www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - Flagshit concurrency

reply Gilbert Dawson <why needed.com> writes:
Hello

I'm new to D. I've been studying new languages for new concurrency ideas. I'd
like to implement a small and lightweight Lighttpd replacement and replace the
scripting with MiniD. One has heard in many places that D supports a new idea
of "flagship message passing". So I'd like to use this with async i/o and one
threaded server process. Is it suitable for this?
May 15 2011
parent reply Robert Clipsham <robert octarineparrot.com> writes:
On 15/05/2011 10:08, Gilbert Dawson wrote:

 Hello

 I'm new to D.


Welcome :)


 I've been studying new languages for new concurrency
 ideas. I'd like to implement a small and lightweight Lighttpd
 replacement and replace the scripting with MiniD. One has heard in
 many places that D supports a new idea of "flagship message passing".


You've hit your first roadblock here - MiniD is written in D1/Tango, the 
concurrency model you speak of is D2 only (not that it couldn't be 
implemented in D1). Your options are then - port MiniD to D2 - that will 
take a while, it's heavily dependent on Tango, which is D1 only 
(currently, D2 support is underway it seems); Use another scripting 
language that has D bindings; Use D - dmd is fast enough that you could 
in fact use D as a scripting language.


 So I'd like to use this with async i/o and one threaded server
 process. Is it suitable for this?


I believe someone is working on async i/o as a GSoC project. I don't 
believe support for it exists in Phobos currently - I could be wrong. 
Again, you could use bindings to C for this.

-- 
Robert
http://octarineparrot.com/
May 15 2011
next sibling parent Nick_B <nick.NOSPAMbarbalich gmail.com> writes:
On 16/05/2011 5:49 a.m., Robert Clipsham wrote:

 On 15/05/2011 10:08, Gilbert Dawson wrote:

 Hello

 I'm new to D.


 Welcome :)


 I've been studying new languages for new concurrency
 ideas. I'd like to implement a small and lightweight Lighttpd
 replacement and replace the scripting with MiniD. One has heard in
 many places that D supports a new idea of "flagship message passing".


 You've hit your first roadblock here - MiniD is written in D1/Tango, the
 concurrency model you speak of is D2 only (not that it couldn't be
 implemented in D1).


if you want to learn d1 & tango I suggest that you get the book:

http://www.amazon.com/Learn-Tango-Macleod-Kris-Bell/dp/1590599608

and the community can be found here:

www.dsource.org/projects/tango

Tango is in the process of being ported to D2.

cheers
Nick
May 16 2011
prev sibling parent Jonas Drewsen <jdrewsen nospam.com> writes:
On 15/05/11 19.49, Robert Clipsham wrote:

 On 15/05/2011 10:08, Gilbert Dawson wrote:

 Hello

 I'm new to D.


 Welcome :)


 I've been studying new languages for new concurrency
 ideas. I'd like to implement a small and lightweight Lighttpd
 replacement and replace the scripting with MiniD. One has heard in
 many places that D supports a new idea of "flagship message passing".


 You've hit your first roadblock here - MiniD is written in D1/Tango, the
 concurrency model you speak of is D2 only (not that it couldn't be
 implemented in D1). Your options are then - port MiniD to D2 - that will
 take a while, it's heavily dependent on Tango, which is D1 only
 (currently, D2 support is underway it seems); Use another scripting
 language that has D bindings; Use D - dmd is fast enough that you could
 in fact use D as a scripting language.


 So I'd like to use this with async i/o and one threaded server
 process. Is it suitable for this?


 I believe someone is working on async i/o as a GSoC project. I don't
 believe support for it exists in Phobos currently - I could be wrong.
 Again, you could use bindings to C for this.


Noone is working on the async I/O project for GSoC unfortunately since 
we did not get enough GSoC slots assigned.

Message passing is mostly used for multithreaded use. Since your app is 
single threaded I do not believe message passing is the way to go. A 
pure async reactor based design would probably give the best performance.

/Jonas
May 16 2011