From athas at sigkill.dk  Tue Jun  6 23:58:49 2006
From: athas at sigkill.dk (Troels Henriksen)
Date: Wed, 07 Jun 2006 01:58:49 +0200
Subject: [climacs-devel] paredit.lisp, regularization of motion
	commands, and more
In-Reply-To: <20060508022636.8A0909800B@pluto.mumble.net> (Taylor R.
	Campbell's message of "Mon, 8 May 2006 02:26:36 +0000")
References: <20060508022636.8A0909800B@pluto.mumble.net>
Message-ID: <8764jdygxy.fsf@sigkill.dk>

"Taylor R. Campbell" <campbell at mumble.net> writes:

> Any thoughts & comments on any of these ideas?

I like these ideas and have begun experimenting with them in my
Climacs tree (don't worry, I won't commit anything until it's stable
and reviewed here). It seems to work very well and does make it
simpler to manage the growing number of commands in Climacs. There is
one thing from this proposal that I do not like, though - I do not
believe the FORWARD-OBJECT generic function should take a syntax as
argument for a number of reasons:

         * FORWARD-OBJECT is a low-level function that should behave
           identically across all syntaxes anyway.
         * FORWARD-OBJECT is part of the buffer protocol, which is
           used in so many places that changing it would probably
           cause wide breakage.
         * FORWARD-OBJECT is used in layers that have no concepts of
           syntaxes, just buffers.

Therefore, FORWARD-OBJECT's interface will be a bit different than the
other movement functions, though this should invisible to the user at
the command-level.

-- 
\  Troels "Athas" Henriksen
/\ - Insert witty signature


From athas at sigkill.dk  Wed Jun  7 13:25:27 2006
From: athas at sigkill.dk (Troels Henriksen)
Date: Wed, 07 Jun 2006 15:25:27 +0200
Subject: [climacs-devel] paredit.lisp, regularization of motion
	commands, and more
In-Reply-To: <8764jdygxy.fsf@sigkill.dk> (Troels Henriksen's message of
	"Wed, 07 Jun 2006 01:58:49 +0200")
References: <20060508022636.8A0909800B@pluto.mumble.net>
	<8764jdygxy.fsf@sigkill.dk>
Message-ID: <87k67tayig.fsf@sigkill.dk>

Troels Henriksen <athas at sigkill.dk> writes:

> I like these ideas and have begun experimenting with them in my
> Climacs tree.

It works nicely so far, but I have run into a few issues that seem to
require refactoring of some parts of Climacs. Here is my analysis:

buffer.lisp implements the buffer protocol which is suitable for our
needs. It provides a basic interface to the buffer implementation and
has no concept of syntaxes, hence, higher-level facilities (indeed,
almost any directly user-oriented functionality) should be implemented
elsewhere.

base.lisp is built on top of the buffer protocol and currently
provides more high-level facilities, such as moving the mark by line,
word, and so forth. However, base.lisp does not have the concept of a
syntax either, meaning that functions such as FORWARD-WORD can not be
(simply) implemented based on what a given syntax considers a
word. Also, base.lisp is used be syntax.lisp to implement the syntax
facilities, so making base.lisp understand syntaxes would probably be
a bit of a kludge. Hence, I propose making base.lisp provide nothing
but simple facilities that are not dependent on subjective terms such
as "word", but instead rely on final definitions such as "line",
"buffer" and "object", all of which are cast in stone by the buffer
protocol. Well, almost. I believe that base.lisp should not export any
functions that move the mark or change the contents of the buffer, but
only functions that can be used to query marks and buffers for
information - for example END-OF-LINE-P, DO-BUFFER-REGION,
END-OF-BUFFER-P, etc. I suggest moving the functions that move the
mark to a file and package ("motion.lisp") that implements Taylor
Campbell's framework for defining motion functions, and which
understands the concept of a syntax, thereby permitting the
implementation of functions such as FORWARD-WORD to use the definition
of a word specified by a given syntax. The functions that modify the
buffer contents should be moved to a file and package ("editing.lisp")
that understands the concept of a syntax, like
motion.lisp. editing.lisp should probably be implemented using the
facilities defined in motion.lisp.

syntax.lisp will also require a bit of modification - the definitions
of the FORWARD-FOO functions should be moved to motion.lisp, and a
number of generic predicates (WORD-CONSTITUENT-P, WHITESPACEP, etc)
should be provided instead; a building block for implementing
syntax-aware motion- and editing-functions.

Any comments?

-- 
\  Troels "Athas" Henriksen
/\ - Insert witty signature


From athas at sigkill.dk  Sat Jun 10 22:05:09 2006
From: athas at sigkill.dk (Troels Henriksen)
Date: Sun, 11 Jun 2006 00:05:09 +0200
Subject: [climacs-devel] paredit.lisp, regularization of motion
	commands, and more
In-Reply-To: <87k67tayig.fsf@sigkill.dk> (Troels Henriksen's message of
	"Wed, 07 Jun 2006 15:25:27 +0200")
References: <20060508022636.8A0909800B@pluto.mumble.net>
	<8764jdygxy.fsf@sigkill.dk> <87k67tayig.fsf@sigkill.dk>
Message-ID: <87r71w1xbe.fsf@sigkill.dk>

Troels Henriksen <athas at sigkill.dk> writes:

> It works nicely so far, but I have run into a few issues that seem to
> require refactoring of some parts of Climacs. Here is my analysis:

And now I can present some code. I am not ready to commit to the tree,
as I am pretty sure there are still some things that do not work
properly (because some functions have changed their signature), but it
compiles cleanly and works for the trivial editing tests I have
performed (including use of the more advanced commands in Lisp
syntax). Also, this patch reworks a number of editing, deletion and
killing functions and commands into a framework similar to what Taylor
Campbell proposed for the motion functions and commands. I am not
quite sure that limit actions are being used properly though.

The patch, which should apply to a fresh Climacs tree, can be found
at: http://sigkill.dk/code/motion-editing-refactoring.diff

Please review and test it, and don't hesitate to criticize the
approach taken.
-- 
\  Troels "Athas" Henriksen
/\ - Insert witty signature


From athas at sigkill.dk  Sat Jun 17 22:35:39 2006
From: athas at sigkill.dk (Troels Henriksen)
Date: Sun, 18 Jun 2006 00:35:39 +0200
Subject: [climacs-devel] Patch for putting Swine in Climacs
Message-ID: <87r71nz9z8.fsf@sigkill.dk>

Attached to this post is a work-in-progress patch to Climacs that
integrates Swine with the Lisp syntax. I have based this on the Swine
from CLIM-desktop, minus the stuff requiring Clouseau or Closure, but
with a few cleanups, and a slight change in the informal protocol used
to query the Lisp for information: the functions now take an "image"
parameter taken from the syntax object. If NIL, the Swine functions
will basically not work. The idea is that code added to climacs.asd
tries to detect whether or not Swank is available, and if it is, loads
lisp-syntax-swank.lisp that sets up Lisp syntax objects to use Swank
for evaluating code, jumping to definitions, etc. (stuff that would be
broken without Swank). This also opens up the possibility of creating
a remote-image-objet for connecting Climacs to a remote Lisp like
SLIME is capable of, but this will require a few more modifications of
how the Lisp syntax works.

This patch is obviously not clean enough to commit yet, but please
comment on it if you feel that the approach taken is suboptimal.

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: climacs-swine.diff
URL: <https://mailman.common-lisp.net/pipermail/climacs-devel/attachments/20060618/13fe31ac/attachment.ksh>
-------------- next part --------------

-- 
\  Troels "Athas" Henriksen
/\ - Insert witty signature

From athas at sigkill.dk  Thu Jun 29 14:42:19 2006
From: athas at sigkill.dk (Troels Henriksen)
Date: Thu, 29 Jun 2006 16:42:19 +0200
Subject: [climacs-devel] Patch for putting Swine in Climacs
In-Reply-To: <87r71nz9z8.fsf@sigkill.dk> (Troels Henriksen's message of
	"Sun, 18 Jun 2006 00:35:39 +0200")
References: <87r71nz9z8.fsf@sigkill.dk>
Message-ID: <871wt89g6s.fsf@sigkill.dk>


I think I have gotten most of the quirks worked out. Unless I hear
complaints, I'm going to commit this patch to the Climacs repo in a
few days: http://sigkill.dk/code/climacs-swine.diff

It's not complete and it's not perfect, but it works as well as the
Swine in CLIM-desktop. When this has been committed I'm going to start
work on making the Lisp syntax less expectant on being able to work
with true Lisp objects, eventually leading to the possibility of
developing a remote-Swank-backend like the one in SLIME.

-- 
\  Troels "Athas" Henriksen
/\ - Insert witty signature