Re: analogy pmc/content with inode/file

At 12:06 PM +0200 10/27/04, Leopold Toetsch wrote:
> Stéphane Payrard <stef@xxxxxxxxxxx> wrote:
> >  I want to make an loose analogy between pmc/their_content and
> >  Unix inodes/files which I use as a mnemonic.
>
> >  PMCs are the equivalent of inodes, each one
> >  reference some content which is the equivalent of a file.
>
> >  C<set_pmc>, when the source and the destination pmcs are of the same type,
> >  is like creating a hard_link to a file. The two pmcs share the
> >  same content.
>
> ,--[ pdd02_vtables.pod ]------------------------------------------
> |       void set_pmc(INTERP, PMC* self, PMC* value)
> |            Sets the value of the PMC in self to the value of the
> |            PMC in value.
> |
> |        void assign_pmc(INTERP, PMC* self, PMC* value)
> |            Sets the value of the PMC in self to the value of the
> |            PMC in value by copying the value.
> `-----------------------------------------------------------------
>
> So probably yes. OTOH having e.g. two array PMCs pointing to the same
> array structure could be rather error prone.
>
> I think, we have to consider first variables vs values. Variables have a
> name and hold values (the PMCs).

Not quite. The PMC is the variable, not the 
value. The contents of the PMC are the value.

> It depends on the HLL, what e.g. an
> assignment is doing:

These aren't all assignment, and the chosen punctuation is really unfortunate.

>   i = j   # Python - both variables refer to the same value

This isn't assignment. This is name rebinding -- 
that is, two names in some namespace refer to the 
same PMC

>   i += k  # mutable objects are changed in place
>           # immutable objects are rebound to hold the new value

This is one of those bizarre mutant cases -- it's really a method call.

>   @a = @b # Perl - copy array

This is assignment, sort of. If you squint hard, 
given the stuff perl's doing behind the scenes.

In all these it's important to note that in most 
languages (certainly the ones we care about) 
objects are *reference* types, not value types. 
That is, there's a level of indirection (likely 
an ObjectReference or just generic PMCReference 
PMC) between the name and the ultimate object 
structure, something we don't do right now.
--
                                Dan

--------------------------------------it's like this-------------------
Dan Sugalski                          even samurai
dan@xxxxxxxxx                         have teddy bears and even
                                      teddy bears get drunk