Re: RFC 180 (v1) Object Class hooks into C<printf>

Hildo Biersma wrote:
> 
> > =head1 ABSTRACT
> >
> > There needs to be a way for an object class to define C<printf> format
> > specifiers for use in formatting objects into strings with C<printf> and
> > C<sprintf>.
> 
> I find myself agreeing with your sentiment, but the approach in this RFC
> is not sufficiently general.  Why only provide hooks for printf, not for
> formats and output disciplines?
> 
> I think a better approach would be that the to-string operator for
> objects should get an optional 'width' parameter.  In normal cases
> (stringification, interpolation), that argument would not be set.  In
> printf and formats, but maybe also by specific file disciplines (e.g.
> the 72-character-wide output file), the width parameter would be set and
> indicate how wide the object may print its data.

Thanks you, I had forgot about formats.  Note that more then just width 
information is needed in the general case (and I should add more
examples
to the RFC to show this).  Take the set of printf specifiers that would
be needed for the Math::BigFloat package.  You would want to define %f,
%e
and %g at least with all the decimal point placement, padding and sign
stuff.  Even formats allow for 'basic print using' picture like fields
(which I believe get turned into printf specifiers internally) that have
more
information then just width.  In addition I intended that arbitrary
object
classes could define unique to the class printf specifiers that contain
anything the writer wants to define between the % and the final letter.

I will write a second version of the RFC over the weekend to address
these issues.

--
Mark Biggar
[email protected]