Named code blockes

[Alex Martelli]

"D-Man" <dsh8290 at rit.edu> wrote in message
news:mailman.988076385.11192.python-list at python.org...
> On Mon, Apr 23, 2001 at 04:49:01PM -0700, James_Althoff at i2.com wrote:
> |
> | <jim response>
> |
> | In Java unnamed (anonymous, inner) classes
> | are used all over the place in Swing-GUI code.
> | Not advocating -- just pointing it out.  :-)
> |
> | </jim response>
>
> IMO that is a kludge to work-around for a significant language
> deficiency -- if functions (and classes) were first-class objects the
> function object itself could be passed as the event handler!

It seems that when one CAN does something, one DOES do it:-).

I have never agreed with the Borland/Microsoft position that
the lack of first-class callables is a "huge gaping hole" in
Java (the same excellent software architect who did Delphi
for Borland then moved to Microsoft to do similar frameworks
for them -- first for Java enriched with first class callables
called "delegates", then for C# who also has such constructs --
may be responsible for both firms' "position" in this:-).  I
think first-class callables are nice-to-have, and I'm glad
Python and C++ give them to me, but when I worked in Java I
didn't really feel deprived at all, particularly not after
they introduced inner classes.

OTOH, leaving inner-class instantiation anonymous always rather
grated on me.  In my limited experience tutoring/mentoring others
in Java, I kept seeing misunderstandings based on that -- people
who thought they were instantiating an interface, rather than an
anonymous inner class implementing that interface -- not to
mention the trouble explaining to debuggers what anonymous class
you actually MEANT:-).  My employer ended up not pursuing Java
seriously enough to warrant coding guidelines, else I would have
pushed for naming to be recommended here.

I do have an example where I'm _exceedingly_ glad I can leave
something unnamed, although it *DOES* keep giving debugger
problems -- instantiations of C++ templates, particularly
method-templates of a class, and PARTICULARLY when the compiler
is deducing argument types on my behalf.  In this case, if I
had to give names to the instantiations, the type-deduction
work of the compiler would be for naught... so there is a
very specific reason for "leaving nameless"!  (It _does_ give
problems too, but, here, having to name WOULD substantially
diminish power -- so, there IS a tradeoff of some sort).


Alex