Re: core.async: Unbound channels

[Ernesto Garcia]

Thanks Alex!

Correct, the channel implementation takes care that "transduced" channels 
always pass elements through the transducer and the buffer. Also, a 
FixedBuffer allows running out of limit for those cases, see this example 
with a FixedBuffer of size 1 making space for 4 elements:

(def c (chan 1 (mapcat seq)))
=> #'psdk.hack-config/c

(put! c "hola")
=> true

(-> c .buf .buf)
=> (\a \l \o \h)

However, the blocking semantics of a channel change if a buffer is 
enforced. You can't have a "transduced" channel for which all puts will 
block for a take. Not sure if this is too limiting for any practical case.

For the case of expanding transducers, I am not sure if there would be 
sensible semantics for channel operations without a buffer.

-- 
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
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/clojure/a652870b-59c2-40f9-b4e0-7387e87de204%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: core.async: Unbound channels

[Alex Miller]

Expanding transducers (like mapcat) can produce multiple output values per 
input value, and those have to have someplace to go.

-- 
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
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/clojure/1be8a455-f7c9-4e79-8906-d49ab6a09aef%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: core.async: Unbound channels

[Ernesto Garcia]

I see. Bufferless channels are meant to be used within the core.async 
threading architecture, where there will be a limited number of blocked 
puts and takes. At the boundaries, channels with dropping or sliding 
windows can be used for limiting work.

So, my original question actually turns into: Why do channels with a 
transducer need a buffer? Is it just a limitation of the implementation, or 
does it have a conceptual reason behind 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
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/clojure/a4201a4b-7d84-4457-a857-ed6e252f6b0e%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: core.async: Unbound channels

[Matching Socks]

"Effective" is in the eye of the beholder.  The 1024 limit helps surface 
bugs wherein more than a thousand threads are blocked for lack of a certain 
channel's buffer space.  But the 1024 limit does not pertain if 1 thread 
would like to do thousands of puts for which there is no buffer space.  In 
the latter case, that thread is in a loop that does 1 put at a time, and 
therefore only 1 put, at any given moment, counts against the 1024 limit.

-- 
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
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/clojure/844b72c4-71a3-43a9-b3bc-41c56922f3b2%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: core.async: Unbound channels

[Ernesto Garcia]

On Thursday, July 4, 2019 at 4:24:33 PM UTC+2, Matching Socks wrote:
>
> Ernesto, you may be interested in the informative response to this 
> enhancement request, https://clojure.atlassian.net/browse/ASYNC-23 
> ,
>  
> "Support channel buffers of unlimited size".
>

An example: Here a function that makes a request, and returns a channel 
with responses with the same referenceId as the request. The function taps 
a mult-channel, but I don't now what buffer to use.I would expect that the 
original mult-channel already implements some buffering limits.

(defn request [client events-mult-channel req]
  (let [ref-id (.getReference req)
result-channel (chan 16 (filter #(= ref-id (.getReference %]
(tap events-mult-channel result-channel)
(.send client req)
result-channel))

 

> Anyway, if you do not want to think very hard about buffer size, you can 
> specify a size of 1.  It does not limit the number of items the producer 
> may put on the channel, but it affects how soon (and how often) the 
> producer's puts must wait for a consumer.
>

I see, puts start to block when the buffer is full or when no buffer. Isn't 
the 1024-limit on blocked puts an effective limit, like a buffer size?

-- 
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
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/clojure/205bb066-7014-41ec-b552-153b1331143c%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: core.async: Unbound channels

[Matching Socks]

Ernesto, you may be interested in the informative response to this 
enhancement request, https://clojure.atlassian.net/browse/ASYNC-23, 
"Support channel buffers of unlimited size".  Anyway, if you do not want to 
think very hard about buffer size, you can specify a size of 1.  It does 
not limit the number of items the producer may put on the channel, but it 
affects how soon (and how often) the producer's puts must wait for a 
consumer.

-- 
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
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/clojure/f79addd3-794d-4a6a-ba45-550a2b75e389%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: core.async: Unbound channels

[Ernesto Garcia]

Thanks for your response, it is important to know. (Sorry for my lexical 
typo: *unbound**ed*. I didn't realize it derives from the verb *bound*, not 
*bind*!)

My question on channel boundaries still holds though. Why the enforcement 
of boundaries *always*?

On Wednesday, July 3, 2019 at 5:16:31 PM UTC+2, Ghadi Shayban wrote:
>
> (chan) is not a channel with an unbounded buffer. It is a channel with 
> *no* buffer and needs to rendezvous putters and takers 1-to-1.  
> (Additionally it will throw an exception if more than 1024 takers or 
> putters are enqueued waiting)
>
> On Wednesday, July 3, 2019 at 7:14:46 AM UTC-4, Ernesto Garcia wrote:
>>
>> You can create a unbound channel with (chan), but not if you use a 
>> transducer; (chan nil (filter odd?)) will raise an error that no buffer 
>> is provided. Why is this the case?
>>
>> Why the enforcement of all channels to be bound? In a program, there will 
>> be channels that propagate to other channels, so only channels at the 
>> boundaries would require to be bound?
>>
>> Channel limits are also much dependent on the particular process and 
>> environment. How would we write generic code that creates channels, if 
>> those need to be bound to limits unknown?
>>
>> Thanks,
>> Ernesto
>>
>

-- 
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
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/clojure/7f9f5efe-883f-47e4-b2a5-a457fb7444ed%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: core.async: Unbound channels

[Ghadi Shayban]

(chan) is not a channel with an unbounded buffer. It is a channel with *no* 
buffer and needs to rendezvous putters and takers 1-to-1.  (Additionally it 
will throw an exception if more than 1024 takers or putters are enqueued 
waiting)

On Wednesday, July 3, 2019 at 7:14:46 AM UTC-4, Ernesto Garcia wrote:
>
> You can create a unbound channel with (chan), but not if you use a 
> transducer; (chan nil (filter odd?)) will raise an error that no buffer 
> is provided. Why is this the case?
>
> Why the enforcement of all channels to be bound? In a program, there will 
> be channels that propagate to other channels, so only channels at the 
> boundaries would require to be bound?
>
> Channel limits are also much dependent on the particular process and 
> environment. How would we write generic code that creates channels, if 
> those need to be bound to limits unknown?
>
> Thanks,
> Ernesto
>

-- 
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
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/clojure/12cf0bf8-dafd-429b-a2d2-3875f579f1c1%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

core.async: Unbound channels

[Ernesto Garcia]

You can create a unbound channel with (chan), but not if you use a 
transducer; (chan nil (filter odd?)) will raise an error that no buffer is 
provided. Why is this the case?

Why the enforcement of all channels to be bound? In a program, there will 
be channels that propagate to other channels, so only channels at the 
boundaries would require to be bound?

Channel limits are also much dependent on the particular process and 
environment. How would we write generic code that creates channels, if 
those need to be bound to limits unknown?

Thanks,
Ernesto

-- 
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
--- 
You received this message because you are subscribed to the Google Groups 
"Clojure" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to clojure+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/clojure/0fb4746a-ff0a-457a-99fd-8763ab526efd%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.