[LLVMdev] Iterator protocols

[Talin]

This is related to the general question of efficiency of unwinds. I'm 
mulling over whether to use the Java-style or Python-style iterator 
protocol for my language. The Python style is to have a special 
exception (StopIteration) that is thrown when the end of the sequence is 
reached. The Java style is to have a separate "hasNext" method on the 
iterator object that says whether or not the sequence is finished.

So the question is, what's the trade-off. In most languages that support 
exceptions, you tend to think of exceptions as expensive operations that 
should only be thrown if something truly "exceptional" happens. OTOH, 
the Java case is also made worse by the fact that a large part of the 
time you'll be using the more expensive interface dispatching, rather 
than simple vtable dispatching.

I would imagine that for smaller sequences the Java protocol wins 
(because of the expense of unwinding), whereas for longer sequences the 
per-item overhead dominates. The question is, what's the cutoff point 
likely to be - how expensive is an unwind compared to, say, a regular 
function call. (I haven't tried to measure it because I'm not yet sure 
what I am measuring.) I'm just curious if anyone has an opinion on this.

-- Talin