On Friday, February 6, 2015 at 11:34:30 AM UTC-6, Jason Bertsche wrote:
We've definitely discussed the idea of bag/multiset support in
NetLogo before, but, as was already noted in this thread, it would
be a substantial thing to introduce. In this particular
situation, it seems like the use case for multisets is not
terribly difficult to work around, but I'd be interested to hear
about other cases in which people might find multisets useful. If
there's enough need for it, multiset support could /possibly/ be
added, but my intuition is that the need didn't seem to arise in
the first 15 years of NetLogo's history, so what would cause it to
arise now?
Maybe it's significant that it wasn't until two years ago that there
was enough motivation for someone (Nicolas) to introduce the *rnd*
extension with more flexible agent selection options. To me those
functions seem very valuable, but the NetLogo community got by without
them for a long time. (I didn't need one until yesterday!)
Then again, maybe there's been a qualitative change, in recent years,
in the NetLogo community and the uses to which NetLogo is being put.
(An emergent phenomenon!) So maybe the past isn't a reliable
indicator of future needs.
I suspect it might also be opposed on grounds of complicating the
language.
I can easily understand that it would complicate the source code, and
that that might be enough to make it not worth adding "agentbags".
Does it complicate the NetLogo language, though, if agentset
primitives knew how to deal with "agentbags". Maybe there's something
I'm not seeing.
Either way, I think it's an interesting idea worth discussing.
On 02/06/2015 11:18 AM, Bryan Head wrote:
Glad you think that will work for you. I should note that if you
want to reproduce the random ordering behavior of `ask` and
`with`, just put `shuffle` before `agent-list`. For instance
`foreach shuffle agent-list [ ... ]` and `map [...] shuffle
agent-list`. You'll take a performance hit, but then they will
behave more like the agentset primitives, and you won't see
artifacts cropping up from using the same order every time.
On Fri Feb 06 2015 at 11:11:13 AM Marshall <[email protected]
<javascript:>> wrote:
Thanks Bryan--That's a good solution. Hadn't thought of it.
(Seems obvious, now!) I'll probably do that. And I
appreciate knowing that uniqueness is deeply embedded in the
nature of agentsets.
Roman, you're right--I should have called that what I was
proposing an "agentmultiset".
On Friday, February 6, 2015 at 10:30:05 AM UTC-6, Bryan Head
wrote:
Hi Marshall,
Writing you're own ask-procedure that operates on lists
could be a pretty easy workaround. For instance:
to ask-list [ agent-list commands ]
foreach agent-list [ ask ? [ run commands ] ]
end
which you then call like:
ask-list agents-with-repeats task [ do-stuff ]
Where `agents-with-repeats` is your list of agents. Note
the `task` primitive is unfortunately required. Besides
that, this should pretty much be a drop-in replacement
for `ask` after you switch to using a list. `of` could be
similarly transformed:
to-report of-list [ agent-list reporter ]
report map [ [ runresult reporter ] of ? ] agent-list
end
Called like: of-list agents-with-repeats task [
turtle-variable ]
Now `with`:
to-report with-list [ agent-list predicate ]
report filter [ [ runresult reporter ] of ? ] agent-list
end
Called like: with-list agents-with-repeats task [
turtle-variable = 5 ]
Besides reordering arguments and requiring `task`, these
should pretty much be drop-in replacements for their
agentset counterparts. As you said, the uniqueness of
agents in agentsets is quite baked in.
Hope that helps!
Bryan
On Fri Feb 06 2015 at 10:00:48 AM Marshall
<[email protected]> wrote:
A model I'm working on includes a series of functions
that implement a random choice of turtles that will
send messages to another turtle. Recently, I decided
it might be better to allow the selection of senders
to be random with repeats. Nicolas Payette's rnd
extension <https://github.com/NetLogo/Rnd-Extension>
provides a convenient function that provides this
functionality, returning a list that may contain
repeats:|||weighted-n-of-with-repeats|. (Thanks Nicolas!)
However, converting a list of turtles with repeats
into a turtleset loses the repeats; agentsets contain
only unique elements. So if I want to allow repeats
in the turtles that send messages, I have to rewrite
a small but significant bit of code in different
functions, replacing ask's with loops, etc.
Question: Might it be useful to allow a new kind of
agentset that allows repeats? It would be useful to
me in this situation, but I know that the idea
violates longstanding assumptions about agentsets,
and I suspect that it would also require a lot of
changes to the NetLogo source to implement.
I thought I'd raise it as a question, anyway, to see
what others think.
--
You received this message because you are subscribed
to the Google Groups "netlogo-devel" group.
To unsubscribe from this group and stop receiving
emails from it, send an email to
[email protected].
For more options, visit
https://groups.google.com/d/optout
<https://groups.google.com/d/optout>.
--
You received this message because you are subscribed to the
Google Groups "netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from
it, send an email to [email protected]
<javascript:>.
For more options, visit https://groups.google.com/d/optout
<https://groups.google.com/d/optout>.
--
You received this message because you are subscribed to the
Google Groups "netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it,
send an email to [email protected] <javascript:>.
For more options, visit https://groups.google.com/d/optout
<https://groups.google.com/d/optout>.
--
You received this message because you are subscribed to the Google
Groups "netlogo-devel" group.
To unsubscribe from this group and stop receiving emails from it, send
an email to [email protected]
<mailto:[email protected]>.
For more options, visit https://groups.google.com/d/optout.
