Re: [Parseperl-discuss] PPI + Storable = ?

[Adam K. <ad...@ph...>]

Yes.

PPI uses a weakref-based tree system.

Although the objects themselves are built using a normal non-recursive=20
structure, they look UP the tree by using the %PPI::Element::PARENT=20
hash, which contains a weak reference to their parent Node keyed by the=20
Scalar::Util::refaddr of the child object.

There's an entry in PARENT for every single child Element object.

So far as I'm aware, this is pretty much the ideal way of implementing=20
trees structures in Perl. It's fast as hell (and faster once we get a XS=20
version in PPI::XS), is completely immune from circular reference leaks,=20
and it even cleans itself properly when PPI::Document objects are=20
destroyed. It also handily makes everything look like plain trees in the=20
debugger.

The DESTROY method for each element cleans up it's entry in PARENT,=20
keeping the hash from exploding in size.

However, whenever you manipulate PDOM tree you have to keep the PARENT=20
refs up to date. Most of the time this is done for you transparently.

Take a look at the implementation of PPI::Document::clone.

It does a fast Clone::clone, and then walk the PDOM tree relinking all=20
of the children to the parents in the cloned copy.

In order to make PPI::Documents natively Storable-friendly, we probably=20
need a Storable Hook method to do the same sort of relinking process=20
when we thaw out the document.

Write me a quick 17_storable.t test script (just copy one of the other=20
smaller test scripts) and attach it to a rt.cpan.org bug report, and=20
I'll handle the implementation for you. I probably should refactor that=20
bit a little anyway so it uses the same code as ->clone.

Adam K

Johan Lindstr=F6m wrote:
> I'm trying to cache parsed PPI::Document objects by serializing them=20
> with Storable, but I see strange things. The objects before/after look=20
> the same, but e.g. a $node->statement doesn't work with the thawed obje=
ct.
>=20
> Is there any known reason it wouldn't "just work" to freeze/thaw a=20
> PPI::Document with it's nodes etc?
>=20
>=20
> /J
>=20
>=20
>=20
> -------------------------------------------------------
> This SF.Net email is sponsored by the 'Do More With Dual!' webinar=20
> happening
> July 14 at 8am PDT/11am EDT. We invite you to explore the latest in dua=
l
> core and dual graphics technology at this free one hour event hosted by=
=20
> HP, AMD, and NVIDIA.  To register visit http://www.hp.com/go/dualwebina=
r
> _______________________________________________
> Parseperl-discuss mailing list
> Par...@li...
> https://lists.sourceforge.net/lists/listinfo/parseperl-discuss