The first thing I ever implemented with promises was dataflow
programming in this style. A good read is the chapter on declarative
concurrency in Peter van Roy's Concepts, Techniques and Models of
Computer Programming. Oz has the concept of concurrent
single-assignment variables, which are much the same as promises.
All you need is this:
(defn deref-seq [x]
(proxy [clojure.lang.ASeq] []
(first ([] (first @x)))
(next ([] (next @x)))))
This works with anything that derefs to a seqable value. For example,
here's how you'd implement your own version of lazy-seq in terms of
the delay form:
(defmacro lazy-seq* [& body]
`(deref-seq (delay (do ~...@body))))
Here's a basic example with promise sequences:
user> (def p (promise))
#'user/p
user> (future (doseq [x (promise-seq p)] (println "New element:" x)))
#<core$future_call$reify__2...@dbb638: :pending>
user> (def q (promise))
#'user/q
user> (do (deliver p (cons 1 (promise-seq q))) nil)
nil
;; 'New element: 1' is printed by future.
user> (do (deliver q (cons 2 nil)) nil)
nil
;; 'New element: 2' is printed by future.
(In case you're wondering, the (do ... nil) is intended to prevent the
REPL from printing (and hence implicitly dereferencing) the tail
promises. This is annoying quirk of the current printer
implementation. When I first played with promise sequences, I changed
print-method for promises so this doesn't happen.)
With that in place, you can connect a pair of threads with a matched
pair of promises. Indeed, it's easy enough to abstract away most of
the scaffolding for dataflow programming with promise sequences so it
looks much as it does in a language like Oz.
-Per
On Thu, Apr 15, 2010 at 7:10 PM, alux <alu...@googlemail.com> wrote:
> Hello Anniepoo,
>
> thank you for the answer. I t made me think in new directions,
> splendid!
>
> Its streams, I completely agree. What I still try is mapping this to
> the ubiquitous Clojure-sequences. And what I came up with is seeing
> the stream as blocking lazy sequence of strings (the lines), that can
> then be transformed into blocking lazy sequence of messages (that may
> be multiline, I didn't mentioned that before).
>
> And now I'm not sure whether the blocking permits to be a sequence. As
> far as I understand, blocking (in the Clojure world) should be
> restricted to references. And if head and tail of a "seq" are
> references, they dont fit the interface anymore, or do they.
>
> Still thinking ;-)
>
> Kind regards, alux
>
> On 14 Apr., 22:41, Anniepoo <annie6...@yahoo.com> wrote:
>> what you want is just a stream in each direction.
>>
>> Bob: So, how do you feel about the Smith contract?
>> Fred: It's not a good deal for us
>>
>> Fred: Do you think we'll have 4 engineers working on it?
>> Bob: Turn left here
>> Bob: No, more like 5
>>
>> Fred: at the light?
>> Bob: Yes
>>
>> I'd make a wrapper around a socket connection. If you want an async
>> call,
>> like Fred's last question "at the light?" then have a different send
>> access that takes two args,
>> (ask-bob "at the light?" (fn [response] ... handle the
>> response....))
>>
>> as opposed to
>> (tell-bob "blah blah")
>> that doesn't get a response
>> (after all, you presumably need to know what question Bob is answering
>> 'yes' to)
>>
>> ask-bob sends bob an id along with the question, and shoves the
>> function into a map with id as key
>> when bob sends a response the function gets called
>>
>> The recieving end is a good candidate for a multimethod.
>> the dispatch function would somehow parse the message for type
>> and figure out which handler handles it.
>
> --
> You received this message because you are subscribed to the Google
> Groups "Clojure" group.
> To post to this group, send email to clojure@googlegroups.com
> Note that posts from new members are moderated - please be patient with your
> first post.
> To unsubscribe from this group, send email to
> clojure+unsubscr...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/clojure?hl=en
>
> To unsubscribe, reply using "remove me" as the subject.
>
--
You received this message because you are subscribed to the Google
Groups "Clojure" group.
To post to this group, send email to clojure@googlegroups.com
Note that posts from new members are moderated - please be patient with your
first post.
To unsubscribe from this group, send email to
clojure+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/clojure?hl=en
