http://source.lukas-renggli.ch/continuations/Generator-lr.3.mcz

Lukas

On 10 February 2010 06:49, Colin Putney <cputney at wiresong.ca> wrote:
>
> On 2010-02-09, at 9:27 PM, Andreas Raab wrote:
>
>> Folks -
>>
>> One thing I really like about Python is Generators (http://docs.python.org/tutorial/classes.html#generators). They allow code to provide a stream (iterator) interface where it's not easy to provide the interface explicitly. I.e., it would allow one to provide a stream interface to an arbitrary computation like:
>>
>> gen := Generator on: [:yield| 3 to: 1 by: -1 do:[:i| yield value: i]].
>> gen next. "=>3"
>> gen next. "=>2"
>> gen next. "=>1"
>> gen next. "=>nil"
>>
>> This *ought* to be really simple to do in Smalltalk by using blocks and some voodoo in the execution engine. Has anyone ever done that?
>
> Generators are essentially coroutines, which makes them pretty easy to implement in Smalltalk - thisContext is all you need. They've been implemented a few times to my knowledge. GST has an implementation that is pretty similar to what Python provides:
>
> http://smalltalk.gnu.org/faq/47
>
> Also, Stephen Pair posted an implementation here in December:
>
> http://n4.nabble.com/ANN-CoroutineReadStream-tt963004.html#a963004
>
> I've done stuff like this for testing. My MockLibrary package provides a class called Mocket ("mock socket") that uses coroutines to make it easier to test network code. It's not a "generator" exactly, but it does simulate reading data off a socket by jumping back into the test case and letting it generate the data.
>
> Colin
>



-- 
Lukas Renggli
http://www.lukas-renggli.ch