[ruby-talk:22030] Re: Marshal won't dump a Proc

In article <[e-mail]>, "Yukihiro
Matsumoto" <[e-mail]> wrote:


> Procs, Bindings and Continuations contain references to C stack
> information which is not portable.  Methods have references to C
> function.  All of above information is not portable.

Thanks to matz and Will for those explanations.  The Proc I wanted to
dump/restore is a block that's provided when an instance of my class is
created.  It specifies how to get the nth element of a sequence and hence
can be pretty much anything.  Hence, asking the user to write it as a
string that's going to be eval-ed would make using my class a little
messy.

I've provided a mechanism whereby the user can ask for the current set of
values to be saved between sessions.  This is useful for sequences where
the calculation time is significant, eg generating prime numbers.

I then thought it would be nice to be able to store the Proc that defined
how to get the next element, so then you could load one of these
sequences without having to have the original code.  Ie, the dump file
would carry all the information required to continue from where the last
program that used that sequence left off.

I can see now that storing a Proc would be much more complex than I'd
originally thought.  While the block of code I want to save is generally
going to be self-contained, I guess there's no way for the serialisation
of a Proc object to know that.

In any case, I guess that even the simplest piece of code, eg "{ |i| i
}", uses Object and serialising that could presumably be quite complex,
especially since someone could have used the dynamic features of ruby to
modify Object by adding or changing methods.

Nothing's ever as simple as it seems, is it :-) ??

What I might be able to do is provide two means of constructing the
object, one by attaching a block and the other by passing a string that's
eval-ed into a Proc for run-time use and dumped/restored  for persistence.

Thanks again for the lucid explanations.