[Haskell-cafe] Inputs to classic FRP: unsafeInterleaveIO/unsafePerformIO

[Heinrich Apfelmus]

Edward Amsden wrote:
> As far as I can tell, with classic FRP implementations (those which
> use behaviors as a first-class abstraction), the only way to create a
> behavior or
> event based on some external input (for instance keypresses or
> microphone input) is to do something with unsafePerformIO or
> unsafeInterleaveIO. A behavior is a value, which when evaluated at a
> specific time would have to either block its evaluation until input
> could be read, or check the input at that particular time.
> 
> Is there any other way of implementing external behaviors besides that?

Yes, there are other ways, see for example the implementation here: 
http://tinyurl.com/frp-automaton . This is essentially a pure variant of 
Ryan's implementation.

My implementation has a serious problem, namely that sharing is lost. I 
think this is the case for Ryan's implementation as well. The state of a 
behavior will be duplicated and updates multiple times. This can be 
fixed by observing sharing, of course.

I'm currently working on a push-driven FRP implementation. (Though I'm 
getting second thoughts as to whether the small increase in efficiency 
is worth the implementation cost.) See also

http://apfelmus.nfshost.com/blog/2011/04/24-frp-push-driven-sharing.html



Best regards,
Heinrich Apfelmus

--
http://apfelmus.nfshost.com