Re: [stack] Generators (was: Concatenative Hardware)

Sorry for not changing the subject line earlier...

On Nov 2, 2007, at 19:26 , William Tanksley, Jr wrote:

> Don Groves <[email protected]> wrote:
>> A generator is a persistent function. When activated, it pushes a
>> copy of
>> itself on the stack. The copy performs the generative function and
>> remains
>> on the stack until specifically dropped. Each time it comes to the
>> top of the stack it generates a value according to its definition.
>
> "Comes to the top of the stack" is unclear. It would make more sense
> to say that every time GENERATE is called on it, it produces the next
> value. (You could also overload UNCONS to work that way... but that
> seems dangerous.)

Good point! I could just use the I or DIP combinator to execute a  
generator.
This would make it more generally useful.


>> Example: [1 2 ...] defgen Z+
>> Z+ is now a generator a copy of which will generate positive  
>> integers as
>> long as it remains active (until dropped from the stack).
>
> As long as you're defining new syntax, why not have "[x ...]" be
> defined in the language as a "generator quotation", so you don't need
> 'defgen'? Just use 'const' or whatever.

That was my original design and I'll return to that, thanks!


>> stack:
>> => Z+
>> yields -
>> stack:[2 3 ...] 1
>> =>
>
>> The next time it comes to the top of the stack, it will generate a 2,
>> thusly -
>> stack:[3 4 ...] 2
>> =>
>
> Specifically, does this mean that:
>
> => Z+ SWAP SWAP
> yields -
> stack:1 2 [4 5 ...] 3
>
> ?
>
> If so, is there any way to drop a generator from the stack?

I was thinking of having a dropY instruction, but then that couldn't be
defined as swap drop could it?  This problem is also solved by using
a combinator to execute a generator.


>> This behaviour continues until the user/program decides to stop it.
>
> How?
>
>> Thus,
>> the ellipsis generator, ..., can mimic an infinite list. Z+ is
>> included in the
>> ACL standard library.
>
> Obviously the ellipsis generator is either very clever with lots of
> special cases, or it can only handle additive sequences.

That depends on the definition. [0 -1 ...] would generate negative
sequences, for example. I could also include an operator in the  
definition
and make it more group-like. Nothing in ACL is cast in concrete yet  
which
is one reason I appreciate your (and other's) comments and suggestions
so much!


>> ... and so on until 99 is produced, at which point the generator copy
>> self-destructs.
>
> Without being dropped? How will client code know that their stack's
> been changed?

One thing I forgot to mention is that a finite generator acts like an  
empty
list after generating its last value so a null value will be  
generated when
the client code asks for the next value. I suppose I could require  
the client
code to drop the generator at that point...
--
Don


>
>> Don Groves
>
> -Billy
>
>
>
> Yahoo! Groups Links
>
>
>
>