The McCLIM developers are happy to release version 0.9.2 of McCLIM. It
was tested on SBCL (both threaded and unthreaded), OpenMCL, CMUCL,
Scieneer CL, and Allegro Common Lisp (ANSI mode only).
Get the tarball at
<http://common-lisp.net/project/mcclim/downloads/mcclim-0.9.2.tar.gz>
or install McCLIM via asdf-install.
We are looking forward to your comments and bug reports. Please send
them to ············@common-lisp.net. The list of currently known bugs
can be found at <http://mcclim.cliki.net/Bug>.
Have fun using McCLIM (and build good things with it),
The McCLIM developers.
RELEASE NOTES FOR McCLIM 0.9.2, "Laetare Sunday":
Compatibility
=============
This release works on CMUCL, SBCL, CLISP, OpenMCL, Allegro CL,
LispWorks, and the Scieneer CL, using the CLX X Window bindings.
Changes to the Install Process
==============================
Implementation-specific INSTALL.* files were removed. Generic and
implementation-specific Installation instructions were improved and
merged into the file INSTALL.
This release requires the "spatial-trees" library by Christophe
Rhodes. Get it via asdf-install or at http://cliki.net/spatial-trees.
Changes to Backends
===================
Copy & Paste code in the CLX backend was improved and should now
adhere more strictly to ICCCM.
Support for connecting to a ssh-forwarded display was restored.
Several unused parts (marked with #+unicode) of the CLX backend were
removed, thus restoring buildability on installations of clisp that
have the unicode feature turned on.
Double buffering for panes was implemented. To use it, create panes
with the :double-buffering t initarg.
There is now rudimentary support for entering non-Ascii characters from
X11 ports using SBCL CLX (a.k.a. telent CLX).
McCLIM ships experimental support for TrueType font rendering using
the FreeType libraries and the free Bitstream Vera fonts. To use it,
link Experimental/freetype/mcclim-freetype.asd to one of your
asdf:*central-registry* directories and load the "MCCLIM-FREETYPE"
system.
An experimental "Null" backend was added that should allow testing of
CLIM functionality without requiring a GUI environment to run.
Changes to the Documentation
============================
A new chapter on contributed applications was added.
Several new figures and examples were added to the manual
Clemens Fruhwirth added a CLIM tutorial paper called "A Guided Tour to
CLIM". It is available in Doc/Guided-Tour/.
Changes to Contributed Applications and Examples
================================================
New application: A CLIM Debugger (by Peter Mechlenborg). It resides in
Apps/Debugger/.
New application: Functional-Geometry by Frank Buss and Rainer
Joswig. It resides in Apps/Functional-Geometry/.
The Inspector now is now able to disassemble functions and inspect
pathnames.
The Listener can now produce vertically-aligned graphs.
The Scigraph application now builds on SBCL again.
A demo for drag-and-drop-translators was added.
Further additions to McCLIM
===========================
There is now a test suite, located in Tests/. It contains tests for
regions, bounding rectangles, transformations, commands, and the
PostScript backend. With the addition of the Null backend, we hope to
add several more tests for more chapters of the CLIM spec.
New Extension "conditional-commands": allows activation/deactivation
of commands when other commands are invoked. It resides in
Extensions/conditional-commands/.
Status of the CLIM 2 Spec Implementation
========================================
Here is a list of what we think works, organized by chapters and
sections of the CLIM 2 specification.
Chapter 3 Regions
Mostly finished. There are some troublesome parts of the
specification that may not be implemented for all possible
regions, for instance region-contains-region-p. There may not
be an efficient way of implementing this function for all kinds
of regions.
Chapter 4, Bounding rectangles
Finished
Chapter 5, Affine transformations
Finished
Chapter 6, Overview of window facilities
Finished
Chapter 7, Properties of sheets
Finished, though the correct behavior of sheet transformations may
not have been tested.
Chapter 8, Sheet protocols
Finished
Chapter 9, Ports, Grafts, and Mirrored sheets
Finished
Chapter 10, Sheet and medium output facilities
Finished
Chapter 11, Text styles
Mostly complete.
There is now experimental support for device font text styles (via
make-device-font-text-style) for the CLX, PostScript, and
CLX+FreeType backends.
Chapter 12, Graphics
Finished
Chapter 13, Drawing in Color
I am note sure about the state of this. I thought we were doing
only full opacity and full transparency, but I see traces of more
general designs.
Chapter 14, General Designs
The composition of designs is not supported. We do support regions
as designs.
Chapter 15, Extended Stream Output
Extended output streams are fully supported.
Chapter 16, Output Recording
Output recording is mostly implemented.
This release ships with a standard-tree-output-record type for the
first time. The tree output record type speeds up point- and
region-based queries, but slows down insertion of output records
by a bit.
make-design-from-output-record is not implemented. *Note*: the
coordinates in output records are relative to the stream. This is
in conformance with the Spec, but not necessarily compatible with
other CLIM implementations.
Chapter 17, Table Formatting
Table formatting is completely implemented.
Chapter 18, Graph Formatting
Graph formatting is fully implemented. The :hash-table argument
to format-graph-from-roots is ignored.
Support for a :dag graph type was added, as was support for
vertically oriented graphs and support for the :arc-drawer
argument to format-graph-from-roots.
Chapter 19, Bordered Output
Bordered output is fully supported.
Chapter 20, Text Formatting
With the exception of the :after-line-break-initially argument to
filling-output, this chapter is fully implemented.
Chapter 21 Incremental Redisplay
The updating-output interface to incremental redisplay is
implemented. McCLIM makes no effort to move i.e., bitblit, output
records; they are always erased and redrawn if their position
changes. This is much more compatible with support for partial
transparency. The :x, :y, :parent-x and :parent-y arguments to
redisplay-output-record are ignored. McCLIM follows the spirit of
21.3 "Incremental Redisplay Protocol", but we have not tried very
hard to implement the vague description in the
Spec. augment-draw-set, note-output-record-child-changed and
propagate-output-record-changes-p are not implemented.
Incremental redisplay in McCLIM may still suffer from performance
problems, despite the presence of spatially-organized compound
output record types.
Chapter 22, Extended Stream Input
The implementation of extended input streams is quite
complete. (setf* pointer-position) is not implemented. There is no
stream numeric argument, so that slot of the accelerator-gesture
condition is always 1.
Chapter 23 Presentation Types
Most of the literal specification of this chapter is
implemented. Specific accept and present presentation methods for some
types are not implemented, so the default method may be
surprising.
The output record bounding rectangle is always used or highlighting
and pointer testing.
presentation-default-processor is not implemented.
The presentation method mechanism supports all method
combinations. The body of a presentation method is surrounded
with a block of the same name as the presentation method, not just
the magic internal name. The method by which presentation type
parameters and options are decoded for the method bodies is a bit
different from real CLIM. In particular, you cannot refer to the
type parameters and options in the lambda list of the method.
The NIL value of presentation-single-box is now supported.
Presentation type histories are now partially implemented. The
gesture C-M-y should recall the last entered presentation.
define-drag-and-drop-translator is now implemented.
Chapter 24 Input Editing and Completion Facilities
with-input-editor-typeout is not implemented.
The noise strings produced by input-editor-format and the strings
produced by presentation-replace-input are not read-only. This
could lead to interesting "issues" if the user edits them.
Only a few of the suggested editing commands are implemented. An
additional command that is implemented is control-meta-B, which
drops into the debugger. add-input-editor-command is not
implemented.
with-accept-help is not implemented.
Chapter 25 Menu Facilities
The protocol is implemented, but McCLIM doesn't use it to draw
command table menus.
Chapter 26 Dialog Facilities
McCLIM contains a basic, somewhat buggy implementation of
accepting-values. There is little user feedback as to what has
been accepted in a dialog. The user has to press the "OK" button
to exit the dialog; there are no short cuts. There are no special
accept-present-default methods for member or subset presentation
types. Command-buttons are not implemented. There is no
gadget-based implementation of accepting-values.
The internal structure of accepting-values should be "culturally
compatible" with real CLIM; if you have some spiffy hack, check
the source.
:own-window is now supported in accepting-values.
Chapter 27 Command Processing
command-line-complete-input is not implemented (the
functionality does exist in the accept method for command-name).
display-command-table-menu and menu-choose-command-from-table are
not implemented. Menu-command-parser is not implemented, though the
functionality obviously is. Nothing is done about partial menu
commands. There is no support for numeric arguments.
The command-or-form presentation type is not implemented.
Chapter 28 Application Frames
raise-frame, bury-frame and notify-user are not implemented.
:accept-values panes are not implemented.
frame-maintain-presentation-histories is not implemented.
frame-drag-and-drop-feedback and frame-drag-and-drop-highlighting
are now implemented.
execute-frame-command ignores the possibility that frame and the
current frame might be different.
display-command-menu isn't implemented.
Chapter 29 Panes
Due to the way the space-allocation protocol is implemented, it is
not easy to create application-specific layout-panes. Client code
needs to know about :AROUND methods to compose-space, but they are
not mentioned in the spec.
restraining-pane is partially implemented.
Chapter 30 Gadgets
This chapter is implemented.
with-output-as-gadget is not quite working yet, but it was
improved since the last release.
McCLIM developers <············@common-lisp.net> writes:
> This release ships with a standard-tree-output-record type for the
> first time. The tree output record type speeds up point- and
> region-based queries, but slows down insertion of output records
> by a bit.
This was one of the objectives of one of the Summer of Code projects I
proposed [1] last year; as with the x86 calling convention project,
since some work has been done on it, and for want of a better place
for a summary, it's probably worth reporting our findings here.
Briefly, output recording is the ability of the CLIM engine to
remember where output has been drawn, and additionally what the output
is; presentations, built atop of output recording, additionally
associate a lisp object with the output for later use in commands
initiated with the pointer.
The simplest possible way of implementing this memory, a linear data
structure such as a list or adjustable vector, results in linear time
for searching: the common questions to be asked are "what output
records intersect this region?" and "what output records are under
this pointer position?", and answering those questions with a
sequence-output-record fundamentally takes linear time, as all
individual records must be queried.
Andreas Fuchs, developer of beirc (an IRC client using mcclim, [2]),
found that for his application the linear scaling for search was
causing slowdown problems at typical scrollback-buffer size. He
therefore implemented tree-output-records in terms of R-trees [3],
which have usual-case (a slightly sloppy phrase for a slightly
ill-defined concept) logarithmic search time in the number of records.
This loses the previous amortized constant-time insertion cost of a
list or adjustable vector, and the constant factors are probably
different, too (see [4]), but the scalability of beirc at least was
noticeably improved.
There remains work to be done; McCLIM is a two-dimensional
environment, while the spatial-trees library is for general
dimensionality; this generality (needless from McCLIM's point of
view), as well as some invariant-checking code in spatial-trees,
contribute to the overhead of r-tree output-records being larger than
necessary [5]. While some initial measurements have been made of
interactive performance, the evidence so far is anecdotal over how
large a benefit can be felt; additionally, we don't yet have a solid
feel for how best to structure applications to get the best benefit
from the various forms of compound output records (sequence, tree)
available.
As mentioned in the release announcement I'm replying to, this
functionality is in the McCLIM 0.9.2 release.
Christophe
[1] http://www.alphageeksinc.com/cgi-bin/lispnyc.cgi?GuiSpatialIndexing
[2] http://common-lisp.net/project/beirc/
[3] http://www.rtreeportal.org/
[4] http://paste.lisp.org/display/17628
http://paste.lisp.org/display/17633
[5] http://paste.lisp.org/display/17691