develooper Front page | perl.perl5.porters | Postings from September 2016

Re: pounding the smartmatch drum

Thread Previous | Thread Next
September 9, 2016 12:30
Re: pounding the smartmatch drum
Message ID:
Aristotle Pagaltzis wrote:
>Then again, if we split LHS from RHS overloads and LHS overloads become
>rare then that problem should be substantially contained.

It wouldn't be all that rare.  Marking the feature as being only for
autodie's use wouldn't work, especially if there's no intent to actually
remove it later.  And even if it is only used by autodie, that doesn't
render it contained.

If there are lhs overloads, then all the matching rules that are subject
to them are just attractive nuisances.  In Tony's plan I think that
would be all except for the one with an overloaded object on the rhs.
Since your objection to breaking code using given($@) with autodie
applies to any use of given, presumably you're actually arguing for us
to make no change at all to smartmatching semantics, in which case there
are even more.

>        anyone just reading the autodie POD would even today not know to
>assume anything amiss with the given($@) interface.

That's true.  They'll learn that there's a problem quite soon, once they
run some code following that pattern, by virtue of the "experimental"
warning.  But there's effectively a conflict between the warning and
the autodie doc.  Even if we don't resolve smartmatch now, we should at
least put something in the autodie and autodie::exception documentation
warning about the unstable nature of the mechanism.


Thread Previous | Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About