On Mon, Oct 27, 2003 at 04:50:18PM +0100, Steffen Goeldner wrote:
> Tim Bunce wrote:
> >
> > On Fri, Oct 24, 2003 at 11:00:55AM +0200, Steffen Goeldner wrote:
> > > Is there any value in returning a value (pun intended)
> > > from STORE? E.g.:
> > >
> > > sub STORE {
> > > ...
> > > return 1;
> > >
> > > For dbd_db_STORE_attrib(), DBI::DBD states:
> > >
> > > The return value is TRUE if you have handled the
> > > attribute or FALSE otherwise.
> >
> > That's needed so the Driver.xst code knows if it should
> > then call SUPER::STORE.
> >
> > > Does the same rule apply to STORE()? If so, where
> > > is the returned value used?
> >
> > The only way for an application to get the return value is by calling
> > STORE explicitly. I guess that might be useful in some situations.
>
> Three cases cross my mind:
>
> 1) Return nothing (or an arbitrary value), i.e. the caller shouldn't
> use that value.
>
> 2) Return a boolean value, indicating success or failure.
>
> 3) Return the new value.
>
> Tim, can you settle on one of these cases?
I think it has to be 1 unless the driver docs indicate otherwise.
> If so, I'd like to include your recommendation into DBI::DBD.
Thanks.
> BTW: Case 1, something like that:
>
> sub STORE {
> ...
> $dbh->SUPER::STORE($attr, $val);
> return;
> }
>
> seems to collide with current practice:
>
> sub begin_work {
> ...
> $dbh->STORE('BegunWork', 1);
> }
>
> $rc = $dbh->begin_work or die $dbh->errstr;
That's a bug. I've changed begin_work to return 1 instead of falling
off the end.
> P.S.: Funny, a similar question entered the p5p mailing-list
>
> <http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2003-10/msg01249.html>
Here's the proof from a trace (with a few lines deleted):
$ DBI_TRACE=2 perl -MDBI -e '$a = DBI->install_driver("mysql")->{foo}=1'
-> DBI->install_driver(mysql) for freebsd perl=5.006001 pid=24349 ruid=307 euid=307
<- install_driver= DBI::dr=HASH(0x81ba9b0)
-> STORE in DBD::_::common for DBD::mysql::dr (DBI::dr=HASH(0x810709c)~INNER 'foo'
1)
<- STORE= '' at -e line 1
-> FETCH in DBD::_::common for DBD::mysql::dr (DBI::dr=HASH(0x810709c)~INNER 'foo')
<- FETCH= undef at -e line 1
note that the FETCH is skipped in a void context:
$ DBI_TRACE=2 perl -MDBI -e 'DBI->install_driver("mysql")->{foo}=1'
-> DBI->install_driver(mysql) for freebsd perl=5.006001 pid=24348 ruid=307 euid=307
<- install_driver= DBI::dr=HASH(0x81ba9b0)
-> STORE in DBD::_::common for DBD::mysql::dr (DBI::dr=HASH(0x81070b4)~INNER 'foo'
1)
<- STORE= '' at -e line 1
I should add some notes about that to the docs. It has interesting
knock-on issues in relation to a) testing an attribute assignment,
and b) performance impact.
Thanks.
Tim.
