[asterisk-dev] PREDIGITS for WaitExten

[Tilghman Lesher]

On Tuesday 01 April 2008 09:55:29 Timo Teräs wrote:
> Tilghman Lesher wrote:
> > On Tuesday 01 April 2008 03:08:51 Timo Teräs wrote:
> >> This way the overlapped dialling would work properly over IAX. E.g. if
> >> all digits are not received, the WaitExten() would wait for additional
> >> digits; and not just ignore the already received digits, what it does
> >> now by default. (Eg. in some cases the PBX might have sent 1230 in the
> >> first place. Without the _123! the call would be dropped with invalid
> >> extension. With only WaitExten(); it would ignore the first received
> >> zero.)
> >
> > You can do this already with dialplan logic:
> >
> > exten => _123000,1,Verbose(2,000 selected)
> > exten => _123001,1,Verbose(2,001 selected)
> > exten => _123[1-9]!,1,Dial(Zap/g1/${EXTEN:3})
> > exten => _123!,1,Set(PREDIGITS=${EXTEN})
> > exten => _123!,n,Set(origcontext=${CONTEXT})
> > exten => _123!,n,WaitExten(getmoredigits)
> >
> > exten => 123456,1,Verbose(2,Something else)
> >
> > [getmoredigits]
> > exten => 456,1,Goto(${origcontext},${PREDIGITS}${EXTEN},1)
>
> This would not work. WaitExten does not take context.

Oh, sorry.  But you can do something very similar...
...
exten => _123!,n,Goto(getmoredigits,s,1)

[getmoredigits]
exten => s,1,WaitExten(10)
exten => _X!,1,Goto(${origcontext},${PREDIGITS}${EXTEN},1)
exten => t,1,Goto(${origcontext},i,1)

Same idea, different code.  This will continue until your original context
decides that the extension is complete or invalid.

> It's way cleaner to get WaitExten accept PREDIGITS. Or alternatively I
> could write a new application that basically is
> WaitAdditionalDigitsForExten that would wait for additional digits until
> another extension matches or timeout comes.

I really don't think WaitExten needs to be altered in that way.

> The point is that in above example. If PBX first sent '123' it would
> go to WaitExten. Then when additional (overlapped) digit of '1' would come
> it would immediate match the _123[1-9]! rule and dial out. Or if the
> overlapped digit was '0' it would wait for more digits and getting another
> two of '0's would then match 123000. And the same rules would also work
> if in the first place was received a full number like "1239321" or
> "123000".
>
> So I would want to make _123! act as if it is known that _123 is a known
> prefix, but it needs more digits to make a routing decision.

The way I would implement this is as an application which returned a special
value to indicate that the extension was not yet complete and the pbx routines
in main/pbx.c should continue waiting for more digits.  Currently, the only
valid return values are -1, 0, '*', '#', and '0' through '9'.  Perhaps you
could return 128 (specifically chosen, since it's the first positive integer
after the full set of ASCII) to indicate that the extension is incomplete and
needs more digits.

> Sort of what chan_zap does when overlapdial=yes and immediate=no. But it
> should also work when the call is coming from IAX (or any other) context.
>
> >> I have a rough patch for this (I didn't test it), but it gives a general
> >> idea what is needed to achieve this. If you think this would be
> >> acceptable, I can make tested patch against 1.6 branch and/or trunk (and
> >> sign the license stuff + send the patch to bugs db).
> >
> > I can't look at your patch until it's posted on the bugtracker, with the
> > appropriate contributor license filed.
>
> Ok. I'll test it a bit, maybe polish it and send it on bugtracker. I signed
> the license earlier today.

Looking forward to seeing it.

-- 
Tilghman