On Sat, 4 Jun 2016, Frank Vanoni wrote:> Another possible approach to blacklist two (or more) specific callers > (098765432 and 012345678 as example) > > exten => _+x.,1,Gosub(blacklist,s,1) > exten => _+x.,n,.... > exten => black,1,playback(tt-monkeys) > > In blacklist.conf > > exten => s/098765432,1,Goto(black,1) > exten => s/012345678,1,Goto(black,1)Using a 'goto' to exit from a gosub is a bad idea. A better idea would be to set a channel variable and check it's value after the return, in the calling context. Also, can a 'goto' in a subroutine reference an extension in the calling context? Seems weird, but 'dialplan' is a weird language :) -- Thanks in advance, ------------------------------------------------------------------------- Steve Edwards sedwards at sedwards.com Voice: +1-760-468-3867 PST https://www.linkedin.com/in/steve-edwards-4244281
On Sat, 2016-06-04 at 15:19 -0700, Steve Edwards wrote:> Using a 'goto' to exit from a gosub is a bad idea.Why?> A better idea would be > to set a channel variable and check it's value after the return, in the > calling context.The idea is to update the blacklist.conf whenever I want to add or remove a specific number or an entire area code and leave the extensions.conf untouched and to avoid complex regular expressions.> Also, can a 'goto' in a subroutine reference an extension in the calling > context? Seems weird, but 'dialplan' is a weird language :)Well... I'm not an expert and my approach is by "trial and error". It works perfectly. :-) Frank
On Mon, 6 Jun 2016, Frank Vanoni wrote:> On Sat, 2016-06-04 at 15:19 -0700, Steve Edwards wrote: > >> Using a 'goto' to exit from a gosub is a bad idea. > > Why?The purpose of a subroutine (code that is entered by a gosub and exited by a return) is to allow the creation of easily reusable code. It 'packages' complex or frequently used code into a nice little black box. You don't need to know all the details of what happens in the box and you can use this little box throughout your code without having to tell it where to continue when it has finished -- it just 'knows' by virtue of it's implementation. A 'gosub' is implemented (in most languages) by pushing the address of the instruction following the gosub onto the stack. When the return is executed, this address is popped off the stack and loaded into the program counter. Using a 'goto' to exit from a subroutine does not 'pop' off the return address. If this cycle of pushing addresses and not popping off addresses is repeated enough, you may run out of stack space. Think how complex and difficult to maintain your dialplan would be if you had to tell each application (dial(), playback(), set(), etc) where to go when it was finished. Even worse, imagine if each application had a fixed "I'm finished" address (priority or label). Most programmers consider 'goto' to be 'evil' because it allows (encourages?) lazy design leading to difficult to maintain/reuse code. (Google 'why is goto bad'.) Some languages do not even include a 'goto' by design to encourage programmers to write better code.>> A better idea would be to set a channel variable and check it's value >> after the return, in the calling context. > > The idea is to update the blacklist.conf whenever I want to add or > remove a specific number or an entire area code and leave the > extensions.conf untouched and to avoid complex regular expressions.The idea is fine. The implementation is flawed. It should be implemented as a subroutine (or AGI) and return the success or failure as a channel variable. This will result in an 'easier to comprehend' and more maintainable dialplan. This 'design pattern' (a subroutine) would allow you to reuse this same 'black box' in other parts of your dialplan. Think of 'the next guy' -- which may be you in a couple of months when the 'finer details' of your implementation fade. If you jump all around your dialplan it gets very hard to comprehend. If you can see that you execute a little black box and then do something based on an intuitively named channel variable the design and intent is obvious.> Well... I'm not an expert and my approach is by "trial and error". It > works perfectly. :-)A 'better' approach is to learn from the mistakes of others. I suspect it 'works perfectly' until it's been running long enough to cause difficult to diagnose problems. -- Thanks in advance, ------------------------------------------------------------------------- Steve Edwards sedwards at sedwards.com Voice: +1-760-468-3867 PST https://www.linkedin.com/in/steve-edwards-4244281