF that failed because of a timout on a > > lock, shouldn't the transaction fail? Or a select into? Either of those > > should make a transaction fail, and they're just selects. > Ok.Any lock or update,delete, insert (and all ddl command) start transaction > (select for update, too), but simple select no.Select don't change data and no > transaction - this process cannot lost consistency (any command with error > too). At least in serializable isolation level you'll probably get different results if a transaction commits between those two selects based on whether a transaction is started or not. Should two serializable selects in the same session see the same snapshot when autocommit is off?
On Tuesday 10 September 2002 11:50 pm, Stephan Szabo wrote: > On Tue, 10 Sep 2002, snpe wrote: > > On Tuesday 10 September 2002 07:46 pm, scott.marlowe wrote: > > > What if it's a select for update? IF that failed because of a timout > > > on a lock, shouldn't the transaction fail? Or a select into? Either > > > of those should make a transaction fail, and they're just selects. > > > > Ok.Any lock or update,delete, insert (and all ddl command) start > > transaction (select for update, too), but simple select no.Select don't > > change data and no transaction - this process cannot lost consistency > > (any command with error too). > > At least in serializable isolation level you'll probably get different > results if a transaction commits between those two selects based on > whether a transaction is started or not. Should two serializable selects > in the same session see the same snapshot when autocommit is off? It is session, not transaction.My select don't change data and this is not transaction. My abother question, agian : why error (bad typing) start transaction ? regards haris peco
On Wed, 11 Sep 2002, snpe wrote:
> On Tuesday 10 September 2002 11:50 pm, Stephan Szabo wrote:
> > On Tue, 10 Sep 2002, snpe wrote:
> > > On Tuesday 10 September 2002 07:46 pm, scott.marlowe wrote:
> > > > What if it's a select for update? IF that failed because of a timout
> > > > on a lock, shouldn't the transaction fail? Or a select into? Either
> > > > of those should make a transaction fail, and they're just selects.
> > >
> > > Ok.Any lock or update,delete, insert (and all ddl command) start
> > > transaction (select for update, too), but simple select no.Select don't
> > > change data and no transaction - this process cannot lost consistency
> > > (any command with error too).
> >
> > At least in serializable isolation level you'll probably get different
> > results if a transaction commits between those two selects based on
> > whether a transaction is started or not. Should two serializable selects
> > in the same session see the same snapshot when autocommit is off?
> It is session, not transaction.My select don't change data and this is not
> transaction.
We're going around in circles.
Does it matter if data is changed? I don't think so, since at least in
serializable isolation level the snapshot that is seen depends on whether
you're in a transaction or not, and given autocommit=off I believe that
you should get a consistent snapshot between them.
If you believe it should matter, you need to give a reason. I don't
think it's a spec reason given that my sql92 spec draft says:
"The following SQL-statements are transaction initiating SQL-
statements, i.e., if there is no current transaction, and a
statement of this class is executed, a transaction is initiated:
... o <select statement: single row>
o <direct select statement: multiple rows>"
unless it changed.
There might be a compatibility reason, if so, with what and is it stronger
than reasons to start a transaction.
There might be another logical reason, if so, what is it and why does
it matter?
> My abother question, agian : why error (bad typing) start transaction ?
That depends. Given the way the spec is worded, it says nothing about
other statements, so we need to decide those ourselves. I don't see
anything that implies that a select statement that errors would be
any different than a select statement that doesn't as far as starting
a transaction goes in my sql92 spec draft. If you were to type in
foo as a command, I could see a case that maybe that shouldn't be
transaction initiating, but afair that wasn't the case you had, you
had a select command against an invalid table name.
On Wednesday 11 September 2002 01:25 am, Stephan Szabo wrote: > On Wed, 11 Sep 2002, snpe wrote: > > On Tuesday 10 September 2002 11:50 pm, Stephan Szabo wrote: > > > On Tue, 10 Sep 2002, snpe wrote: > > > > On Tuesday 10 September 2002 07:46 pm, scott.marlowe wrote: > > > > > What if it's a select for update? IF that failed because of a > > > > > timout on a lock, shouldn't the transaction fail? Or a select > > > > > into? Either of those should make a transaction fail, and they're > > > > > just selects. > > > > > > > > Ok.Any lock or update,delete, insert (and all ddl command) start > > > > transaction (select for update, too), but simple select no.Select > > > > don't change data and no transaction - this process cannot lost > > > > consistency (any command with error too). > > > > > > At least in serializable isolation level you'll probably get different > > > results if a transaction commits between those two selects based on > > > whether a transaction is started or not. Should two serializable > > > selects in the same session see the same snapshot when autocommit is > > > off? > > > > It is session, not transaction.My select don't change data and this is > > not transaction. > > We're going around in circles. > > Does it matter if data is changed? I don't think so, since at least in > serializable isolation level the snapshot that is seen depends on whether > you're in a transaction or not, and given autocommit=off I believe that > you should get a consistent snapshot between them. > > If you believe it should matter, you need to give a reason. I don't > think it's a spec reason given that my sql92 spec draft says: > > "The following SQL-statements are transaction initiating SQL- > statements, i.e., if there is no current transaction, and a > statement of this class is executed, a transaction is initiated: > ... > o <select statement: single row> > > o <direct select statement: multiple rows>" > unless it changed. > > There might be a compatibility reason, if so, with what and is it stronger > than reasons to start a transaction. > > There might be another logical reason, if so, what is it and why does > it matter? > > > My abother question, agian : why error (bad typing) start transaction ? > > That depends. Given the way the spec is worded, it says nothing about > other statements, so we need to decide those ourselves. I don't see > anything that implies that a select statement that errors would be > any different than a select statement that doesn't as far as starting > a transaction goes in my sql92 spec draft. If you were to type in > foo as a command, I could see a case that maybe that shouldn't be > transaction initiating, but afair that wasn't the case you had, you > had a select command against an invalid table name. yes, we're going around in circles. Ok.I agreed (I think because Oracle do different) Transaction start I type invalid command I correct command I get error Why.If is it transactin, why I get error I want continue. I am see this error with JDeveloper (work with Oracle, DB2 an SQL Server) It is not matter for me transaction or not.I get error for correct command after invalid I am sorry if I am confused.English is not my language. regards Haris Peco
On Wed, 11 Sep 2002, snpe wrote: > yes, we're going around in circles. > > Ok.I agreed (I think because Oracle do different) > Transaction start > I type invalid command > I correct command > I get error > > Why.If is it transactin, why I get error > I want continue. > I am see this error with JDeveloper (work with Oracle, DB2 an SQL Server) Right, that's a separate issue (I alluded to it earlier, but wasn't sure that's what you were interested in). PostgreSQL treats all errors as unrecoverable. It may be a little loose about immediately rolling back due to the fact that historically autocommit was on and it seemed better to not go into autocommit mode after the error. I doubt that 7.3 is going to change that behavior, but a case might be made that when autocommit is off the error immediately causes a rollback and new transaction will start upon the next statement (that would normally start a transaction). At some point in the future, you'll probably be able to do nested transactions or savepoints or error recovery and this will all be moot. > It is not matter for me transaction or not.I get error for correct command > after invalid
On Wednesday 11 September 2002 02:09 am, Stephan Szabo wrote: > On Wed, 11 Sep 2002, snpe wrote: > > yes, we're going around in circles. > > > > Ok.I agreed (I think because Oracle do different) > > Transaction start > > I type invalid command > > I correct command > > I get error > > > > Why.If is it transactin, why I get error > > I want continue. > > I am see this error with JDeveloper (work with Oracle, DB2 an SQL Server) > > Right, that's a separate issue (I alluded to it earlier, but wasn't sure > that's what you were interested in). PostgreSQL treats all errors as > unrecoverable. It may be a little loose about immediately rolling back > due to the fact that historically autocommit was on and it seemed better > to not go into autocommit mode after the error. > > I doubt that 7.3 is going to change that behavior, but a case might be > made that when autocommit is off the error immediately causes a rollback > and new transaction will start upon the next statement (that would > normally start a transaction). > Why rollback.This is error (typing error).Nothing happen. I think that we need clear set : what is start transaction ? I think that transaction start with change data in database (what don't change data this start not transaction. Oracle dot this and I think that is correct)) P.S when I can find SQL 99 specification ? regards
On Wed, 11 Sep 2002, snpe wrote: > On Wednesday 11 September 2002 02:09 am, Stephan Szabo wrote: > > On Wed, 11 Sep 2002, snpe wrote: > > > yes, we're going around in circles. > > > > > > Ok.I agreed (I think because Oracle do different) > > > Transaction start > > > I type invalid command > > > I correct command > > > I get error > > > > > > Why.If is it transactin, why I get error > > > I want continue. > > > I am see this error with JDeveloper (work with Oracle, DB2 an SQL Server) > > > > Right, that's a separate issue (I alluded to it earlier, but wasn't sure > > that's what you were interested in). PostgreSQL treats all errors as > > unrecoverable. It may be a little loose about immediately rolling back > > due to the fact that historically autocommit was on and it seemed better > > to not go into autocommit mode after the error. > > > > I doubt that 7.3 is going to change that behavior, but a case might be > > made that when autocommit is off the error immediately causes a rollback > > and new transaction will start upon the next statement (that would > > normally start a transaction). > > > > Why rollback.This is error (typing error).Nothing happen. Postgresql currently has no real notion of a recoverable error. In the case of the error you had, probably nothing bad would happen if it continued, but what if that was a unique constraint violation? Continuing would currently probably let you see the table in an invalid state. > I think that we need clear set : what is start transaction ? > I think that transaction start with change data in database > (what don't change data this start not transaction. > Oracle dot this and I think that is correct)) I disagree because I think that two serializable select statements in autocommit=off (without a commit or rollback of course) should see the same snapshot. I'm trying to find something either way in a pdf copy of sql99. The multiple row select has gotten hidden somewhere, so it's possible that it's not, but all of opening a cursor, fetching from a cursor and the single row select syntax are labeled as transaction initiating.
On Tue, 10 Sep 2002, Tom Lane wrote: > As of CVS tip, SET commands *do* initiate transactions > if you have autocommit off. By your reading of Date, this is not > spec compliant for certain SET variables: a SET not already within > a transaction should not start a transaction block, at least for the > variables mentioned above. It occurs to me that it'd be reasonable > to make it act that way for all SET variables. I agree. SET variables are normally related to the behaviour of a session, not information stored in the database. And your autocommit example shows why having them start a transaction is a problem. But there were some issues with rolling back and SET commands, weren't there? I remember a long discussion about this that I'm not sure I want to go back to. :-) > > Nor, of course, are the nonexecutable statements DECLARE CURSOR, > > DECLAR LOCAL TEMPORARY TABLE, BEGIN DECLARE SECTION, SEND DECLARE > > SECTIONS, and WHENEVER. > > Hmm. I think the spec's notion of DECLARE must be different from ours. > Our implementation of DECLARE CURSOR both declares and opens the cursor, > and as such it *must* be transaction-initiating; else it's useless. Well, I'm not going to go chase it down right now, but ISTR that DECLAREing a cursor just allocates a variable name or the storage for it or something like that; it doesn't actually create an active cursor. cjs -- Curt Sampson <cjs@cynic.net> +81 90 7737 2974 http://www.netbsd.org Don't you know, in this new Dark Age, we're all light. --XTC
On Tue, 10 Sep 2002, Barry Lind wrote: > I am waiting for this thread to conclude before deciding exactly what to > do for the jdbc driver for 7.3. While using the 'set autocommit true' > syntax is nice when talking to a 7.3 server, the jdbc driver also needs > to be backwardly compatible with 7.2 and 7.1 servers. Can you not check the server's version on connect? It would be ideal if the JDBC driver, without modification, ran all tests properly against 7.3, 7.2 and 7.1. cjs -- Curt Sampson <cjs@cynic.net> +81 90 7737 2974 http://www.netbsd.org Don't you know, in this new Dark Age, we're all light. --XTC
On Tue, 10 Sep 2002, Bruce Momjian wrote: > Do we want to say "With autocommit off, SET will be in it's own > transaction if it appears before any non-SET command", and "SETs are > rolled back except if autocommit off and they appear before any > non-SET"? Not really, I don't think. But I'm starting to wonder if we should re-think all SET commands being rolled back if a transaction fails. Some don't seem to make sense, such as having SET AUTOCOMMIT or SET SESSION AUTHORIZATION roll back. cjs -- Curt Sampson <cjs@cynic.net> +81 90 7737 2974 http://www.netbsd.org Don't you know, in this new Dark Age, we're all light. --XTC
Curt Sampson wrote: > On Tue, 10 Sep 2002, Bruce Momjian wrote: > > > Do we want to say "With autocommit off, SET will be in it's own > > transaction if it appears before any non-SET command", and "SETs are > > rolled back except if autocommit off and they appear before any > > non-SET"? > > Not really, I don't think. > > But I'm starting to wonder if we should re-think all SET commands being > rolled back if a transaction fails. Some don't seem to make sense, such > as having SET AUTOCOMMIT or SET SESSION AUTHORIZATION roll back. Yes, but the question is whether it is better to be consistent and roll them all back, or to pick and choose which ones to roll back. Consistency is nice. -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073
On Tue, 2002-09-10 at 21:44, Curt Sampson wrote: > But there were some issues with rolling back and SET commands, > weren't there? I remember a long discussion about this that I'm > not sure I want to go back to. :-) So.. Unless explicitly requested, a SET command should have immediate effect? The other constrictive value I can think of is search_path. -- Must be transaction safe BEGIN; CREATE SCHEMA <newschema>; SET search_path = <newschema>; ROLLBACK; CREATE TABLE... -- This should be ok BEGIN; SET autocommit = on; INSERT ... COMMIT; -- SET takes place on commit, as it was an explicit transaction -- This is requested behavior SET autocommit = off; SET autocommit = on; INSERT... -- immediate effect, since autocommit is on -- This gets interesting be ok as the schema must exist SET autocommit = off; CREATE SCHEMA <newschema>; SET search_path = <newschema>; -- implicit commit here? ROLLBACK; CREATE TABLE ... -- search_path must roll back or schema must have been created -- Similar to the above SET autocommit = off; CREATE TABLE ... SET autocommit = on; -- implicit commit here? ROLLBACK; -- Does this rollback anything? -- Was CREATE TABLE committed with the second SET statement? > Well, I'm not going to go chase it down right now, but ISTR that > DECLAREing a cursor just allocates a variable name or the storage for it > or something like that; it doesn't actually create an active cursor. Indeed, this is how the cursor is able to cross transactions. It is closed at transaction commit, and re-created in next use. 4.29: For every <declare cursor> in an SQL-client module, a cursor is effectively created when an SQLtransaction (see Subclause 4.32, SQL-transactions ) referencing the SQL-client module is initiated. -- Rod Taylor
Curt, Yes I can check the server version on connect. In fact that is what the driver already does. However I can't check the version and then based on the version call set autocommit true in one round trip to the server. Since many people don't use connection pools, I am reluctant to add the overhead of an extra roundtrip to the database to set a variable that for most people will already be set to true. It would be ideal if I could in one hit to the database determine the server version and conditionally call set autocommit based on the version at the same time. thanks, --Barry Curt Sampson wrote: > On Tue, 10 Sep 2002, Barry Lind wrote: > > >>I am waiting for this thread to conclude before deciding exactly what to >>do for the jdbc driver for 7.3. While using the 'set autocommit true' >>syntax is nice when talking to a 7.3 server, the jdbc driver also needs >>to be backwardly compatible with 7.2 and 7.1 servers. > > > Can you not check the server's version on connect? > > It would be ideal if the JDBC driver, without modification, ran > all tests properly against 7.3, 7.2 and 7.1. > > cjs
On Wed, 11 Sep 2002, snpe wrote: > On Wednesday 11 September 2002 02:09 am, Stephan Szabo wrote: > > On Wed, 11 Sep 2002, snpe wrote: > > > yes, we're going around in circles. > > > > > > Ok.I agreed (I think because Oracle do different) > > > Transaction start > > > I type invalid command > > > I correct command > > > I get error > > > > > > Why.If is it transactin, why I get error > > > I want continue. > > > I am see this error with JDeveloper (work with Oracle, DB2 an SQL Server) > > > > Right, that's a separate issue (I alluded to it earlier, but wasn't sure > > that's what you were interested in). PostgreSQL treats all errors as > > unrecoverable. It may be a little loose about immediately rolling back > > due to the fact that historically autocommit was on and it seemed better > > to not go into autocommit mode after the error. > > > > I doubt that 7.3 is going to change that behavior, but a case might be > > made that when autocommit is off the error immediately causes a rollback > > and new transaction will start upon the next statement (that would > > normally start a transaction). > > > > Why rollback.This is error (typing error).Nothing happen. > I think that we need clear set : what is start transaction ? > I think that transaction start with change data in database > (what don't change data this start not transaction. Another interesting case for a select is, what about select func(x) from table; Does func() have any side effects that might change data? At what point do we decide that the statement needs a transaction?
On Tue, 10 Sep 2002, Barry Lind wrote: > Yes I can check the server version on connect. In fact that is what the > driver already does. However I can't check the version and then based > on the version call set autocommit true in one round trip to the server. > Since many people don't use connection pools, I am reluctant to add > the overhead of an extra roundtrip to the database to set a variable > that for most people will already be set to true. It would be ideal if > I could in one hit to the database determine the server version and > conditionally call set autocommit based on the version at the same time. Hmm. I don't think that there's any real way to avoid a second round trip now, but one thing we might do with 7.3 would be to add a standard stored procedure that will deal with setting appropriate variables and suchlike, and returning the version number and any other information that the JDBC driver needs. (Maybe it can return a key/value table.) That way, once we desupport 7.2 in the far future, we can reduce this to one round trip. Or perhaps we we could try to execute that stored procedure and, if it fails, create it. (Or, if creating it fails, do things the hard way.) That way the first connection you make where the SP is not there you have the overhead of adding it, but all connections after that can use it. (I assume you'd grant all rights to it to the general public.) And it could return its own version so that newer drivers could upgrade it if necessary. Or maybe just have a differently-named one for each version of the driver. This is a bit kludgy, but also sort of elegant, if you think about it.... On the other hand, perhaps we should just live with two round trips. So long as we've got command batching at some point, we can get the version, and then send all the setup commands we need as a single batch after that. cjs -- Curt Sampson <cjs@cynic.net> +81 90 7737 2974 http://www.netbsd.org Don't you know, in this new Dark Age, we're all light. --XTC
On Wednesday 11 September 2002 04:58 am, Stephan Szabo wrote: > On Wed, 11 Sep 2002, snpe wrote: > > On Wednesday 11 September 2002 02:09 am, Stephan Szabo wrote: > > > On Wed, 11 Sep 2002, snpe wrote: > > > > yes, we're going around in circles. > > > > > > > > Ok.I agreed (I think because Oracle do different) > > > > Transaction start > > > > I type invalid command > > > > I correct command > > > > I get error > > > > > > > > Why.If is it transactin, why I get error > > > > I want continue. > > > > I am see this error with JDeveloper (work with Oracle, DB2 an SQL > > > > Server) > > > > > > Right, that's a separate issue (I alluded to it earlier, but wasn't > > > sure that's what you were interested in). PostgreSQL treats all errors > > > as unrecoverable. It may be a little loose about immediately rolling > > > back due to the fact that historically autocommit was on and it seemed > > > better to not go into autocommit mode after the error. > > > > > > I doubt that 7.3 is going to change that behavior, but a case might be > > > made that when autocommit is off the error immediately causes a > > > rollback and new transaction will start upon the next statement (that > > > would normally start a transaction). > > > > Why rollback.This is error (typing error).Nothing happen. > > I think that we need clear set : what is start transaction ? > > I think that transaction start with change data in database > > (what don't change data this start not transaction. > > Another interesting case for a select is, what about > select func(x) from table; > Does func() have any side effects that might change data? > At what point do we decide that the statement needs a > transaction? Function in select list mustn't change any data. What if function change data in from clause ?
> > > Why rollback.This is error (typing error).Nothing happen. > > > I think that we need clear set : what is start transaction ? > > > I think that transaction start with change data in database > > > (what don't change data this start not transaction. > > > > Another interesting case for a select is, what about > > select func(x) from table; > > Does func() have any side effects that might change data? > > At what point do we decide that the statement needs a > > transaction? > Function in select list mustn't change any data. > What if function change data in from clause ? Why can't the function change data? I've done this one a number of times through views to log the user pulling out information from the system, and what it was at the time (time sensitive data). -- Rod Taylor
On Wednesday 11 September 2002 03:14 am, Stephan Szabo wrote: > On Wed, 11 Sep 2002, snpe wrote: > > On Wednesday 11 September 2002 02:09 am, Stephan Szabo wrote: > > > On Wed, 11 Sep 2002, snpe wrote: > > > > yes, we're going around in circles. > > > > > > > > Ok.I agreed (I think because Oracle do different) > > > > Transaction start > > > > I type invalid command > > > > I correct command > > > > I get error > > > > > > > > Why.If is it transactin, why I get error > > > > I want continue. > > > > I am see this error with JDeveloper (work with Oracle, DB2 an SQL > > > > Server) > > > > > > Right, that's a separate issue (I alluded to it earlier, but wasn't > > > sure that's what you were interested in). PostgreSQL treats all errors > > > as unrecoverable. It may be a little loose about immediately rolling > > > back due to the fact that historically autocommit was on and it seemed > > > better to not go into autocommit mode after the error. > > > > > > I doubt that 7.3 is going to change that behavior, but a case might be > > > made that when autocommit is off the error immediately causes a > > > rollback and new transaction will start upon the next statement (that > > > would normally start a transaction). > > > > Why rollback.This is error (typing error).Nothing happen. > > Postgresql currently has no real notion of a recoverable error. > In the case of the error you had, probably nothing bad would happen > if it continued, but what if that was a unique constraint violation? > Continuing would currently probably let you see the table in an > invalid state. > If decision (transaction or not) is after parser (before execute) this isn't problem. I don't know when postgresql make decision, but that is best after parser. I parser find error simple return error and nothing happen > > I think that we need clear set : what is start transaction ? > > I think that transaction start with change data in database > > (what don't change data this start not transaction. > > Oracle dot this and I think that is correct)) > > I disagree because I think that two serializable select statements > in autocommit=off (without a commit or rollback of course) should > see the same snapshot. > Question ? All select in one transaction return same data - no matter if any change and commit data ? > I'm trying to find something either way in a pdf copy of sql99. > The multiple row select has gotten hidden somewhere, so it's possible > that it's not, but all of opening a cursor, fetching from a cursor > and the single row select syntax are labeled as transaction initiating. Can I find sql99 spec anywhere ? Thanks
On Wed, 11 Sep 2002, snpe wrote: > On Wednesday 11 September 2002 04:58 am, Stephan Szabo wrote: > > On Wed, 11 Sep 2002, snpe wrote: > > > On Wednesday 11 September 2002 02:09 am, Stephan Szabo wrote: > > > > On Wed, 11 Sep 2002, snpe wrote: > > > > > yes, we're going around in circles. > > > > > > > > > > Ok.I agreed (I think because Oracle do different) > > > > > Transaction start > > > > > I type invalid command > > > > > I correct command > > > > > I get error > > > > > > > > > > Why.If is it transactin, why I get error > > > > > I want continue. > > > > > I am see this error with JDeveloper (work with Oracle, DB2 an SQL > > > > > Server) > > > > > > > > Right, that's a separate issue (I alluded to it earlier, but wasn't > > > > sure that's what you were interested in). PostgreSQL treats all errors > > > > as unrecoverable. It may be a little loose about immediately rolling > > > > back due to the fact that historically autocommit was on and it seemed > > > > better to not go into autocommit mode after the error. > > > > > > > > I doubt that 7.3 is going to change that behavior, but a case might be > > > > made that when autocommit is off the error immediately causes a > > > > rollback and new transaction will start upon the next statement (that > > > > would normally start a transaction). > > > > > > Why rollback.This is error (typing error).Nothing happen. > > > I think that we need clear set : what is start transaction ? > > > I think that transaction start with change data in database > > > (what don't change data this start not transaction. > > > > Another interesting case for a select is, what about > > select func(x) from table; > > Does func() have any side effects that might change data? > > At what point do we decide that the statement needs a > > transaction? > Function in select list mustn't change any data. > What if function change data in from clause ? There is no such restriction. The behavior is not necessarily well defined in all cases, but postgresql certainly doesn't require that the functions not change data especially given that postgresql takes: select func(); as the way to call to func(); Example session from 7.3 just pre-beta included below. ---- sszabo=# create table b(a int); CREATE TABLE sszabo=# create table a(a int); CREATE TABLE sszabo=# create function f(int) returns int as 'insert into b values ($1); select $1;' language 'sql'; CREATE FUNCTION sszabo=# insert into a values (1); INSERT 17010 1 sszabo=# select f(a) from a;f ---1 (1 row) sszabo=# select * from b;a ---1 (1 row)
On Wednesday 11 September 2002 02:38 pm, Rod Taylor wrote: > > > > Why rollback.This is error (typing error).Nothing happen. > > > > I think that we need clear set : what is start transaction ? > > > > I think that transaction start with change data in database > > > > (what don't change data this start not transaction. > > > > > > Another interesting case for a select is, what about > > > select func(x) from table; > > > Does func() have any side effects that might change data? > > > At what point do we decide that the statement needs a > > > transaction? > > > > Function in select list mustn't change any data. > > What if function change data in from clause ? > > Why can't the function change data? I've done this one a number of > times through views to log the user pulling out information from the > system, and what it was at the time (time sensitive data). Scenario : Func change data in table in form clause I fetch 3 (after row 1 and 2) and then change row 1 What result expect ?
On Wed, 11 Sep 2002, snpe wrote: > On Wednesday 11 September 2002 03:14 am, Stephan Szabo wrote: > > On Wed, 11 Sep 2002, snpe wrote: > > > On Wednesday 11 September 2002 02:09 am, Stephan Szabo wrote: > > > > On Wed, 11 Sep 2002, snpe wrote: > > > > > yes, we're going around in circles. > > > > > > > > > > Ok.I agreed (I think because Oracle do different) > > > > > Transaction start > > > > > I type invalid command > > > > > I correct command > > > > > I get error > > > > > > > > > > Why.If is it transactin, why I get error > > > > > I want continue. > > > > > I am see this error with JDeveloper (work with Oracle, DB2 an SQL > > > > > Server) > > > > > > > > Right, that's a separate issue (I alluded to it earlier, but wasn't > > > > sure that's what you were interested in). PostgreSQL treats all errors > > > > as unrecoverable. It may be a little loose about immediately rolling > > > > back due to the fact that historically autocommit was on and it seemed > > > > better to not go into autocommit mode after the error. > > > > > > > > I doubt that 7.3 is going to change that behavior, but a case might be > > > > made that when autocommit is off the error immediately causes a > > > > rollback and new transaction will start upon the next statement (that > > > > would normally start a transaction). > > > > > > Why rollback.This is error (typing error).Nothing happen. > > > > Postgresql currently has no real notion of a recoverable error. > > In the case of the error you had, probably nothing bad would happen > > if it continued, but what if that was a unique constraint violation? > > Continuing would currently probably let you see the table in an > > invalid state. > > > If decision (transaction or not) is after parser (before execute) this isn't > problem. > I don't know when postgresql make decision, but that is best after parser. > I parser find error simple return error and nothing happen Are you saying that it's okay for: insert into nonexistant values (3); and insert into existant values (3); where 3 is invalid for existant to work differently? I think that'd be tough to get past some people, but you might want to write a proposal for why it should act that way. (Don't expect anything for 7.3, but 7.4's devel will start sometime.) > > > I think that we need clear set : what is start transaction ? > > > I think that transaction start with change data in database > > > (what don't change data this start not transaction. > > > Oracle dot this and I think that is correct)) > > > > I disagree because I think that two serializable select statements > > in autocommit=off (without a commit or rollback of course) should > > see the same snapshot. > > > Question ? > All select in one transaction return same data - no matter if any change and > commit data ? It depends on the isolation level of the transaction I believe. This sequence in read committed (in postgresql) and serializable give different results. T1: begin; T1: select * from a; T2: begin; T2: insert into a values (3); T2: commit; T1: select * from a; In serializable mode, you can't get "non-repeatable read" effects: SQL-transaction T1 reads a row. SQL-transaction T2 then modifies or deletes that row and performs a COMMIT. If T1 then attempts to reread the row, it may receive the modified value of discover that the row has been deleted.
On Wed, 11 Sep 2002, snpe wrote: > On Wednesday 11 September 2002 02:38 pm, Rod Taylor wrote: > > > > > Why rollback.This is error (typing error).Nothing happen. > > > > > I think that we need clear set : what is start transaction ? > > > > > I think that transaction start with change data in database > > > > > (what don't change data this start not transaction. > > > > > > > > Another interesting case for a select is, what about > > > > select func(x) from table; > > > > Does func() have any side effects that might change data? > > > > At what point do we decide that the statement needs a > > > > transaction? > > > > > > Function in select list mustn't change any data. > > > What if function change data in from clause ? > > > > Why can't the function change data? I've done this one a number of > > times through views to log the user pulling out information from the > > system, and what it was at the time (time sensitive data). > Scenario : > Func change data in table in form clause > I fetch 3 (after row 1 and 2) and then change row 1 > What result expect ? Just because the behavior is sometimes undefined by the spec doesn't mean the construct should be disallowed. Grouped character string columns also could have implementation-dependent behavior (which never needs to be specified), but we don't disallow that either.
On Wednesday 11 September 2002 02:55 pm, Stephan Szabo wrote: > On Wed, 11 Sep 2002, snpe wrote: > > On Wednesday 11 September 2002 03:14 am, Stephan Szabo wrote: > > > On Wed, 11 Sep 2002, snpe wrote: > > > > On Wednesday 11 September 2002 02:09 am, Stephan Szabo wrote: > > > > > On Wed, 11 Sep 2002, snpe wrote: > > > > > > yes, we're going around in circles. > > > > > > > > > > > > Ok.I agreed (I think because Oracle do different) > > > > > > Transaction start > > > > > > I type invalid command > > > > > > I correct command > > > > > > I get error > > > > > > > > > > > > Why.If is it transactin, why I get error > > > > > > I want continue. > > > > > > I am see this error with JDeveloper (work with Oracle, DB2 an SQL > > > > > > Server) > > > > > > > > > > Right, that's a separate issue (I alluded to it earlier, but wasn't > > > > > sure that's what you were interested in). PostgreSQL treats all > > > > > errors as unrecoverable. It may be a little loose about > > > > > immediately rolling back due to the fact that historically > > > > > autocommit was on and it seemed better to not go into autocommit > > > > > mode after the error. > > > > > > > > > > I doubt that 7.3 is going to change that behavior, but a case might > > > > > be made that when autocommit is off the error immediately causes a > > > > > rollback and new transaction will start upon the next statement > > > > > (that would normally start a transaction). > > > > > > > > Why rollback.This is error (typing error).Nothing happen. > > > > > > Postgresql currently has no real notion of a recoverable error. > > > In the case of the error you had, probably nothing bad would happen > > > if it continued, but what if that was a unique constraint violation? > > > Continuing would currently probably let you see the table in an > > > invalid state. > > > > If decision (transaction or not) is after parser (before execute) this > > isn't problem. > > I don't know when postgresql make decision, but that is best after > > parser. I parser find error simple return error and nothing happen > > Are you saying that it's okay for: > insert into nonexistant values (3); > and > insert into existant values (3); > where 3 is invalid for existant to work > differently? > I think that'd be tough to get past some people, but you might > want to write a proposal for why it should act that way. (Don't > expect anything for 7.3, but 7.4's devel will start sometime.) > I don't understand all, but when I tell 'error' I think "syntax error" If error is contraint error again nothin change, only error return > > > > I think that we need clear set : what is start transaction ? > > > > I think that transaction start with change data in database > > > > (what don't change data this start not transaction. > > > > Oracle dot this and I think that is correct)) > > > > > > I disagree because I think that two serializable select statements > > > in autocommit=off (without a commit or rollback of course) should > > > see the same snapshot. > > > > Question ? > > All select in one transaction return same data - no matter if any change > > and commit data ? > > It depends on the isolation level of the transaction I believe. > This sequence in read committed (in postgresql) and serializable give > different results. > > T1: begin; > T1: select * from a; > T2: begin; > T2: insert into a values (3); > T2: commit; > T1: select * from a; > > In serializable mode, you can't get "non-repeatable read" effects: > SQL-transaction T1 reads a row. SQL-transaction T2 then modifies > or deletes that row and performs a COMMIT. If T1 then attempts to > reread the row, it may receive the modified value of discover that the > row has been deleted. If serialization strict connect with transaction then ok. haris peco
On Wed, 11 Sep 2002, snpe wrote: > On Wednesday 11 September 2002 02:55 pm, Stephan Szabo wrote: > > On Wed, 11 Sep 2002, snpe wrote: > > > > > > If decision (transaction or not) is after parser (before execute) this > > > isn't problem. > > > I don't know when postgresql make decision, but that is best after > > > parser. I parser find error simple return error and nothing happen > > > > Are you saying that it's okay for: > > insert into nonexistant values (3); > > and > > insert into existant values (3); > > where 3 is invalid for existant to work > > differently? > > I think that'd be tough to get past some people, but you might > > want to write a proposal for why it should act that way. (Don't > > expect anything for 7.3, but 7.4's devel will start sometime.) > > > I don't understand all, but when I tell 'error' I think "syntax error" > If error is contraint error again nothin change, only error return I don't understand what you mean here. Are you saying that both of those queries should not start transactions? Then that wouldn't be starting between the parser and execute since you won't know that the row violates a constraint until execution time. > > > > I disagree because I think that two serializable select statements > > > > in autocommit=off (without a commit or rollback of course) should > > > > see the same snapshot. > > > > > > Question ? > > > All select in one transaction return same data - no matter if any change > > > and commit data ? > > > > It depends on the isolation level of the transaction I believe. > > This sequence in read committed (in postgresql) and serializable give > > different results. > > > > T1: begin; > > T1: select * from a; > > T2: begin; > > T2: insert into a values (3); > > T2: commit; > > T1: select * from a; > > > > In serializable mode, you can't get "non-repeatable read" effects: > > SQL-transaction T1 reads a row. SQL-transaction T2 then modifies > > or deletes that row and performs a COMMIT. If T1 then attempts to > > reread the row, it may receive the modified value of discover that the > > row has been deleted. > If serialization strict connect with transaction then ok. I again am not sure I understand, are you saying that under serializable select should start a transaction but it shouldn't under read committed? That seems like a bad idea to me, either it should or it shouldn't in my opinion. Perhaps it'd be better if you wrote up what you think it should do in all these cases and then we could look at them as a whole. (Cases I can see right now are, select under serializable, select under read committed, garbage command, select to non existant table, insert to non existant table, insert that fails due to unique constraint, insert that fails due to exception raised by a before trigger, insert that fails due to exception raised by an after trigger, insert that does nothing due to before trigger, update that fails due to any of those after some rows have already successfully been modified and probably some others).
On Wednesday 11 September 2002 06:11 pm, Stephan Szabo wrote: > On Wed, 11 Sep 2002, snpe wrote: > > On Wednesday 11 September 2002 02:55 pm, Stephan Szabo wrote: > > > On Wed, 11 Sep 2002, snpe wrote: > > > > If decision (transaction or not) is after parser (before execute) > > > > this isn't problem. > > > > I don't know when postgresql make decision, but that is best after > > > > parser. I parser find error simple return error and nothing happen > > > > > > Are you saying that it's okay for: > > > insert into nonexistant values (3); > > > and > > > insert into existant values (3); > > > where 3 is invalid for existant to work > > > differently? > > > I think that'd be tough to get past some people, but you might > > > want to write a proposal for why it should act that way. (Don't > > > expect anything for 7.3, but 7.4's devel will start sometime.) > > > > I don't understand all, but when I tell 'error' I think "syntax error" > > If error is contraint error again nothin change, only error return > > I don't understand what you mean here. Are you saying that both of > those queries should not start transactions? Then that wouldn't > be starting between the parser and execute since you won't know that > the row violates a constraint until execution time. > > > > > > I disagree because I think that two serializable select statements > > > > > in autocommit=off (without a commit or rollback of course) should > > > > > see the same snapshot. > > > > > > > > Question ? > > > > All select in one transaction return same data - no matter if any > > > > change and commit data ? > > > > > > It depends on the isolation level of the transaction I believe. > > > This sequence in read committed (in postgresql) and serializable give > > > different results. > > > > > > T1: begin; > > > T1: select * from a; > > > T2: begin; > > > T2: insert into a values (3); > > > T2: commit; > > > T1: select * from a; > > > > > > In serializable mode, you can't get "non-repeatable read" effects: > > > SQL-transaction T1 reads a row. SQL-transaction T2 then modifies > > > or deletes that row and performs a COMMIT. If T1 then attempts to > > > reread the row, it may receive the modified value of discover that the > > > row has been deleted. > > > > If serialization strict connect with transaction then ok. > > I again am not sure I understand, are you saying that under serializable > select should start a transaction but it shouldn't under read committed? > That seems like a bad idea to me, either it should or it shouldn't in > my opinion. > > Perhaps it'd be better if you wrote up what you think it should do in > all these cases and then we could look at them as a whole. > (Cases I can see right now are, select under serializable, select under > read committed, garbage command, select to non existant table, > insert to non existant table, insert that fails due to unique constraint, > insert that fails due to exception raised by a before trigger, > insert that fails due to exception raised by an after trigger, > insert that does nothing due to before trigger, update that fails > due to any of those after some rows have already successfully been > modified and probably some others). One question first ? What mean ? ERROR: current transaction is aborted, queries ignored until end of transaction block I am tried next (autocommit=true in postgresql.conf) 1. begin; 2. select * from tab; query work 3. show t; -- force stupid syntax error 4. select * from tab; ERROR: current transaction is aborted, queries ignored until end of transaction block 5.end; 6. select * from tab; query work I must rollback or commit transaction when I make stupid syntax error. This is same with autocommit=false It is maybe ok, I don't know. For rest is ok (if level serializable select start transaction) Thanks
On Wed, 11 Sep 2002, snpe wrote: > On Wednesday 11 September 2002 06:11 pm, Stephan Szabo wrote: > > > I again am not sure I understand, are you saying that under serializable > > select should start a transaction but it shouldn't under read committed? > > That seems like a bad idea to me, either it should or it shouldn't in > > my opinion. > > > > Perhaps it'd be better if you wrote up what you think it should do in > > all these cases and then we could look at them as a whole. > > (Cases I can see right now are, select under serializable, select under > > read committed, garbage command, select to non existant table, > > insert to non existant table, insert that fails due to unique constraint, > > insert that fails due to exception raised by a before trigger, > > insert that fails due to exception raised by an after trigger, > > insert that does nothing due to before trigger, update that fails > > due to any of those after some rows have already successfully been > > modified and probably some others). > > One question first ? > > What mean ? > ERROR: current transaction is aborted, queries ignored until end of > transaction block > I am tried next (autocommit=true in postgresql.conf) The transaction has encountered an unrecoverable error (remember, all errors are currently considered unrecoverable) and the transaction is in a potentially unsafe state. > 1. begin; > 2. select * from tab; > query work > 3. show t; -- force stupid syntax error > 4. select * from tab; > ERROR: current transaction is aborted, queries ignored until end of > transaction block > 5.end; > 6. select * from tab; > query work > > I must rollback or commit transaction when I make stupid syntax error. Note that even with end you get effectively a rollback in this case since the transaction as a whole ended in an error state. > This is same with autocommit=false > It is maybe ok, I don't know. Well, at least until we have savepoints or nested transactions, there's only a limited amount of freedom in the implementation. > For rest is ok (if level serializable select start transaction) Like I said above, having the transaction starting of select being dependent on the isolation level variable sounds like a bad idea. In addition that still doesn't deal with select statements with side effects.