[663] TAP/Parser.pm - next_nonblock() method

[Eric Wilhelm]

# from Andy Armstrong
# on Wednesday 03 October 2007 02:03:

>> TAP/Parser.pm                  - next_nonblock() method
>> TAP/Parser/Iterator/Process.pm - is_ready() to support nonblocking
>
>Please don't. We already have a way of polling the iterators using  
>select.

But that requires you to reach-around the Parser to get the handles out 
of the Process object, then stick them in your own Select object ala 
T::P::Multiplexer.pm.  While I'll grant that this runs faster, I would 
hate for that to be the only way to have a nonblocking parser.

Granted that next_nonblock() should not assume a select object in the 
base class, but this requires a bigger change to the Parser, not just 
reverting my change and saying "we don't need it."

In Multiplexer, you've had to add duplicate code to support Source 
objects which have no filehandle.  That says to me that the API needs 
to be normalized.  Even if Multiplexer doesn't want to do it *that* way 
for speed reasons, having no next_nonblock() requires more 
documentation than having it, thus making T::H harder to use where 
blocking is a concern.

--Eric
-- 
"Because understanding simplicity is complicated."
--Eric Raymond
---------------------------------------------------
    http://scratchcomputing.com
---------------------------------------------------