Nabble - SWI Prolog

Re: bagof/3 in 5.8.0

2009-11-10T00:21:05Z

On Tue, 2009-11-10 at 08:59 +0100, Bart Demoen wrote:
> Have you looked at the bagof/setof implementation in XSB ?

No. The previous rewrite was modeled after YAP, with
introduction of some more complexity and bugs :-) Before
we all try to analyse how XSB's work while I guess you already
did so, it might be wise to add a few words how it is done
differently and what advantage that can bring :-)

Cheers --- Jan

> Bart Demoen
>
> Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: bagof/3 in 5.8.0

2009-11-09T23:59:01Z

Have you looked at the bagof/setof implementation in XSB ?

Bart Demoen

Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: bagof/3 in 5.8.0

2009-11-09T16:15:24Z

Jan Wielemaker:
>I agree that in the ideal world current subsumes_chk/2 should have been
>called subsumes/2. I don't think we can change that anymore. I'm
>happy to drop subsumes/2 from a standard proposal. I'm also happy
>do drop the interaction with the occurs-check (I'll have a look at
>that now).

So we (both) agree to consider subsumes_chk/2 now. We also agree that
subsumes_chk/2 should not consider constraints? And that

?- freeze(X, nimporte), subsumes_chk(_,X).

succeeds instead trying to execute nimporte? Idem

?- freeze(X, nimporte), subsumes_chk(X,X).

> Otherwise, in normally I follow standardization.

Please note that there is nothing definitive about these predicates at
all. We had one discussion about core (N208) so far in one meeting.
It would not be the first time that seemingly consistent proposals
then are reconsidered again and again. And - always - this has been
for good.

The only thing definitive is the ballot (when it is passed, or fails
too often...).

>Oh, and I very much welcome a standardization forum disconnected from
>vendor/personal pages. Its good that there is a lot of discussion here,
>but it would be better that if the topic is not specific to SWI-Prolog
>that it is discussed in a place where other vendors listen and can feel
>`at home'.

Prior to going out into the large, it is helpful to discuss things in
the small - and SWI users are very constructive. It does not make
much sense to discuss things within WG17 if SWI people find all kinds
of holes. After all, good quality of a proposal helps a lot. Think
of number_codes/2 ...

And so far, I got two (2) constructive replies from the standard
mailing list when collecting informations about the current practice
of evaluable functors and built-in predicates and one (1) complaint.
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: bagof/3 in 5.8.0

2009-11-09T15:11:37Z

Jan Wielemaker:
>I removed the reordering argument. If I understand things correctly
>whether or not reordering is needed will be a decision for all bags,
>not something to be done on bag-by-bag basis.

There is need for resorting, if the findall found free variables
shared between the Witness and the Solution and those variables are of
relevance in the sort/2 such that when binding the free variables to
one instance, the list is no longer sorted. So this is a very complex
dynamic property. The easiest way to find out that property is to see
if resorting will change anything :-).

The patch I sent you (got it per mail?) should fix the (in my opinion)
only reasonable case, which is where the free variables are all
ground.

http://www.complang.tuwien.ac.at/ulrich/swi-prolog/git/0002-ENHANCED-Avoid-double-sorting-in-setof-3-for-the-com.patch

> '$e_free_variables'(Templ^Goal, Vars),
> ( Vars == v

I could have sworn that '$e_free_variables'/2 produces []! How
suggestive code can be.
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: bagof/3 in 5.8.0

2009-11-09T14:37:16Z

Bart Demoen:
>ps. ever seen bagof/3 goals with an underscore (explicit anonymous
> variable) in the second argument ? in other contexts, _ means "I am
> not interested in that part of the answer" - in bagof/3 it means

To finish your sentence: in both setof/3 and bagof/3 anonymous
variables will cause to enumerate alternate (and thus invisible)
solutions. This is not intuitive to Prolog programmers. In fact, I
would like to see one case where an _ is used for good within setof/3
or bagof/3. I mean in the second argument, indeed. The idiom
setof(t, Goal, _) is very useful.

There is one remedy to this: library(lambda)! With it, _ gets its
meaning back! See the last example at:

http://www.complang.tuwien.ac.at/ulrich/Prolog-inedit/ISO-Hiord#setof

In the long term lambdas should be compiled much like
library(apply_macros). At least I made sure that they can be compiled
- for the meaningful cases. Then, there would be no overhead at all.

(Sorry for this advertisement, but every time I post this I get some
response form people who have not seen it.)
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: enabling clpfd in a program

2009-11-09T12:12:26Z

> Current Problem: cannot load the clpfd library
>
> */
>
> use_module(library(clpfd)).

try

:- use_module(library(clpfd)).

Cheers

Bart Demoen

Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

enabling clpfd in a program

2009-11-09T12:03:18Z

In the program below, I cannot load the clpfd library for use in the
program. But I want to use the #= operator so I must load it. Please
advise on how to get the module import statement working:

/*

Exercise:
http://cs.union.edu/~striegnk/learn-prolog-now/html/node26.html#subsec.l3.ex2

Binary functor naming convention:

F(X,Y) means X is an F of Y

example:

child(X,Y) means X is a child of Y

Current Problem: cannot load the clpfd library

*/

use_module(library(clpfd)).

descendant(X,Y,1) :- child(X,Y).

descendant(X,Y,N) :-
N #= N1 + N2,
descendant(X1, Y, N1),
descendant(X, X1, N2),

child(c, assembler).
child(cplus, c).
child(d, cplus).
child(perl, c).
child(perl, shell).

parent(X,Y) :- child(Y,X).
grandparent(X,Y) :- descendant(X,Y,2).
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Compiling the stack shifter

2009-11-09T04:54:09Z

Hello,

On Mon, Nov 9, 2009 at 21:36, Jan Wielemaker <J.Wielemaker@...> wrote:
>
> Works for me, unless you followed the option to set -DSECURE_GC. This
> makes GC *much* slower and for compiling CHR it really takes long (30
> minutes or so) to compile in this mode.

Indeed, I was using -DSECURE_GC...

> The current version seems pretty stable. I advice compiling with the
> normal flags until you get a crash. Then you can recompile the Prolog
> kernel by editing src/Makefile and set
>
> COFLAGS=-gdwarf-2 -g3 -DSECURE_GC -fno-strict-aliasing
>
> Now, run make clean && make && make install in src only. If this
> makes your system too slow, drop -DSECURE_GC. If you get a crash,
> I can give some hints for debugging.

Thank you for the hint :-)

Regards,

--
Nicolas
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Ann: plosc 0.2 - OSC library

2009-11-09T04:40:52Z

Hi all,
Just wanted to say that a new version of the OSC library
is now available from the GIT repository at

http://prolog.cs.vu.nl/git/contrib/plosc.git

The chief addition since the initial version is the ability
to receive OSC messages and their arguments and to
process them in Prolog, eg

...
osc_mk_server(7770, S),
osc_add_handler(S, '/foo', [string(_), float(_)], foo),
osc_start_server(S).
...

foo(Message, [string(S), float(X)]) :-
format('received ~w: ~w, ~w\n', [Message,S,X]).

I'm using liblo to implement the OSC server, which ensures
that time-stamped messages are properly scheduled and
therefore not dispatched to their Prolog handlers until the
proper time.

cheers,
Samer.

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Compiling the stack shifter

2009-11-09T04:36:05Z

On Monday 09 November 2009 01:13:53 pm Nicolas Pelletier wrote:

> Hello,
>
> While compiling the stack shifter, the compilation seems to get stuck
> on chr. The last trace I get is:
>
> ../pl.sh -L50M -G50M -T20M -p chr=. -q -f ./chr_swi_bootstrap.pl \
> -g
> "chr_compile_step4('guard_entailment.chr','guard_entailment.pl'),halt"
> \
> -t 'halt(1)'
>
> ... and from that point on, the compilation process takes up 100% cpu
> and does not seem to make any progress. Has anyone already seen this?

Works for me, unless you followed the option to set -DSECURE_GC. This
makes GC *much* slower and for compiling CHR it really takes long (30
minutes or so) to compile in this mode.

The current version seems pretty stable. I advice compiling with the
normal flags until you get a crash. Then you can recompile the Prolog
kernel by editing src/Makefile and set

COFLAGS=-gdwarf-2 -g3 -DSECURE_GC -fno-strict-aliasing

Now, run make clean && make && make install in src only. If this
makes your system too slow, drop -DSECURE_GC. If you get a crash,
I can give some hints for debugging.

--- Jan

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Compiling the stack shifter

2009-11-09T04:13:53Z

Hello,

While compiling the stack shifter, the compilation seems to get stuck
on chr. The last trace I get is:

../pl.sh -L50M -G50M -T20M -p chr=. -q -f ./chr_swi_bootstrap.pl \
-g
"chr_compile_step4('guard_entailment.chr','guard_entailment.pl'),halt"
\
-t 'halt(1)'

... and from that point on, the compilation process takes up 100% cpu
and does not seem to make any progress. Has anyone already seen this?

Thanks in advance for your help,

--
Nicolas
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Ann: Windows binary previews (fixed)

2009-11-09T02:13:58Z

Jacques,

I can't really comment on that. I guess it is an issue between the VC9
runtime library and the Pascal runtime system. But for what and how to
debug it and hack around it, I guess you need a Windows/VC*/Pascal
expert. I'm none of those. I'd start searching the internet for issues
embedding multi-threaded VC9 components into Pascal. I don't really
see how this has anything to do with Prolog itself.

Sorry --- Jan

On Sunday 08 November 2009 06:38:06 pm j.gressier wrote:

> I now get this message when I run the program (XP 32 bits) :
>
> ------------------ message
> Runtime error!
> Program c:\..........myprogram.exe
> R6034
> An application has made an attempt to load the C runtime library
> incorrectly. ------------------ end message
>
> Swi-prolog is embedded.
>
> It is initialized by (Pascal code):
> argv[0]:=StrNew(PChar('libpl.dll'));
> argv[1]:=StrNew(PChar('-x'));
> argv[2]:=StrNew(PChar(lib));
> if not(( PL_initialise(3, argv) = 1 )) then showmessage ('Impossible
> d''initialiser l''environnement');
>
> It crashes in PL_initialise.
> PL_initialise is declared thus :
>
> function PL_initialise(argc: Integer; argv: array of PChar): Integer;
> cdecl; external 'libpl.dll';
>
> Best reagards
> Jacques Gressier

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Ann: Windows binary previews (fixed)

2009-11-08T09:38:06Z

Hi,

I now get this message when I run the program (XP 32 bits) :

------------------ message
Runtime error!
Program c:\..........myprogram.exe
R6034
An application has made an attempt to load the C runtime library incorrectly.
------------------ end message

Swi-prolog is embedded.

It is initialized by (Pascal code):
    argv[0]:=StrNew(PChar('libpl.dll'));
    argv[1]:=StrNew(PChar('-x'));
    argv[2]:=StrNew(PChar(lib));
    if not(( PL_initialise(3, argv) = 1 ))    then showmessage ('Impossible d''initialiser l''environnement');

It crashes in PL_initialise.
PL_initialise is declared thus :

function PL_initialise(argc: Integer; argv: array of PChar): Integer; cdecl; external 'libpl.dll';

Best reagards
Jacques Gressier

> Message du 08/11/09 18:06

> De : "Jan Wielemaker"
> A : "Joel Foutelet"
> Copie à : "SWI-Prolog mailinglist"
> Objet : Re: [SWIPL] Ann: Windows binary previews (fixed)
>
>
> On Sun, 2009-11-08 at 15:16 +0100, Joel Foutelet wrote:
> > Hi Jan
> >
> > I must say that today, after the restart of my PC I have not this problem
> > anymore.
> > May be my machine was tired ...
>
> Happens with Windows machines :-)
>
> More news: library(time) was broken due to a bug in the VC9 runtime
> library (calling signal()) with anything except for a few reserved
> signal numbers crashes the program.
>
> So, the Windows version now uses the same scheduling thread as the
> other platforms, removing the alarm() limits and this library should
> work in both the 64-bit and 32-bits versions.
>
> The installers are replaced.
>
> Cheers --- Jan
>
> _______________________________________________
> SWI-Prolog mailing list
> SWI-Prolog@...
> https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog
>
>

Cordialement
Jacques Gressier
mail : j.gressier@...

site : http://perso.wanadoo.fr/jgressier

Gratuite, garantie à vie et déjà utilisée par des millions d'internautes...
vous aussi, pour votre adresse e-mail, choisissez laposte.net.

Laposte.net, bien + qu'une messagerie

AIPR-10 Call for papers

2009-11-08T09:12:07Z

AIPR-10 Call for papers

The 2010 International Conference on Artificial Intelligence and Pattern Recognition (AIPR-10) will be held during 12-14 of July 2010 in Orlando, FL, USA. AIPR is an important event in the areas of Artificial Intelligence (AI) as well as Pattern Recognition (PR) and focuses on all areas of AI, PR and related topics. The conference will be held at the same time and location where several other major international conferences will be taking place. The conference will be held as part of 2010 multi-conference (MULTICONF-10)_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Ann: Windows binary previews (fixed)

2009-11-08T09:04:12Z

On Sun, 2009-11-08 at 15:16 +0100, Joel Foutelet wrote:
> Hi Jan
>
> I must say that today, after the restart of my PC I have not this problem
> anymore.
> May be my machine was tired ...

Happens with Windows machines :-)

More news: library(time) was broken due to a bug in the VC9 runtime
library (calling signal()) with anything except for a few reserved
signal numbers crashes the program.

So, the Windows version now uses the same scheduling thread as the
other platforms, removing the alarm() limits and this library should
work in both the 64-bit and 32-bits versions.

The installers are replaced.

Cheers --- Jan

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Ann: Windows binary previews (fixed)

2009-11-08T06:16:42Z

Hi Jan

I must say that today, after the restart of my PC I have not this problem
anymore.
May be my machine was tired ...

Joël

----- Original Message -----
From: "Jan Wielemaker" <J.Wielemaker@...>
To: "Joel Foutelet" <joel.foutelet@...>
Cc: "SWI-Prolog mailinglist" <swi-prolog@...>
Sent: Sunday, November 08, 2009 2:42 PM
Subject: Re: [SWIPL] Ann: Windows binary previews (fixed)

> Hi Joël,
>
> I have tried hard to reproduce this, but I failed. Does it have
> any relation to the file you want to edit? In particular, does
> the filename contain non-ASCII characters?
>
> Anyone else who can confirm this problem or confirm that the
> edit ... menu works fine?
>
> P.s. I assume that we are talking about the 32-bit version.
>
> Thanks --- Jan
>
> On Sat, 2009-11-07 at 12:23 +0100, Joel Foutelet wrote:
>> Hi Jan
>>
>> Still errors :
>> Start pl.
>> First try to edit a file -> nothing (but no error writen).
>> Second try -> the editing window is shown with the first file I choosed
>> and
>> in the same moment the OpenFile dialog box is shown (sorry for my bad
>> english).
>>
>> It''s not always the same,
>> Start Pl
>> try to edit a file -> Ok
>>
>>
>> Another try
>> Start Pl
>> try to edit a file -> Ok after 3/4 seconds.
>>
>> Joël
>>
>> ----- Original Message -----
>> From: "Jan Wielemaker" <J.Wielemaker@...>
>> To: "Joel Foutelet" <joel.foutelet@...>
>> Cc: "SWI-Prolog mailinglist" <swi-prolog@...>
>> Sent: Saturday, November 07, 2009 11:12 AM
>> Subject: Re: [SWIPL] Ann: Windows binary previews (fixed)
>>
>>
>> > Hi,
>> >
>> > I've implemented a work-around and uploaded a new installer. Please
>> > give it a try (check that the installer gives nov7 as production date
>> > and you don't pick a version from some cache).
>> >
>> > Cheers --- Jan
>> >
>> > On Fri, 2009-11-06 at 22:18 +0100, Joel Foutelet wrote:
>> >> Sorry my first message was cut:
>> >> My OS is Windows Vistal Integral 32 bits.
>> >>
>> >> Error messages are :
>> >>
>> >> 1 ?- [PCE system error: adt\date.c:623: Assertion failed:
>> >> sizeof(time_t)
>
> _______________________________________________
> SWI-Prolog mailing list
> SWI-Prolog@...
> https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog
>

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Ann: Windows binary previews (fixed)

2009-11-08T05:42:29Z

Hi Joël,

I have tried hard to reproduce this, but I failed. Does it have
any relation to the file you want to edit? In particular, does
the filename contain non-ASCII characters?

Anyone else who can confirm this problem or confirm that the
edit ... menu works fine?

P.s. I assume that we are talking about the 32-bit version.

Thanks --- Jan

On Sat, 2009-11-07 at 12:23 +0100, Joel Foutelet wrote:

> Hi Jan
>
> Still errors :
> Start pl.
> First try to edit a file -> nothing (but no error writen).
> Second try -> the editing window is shown with the first file I choosed and
> in the same moment the OpenFile dialog box is shown (sorry for my bad
> english).
>
> It''s not always the same,
> Start Pl
> try to edit a file -> Ok
>
>
> Another try
> Start Pl
> try to edit a file -> Ok after 3/4 seconds.
>
> Joël
>
> ----- Original Message -----
> From: "Jan Wielemaker" <J.Wielemaker@...>
> To: "Joel Foutelet" <joel.foutelet@...>
> Cc: "SWI-Prolog mailinglist" <swi-prolog@...>
> Sent: Saturday, November 07, 2009 11:12 AM
> Subject: Re: [SWIPL] Ann: Windows binary previews (fixed)
>
>
> > Hi,
> >
> > I've implemented a work-around and uploaded a new installer. Please
> > give it a try (check that the installer gives nov7 as production date
> > and you don't pick a version from some cache).
> >
> > Cheers --- Jan
> >
> > On Fri, 2009-11-06 at 22:18 +0100, Joel Foutelet wrote:
> >> Sorry my first message was cut:
> >> My OS is Windows Vistal Integral 32 bits.
> >>
> >> Error messages are :
> >>
> >> 1 ?- [PCE system error: adt\date.c:623: Assertion failed: sizeof(time_t)

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Ann: Windows binary previews (fixed)

2009-11-07T03:23:16Z

Hi Jan

Still errors :
Start pl.
First try to edit a file -> nothing (but no error writen).
Second try -> the editing window is shown with the first file I choosed and
in the same moment the OpenFile dialog box is shown (sorry for my bad
english).

It''s not always the same,
Start Pl
try to edit a file -> Ok

Another try
Start Pl
try to edit a file -> Ok after 3/4 seconds.

Joël

----- Original Message -----
From: "Jan Wielemaker" <J.Wielemaker@...>
To: "Joel Foutelet" <joel.foutelet@...>
Cc: "SWI-Prolog mailinglist" <swi-prolog@...>
Sent: Saturday, November 07, 2009 11:12 AM
Subject: Re: [SWIPL] Ann: Windows binary previews (fixed)

> Hi,
>
> I've implemented a work-around and uploaded a new installer. Please
> give it a try (check that the installer gives nov7 as production date
> and you don't pick a version from some cache).
>
> Cheers --- Jan
>
> On Fri, 2009-11-06 at 22:18 +0100, Joel Foutelet wrote:
>> Sorry my first message was cut:
>> My OS is Windows Vistal Integral 32 bits.
>>
>> Error messages are :
>>
>> 1 ?- [PCE system error: adt\date.c:623: Assertion failed: sizeof(time_t)
>> <=
>> sizeof(Any)
>> Stack:
>> [ 6] M @48588304/file <-time(@default/constant)
>> [ 5] M @48588268/emacs_buffer ->slot(time_stamp, @48588304?time)
>> [ 4] M @48588268/emacs_buffer ->file(@48588304/file)
>> [ 3] M @48588268/emacs_buffer ->initialise(@48588304/file,
>> @default/constant)
>> [ 2]
>> [ 1] M
>> @emacs/emacs ->goto_source_location(@48587055/source_location, (nil))
>> ]
>> Requesting host to dump stack ...
>> [20] send(@48588268/emacs_buffer, slot(time_stamp,
>> @48588304/file?time))
>> [19] Send-method on @48588268/emacs_buffer:
>> emacs_buffer->file(@48588304/file)
>> [18] '$c_call_prolog'
>> <Alien goal>
>> [17] send(@48588268/emacs_buffer, file(@48588304/file))
>> [16] Send-method on @48588268/emacs_buffer:
>> emacs_buffer->initialise(@48588304/file, @default/constant)
>> [15] '$c_call_prolog'
>> <Alien goal>
>> [14] new(_G2003,
>> emacs_buffer('d:/developpements/prolog/divers/knights.pl'))
>> [13] Send-method on @emacs/emacs:
>> emacs->goto_source_location(@48587055/source_location, @default/constant)
>> [12] '$c_call_prolog'
>> <Alien goal>
>> [11] send(@emacs/emacs,
>> goto_source_location(source_location('d:/developpements/prolog/divers/knights.pl')))
>>
>> Joël
>>
>> ----- Original Message -----
>> From: "Jan Wielemaker" <J.Wielemaker@...>
>> To: "SWI-Prolog mailinglist" <swi-prolog@...>
>> Sent: Friday, November 06, 2009 5:16 PM
>> Subject: Re: [SWIPL] Ann: Windows binary previews
>>
>>
>> > On Fri, 2009-11-06 at 17:04 +0100, Jan Wielemaker wrote:
>> >
>> >> Please test this version, notably to verify that it hasn't been
>> >> miscompiled by VC9 and that the packaging is still correct. I am
>> >> also curious to know whether this version is faster; before, the
>> >> Win32 version was compiled with VC6 and the Win64 version with
>> >> VC8.
>> >
>> > I'm not too lazy to test myself, but I do not have any physical
>> > MS-Windows machines; only a few virtual ones under VMWARE server
>> > and timing pretty inaccurate in virtual machines.
>> >
>> > Cheers --- Jan
>> >
>> > _______________________________________________
>> > SWI-Prolog mailing list
>> > SWI-Prolog@...
>> > https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog
>> >
>>
>> _______________________________________________
>> SWI-Prolog mailing list
>> SWI-Prolog@...
>> https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog
>
>

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: alarm

2009-11-07T02:18:11Z

On Sat, 2009-11-07 at 10:18 +0900, Nicolas Pelletier wrote:

> Hello,
>
> On Sat, Nov 7, 2009 at 02:50, PF <paul.i.fodor@...> wrote:
> > Sorry, I sent my post with the wrong header. My post has nothing to do with
> > Jan's post.
> > Paul.
> > On Fri, Nov 6, 2009 at 12:47 PM, PF <paul.i.fodor@...> wrote:
> >>
> >> Hi,
> >> I'm using alarm from the C time library to call a predicate in the future
> >> and I get the error:
> >> ERROR: Not enough resources: timers
> >> How can I increase this number of timers? Is it set by the OS? Is it a
> >> constant in the configuration of SWI? It is a constant set in the SWI
> >> sources and should I compile SWI to change it?
>
> Under Unix/Linux, timers are implemented using signals (SIGALRM). The
> OS sets a limit to the number of pending signals for a process, which
> you can get and set on the C side of things with getrlimit and
> setrlimit. Your shell will typically have a built-in command called
> 'ulimit' which wraps the call to getrlimit/setrlimit. So you will have
> to call 'ulimit' from the shell where you start SWI (if that is how
> you start it).
>
> Note however that these limits are "soft limits": you cannot set them
> arbitrarily high, the OS has a hard limit it will never cross (barring
> recompilation). In my experience, the bottom-line is: if you run out
> of signals/file descriptors/etc... then this is a sign the design is
> wrong. As for scheduling events, using one "timer manager thread"
> (e.g. only ever one signal) which will use alarm, and then wake up the
> other threads/processes by sending them a message scales well, and
> gives a better overall time behaviour (e.g. more predictible, fewer
> interruptions for the OS) than having every thread use alarm itself.

It is probably a good idea to share the same implementation. At some
points (before threading), the library used SIGALARM on Unix and MM
timers on Windows. With threading, the SIGALARM option no longer
worked, so the Unix version now uses a separate thread to do all
the scheduling. The windows version still uses the MM timers,
but these are pretty limited.

Such a change would certainly simplify the implementation :-)

Cheers --- Jan

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Ann: Windows binary previews (fixed)

2009-11-07T02:12:34Z

Hi,

I've implemented a work-around and uploaded a new installer. Please
give it a try (check that the installer gives nov7 as production date
and you don't pick a version from some cache).

Cheers --- Jan

On Fri, 2009-11-06 at 22:18 +0100, Joel Foutelet wrote:

> Sorry my first message was cut:
> My OS is Windows Vistal Integral 32 bits.
>
> Error messages are :
>
> 1 ?- [PCE system error: adt\date.c:623: Assertion failed: sizeof(time_t) <=
> sizeof(Any)
> Stack:
> [ 6] M @48588304/file <-time(@default/constant)
> [ 5] M @48588268/emacs_buffer ->slot(time_stamp, @48588304?time)
> [ 4] M @48588268/emacs_buffer ->file(@48588304/file)
> [ 3] M @48588268/emacs_buffer ->initialise(@48588304/file,
> @default/constant)
> [ 2]
> [ 1] M
> @emacs/emacs ->goto_source_location(@48587055/source_location, (nil))
> ]
> Requesting host to dump stack ...
> [20] send(@48588268/emacs_buffer, slot(time_stamp,
> @48588304/file?time))
> [19] Send-method on @48588268/emacs_buffer:
> emacs_buffer->file(@48588304/file)
> [18] '$c_call_prolog'
> <Alien goal>
> [17] send(@48588268/emacs_buffer, file(@48588304/file))
> [16] Send-method on @48588268/emacs_buffer:
> emacs_buffer->initialise(@48588304/file, @default/constant)
> [15] '$c_call_prolog'
> <Alien goal>
> [14] new(_G2003,
> emacs_buffer('d:/developpements/prolog/divers/knights.pl'))
> [13] Send-method on @emacs/emacs:
> emacs->goto_source_location(@48587055/source_location, @default/constant)
> [12] '$c_call_prolog'
> <Alien goal>
> [11] send(@emacs/emacs,
> goto_source_location(source_location('d:/developpements/prolog/divers/knights.pl')))
>
> Joël
>
> ----- Original Message -----
> From: "Jan Wielemaker" <J.Wielemaker@...>
> To: "SWI-Prolog mailinglist" <swi-prolog@...>
> Sent: Friday, November 06, 2009 5:16 PM
> Subject: Re: [SWIPL] Ann: Windows binary previews
>
>
> > On Fri, 2009-11-06 at 17:04 +0100, Jan Wielemaker wrote:
> >
> >> Please test this version, notably to verify that it hasn't been
> >> miscompiled by VC9 and that the packaging is still correct. I am
> >> also curious to know whether this version is faster; before, the
> >> Win32 version was compiled with VC6 and the Win64 version with
> >> VC8.
> >
> > I'm not too lazy to test myself, but I do not have any physical
> > MS-Windows machines; only a few virtual ones under VMWARE server
> > and timing pretty inaccurate in virtual machines.
> >
> > Cheers --- Jan
> >
> > _______________________________________________
> > SWI-Prolog mailing list
> > SWI-Prolog@...
> > https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog
> >
>
> _______________________________________________
> SWI-Prolog mailing list
> SWI-Prolog@...
> https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Ann: Windows binary previews

2009-11-07T01:15:08Z

On Fri, 2009-11-06 at 22:18 +0100, Joel Foutelet wrote:
> Sorry my first message was cut:
> My OS is Windows Vistal Integral 32 bits.
>
> Error messages are :

Thanks for the complete one. At least this makes the issue clear.
Apparently Microsoft changed time_t; I guess to be 64-bits. That is a
bit nasty to fix in XPCE's architecture, but I'll find a way around.

I guess the quickest way around is to cast to intptr_t. That will do
until 2038 :-)

--- Jan

> 1 ?- [PCE system error: adt\date.c:623: Assertion failed: sizeof(time_t) <=
> sizeof(Any)
> Stack:
> [ 6] M @48588304/file <-time(@default/constant)
> [ 5] M @48588268/emacs_buffer ->slot(time_stamp, @48588304?time)
> [ 4] M @48588268/emacs_buffer ->file(@48588304/file)
> [ 3] M @48588268/emacs_buffer ->initialise(@48588304/file,
> @default/constant)
> [ 2]
> [ 1] M
> @emacs/emacs ->goto_source_location(@48587055/source_location, (nil))
> ]
> Requesting host to dump stack ...
> [20] send(@48588268/emacs_buffer, slot(time_stamp,
> @48588304/file?time))
> [19] Send-method on @48588268/emacs_buffer:
> emacs_buffer->file(@48588304/file)
> [18] '$c_call_prolog'
> <Alien goal>
> [17] send(@48588268/emacs_buffer, file(@48588304/file))
> [16] Send-method on @48588268/emacs_buffer:
> emacs_buffer->initialise(@48588304/file, @default/constant)
> [15] '$c_call_prolog'
> <Alien goal>
> [14] new(_G2003,
> emacs_buffer('d:/developpements/prolog/divers/knights.pl'))
> [13] Send-method on @emacs/emacs:
> emacs->goto_source_location(@48587055/source_location, @default/constant)
> [12] '$c_call_prolog'
> <Alien goal>
> [11] send(@emacs/emacs,
> goto_source_location(source_location('d:/developpements/prolog/divers/knights.pl')))
>
> Joël
>
> ----- Original Message -----
> From: "Jan Wielemaker" <J.Wielemaker@...>
> To: "SWI-Prolog mailinglist" <swi-prolog@...>
> Sent: Friday, November 06, 2009 5:16 PM
> Subject: Re: [SWIPL] Ann: Windows binary previews
>
>
> > On Fri, 2009-11-06 at 17:04 +0100, Jan Wielemaker wrote:
> >
> >> Please test this version, notably to verify that it hasn't been
> >> miscompiled by VC9 and that the packaging is still correct. I am
> >> also curious to know whether this version is faster; before, the
> >> Win32 version was compiled with VC6 and the Win64 version with
> >> VC8.
> >
> > I'm not too lazy to test myself, but I do not have any physical
> > MS-Windows machines; only a few virtual ones under VMWARE server
> > and timing pretty inaccurate in virtual machines.
> >
> > Cheers --- Jan
> >
> > _______________________________________________
> > SWI-Prolog mailing list
> > SWI-Prolog@...
> > https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog
> >
>
> _______________________________________________
> SWI-Prolog mailing list
> SWI-Prolog@...
> https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: alarm

2009-11-06T17:18:45Z

Hello,

On Sat, Nov 7, 2009 at 02:50, PF <paul.i.fodor@...> wrote:

> Sorry, I sent my post with the wrong header. My post has nothing to do with
> Jan's post.
> Paul.
> On Fri, Nov 6, 2009 at 12:47 PM, PF <paul.i.fodor@...> wrote:
>>
>> Hi,
>> I'm using alarm from the C time library to call a predicate in the future
>> and I get the error:
>> ERROR: Not enough resources: timers
>> How can I increase this number of timers? Is it set by the OS? Is it a
>> constant in the configuration of SWI? It is a constant set in the SWI
>> sources and should I compile SWI to change it?

Under Unix/Linux, timers are implemented using signals (SIGALRM). The
OS sets a limit to the number of pending signals for a process, which
you can get and set on the C side of things with getrlimit and
setrlimit. Your shell will typically have a built-in command called
'ulimit' which wraps the call to getrlimit/setrlimit. So you will have
to call 'ulimit' from the shell where you start SWI (if that is how
you start it).

Note however that these limits are "soft limits": you cannot set them
arbitrarily high, the OS has a hard limit it will never cross (barring
recompilation). In my experience, the bottom-line is: if you run out
of signals/file descriptors/etc... then this is a sign the design is
wrong. As for scheduling events, using one "timer manager thread"
(e.g. only ever one signal) which will use alarm, and then wake up the
other threads/processes by sending them a message scales well, and
gives a better overall time behaviour (e.g. more predictible, fewer
interruptions for the OS) than having every thread use alarm itself.

Hope this helps,

--
Nicolas
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Ann: Windows binary previews

2009-11-06T13:18:55Z

Sorry my first message was cut:
My OS is Windows Vistal Integral 32 bits.

Error messages are :

1 ?- [PCE system error: adt\date.c:623: Assertion failed: sizeof(time_t) <=
sizeof(Any)
Stack:
[ 6] M @48588304/file <-time(@default/constant)
[ 5] M @48588268/emacs_buffer ->slot(time_stamp, @48588304?time)
[ 4] M @48588268/emacs_buffer ->file(@48588304/file)
[ 3] M @48588268/emacs_buffer ->initialise(@48588304/file,
@default/constant)
[ 2]
[ 1] M
@emacs/emacs ->goto_source_location(@48587055/source_location, (nil))
]
Requesting host to dump stack ...
[20] send(@48588268/emacs_buffer, slot(time_stamp,
@48588304/file?time))
[19] Send-method on @48588268/emacs_buffer:
emacs_buffer->file(@48588304/file)
[18] '$c_call_prolog'
<Alien goal>
[17] send(@48588268/emacs_buffer, file(@48588304/file))
[16] Send-method on @48588268/emacs_buffer:
emacs_buffer->initialise(@48588304/file, @default/constant)
[15] '$c_call_prolog'
<Alien goal>
[14] new(_G2003,
emacs_buffer('d:/developpements/prolog/divers/knights.pl'))
[13] Send-method on @emacs/emacs:
emacs->goto_source_location(@48587055/source_location, @default/constant)
[12] '$c_call_prolog'
<Alien goal>
[11] send(@emacs/emacs,
goto_source_location(source_location('d:/developpements/prolog/divers/knights.pl')))

Joël

----- Original Message -----
From: "Jan Wielemaker" <J.Wielemaker@...>
To: "SWI-Prolog mailinglist" <swi-prolog@...>
Sent: Friday, November 06, 2009 5:16 PM
Subject: Re: [SWIPL] Ann: Windows binary previews

> On Fri, 2009-11-06 at 17:04 +0100, Jan Wielemaker wrote:
>
>> Please test this version, notably to verify that it hasn't been
>> miscompiled by VC9 and that the packaging is still correct. I am
>> also curious to know whether this version is faster; before, the
>> Win32 version was compiled with VC6 and the Win64 version with
>> VC8.
>
> I'm not too lazy to test myself, but I do not have any physical
> MS-Windows machines; only a few virtual ones under VMWARE server
> and timing pretty inaccurate in virtual machines.
>
> Cheers --- Jan
>
> _______________________________________________
> SWI-Prolog mailing list
> SWI-Prolog@...
> https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog
>

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: bagof/3 in 5.8.0

2009-11-06T13:16:20Z

Thank you very much for the crystal clear explanation, professor; I wish Jan could include this post of yours (somewhat edited, of course) in the documentation of bagof; you see, the problem with swip documentation is that it pre-supposes that everyone understands the semantics of every predicate published thereof. SWIP is one of the best software on the planet and contributions like yours could make it better understood (as it is a logical reasoning environment, on top of a logical language).

Here are some facts:

I used bagof too many times without caring that much about the 2nd argument
When in doubt, I marked unimportant(!) data as free variables
And then, when the code changed, I debugged (and debugged); and most of the hard stuff revealed themselves to be there
And of course, like everybody else, I used underscores in the 2nd argument quite a few times (as it was obvious to me what I meant)

Now, to your questions:

Why have I used it? Because it happened to be there and seemed to be what both it's name and (scant) documentation indicated.

Why I thought that was the best that I could do? Because it happened to be there ... and I did not think I could do any better myself.

Have I ever doubted that? Nope, I doubted myself.

Have I changed my opinion? Now that you mention it, yes, effectively.

But not thoroughly, as your "ps" leaves something to be desired à la Fermat.

Regards,

Ceyhun

On Thu, Nov 5, 2009 at 10:21 PM, Bart Demoen <bart.demoen@...> wrote:

Ceyhun Ciper wrote:

> Why do you avoid bagof/3 like the plague?

It is crucial to use bagof/3 only when you know exactly which
variables it contains - also deep inside its second argument. If you
are not sure, or if things change, you will end up writing stuff with
^ in the goal, just for your protection, even when not necessary
initially.

Suppose you use bagof/3 on a ADT for trees, and for some reason the
implementation of the ADT suddenly uses a different internal
representation introducing variables (that could be bound meaningfully
as well, for instance the extra slot in a node indicates the depth of
the subtree, but as long as it has not been computed, it is variable;
once computed it is a fixed integer) you are probably in deep problems.

Unless you wrote all your bagof/3 calls like

bagof(Something,Tree^computesomething(Tree,Something),Somethings)

in which case you could have used findall/3 (followed by a test Somethings==[])

I am sure you can see some flaw in what I wrote, but most human
programmers don't off hand - you will need some thought. Well, that's
enough for me :-) Because my main objection against bagof/3 is that it
is just too complicated to be used in programs, especially in programs
that others need to read.

Also, consider that SWI's implementation of bagof/3 is (was by now)
buggy, and that this went unnoticed for many years. Something about
bagof/3, some feature of bagof/3 must be unsuited to programmers: this
feature is not used, or the bug would not have remained unnoticed for
so long. This feature is related to variables.

Note that Ulrich wrote
> (binding B to 1 and not binding B at all - how irritating!).
I usually misunderstand Ulrich, but this time, I feel confident that
he is not making a nice comment towards the spec of bagof/3.

Finally, my programming experience is limited like everyone's, but I
get away with findall/3 almost always. In the few cases I need
something like a findall/3 that retains the identity of variables
(something that bagof/3 can be used for), I prefer to write my own
bagging predicate.

Ceyhun Ciper writes:

> I really don't know where and/or when I should or shouldn't use it

This sounds like you use bagof/3 at least sometimes - why don't you
tell us why you have used it, why you thought that was the best you
could do, whether you ever doubted that, and whether you have changed
your opinion ?

Cheers

Bart Demoen

ps. ever seen bagof/3 goals with an underscore (explicit anonymous
variable) in the second argument ? in other contexts, _ means "I am
not interested in that part of the answer" - in bagof/3 it means

Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm

Re: Ann: Windows binary previews

2009-11-06T09:56:25Z

Hi

I can't edit files : here are the error messages :

% library(win_menu) compiled into win_menu 0.02 sec, 11,760 bytes
% library(swi_hooks) compiled into pce_swi_hooks 0.00 sec, 2,024 bytes
% The graphical front-end will be used for subsequent tracing
% c:/emacs/pl.ini compiled 0.13 sec, 767,860 bytes
XPCE 6.6.66, July 2009 for Win32: NT,2000,XP
Copyright (C) 1993-2009 University of Amsterdam.
XPCE comes with ABSOLUTELY NO WARRANTY. This is free software,
and you are welcome to redistribute it under certain conditions.
The host-language is SWI-Prolog version 5.9.0

For HELP on prolog, please type help. or apropos(topic).
on xpce, please type manpce.

1 ?- [PCE system error: adt\date.

Joël

----- Original Message -----
From: "Jan Wielemaker" <J.Wielemaker@...>
To: "SWI-Prolog mailinglist" <swi-prolog@...>
Sent: Friday, November 06, 2009 5:04 PM
Subject: [SWIPL] Ann: Windows binary previews

> Hi,
>
> The development download page (http://www.swi-prolog.org/download/devel)
> now contains new Windows installers created from the current GIT
> version. They are labeled 5.9.0, but are in fact previews for what
> is probably close to 5.8.1.
>
> Both the Win32 and Win64 installers have been compiled using Microsoft
> VC9 on WinXP-64 using Degski's config.cmd script for MS-Windows.
>
> Please test this version, notably to verify that it hasn't been
> miscompiled by VC9 and that the packaging is still correct. I am
> also curious to know whether this version is faster; before, the
> Win32 version was compiled with VC6 and the Win64 version with
> VC8.
>
> Besides the new compiler, there are several enhancements in this
> version, including a bagof/3 that makes some people happy :-)
>
> Enjoy --- Jan
>
>
>
>
> _______________________________________________
> SWI-Prolog mailing list
> SWI-Prolog@...
> https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog
>

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

alarm

2009-11-06T09:50:29Z

Sorry, I sent my post with the wrong header. My post has nothing to do with Jan's post.

Paul.

On Fri, Nov 6, 2009 at 12:47 PM, PF <paul.i.fodor@...> wrote:

Hi,
I'm using alarm from the C time library to call a predicate in the future and I get the error:

ERROR: Not enough resources: timers
How can I increase this number of timers? Is it set by the OS? Is it a constant in the configuration of SWI? It is a constant set in the SWI sources and should I compile SWI to change it?

Thank you,
Paul.
pl -g "['etalis_src/etalis.P'],set_etalis_flag(logging,off),['test_60.P'],test('test_data_010k.P'),statistics."
% parser.P compiled 0.00 sec, 8,092 bytes
% binarizer.P compiled 0.00 sec, 6,724 bytes
% compiler.P compiled 0.02 sec, 24,528 bytes
% executor.P compiled 0.00 sec, 9,336 bytes
% utils.P compiled 0.02 sec, 4,496 bytes
% flags.P compiled 0.00 sec, 1,632 bytes
% storage.P compiled 0.00 sec, 7,172 bytes
% etalis_src/etalis.P compiled 0.03 sec, 70,332 bytes
% test_60.P compiled 0.00 sec, 1,436 bytes
% test_data_010k.P compiled 0.25 sec, 1,200,440 bytes
event(alarm(exception_stock(1, 308.1, 904), 1))
event(alarm(exception_stock(1, 316.1, 709), 1))
event(alarm(exception_stock(1, 277.0, 773), 1))
event(alarm(exception_stock(1, 237.1, 946), 1))
event(alarm(exception_stock(1, 183.0, 807), 1))
event(alarm(exception_stock(1, 125.1, 938), 1))
event(alarm(exception_stock(1, 289.9, 671), 1))
event(alarm(exception_stock(1, 200.9, 866), 1))
event(alarm(exception_stock(1, 380.1, 908), 1))
event(alarm(exception_stock(1, 18.1, 910), 1))
event(alarm(exception_stock(1, 315.0, 971), 1))
event(alarm(exception_stock(1, 369.0, 808), 1))
event(alarm(exception_stock(1, 244.1, 730), 1))
event(alarm(exception_stock(1, 56.1, 842), 1))
event(alarm(exception_stock(1, 77.1, 618), 1))
event(alarm(exception_stock(1, 67.0, 842), 1))
event(alarm(exception_stock(1, 322.9, 688), 1))

ERROR: Prolog initialisation failed:
ERROR: Not enough resources: timers

Re: Ann: Windows binary previews

2009-11-06T09:47:50Z

Hi,

I'm using alarm from the C time library to call a predicate in the future and I get the error:

ERROR: Not enough resources: timers

How can I increase this number of timers? Is it set by the OS? Is it a constant in the configuration of SWI? It is a constant set in the SWI sources and should I compile SWI to change it?

Thank you,

Paul.

pl -g "['etalis_src/etalis.P'],set_etalis_flag(logging,off),['test_60.P'],test('test_data_010k.P'),statistics."
% parser.P compiled 0.00 sec, 8,092 bytes
% binarizer.P compiled 0.00 sec, 6,724 bytes
% compiler.P compiled 0.02 sec, 24,528 bytes
% executor.P compiled 0.00 sec, 9,336 bytes
% utils.P compiled 0.02 sec, 4,496 bytes
% flags.P compiled 0.00 sec, 1,632 bytes
% storage.P compiled 0.00 sec, 7,172 bytes
% etalis_src/etalis.P compiled 0.03 sec, 70,332 bytes
% test_60.P compiled 0.00 sec, 1,436 bytes
% test_data_010k.P compiled 0.25 sec, 1,200,440 bytes
event(alarm(exception_stock(1, 308.1, 904), 1))
event(alarm(exception_stock(1, 316.1, 709), 1))
event(alarm(exception_stock(1, 277.0, 773), 1))
event(alarm(exception_stock(1, 237.1, 946), 1))
event(alarm(exception_stock(1, 183.0, 807), 1))
event(alarm(exception_stock(1, 125.1, 938), 1))
event(alarm(exception_stock(1, 289.9, 671), 1))
event(alarm(exception_stock(1, 200.9, 866), 1))
event(alarm(exception_stock(1, 380.1, 908), 1))
event(alarm(exception_stock(1, 18.1, 910), 1))
event(alarm(exception_stock(1, 315.0, 971), 1))
event(alarm(exception_stock(1, 369.0, 808), 1))
event(alarm(exception_stock(1, 244.1, 730), 1))
event(alarm(exception_stock(1, 56.1, 842), 1))
event(alarm(exception_stock(1, 77.1, 618), 1))
event(alarm(exception_stock(1, 67.0, 842), 1))
event(alarm(exception_stock(1, 322.9, 688), 1))

ERROR: Prolog initialisation failed:
ERROR: Not enough resources: timers

Re: Ann: Windows binary previews

2009-11-06T08:16:03Z

On Fri, 2009-11-06 at 17:04 +0100, Jan Wielemaker wrote:

> Please test this version, notably to verify that it hasn't been
> miscompiled by VC9 and that the packaging is still correct. I am
> also curious to know whether this version is faster; before, the
> Win32 version was compiled with VC6 and the Win64 version with
> VC8.

I'm not too lazy to test myself, but I do not have any physical
MS-Windows machines; only a few virtual ones under VMWARE server
and timing pretty inaccurate in virtual machines.

Cheers --- Jan

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Ann: Windows binary previews

2009-11-06T08:04:31Z

Hi,

The development download page (http://www.swi-prolog.org/download/devel)
now contains new Windows installers created from the current GIT
version. They are labeled 5.9.0, but are in fact previews for what
is probably close to 5.8.1.

Both the Win32 and Win64 installers have been compiled using Microsoft
VC9 on WinXP-64 using Degski's config.cmd script for MS-Windows.

Please test this version, notably to verify that it hasn't been
miscompiled by VC9 and that the packaging is still correct. I am
also curious to know whether this version is faster; before, the
Win32 version was compiled with VC6 and the Win64 version with
VC8.

Besides the new compiler, there are several enhancements in this
version, including a bagof/3 that makes some people happy :-)

Enjoy --- Jan

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: news system based on prolog

2009-11-05T14:56:40Z

On Nov 6, 2009, at 5:19 AM, Alan Baljeu wrote:

>> I am using binary relations.
>>
> example:
>> article(1,'Text of the article').
>> article_lang(1,'en').
>> article_id(1,123).
>> article_category(1,2).
>
> That will work, and there are many ways to do this. If you expect
> your data will be complete like typical databases, you might prefer:
>
> article_lang(Uid, Lang) :-
> article(Uid, _Text, Lang, _Id, _Category).
>
> article_id(Uid, Id) :-
> article(Uid, Text, Lang, Id, Category).
>
> article_text(Uid, Text) :-
> article(Uid, Text, Lang, Id, Category).
>
> article_category(Uid, Category) :-
> article(Uid, Text, Lang, Id, Category).

Given a relation R with key K and other data A B C
(so there are functional dependencies R : K -> A, R : K -> B, R : K ->
C)
this can be factored as

R_ABC(K, A, B, C)
R_AB(K, A, B) + R_C(K, C)
R_AC(K, A, C) + R_B(K, B)
R_BC(K, B, C) + R_A(K, A)
R_A(K, A) + R_B(K, B) + R_C(K, C)

(This has nothing to do with Prolog. It should be covered in any good
3rd year data base paper. Date and Darwen have given strong arguments
why attributes that might be missing should be split off into separate
tables.)

All of these are logically equivalent. Using >< for the "bow tie"
operator of relational algebra,
R_ABC = R_AB >< R_C = R_AC >< R_B = R_BC >< R_A = R_A >< R_B >< R_C
R_AB = R_A >< R_B
R_AC = R_A >< R_C
R_BC = R_B >< R_C

They are not equally efficient, but which approach is better
DEPENDS ON WHAT YOU WANT TO DO WITH THE INFORMATION.

If you choose to store R_ABC as a base table and define the others
as views (all of which will be projections), then any query is going
to fetch (in the case of a data base, from disc; in the case of
Prolog, by copying from the clause store to the data structure heap)
the whole thing in one go. If you almost always need the whole tuple,
this is a good thing. There's one lookup, one fetch, and everything
fetched is useful. If you commonly need only one of the attributes,
it's a bad thing. It fetches much more than you need.

One of the interesting things in the data base world in recent years
has been the increasing interest in column stores, where what is
described to the data base system as a more or less conventional
table is actually _stored_ column-wise instead of row-wise, and
the query optimiser ensures that only the columns that are actually
needed for a particular query are fetched.

So if you almost always need only one of the attributes at a time,
you are probably better of storing the factorised form. If you
need some of the attributes together (say A and B) but not the
other, choose a semi-factored form.

The really interesting question comes when you have queries where
the key is not known. Suppose we want
project[A]( select[B=2, K>5] R_ABC )
With full tuples stored, this can be
for each tuple R_ABC(K, A, B, C)
if B = 2 and K > 5
add (A) to the set of results
With fully factored form, this can be
for each tuple R_B(K, B)
if B = 2 and K > 5
for each tuple R_A(K, A) % there's only one!
add (A) to the set of results

(If memory serves me correctly, David Warren's query optimiser
for CHAT-80 did this kind of thing.)

My advice is
- start by checking which attributes might be missing.
those attributes must be in separate tables.
NULL is evil in SQL. In Prolog, it's deadly dangerous.
- next, write your code using the fully factored form,
on the grounds that it is MUCH easier to get
message_sender(Message_ID, Sender)
right than it is to get
message(Message_ID, _, _, _, Sender, _, _, _)
right.
- when you have a clear idea of what you need to do
(and this includes doing some actual profiling!),
then denormalise in whatever way seems good to you.

SWI Prolog's ability to store large chunks of text as atoms really
pays off with something like your problem, because copying atoms
onto the heap is much cheaper than copying lists. However, getting
your code right is still very important.

>> findArt(Str,Art_text,Aid):-
>> article(Aid,Art_text),
>> search_atom(Art_text,Str),
>> article_id(Aid,Article_id).

while ugly, is much easier to write and read than
>>
>
> This could become simply
> article(A,Art_text,B,C,D,E), search_atom(Art_text, Str)

>> But how to write it efficiently in Prolog? We plan to have may be
>> millions of articles connected to diff sections/categories/sites/
>> portals etc.

How big are the articles?
Why not _measure_ (using statistics/0, or otherwise) the size of several
representations of small collections of articles, and use that to
predict how much memory a million-article collection is likely to be?
How much memory will the machine your program runs on have?

Have you considered distributing your program across several machines?

Some kinds of queries (date, size, ID) are easy to express in Prolog,
but free text queries need a different kind of index. For example,
it is very rare for anyone to actually want to search for a *string*
in the text of an article. They're more likely to be interested in
searching for a *word* (or a word with the same stem as the query) or
a phrase &c.

Have you considered putting the text of the articles in an information
retrieval system (such as the free zettair program from Melbourne) and
calling it with free text queries, getting back a list of article IDs,
and doing the rest in Prolog?

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Help with recursion

2009-11-05T14:13:05Z

On Nov 4, 2009, at 6:32 PM, Resinade wrote:

>
> Hi, I'm new to prologe and I'm trying to figure out how to do a
> recursive
> function where i can take 2 bags, such as [1,1,2,3,4,4,4] and
> [1,2,3,3,4,5]
> and add them together and get the answer [1,1,1,2,2,3,3,3,4,4,4,4,5]
> and i'd do an input such as
>
> sumb(A,B,C).
>
> where A is my first bag, B is my second bag and C is my output, any
> help
> would be apreciated.

This sounds like a homework problem.

"sumb" is a TERRIBLE name! Who can tell what it means?
The operation is bag (or multiset) union, so
bag_union(Bag1, Bag2, Union) would stand some chance of not
puzzling your human readers _too_ much.

At first sight, it looks as though SWI's merge/3 built in predicate
does what you want, so

bag_union(Bag1, Bag2, Union) :-
merge(Bag1, Bag2, Union).

Of course, if you to be able to find Bag1 and/or Bag2 given
Union, which is a perfectly reasonable thing for a Prolog
programmer to want, it gets a bit harder.

You specifically asked for help with recursion.
Recursion isn't any different in Prolog than in any other
language. If you haven't read "The Little Lisper" or its
successor "The Little Schemere", this would be a good time.
You might also find http://www.htdp.org/2003-09-26/ helpful.
Those references are about Lisp and Scheme, but many of the
ideas carry over.

The basic questions are
- what argument(s) drive(s) the recursion?
- what is(are) the base case(s)?
- what is(are) the step case(s)?

In this case, I'm going to let the Union argument drive
the recursion. The base case is when the Union is empty.

bag_union([], [], []).

The step cases occur when the Union is not empty.
The first element of the Union must be the first element
of Bag1 or the first element of Bag2, and the remaining
elements of the Union must be the union of what's left.

bag_union([X|Bag1], Bag2, [X|Union]) :-
should_choose(X, Bag1, Bag2),
bag_union(Bag1, Bag2, Union).
bag_union(Bag1, [X|Bag2], [X|Union]) :-
should_choose(X, Bag2, Bag1),
bag_union(Bag1, Bag2, Union).

When should X be chosen from Bag1 instead of Bag2?
If Bag2 is empty, we _can't_ choose from Bag2.
If Bag2 is [Y|Something], we can choose X if X =< Y
(if we are working with numbers) or X @=< Y (if we
are working with terms such as atoms or strings).

should_choose(_, _, []).
should_choose(X, _, [Y|_]) :- X @=< Y.

So now we have it, a correct version of bag union:

bag_union([], [], []).
bag_union([X|Bag1], Bag2, [X|Union]) :-
bag_union_check(Bag2, X),
bag_union(Bag1, Bag2, Union).
bag_union(Bag1, [X|Bag2], [X|Union]) :-
bag_union_check(Bag1, X),
bag_union(Bag1, Bag2, Union).

bag_union_check([], _).
bag_union_check([Y|_], X) :- X @=< Y.

This does actually work in the forward direction.
(Exercise for the reader: why doesn't it work in
the reverse direction?) However, it could be more
efficient. In particular, when X and Y are equal,
it doesn't make any difference which one we pick.
Let's start by unfolding bag_union_check/2.

bag_union([], [], []).
bag_union([X|Bag1], [], [X|Union]) :-
bag_union(Bag1, [], Union).
bag_union([X|Bag1], [Y|Bag2], [X|Union]) :-
X @=< Y,
bag_union(Bag1, [Y|Bag2], Union).
bag_union([], [X|Bag2], [X|Union]) :-
bag_union([], Bag2, Union).
bag_union([Y|Bag1], [X|Bag2], [X|Union]) :-
X @=< Y,
bag_union([Y|Bag1], Bag2, Union).

The only case of overlap is the [_|Bag1], [_|Bag2] clauses.
Let's move them first, and combine them.

bag_union([A|Bag1], [B|Bag2], [C|Union]) :-
( A @=< B ->
C = A,
bag_union(Bag1, [B|Bag2], Union)
;/* A @> B */
C = B,
bag_union([A|Bag1], Bag2, Union)
).
bag_union([X|Bag1], [], [X|Union]) :-
bag_union(Bag1, [], Union).
bag_union([], [], []).
bag_union([], [Y|Bag2], [Y|Union]) :-
bag_union([], Bag2, Union).

We note that the last two clauses are equivalent to
"Union = Bag2 and Union is a list". If we're trusting,
or are using a type checker, we can eliminate that
copying step.

bag_union([A|Bag1], [B|Bag2], [C|Union]) :-
( A @=< B ->
C = A,
bag_union(Bag1, [B|Bag2], Union)
;/* A @> B */
C = B,
bag_union([A|Bag1], Bag2, Union)
).
bag_union([X|Bag1], [], [X|Bag1]).
bag_union([], Bag2, Bag2).

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: news system based on prolog

2009-11-05T12:47:28Z

On Thu, 2009-11-05 at 22:26 +0200, Levan Cheishvili wrote:
> I agree with what you write, and believe that RDF/Sparql may be
> relevant in my case. This is what people say here.

Using the RDF infrastructure might make a lot of sense. It also
provides quite efficient persistency and an easy way to switch
from the memory-based Prolog infrastructure to external database
oriented solutions.

Cheers --- Jan

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: asynchronous callback to prolog from C thread

2009-11-05T12:44:57Z

On Thu, 2009-11-05 at 19:01 +0000, Samer Abdallah wrote:

> I'm working on the other half of my Prolog-OSC module,
> specifically the part that responds to OSC messages. I'm
> using liblo, which provides a relatively simple framework
> for receiving OSC messages by managing a separate server
> thread, which calls back to a registered function when a
> message is received.
>
> The problem I'm having is that when I try to use any of the
> foreign interface functions in the call back function (even
> PL_new_term_refs()) I get a Bus Error.

A first possible reason is that the initiating thread has no
Prolog engine. See PL_create_engine() and friends.

Cheers --- Jan

> To give a flavour of what's going on, to register a callback,
> and start the server thread, I do (in C)
>
> lo_server_thread st = lo_server_thread_new(p, handle_error);
> lo_server_thread_add_method(st, pattern, typespec, prolog_handler,
> (void *)userdata);
> lo_server_thread_start(st);
>
> and the call back looks like this:
>
> static int prolog_handler(const char *path, const char *types,
> lo_arg **argv,
> int argc, lo_message msg, void *user_data)
> {
> term_t term0 = PL_new_term_refs(3);
> printf("message: %s\n",path);
> // do more stuff
> return 1;
> }
>
> Without the call to PL_new_term_ref, this works, but with it
> I get a Bus Error and Prolog crashes.
>
> Is there something I need to do to prepare the server thread?
> Or do I need to create my own server thread and use the
> lower level liblo functions to write the OSC message
> handling loop?
>
> cheers,
> Samer.
>
>
> _______________________________________________
> SWI-Prolog mailing list
> SWI-Prolog@...
> https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: news system based on prolog

2009-11-05T12:26:45Z

I agree with what you write, and believe that RDF/Sparql may be relevant in my case. This is what people say here.

On Thu, Nov 5, 2009 at 6:19 PM, Alan Baljeu <alanbaljeu@...> wrote:

>I am using binary relations.
>

example:
>article(1,'Text of the article').
>article_lang(1,'en').
>article_id(1,123).

>article_category(1,2).

That will work, and there are many ways to do this. If you expect your data will be complete like typical databases, you might prefer:

article_lang(Uid, Lang) :-
article(Uid, _Text, Lang, _Id, _Category).

article_id(Uid, Id) :-
article(Uid, Text, Lang, Id, Category).

article_text(Uid, Text) :-
article(Uid, Text, Lang, Id, Category).

article_category(Uid, Category) :-
article(Uid, Text, Lang, Id, Category).

then:

article(1, 'Text of the article', 'en', 123, 2).
article(2, 'Teuaeo', 'fr', 13, 2).

supplemented by some predicates to access the structure:

You can follow this pattern for all your structures. This is more efficient for queries that acccess several fields concurrently.

>findArt(Str,Art_text,Aid):-

> article(Aid,Art_text),
> search_atom(Art_text,Str),
> article_id(Aid,Article_id).

This could become simply article(A,Art_text,B,C,D,E), search_atom(Art_text, Str)

>But, what will happen if I need to search for inclusion of some string within articles
>of particular portals, sites and categories? And will need to fetch additional properties
>like article_author, article_time etc? Or sort by 2-3 fields and extract just 20 out of all results.

This is trivial to write if you only have one predicate for each concept. Just invoke each predicate and equate matching fields.

> But how to write it efficiently in Prolog? We plan to have may be millions of articles connected to diff sections/categories/sites/portals etc.

SWI indexing is primarily on the first fields of a query, so if your queries are primarily linking records by the key field this will be efficient, basically one linear search for articles, and then filtering those articles by indexed lookups of the other tables.

But I have no experience with large datasets. There are many other approaches to efficiency, and somebody else will have to speak up if this is not quick enough for you.

--
Best Regards
Levan Cheishvili

__________________________________________________________________
Connect with friends from any web browser - no download required. Try the new Yahoo! Canada Messenger for the Web BETA at http://ca.messenger.yahoo.com/webmessengerpromo.php

--
Best Regards
Levan Cheishvili

Re: news system based on prolog

2009-11-05T12:24:11Z

I think you are right. I have downloaded triple20, will try it.

On Thu, Nov 5, 2009 at 6:42 PM, Torbjörn Lager <torbjorn.lager@...> wrote:

If you are going to target SWI-Prolog in particular, I recommend that
you take a close look at the RDF libraries. They will give all that
you need in terms of efficient search, ability to store millions of
"facts", etc.

- Torbjörn

On Thu, Nov 5, 2009 at 4:26 PM, Levan Cheishvili
<levan.cheishvili@...> wrote:
> Hi ALL!
>
> I am participating in the news system project. The "engine" or "data store"
> will be based on prolog.
> As am quite new to prolog and I am asking for your kind advices and
> directions.
>
>
> I am using binary relations.
> Uid - is surrogate key unique all over the DB
> all terms are in form of :
> term(uid,data).
>
> example:
> article(1,'Text of the article').
> article_lang(1,'en').
> article_id(1,123).
>
> 1 - is the surrogate key
>
> article_category(1,2).
> 2 - is a surrogate key of category connected to the article
>
> category(2,'Name of category').
> category_id(2,222). % 222 is id of the category
>
> ....
>
> section(100,'Name of section').
> section_lang(100,'en').
> section_category(100,2).
> ......
>
>
> site(1200,'Name of a site').
> site_section(1200,100).
>
> .....
>
> portal(3000,'Name of portal').
> portal_lang(3000,'en').
> portal_site(3000,2).
>
> ....
> And so on.
>
> My question: is the above Data structure efficient for the news project?
> For example search:
>
> If I want to find all articles which contain some string:
> search_atom(Fullstr,Str):-
> downcase_atom(Fullstr, FullstrL),
> downcase_atom(Str, StrL),
> sub_atom(FullstrL,_,_,_,StrL).
>
>
> findArt(Str,Art_text,Article_id):-
> article(Aid,Art_text),
> search_atom(Art_text,Str),
> article_id(Aid,Article_id).
>
> %this is quite easy and gives me all articles with the string , one by one.
>
> Even if I want to get full list unique articles which contain the string it
> is not a problem.
>
> find_list_of_arts(Str,List):-
> setof(
> Res,
>    (
>     findArt(Str,Art_text,Article_id),
>     Res=[article_id-Article_id, text-Art_text]
>    ),
>    List
> ).
>
> findAll(Str,Result):-
> findall(List,find_list_of_arts(Str,List),Result).
>
> even if there are 30-50000 of atoms, result is generated in acceptable time.
>
> But, what will happen if I need to search for inclusion of some string
> within articles of particular portals, sites and categories? And will need
> to fetch additional properties like article_author, article_time etc? Or
> sort by 2-3 fields and extract just 20 out of all results.
> In SQL it is quite easy to write one sentence or split into several
> statements and will gain memory and speed on big data sets.
>
> But how to write it efficiently in Prolog? We plan to have may be millions
> of articles connected to diff sections/categories/sites/portals etc.
>
> --
> Best Regards
> Levan Cheishvili
>
>

--
Torbjörn Lager
Professor of General and Computational Linguistics
Department of Philosophy, Linguistics and Theory of Science
University of Gothenburg
Box 200, SE-405 30 Gothenburg, Sweden
Phone: +46317864413

--
Best Regards
Levan Cheishvili