Qilang Google Group

typechecking

2009-11-30T23:22:35Z

I've stalled a little in the development. The crucial thing is that I
cannot spend more then
about one hour per day and this thing needs to be though through, so
what is the thing.

To let all code that goes through (tc -) to go through with a weak
type-checking on one need to let types be of an undecidable form, but

Re: einstein riddle

2009-11-26T22:34:15Z

Yep, the define is a pure library function, you should have a basic
DEFUN
as the standard element.

The idea though is to be able to push in functionality in a standard
define construct
very much of the extension mechanism we discussed previously. So there
will be a refactoring
in the end that pushes a lot of the extensions of the def construct

Re: einstein riddle

2009-11-25T06:02:35Z

OK; the timing improvements to the Prolog are very good.

Now regarding your answers to my questions about the new features you
were introducing; these questions were as you recall, how to provide a
formal semantics and to prove type correctness for these
constructions. It seems that you are attempting to answer these

Re: Shen

2009-11-25T05:43:10Z

As I recall, there was a code repository associated with this group
set up by J.T. Gleason and you can use it with no license problems.
Look at [link] and other pages.

Mark

Re: resolving support and internet connection

2009-11-25T05:57:39Z

OK; I have one day off a week and I'm down from the mountains. I'm
working on the internet connection problem. After December 10th I'll
be coming down from the mountains and in civilisation. I'll let you
know how the internet issue works out.

Mark

Shen

2009-11-22T17:29:01Z

Hi,

It was suggested that Shen should be relocated to google code due to
license issues e.g.
QiII not seen as an open license and hence not possible to keep at
giithub, I let BoopCore
still be located at github, because I will dual license that code
because there are people who would work only in QiII and others who

resolving support and internet connection

2009-11-18T09:00:34Z

Just to warn that Lambda Associates may be offline for a temporary
period from next week as I sort out service problems from India. *
This is not a major permanent shutdown * and may not even occur, but
I'm simply warning people that this might happen because resolving
this stuff is more difficult from India.

Re: Qi on Clozure CL

2009-11-14T10:33:35Z

Great,

This is how I try to divide the work.

There is one framework dual licensed under QiII and LPGL that is
the rewrite I've been doing and will live at github under the Shen
project.
Mainly the BoopCore directory.

Here the code transformer and the core pattern matching tool will be
placed.

Re: Qi on Clozure CL

2009-11-13T18:37:30Z

I like this.

2. Dig out how to best do certain things like simple tail call
May try here.

I would try here.

I am not a language designer but this sounds interesting

This would be exciting.

I would enjoy this.

Thanks for the offer.

Regards,
Emeka

Re: Qi on Clozure CL

2009-11-13T10:49:41Z

Can you suggest what area you would like to help with.

Such as (my examples here, please fill in your own if you like so),

1. Answer specific clojure questions
2. Dig out how to best do certain things like simple tail call
optimizations, fast stack implementation for qi-prolog etc.
3. Find out about things that can be hard to implement in clojure in

Re: Qi on Clozure CL

2009-11-12T18:43:28Z

Qi is/was #+CL implementation conditionalized in like 4 places.
I think the trickiest part is below. (now) CCL is ? mostly compiled ?
This did not work in Qi 8.0 -

(DEFUN fix-closures ()
#+CMU (SETQ *closures* '(FUNCTION COMPILED-FUNCTION
EVAL::INTERPRETED-FUNCTION))
#+CLISP (SETQ *closures* '(FUNCTION COMPILED-FUNCTION))

Re: Qi on Clozure CL

2009-11-12T07:09:10Z

Would by any means allow me to participate in it. I won't add much weigh but
I would do my best.

Regards,
Emeka

Re: Qi on Clozure CL

2009-11-11T16:47:16Z

The infrastructure is almost there, I just need a couple of day's mare
and then I will
start to write the translational code

/Stefan

Re: Qi on Clozure CL

2009-11-11T11:42:18Z

Snorgers,

Have you started porting to Clojure?

Regards,
Emeka

Re: Qi on Clozure CL

2009-11-10T21:16:06Z

Hi Nik,

If you would like to try porting it, that's great. You can ask
questions
on the forum and I will try answer your question as good as I can.
I usually check it at least once a day.

Regards
Stefan

Re: Qi on Clozure CL

2009-11-10T18:24:23Z

since in theory CCL is a Common Lisp then it should be a small job to
do. but i don't really know, i'm just theorizing.

Re: Qi on Clozure CL

2009-11-10T20:17:51Z

Nik,
It is time you turn the "you" to "I". It would be exciting following your
work, hope you will start soon.

Regards,
Janus

Re: Qi on Clozure CL

2009-11-10T14:30:34Z

Dear Nik,

Seemingly nobody!
By the way I know you very well!
You are a smart guy and you should be able to do it yourself:-)

Regards
Nik

test-reader.qi

2009-11-05T23:16:45Z

I've now been running same tests and made a couple of examples where
the focus have been in the Yacc part and the segment part of the code
as well as a General Streaming method.

as an example I did a parser for a toy lang. and then created a
modified list stream that
added the capabilities to let the line column and depth be kept in

Re: einstein riddle

2009-11-04T22:28:13Z

This is actually an interesting question (maybe I'm missing the point
but oh well)

consider
def f (u [(m X) Y]) Cont -> ...

now if the input to the function is unbound it will be bound to [X Y]
but this is not enough.
you should bind it to (u [(m X) Y]) and then in the unifications done

Re: einstein riddle

2009-11-01T23:56:44Z

Here is generated lisp for "member"

(DEFUN member (V1466 V1467 V1468)
(LABELS ((g285 ()
(LET ((X1430 (lookup V1467)))
(IF (CONSP X1430)
(LABELS ((g287 ()
(IF
(EQ (PROGN (newframe) (unify V1466 (CAR

Re: einstein riddle

2009-11-01T23:21:13Z

Hi Mark,

Good to see you back and great questions

i) I decided to manage the stack explicitly with a simple array,

ii) I decided to see the impact of an array of cons cells and
allocations of cons cells for
many operations that is done from this stack hence there is a lot
less consing going on

Re: einstein riddle

2009-11-01T17:35:50Z

Just dropped in; looked at the code for Shen. Interesting; some
questions to reflect on though.

1. The 10X speedup sounds great but some details would be good. How
is the speedup gained? What does the generated Lisp code look like?

2. What is the declarative meaning of u? I guess it means 'unifies'.

Re: einstein riddle

2009-10-29T01:12:58Z

Thanks for the explanation.
I had a different impression from Mark's original Shen proposal.
It really clears up the approach you are taking.
I've got a lot to learn before I will have any insights ;-)

Re: einstein riddle

2009-10-27T22:42:02Z

OK, I've replaced define with def in the Lambda.qi file and called it
Lambda.boop.qi
It now can compile itself. The next step is to spend a week writing
tests for it.

The plan is then to start write a typechecker that is turned on all
time that use type driven code
generation to generate target code e.g. lisp and clojure ... it will

Re: einstein riddle

2009-10-26T21:30:10Z

The idea is that Shen will be a version of QI that work on "any" lisp
like platform.

The approach is to write a set of core libs in qi and then make a
compilation
to selected platforms like clojure and CL.

The conclusion I draw from that was that the prolog engine in qi had
to be
rewritten. As a consequence the code is

Re: einstein riddle

2009-10-26T03:17:41Z

Hi Stefan,

I'm new to Qi and I get the impression that you are porting Qi to
Clojure.
I haven't seen any description of the plans for Shen and was wondering
if you are considering the Clojure work the basis of Shen?

I've been away from Lisp and functional programming for quite a while,
but I really like what Mark has done with Qi. When I get up to speed

Re: einstein riddle

2009-10-23T22:08:20Z

Hi,

The previous result was for manually cleaned lisp code aka proof of
method. Now
after making this cleaning as part of the compilation auto-generated
lisp takes 21ms
to solve the Einstein riddle. In the process code has been tested and
bugs squashed.

I also prepared the code in a state so that it can experimented with

Re: einstein riddle

2009-10-11T21:00:35Z

that is pretty exciting news! congratulations :)
sincerely.

einstein riddle

2009-10-10T21:27:54Z

Hi,

I've been busy coding for some time. I now have a working basic
framework. This framework can be used to combine backtracking and
unifications to builed prolog like features. E.g. in my view a natural
building block to spark Shen from.

The last couple of week's I've been focusing in understanding how to

Qi on Clozure CL

2009-10-09T10:08:49Z

Hi folks,

Does somebody work on porting Qi to Clozure CL?

Regards
Nik

Re: Need more eyes on two small Qi-YACC parsers

2009-09-10T18:53:16Z

You should now that the symbols in qi follows the rules of comon-lisp
that
qi package.

You need to work at the character stream level and or work with
strings

/Stefan

Re: Need more eyes on two small Qi-YACC parsers

2009-09-10T00:39:55Z

Thinking of it, I don't even need ::= to be parsed as ::=. My real
intention is to translate stuff. But I'm still confounded by this side-
effect of something I'm not aware of.

From this below, I see that a colon (:) seems to somehow quote what
follows. But I'm really not sure about this. I suppose I'll have to

Re: Need more eyes on two small Qi-YACC parsers

2009-09-09T23:16:15Z

Type ::= to top level and reflect.

Re: Need more eyes on two small Qi-YACC parsers

2009-09-09T21:21:38Z

Since := is part of Qi-yacc syntax; push it into the semantics.

(defcc <foo>
-*- := (if (= -*- ::=) -*- #\Escape);)

Mark

Re: Need more eyes on two small Qi-YACC parsers

2009-09-09T18:08:10Z

Thanks. Second, I want something like this to be recognised:

(defcc <foo>
::=;)
(compile <foo> [::=])
fail!

I think this clashes with Qi-YACC's semantic action's reserved symbol
":=".

I've tried different strategies but haven't been successful in
accepting ::= as valid (see the start of this thread for my trials).

Re: Need more eyes on two small Qi-YACC parsers

2009-09-09T16:59:24Z

That's about right.

Mark

Re: Need more eyes on two small Qi-YACC parsers

2009-09-09T14:53:22Z

Thanks for the answer to my question 1. Have you seen my question 2?

=== 1 ===

So you're telling me that <rule> doesn't send the head of the stuff to
be parsed to <sn>, and the rest of it to <sym>?

(defcc <rule>
<sn> <sym>;)

I thought that the space between <sn> and <sym> in <rule> told the

Re: Need more eyes on two small Qi-YACC parsers

2009-09-09T12:42:07Z

OK, Dan, while I'm waiting for my buddy to turn up.

Well, in your example; the input is

[[1] symbol]

and the head is [1] and not 1. So Qi-Yacc is doing right.

OK; try changing your program to:

(defcc <rule>
[<sn>] <sym>;)

(defcc <sym>
symbol;)

(defcc <sn>
-*- := (if (number? -*-)

Re: Productivity vs Mod Queues

2009-09-09T12:26:37Z

Sorry Dan; I didn't read your message closely enough and posted it as
approved because it was submitted to qilang. I should have noticed
your opening sentence but was reading in a hurry i.e. saw it's not
spam, content seemed ok, so approve. Probably you needed to send this
as a private email.

Re: Productivity vs Mod Queues

2009-09-09T07:23:42Z

Mark Tarver,

What Rich Hackey does is to moderate/delay posts from new members in order
to prevent spams and he also blocks posts from trolls.

Regards,
Emeka

Re: Productivity vs Mod Queues

2009-09-09T05:47:46Z

Right; this is why I unmodded in the first place; so people could get
instant response. ** Also I took a break for 24 hours from all this
having written my epistle, so in this case your post took a bit more
time.** I'm generally a lot faster.

However I will guarantee that your post will appear within 24 hours

Productivity vs Mod Queues

2009-09-09T01:02:34Z

This message is for the moderator, not for Qilang. So I suggest you
delete it after you've read it.

I find it very counter-productive that I have to wait hours (or days?)
before my questions may even be modded-published to the list. And
then, I shall wait hours (or days?) before any answer reaches me. In

Need more eyes on two small Qi-YACC parsers

2009-09-08T19:16:49Z

=== 1 ===
What's wrong with my parser?

(defcc <rule>
<sn> <sym>;)

(defcc <sym>
symbol;)

(defcc <sn>
-*- := (if (number? -*-)
[-*-]
(error "~A is not a number" -*-)))

(compile <rule> [[1] symbol])
error: [1] is not a number

Isn't -*- supposed to be the head of [1], thus 1? And, consequently:

Qi + slime + emacs = love

2009-09-08T16:19:26Z

I finally got slime + emacs working for Qi. If you're not familiar
with slime it eases common lisp development a lot IMHO.

with SBCL running I ran the following:
(require 'swank)
(load "install.lsp")

now the Qi.core has swank loaded and ready to go. In my ~/.emacs I
made:

(defun qi-init-cmd (port-filename coding-system)

Re: restoring direction to qilang

2009-09-08T15:55:47Z

I'm very pleased with this direction you're taking, Mark. That said, I
wish you the best out of your sabbatical in India. :)

restoring direction to qilang

2009-09-08T12:58:21Z

This is a quite important message and concerns the future of qilang,
Qi, Shen and a lot of what has been going on in the last two weeks.
It was going to be a post; but was too long. You can read what I have
to say here. It is about restoring the direction of this group and
Lambda Associates.

Re: Qi question on StackOverflow

2009-09-05T18:47:10Z

Looks like there's really no reason why my idea would be bad. Plus I
get to learn and to answer one of my needs.

You raise one fact: I definitely took that mailing list's post out of
context re: my own purposes. Funny how I didn't *at all* remark the
error of my ways yesterday. :) You're right, it should be mostly a no-

Re: Qi question on StackOverflow

2009-09-05T18:18:52Z

StackOverflow> uhg, another site to logon to.

my only experience was
Inform -> HTML (I chose CL
before I knew of Qi)

With no yacc (lexer) experience, I had a vague idea *
I was going to be writing a Recursive Decent Parser.

SO> using a CFG is a bad idea...

...that was parsing XML in C?

Q&A questions on StackOverflow

2009-09-05T18:02:21Z

Thanks Stefan for your answer on StackOverflow ("SO" thereafter).

Thinking of it, I'm not sure it was very nice of me to ask a question
there and, in some way, require that if you answer it, you do it
there. Sorry for that; I didn't explain why I think it's better for
Qi's/Shen's future, nor did I mention I didn't mind if you prefer to