[REBOL] Re: Object "Serialization"
From: agem:crosswinds at: 23-Feb-2001 0:30
>>>>>>>>>>>>>>>>>> Ursprüngliche Nachricht <<<<<<<<<<<<<<<<<<
Am 22.02.01, 23:32:53, schrieb "GS Jones" <[gjones05--mail--orion--org]> zum
Thema [REBOL] Re: Object "Serialization":
> > > > In a separate thread, a question was asked about serialization of
> > > > objects by Sharriff Aina. I had a similar question in mind and
> was
> > > > hoping for an answer. Since the thread had gotten a bit off
> topic,
> > I'm
> > > > reposting essentially the same question in this separate thread.
> > > >
> > > > To date, when I want to save an object, I've been using 'save.
> Then
> > > > when I need the object back, I use 'read and then 'do the object.
> > > >
> > > > For example:
> > > >
> > > > data: make object! [
> > > > name: "Joe"
> > > > ]
> > > >
> > > > save %my-object.r data
> > > >
> > > > ;... later i need the object ...
> > > > data: read %my-object.r
> > > > data: do data
> > > > print data/name
> > >
> > > use 'load instead of 'read? Then you don't need to 'do the 'data ...
> > > -pekr-
> >
> > Thanks, that is a good point!
> > Scott
> Actually, when I tried it, simply using 'load did not re-initialize
the
> object (unless I am missing some additional subtle point). I played
> with several additional variations, with the same conclusions. So I
> guess the way I've been doing it will have to be good enough. It
still
> seems as though there should be a more seemless and coherent way to
> serialize the object.
data: do %my-object.r
before that you have like
>> a: [make object! []]
== [make object! []]
>> a/1
== make
>> a/2
== object!
So load gives »source«, do »compiles« it.
Note 'save of strings has some bugs before last /core-experimental,
specially with »{». do not:
>> mold "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa{"
== {{aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa{}}
>> load mold
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa{
** Syntax Error: Invalid string --
{aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa{}.
** Where: (line 1)
{aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa{}
because you have to balance »{» or escape it, and there is no escaped
»{» in old rebols.
/core 2.4.40 does it ok:
>> mold "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa{"
==
{^{aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa^^^{^}}
also
>> mold %""
== "%"
>> load mold %""
** Syntax Error: Invalid file -- %
** Near: (line 1) %
and pathes have problems
>> a: first [this/that]
== this/that
>> mold a
** Script Error: this has no value
** Near: mold a
because they get evaluatet by mold.
Don't know if someone want to save pathes?
Then, 'save is similar to copy/deep, so if you have two references to
the same object,
there are two seperate objects after 'do .
And of course if you rely on contexts (hand-made bind), this will not
save correctly.
Knowing this, 'save / 'do is very handy, don't know of a better way.
(hope i have mentioned all?)
But, some while ago someone said he has written a serialisation which
handles some of this problems
and will post it if someone asks?
Volker
