SPUG: Re: IO::All

[Michael R. Wolf]

ced at carios2.ca.boeing.com writes:

> Hello everyone,
>
> I missed Brian's talk unfortunately but was there any mention
> of IO::All's error handling...  was there a demo of subclassing 
> throw() for instance?

In fact, there was a discussion of error handling.  I don't want to
put words in Ingy's mouth, but it went along the lines of...

    "Error handling?  We don't need no stinkin' error handling!"

Humor and playfulness, aside -- I'd call it "error handling lite", or
"go ahead and let it die, if anyone cares, they'll wrap it with an
eval and catch it".

    eval {
      # Ingy code with no error handling goes here...
      }
    warn "Ingy-instigated errors $@" if $@;

I had a brief discussion with Jack about it on a ride home, basically
because Jack had asked a question about it during the talk. Much of
the discussion centered around a historical error-handling conundrum

-- propogate errors up the call stack and let callers check return
   values (pessimistic -- check every call)

-- throw an exception and have callers ignore or propogate it up the
   call stack to anyone who cares (optomistic -- assume sucess, catch
   errors at the "right" level, and ignore them when you can)

-- 
Michael R. Wolf
    All mammals learn by playing!
        MichaelRWolf at att.net