[resent because of my other email address is not subscribed - bah]
[EMAIL PROTECTED] wrote:
>> Is this what you were thinking of?
>
>No. At least, not until B::Deparse works on 100% of perl code. And of
>course, the browser needs to come back to the user and query for the
>names of the arguments and of the method, replace the code with a
>method call, preferably using named parameters, and generally make the
>place look nice.
I think the problem I was trying to solve was more along the lines of how could I
write something to turn a piece of code that relies on globals into a method that
relies on lexicals, without writing a parser.
I'm just playing around now with comparing chunks of the optree to the original code,
and attempting to get enough info back from the
optree to figure out where I need to place the method call - theory being that while
the perl source code may change its variable names fairly significantly, the optree
looks more or less the same.
Of course any milage that I may gain here is subject to extremely hairy conditions in
the first place. Of course at the moment its just
comparing the output to B::Terse - but seeing as B::Terse is a textual
representation of the optree, what works with one should work with the
other.
>And 'Make method' is just one of the refactorings that a good
>refactoring browser should do, need things like 'rename method', 'push
>method to superclass'...
Absolutly - I'm looking at starts rather than ends :-)
>If I can get it to work it's going to be *so* cool.
The other thing I'd really like is a proper class browser. I've hacked together class
browsers in the past, but they always are just a hack, although I've not done anything
since B::Deparse became more clever.
>However, that is a *really* cool approach. Took a while to work out
>what was going on mind, but still, very cool.
It is pretty hairy. What I meant above was 'let Perl worry about
parsing perl, and let me worry about errors that are generated'.
Essentially that is the approach - Damian (uses|used) a similar method for parsing
arguments to attributes in Attribute::Handlers.
>Now, if Robin can just get B::Deparse working all the time... But I
>don't think we'll be seeing that until 5.10...
I've not read p5p for a little while. I was under the impression that
code spat out by B::Deparse was passing any test that wasn't reliant on order of BEGIN
blocks (and therefore things like pragmata where a little bit squew whiff[0]).
Regards,
James
[0] I've always loved this expression. I've never seen it spelled
however, so this may be very wrong.
