Re: Record Lock

[Booth Martin <booth@xxxxxxxxxxxx>]

I believe your problem may originate with the order in which you are 
doing things?  Get the existing number, add 1 to it, and immediately 
update the file with the new number, thereby unlocking the record.  Then 
display it.

If the problem is that the user may cancel out, you can always chain to 
the file, get the new existing number (make sure it is the same as the 
one on the canceled screen), sub 1, and update the file.

As a policy I never leave a file open for a screen program.  Too many 
users go to lunch or get a phone call and someone always gets burned, in 
a non-predictable way.



steema@xxxxxxxxxxxxx wrote:
> There is one program that generates the check number. It is a simple RPG
> that chains the file by company only, adds 1 to the existing number and
> places this on the screen for the user.
>
> I have put in a chain right there to ensure that this new number does not
> already exist, and I have also added a log file to this program.
>
> There is also another program that does checks which I am checking now but
> I beleive it only calls the first program.
>
> I was told that for checks, they don't indicate the division ever, only
> company.
>
> I still want to know what happens to the file lock, when a term. freezes.
>
> Thank you,
>
> Steve
> > > Now I have found out, that someone here thinks it occurs when the users
> > > terminal gets LOCKED. But first of all, wouldn't a record lock remain in
> > > effect or would the CL program that holds the lock in effect deallocate
> > > this lock? yet the check number started to change, did change? its a bit
> > > confusing, but it makes some possible sense.
> > Record locks have nothing at all to do with the creation of the check
> > number.  Furthermore, _object_ locks (ALCOBJ) have nothing to do with
> > the contents of any particular record.  It's hard to tell what you mean
> > by 'terminal gets locked.'  Sounds like a program failure of some sort.
> >
> >  > I am looking at a situation that occassionaly
> >  > occurs, that a check number is allowed to be
> >  > entered that is a dup.
> >
> >  > the key is CO, Div, batch#, check#,
> >  > so you could have the same number and a diff batch.
> >
> > If the key is Co, Div, Batch, Check, then the database designer is
> > explicitly allowing for duplicate check numbers, as long as they occur
> > in a different company or division or batch.
> >
> >  > Or to put it another way, how would you
> >  > design such a program in such a
> >  > way that a dup could NEVER happen?
> >
> > My choice would be to create a new logical file keyed by check number,
> > unique.  That way the database enforces your business rule (no duplicate
> > check numbers.)  However, as Lim warns, this could ripple through the
> > application, causing problems you don't want if there are already
> > 'duplicate' check numbers in the file.
> >
> > As Jerry asked, where do these check numbers come from?  Does the
> > application generate them, or are they input from somewhere (like a
> > customer payment batch?)  If your application generates the check
> > number, there are two tried and true mechanisms for guaranteeing a
> > unique number.  Data area and control file.
> >
> > *lock in dtaara    // locks dtaara; nobody else can read
> > add 1 check
> > out dtaara
> >
> > 'check' chain ctlfile   // locks record; nobody else can chain
> > add 1 check
> > update ctlfile
> >
> >    --buck
> > --
> > This is the RPG programming on the AS400 / iSeries (RPG400-L) mailing list
> > To post a message email: RPG400-L@xxxxxxxxxxxx
> > To subscribe, unsubscribe, or change list options,
> > visit: http://lists.midrange.com/mailman/listinfo/rpg400-l
> > or email: RPG400-L-request@xxxxxxxxxxxx
> > Before posting, please take a moment to review the archives
> > at http://archive.midrange.com/rpg400-l.