Re: Is there a clever way to null out non-void subs?

PS in Advance: I have a bad feeling my mailer just sent out a huge whole 
               of “Out of processes” bounces.  I hope this doesn’t get me
               unsubscribed.  I really am still here.

Zefram <zefram@fysh.org> wrote on Wed, 22 Apr 2015 10:49:39 BST: 

> Debug::Show uses precisely this trick, via B::CallChecker.  D:S itself
> may satisfy your requirement, and if not then it's a model for you to
> follow in writing whatever you actually need.

Awsome. It does. That’s the model I was looking for.

>> Plus you have the problem of wasted evaluation of somebody(42).

> To be clear, the nulling that Debug::Show performs does avoid evaluation
> of the argument expressions.  It excises the entire sub call optree.

Thanks very much, that’s just EXACTLY what I was looking for!  

I hadn’t realized those Perl-based APIs existed, and was a bit queasy that
I might end up having to write it all in XS instead.

There should be something coming out “soon” from LARRYL@CPAN that incorporates
this idea into a more generic flavor of an Exporter class.

The whole point is to be able to use it for other modules whose calls you 
essentially want to #ifdef away at compile time.  This is trivial in C—and
rather not so in Perl.

We have several modules we want to do this with.  For example, this way you
can have a bunch of different types of asserts that don’t cost anything
once they hit production.

The problem is that NYTProf said that 40-45% of my code execution was being
taken up by my very anal^H^H^H^Hreligious assert tests all over.  Those are 
indispensable in development and test, and serve for great documentation.
But they were costing a lot of CPU no matter how effeciently I wrote them.
Stuff like these:

    _assert_all_ascii                  _assert_list_context
    _assert_argc                       _assert_named_parameters
    _assert_argc_max                   _assert_nonnegative_integer
    _assert_argc_min                   _assert_nonnumeric
    _assert_child_status               _assert_nonref
    _assert_class_invocant             _assert_no_wide_characters
    _assert_defined                    _assert_numeric
    _assert_has_nonascii               _assert_numeric_range
    _assert_is_directory               _assert_object_invocant
    _assert_is_identifier              _assert_private_method
    _assert_is_integer                 _assert_protected_method
    _assert_is_opened_handle           _assert_public_method
    _assert_is_plainfile               _assert_reftype
    _assert_is_positive_integer        _assert_strlen                      

There are others.  Possibly I could have tried converting those to XS 
if I hadn’t found a way to “#ifdef them away” so to speak, which your
technique now allows me to do.

Thanks again.

--tom

• Is there a clever way to null out non-void subs? by Tom Christiansen
• Re: Is there a clever way to null out non-void subs? by Eric Brine
• Re: Is there a clever way to null out non-void subs? by Zefram
• Re: Is there a clever way to null out non-void subs? by Tom Christiansen
• Re: Is there a clever way to null out non-void subs? by Aristotle Pagaltzis
• Re: Is there a clever way to null out non-void subs? by Peter Martini
• Re: Is there a clever way to null out non-void subs? by Ricardo Signes