[E-Lang] Why Not Coroutines? (was: Deadlock-free [was
what is good about E?])
Thu, 19 Jul 2001 21:33:34 -0700
Mark S. Miller writes:
>At 07:02 PM Thursday 7/19/01, Richard wrote:
>>You seem to be saying that coroutines can be used to implement user
>>threading. No. *Continuations* (which are provided, eg, by Scheme and
>>Stackless Python) can be used to implement user threading. You mean
>You imply that it's possible to provide coroutines without providing
>continuations. That may be, but I find it hard to imagine. What would the
>API of these coroutine primitives look like?
consider a structured-programming construct
such that "cor" <list of statements> "roc" is a <statement>
(sort of like "begin" ... "end").
then for example
while true do
end of while
would print "a" first time it was run, "b" second time, etc. execution
of a "resume" of course causes the thread of control to exit the
enclosing "cor"-"roc" construct. sort of like "return"
I'm not advocating such a construct, just exhibiting a way of coding up
a coroutine without explicitly naming a continuation, as you requested.
I did what you asked, right?
a saner example API is given on page 316 of Essentials of Prgmng Languages.
that API consists of procs "resume" and "makecoroutine" and is implemented
in terms of call/cc. I bet other Scheme books have it, too