freenode/#lisp - IRC Chatlog
Search
5:40:08
gilberth
Seriously, is typesetting to a PDF document and thus a pile of dead trees still important?
5:40:28
beach
I don't think including additional functionality like that in the standard is the most important matter.
5:40:56
beach
That functionality already exists, and everyone is already using it, so including it solves no problem.
5:41:40
beach
gilberth: No, I am open to other suggestions, but LaTeX is still the only one that includes all the functionality I need.
5:42:43
beach
I think it is way more urgent to eliminate most of the undefined behavior in the Common Lisp HyperSpec.
5:43:41
beach
And *that* is low-hanging fruit because most existing implementations do the right thing.
5:46:44
gilberth
I figure that. I still like scribe through. I figure, that you want to have it be possible to be edited in some sane way.
5:47:55
beach
What I really want is a protocol for manipulating documents, so that we can stop thinking in terms of markup languages. But I think jackdaniel is working on a CLIM-based documentation system, so maybe that problem will be solved.
5:48:50
gilberth
My motivation some twenty years ago to start the web browser was to have a documentation system.
5:54:09
gilberth
Well, I have two "problems". I was not successful on the McCLIM mailing list to get my design idea accepted. And that made me shy. The other is that I am that damn perfectionist.
5:55:52
beach
gilberth: Even if it is not possible to finish everything, there are ways to avoid wasted efforts. I write papers to document ideas, and we extract independent libraries that can be documented and tested separately.
5:56:59
gilberth
beach: Oh yeah. I still need to publish a paper about my regular expression engine. I solved what people where thinking about for twenty-something years.
7:42:28
flip214
gilberth: I'd be interesting in the paper (& code?) as well... Being a Perl Guy since 1994 or so ;/
7:43:32
jackdaniel
I'll have to dig these ideas on the mailing list archive when I'll have some spare time, I've got curious now :)
7:44:00
gilberth
flip214: Perl compatible I did not yet manage. And I don't its semantcs. It is POSIX.
7:46:05
flip214
Still it would be interesting. Once I invested a bit of time (much too little, though) trying to make CL-PPCRE build match functions via a macro,
7:46:28
flip214
so that the compiler can do all its wonderful work instead of "just" calling closures.
7:47:05
jackdaniel
doesn't it define bunch of compiler-macros for things it can optimize at compilation time?
7:47:52
gilberth
flip214: Drop me a one-liner at gilbert@bauhh.de, so that I'll have your email-address.
7:52:26
gilberth
Well. For those that are interested: I compile group capture to Mealy-automata. This is a completely diffrent business.
7:53:41
gilberth
So you get O(1) for group capture. Though 1 might involve a dozend registers being assigned per character read.
9:51:05
Harag
hi, is there a place/method that can be used to run clean up code on a hunchentoot session timeout (if there is such an event)?
11:38:01
jackdaniel
and to not be a nasty person for other server clients preferably on your own acceptor (i.e not acceptor class defined by hunchentoot)
11:54:20
ym
jackdaniel, Hi. Could you, please, explain why parameter symbols (like x and y in :motion-notify) of event-case has fixed names? Wouldn't it be more lispy to have programmer-defined names? Like (:motion-notify (mouse-x mouse-y) ...) and let CLX bind them corresponding on their place (first is x coordinate, second - y, etc) like in defmacro.
11:59:42
jackdaniel
ym: the short answer is because that's how it was specified in clx (I'm not the author, I'm a mere co-maintainer), if you ask for my guess then: I think that event-case was specified that way to mimic with-slots (but with a "event-type" twist)
15:00:17
dtw
It seems that the current stable Debian 9 system has only two CL implementations: SBCL and GCL. Is it just that there are no Debian maintainers for other implementations or does this also tell something about CL upstream development?
15:08:16
edgar-rft
dtw: According to my experience no linux distribution has well maintained CL packages. Linux package systems usually are designed for a C-like infrastructure and are not really suited for Lisp distribution.
15:11:49
dtw
This may affect software distribution. It's easier to tell users of my software to "apt install sbcl" and type "make" than to tell "compile this Lisp implementation first" and then compile my software.
15:11:59
cage_
dtw, in my limited experience with linux distribution Debian has a, relative, good CL team, SBCL is upgraded quickly on sid
15:12:17
dtw
I used to compile SBCL myself but nowadays I have updated it with my Debian (every two years).
15:15:24
loke
dtw: The debian CL libraries are a mess. Old, unsupported and doesn't work with modern software.
15:16:42
cage_
i use quicklisp for libraries but never felt the need to compile sbcl (but, i am on debian testing)
16:29:27
dim
loke: that's the general situation yes, but I've expanded some efforts into fixing that with my ql-to-deb tool and took care of the 60+ indirect dependencies for pgloader
16:30:12
dim
see https://github.com/dimitri/ql-to-deb and https://qa.debian.org/developer.php?login=dim@tapoueh.org ; I'm not so active anymore on the debian front, mostly that's svillemot and Myon doing the work now
16:30:41
dim
loke: yeah that's what I do, and then I have pgloader distributed in debian itself, so I had to make it compliant with the debian contract
16:31:15
dim
any software shipped in debian must be buildable from sources all distributed in debian too
16:31:38
dim
I like this guarantee that debian gives its users, and decided to make it happen for pgloader
16:31:41
loke
ah, sure. But that doesn't mean that those external deps have to be actual debian packages
16:32:02
loke
I guess they can be _build_ dependencies, but they shouldn't be something users have to install.
16:32:40
loke
because the existence of those packages is such a nightmare. I've lost count of the time sI've explained to newcomers that they have to get rid of the borken Debian CL library packages.
16:32:45
dim
for the RPM builds, I'm also providing a source tarball that's built with the Quicklisp bundle facility, making it easier and a single source package
16:33:40
dim
ql-to-deb allows to easily grab what's in Quicklisp and update the debian package with that, so that it should be easy enough to update them monthly
16:33:51
loke
perhaps make a single "quicklisp-build-deps" package that is used during build, but an never be directly installed by a user.
16:34:33
dim
some debian users want to install debian packages for their CL libs, I've met some either on IRC or at ELS, but I would not do that either
16:36:14
loke
dim: yes. I'm not saying they don't exist. But I wouldn't be surprised if the number of users who have had problems with it greatly exceeds the number of user who actually want it.
16:36:50
dim
at least I tried to help, I would still use QL for a CL development environment even on debian