From: Simon Brooke
Subject: Re: In- and Out-of- core editors (was Re: Which one, Lisp or Scheme?)
Date:
Message-ID: <32F7963E.47EC659F@intelligent.co.uk>
Jeff Dalton wrote:
>
> > Essentially, therefore, the design of
> > this function assumes the editing of *text*, not *structure*.
>
> Um, no. Structure editors present code *as text* in pretty much the
> say described for (ed symbol): by pretty-printing.
>
> Since you know this is the case, why do you write as if it were not?
-- Guys, I am not the enemy. We are all on the same side --
In a text editor, any character is individually selectable and
individually changeable. In a structure editor, only complete structural
elements are selectable or changeable. How this is presented -- as
pretty-printed text, or as a three-dimensional graph hanging in space[1]
is irrelevent.
Structure editors may *show* structure as text (or they may not); but
they don't allow you to *manipulate* structure as text. To assert that
two things are alike because their surface representation is alike is to
miss the point entirely, in my opinion.
A system which interprets structure built from the same text but in an
environment separate to the working environment is also not editing
structure in-core; if I have a deep and complex knowledge representation
structure that I'm working on, it isn't necessarily the case that I've
written methods which will allow it to be saved to a file and later
reproduced by the reader. So if the structure creating functions messed
it up, I can either write other functions to get in there and fix it or
I can pull it into an in-core structure editor and fix it by hand. I
can't do that with an out-of-core editor, whether structure or text.
> Well, no. You're also saying that CL is hostile to structure editing,
> or something like that.
Yes, I am, and I believe this is fair criticism. The fact that a system
can be criticised in detail does not mean that it's a bad system; few
human artifacts are completely perfect. It's hostile to structure
editing because it doesn't have a *standard* way of making comments
available in-core. To have such a *standard* way (provided it didn't
have any deleterious side
effects) would involve no cost to people who like text editors, and
substantial benefit to people who like structure editors. So it would
represent an improvement (if, perhaps, an unimportant one)
Simon
[1] yes, it has been done -- for my paper on this, see the proceedings
of the Third Alvey Workshop on Explanation, 1986; my work on
two-dimensional graphical editors for knowledge structures has been
published in half a dozen papers, and, for that matter, I'm by no means
only one to have produced such things.
--
·····@intelligent.co.uk (Simon Brooke)
http://www.intelligent.co.uk/~simon
Morning had broken, and we had run out of gas for the welding torch.
From: Henry Baker
Subject: Re: In- and Out-of- core editors (was Re: Which one, Lisp or Scheme?)
Date:
Message-ID: <hbaker-0402972107090001@10.0.2.1>
In article <·················@intelligent.co.uk>, Simon Brooke
<·····@intelligent.co.uk> wrote:
> -- Guys, I am not the enemy. We are all on the same side --
>
> In a text editor, any character is individually selectable and
> individually changeable. In a structure editor, only complete structural
> elements are selectable or changeable. How this is presented -- as
> pretty-printed text, or as a three-dimensional graph hanging in space[1]
> is irrelevent.
>
> Structure editors may *show* structure as text (or they may not); but
> they don't allow you to *manipulate* structure as text. To assert that
> two things are alike because their surface representation is alike is to
> miss the point entirely, in my opinion.
Actually, a student of mine once built what you would call a structure editor,
in that the internal representation was a tree-structure, but whose 'surface
structure' looked, and could be edited like, text. So there may not be such
a clean distinction.