Nabble - SWI Prolog

Re: Critical bug in zlib library

2009-11-26T05:37:59Z

On Thursday 26 November 2009 02:27:07 pm Roberto Bagnara wrote:

> Jan Wielemaker wrote:
> > On Tue, 2009-11-24 at 18:39 +0100, Roberto Bagnara wrote:
> >> We've found a critical bug in zlib library.
> >>
> >> The predicate test/0 above can be used to trigger the bug:
> >
> > <snip>
> >
> >> pl is blocked in this infinite loop present in function zread
> >> (zlib4pl.c):
> >>
> >> while( (rc=gz_skip_footer(ctx)) == -2 )
> >> { /* TBD: read more */
> >> }
> >>
> >> The "To Be Done" marker seems to imply that the problem is known, but
> >> we wonder if its importance has been underestimated.
> >>
> >> Note that we have found this misbehaviour in a real application that
> >> hangs unexpectedly reading gzipped files of unlucky size.
> >
> > Thanks. Fixed. The patch has been merged into both pl-devel and pl.
>
> Thanks a lot Jan, this was really critical. Another problem we have
> is that SWI-Prolog's gzipped streams accept the `update' and `append'
> open modes, but the implementation, last time I tried, did not seem
> to be correct. So the question is: are these open modes supported
> now, or perhaps will they be supported soon?

I see no trace of such things. I wouldn't even know what it means. You
can't seek around in compressed files easily, so update? Append I guess
should be technically possible, but it is certainly not on my priority
list. If anything, it would be enumerating and opening .zip members.

So, not for me. Of course, if someone extends the library, I'm happy to
include the patches (don't forget the documentation!)

Cheers --- Jan

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

Re: Critical bug in zlib library

2009-11-26T05:27:07Z

Jan Wielemaker wrote:

> On Tue, 2009-11-24 at 18:39 +0100, Roberto Bagnara wrote:
>> We've found a critical bug in zlib library.
>>
>> The predicate test/0 above can be used to trigger the bug:
>
> <snip>
>
>> pl is blocked in this infinite loop present in function zread (zlib4pl.c):
>>
>> while( (rc=gz_skip_footer(ctx)) == -2 )
>> { /* TBD: read more */
>> }
>>
>> The "To Be Done" marker seems to imply that the problem is known, but
>> we wonder if its importance has been underestimated.
>>
>> Note that we have found this misbehaviour in a real application that
>> hangs unexpectedly reading gzipped files of unlucky size.
>
> Thanks. Fixed. The patch has been merged into both pl-devel and pl.

Thanks a lot Jan, this was really critical. Another problem we have
is that SWI-Prolog's gzipped streams accept the `update' and `append'
open modes, but the implementation, last time I tried, did not seem
to be correct. So the question is: are these open modes supported
now, or perhaps will they be supported soon?
Cheers,

Roberto

--
Prof. Roberto Bagnara
Applied Formal Methods Laboratory
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara@...
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Intercepting

2009-11-26T01:36:01Z

On Wednesday 25 November 2009 11:40:56 pm bharat411 wrote:
> Thanks, that works perfectly.
>
> Any suggestions to prevent the program looping when the user closes SWI
> while the program is running? It would help if you knew exactly what input
> is supplied by SWI as part of it's exit routine, so that I could create a
> rule to intercept this input. I just need to break or prevent the loop so
> SWI can shut down properly.

Assuming we are talking about Windows, closing the window sends end-of-file
to the console. So, get0/1 binds its argument to -1.

P.s. instead of using the old get0/1, please us the ISO comliant get_code/1
(which for the sake of this application does the same). Other things to
look at are read_line_to_codes/3 and maybe get_single_char/1.

Cheers --- Jan

> Many Thanks
>
> Bharat
>
> Torbjörn Lager-2 wrote:
> > get0/1 might be what you are looking for, as in:
> >
> > ?- get0(C).
> >
> > C = 10.
> >
> > ?- get0(C).
> >
> > |: 1
> >
> > C = 49.
> >
> > ?- get0(C).
> >
> > |: 2
> >
> > C = 50.
> >
> > - Torbjörn

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

Re: subsumes/2 vs. subsumes_chk/2

2009-11-25T15:32:23Z

Bart Demoen:

>What is the optimality criterium?
>... there is no broad view, just ad hoc examples.

Absolutely! We need to start somehow!

Let's start from the intended uses. As long as a loop checker uses
subsumes_term/2 directly it is safe. Should a constraint appear
accidentally a clean error is produced. A more sophisticated loop
checker could now write:

catch(subsumes_term(G,S), error(subsumption_error,_), Action)

where Action might be a safe default action (mostly probably failure).
Or Action might be a much more sophisticated subsumption checker
taking into account the meaning of the actual constraints involved.

In this manner loop checkers that do not handle constraints at all
would be safe - they would receive an error should a constraint appear
accidentally. Loop checkers that do handle constraints get fast
operation for the simple cases and would invoke Action for the complex
ones.

So we want to maximize the number of cases handled directly by
subsumes_term/2 - taking into account all legitimate uses of
constraints. This is tantamount to asking what does constitute a
legitimate case.

Here is a case I do not consider legitimate:

?- freeze(Y, X = 1), subsumes_term_V1(X,2), ...

Since freeze/2 does not attach any information to X, subsumes_term/2
cannot get that information at all. Not too surprising as freeze/2 is
somewhat the goto of constraints.

Even worse is:

?- freeze(Y, S = s(X)), subsumes_term_V1(X, S).

This case succeeds, because X does not contain any information about
the freeze. Ideally we would have a subsumption_error.

So only constraints that attach to all variables involved can be
considered legitimate.

>subsumes_term_V1:
>
>% Problematic error:
>% ?- dif(X,_), subsumes_term_V1(c,g(X)).
>
>I guess that you consider this problematic because there is no
>instantiation of X for which this would not fail ?

Yes. And this seems to be regardless of the constraint involved. In
this case even freeze(X, write(side_effect(X))) would silently fail.

>Or is it because dif(X,_) does not constrain X ?

Also, ... But in general this second viewpoint depends on the actual
constraints. And we are currently interested to support the general
case.

>[it actually does as long as the implementation of dif is not hidden:
>one can get at the attributes of X and enforce that X becomes equal to
>the _ - I know one isn't supposed to rely on being able to do that,
>but ...]

We are in a twilight zone here. For dif/2: We can observe dif(X,_) by
the toplevel and the primitives it relies on, but I think in general
we rather should consider both answers below as acceptable for dif.

?- dif(X,_).
dif(X, _G405).

?- dif(X,_).
true.

>{I can make similar arguments for rejecting
>% ?- dif(X,_), subsumes_term_V1(X,X).
>as problematic}

How? Here, X is identical. X = X is not an event to trigger
attributes at all.

>% Problematic error:
>% ?- dif(Y,a),subsumes_term_V5(X,Y).
>% would work for dif/2. But in general?

I am hesitating because of errors like

?- Y #> Z, Z #> Y, subsumes_term_V5(X, Y).

Shall they have any meaning? The first two goals ideally should fail
but the convention in constraint solvers is that inconsistencies are
acceptable (as long as they remain visible in answers).

http://www.complang.tuwien.ac.at/ulrich/Prolog-inedit/subsumes_term.pl
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Intercepting

2009-11-25T14:40:55Z

Thanks, that works perfectly.

Any suggestions to prevent the program looping when the user closes SWI while the program is running? It would help if you knew exactly what input is supplied by SWI as part of it's exit routine, so that I could create a rule to intercept this input. I just need to break or prevent the loop so SWI can shut down properly.

Many Thanks

Bharat

Torbjörn Lager-2 wrote:

get0/1 might be what you are looking for, as in:

?- get0(C).
|:
C = 10.

?- get0(C).
|: 1

C = 49.

?- get0(C).
|: 2

C = 50.

- Torbjörn

Re: Intercepting

2009-11-25T12:26:14Z

get0/1 might be what you are looking for, as in:

?- get0(C).
|:
C = 10.

?- get0(C).
|: 1

C = 49.

?- get0(C).
|: 2

C = 50.

- Torbjörn

On Wed, Nov 25, 2009 at 9:03 PM, bharat411 <bharat411@...> wrote:

>
> Hi,
>
> I'm extremely new to SWI-Prolog, Prolog and this mailing list too =) So
> forgive me if this question is a no-brainer (I couldn't find any solutions
> by searching).
>
> I was wondering if there was a way to intercept the <return> key pressed by
> the user, i.e. wait for input, and if it's <return>, do something with it. I
> know I can't use a blank character of some sort because SWI returns the
> 'Unexpected end of clause' error. Any suggestions?
>
> Secondly, my program works off a menu based system, and I have options 1
> through 7 doing something, while entering anything other than these at the
> prompt will output an error message and loop back to the menu. If a user
> were to close SWI using the windows 'X' (top right) while the prompt was
> waiting for input, the menu loop goes on infinitely and SWI doesn't close_______________________________________________

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

Intercepting

2009-11-25T12:03:28Z

Hi,

I'm extremely new to SWI-Prolog, Prolog and this mailing list too =) So forgive me if this question is a no-brainer (I couldn't find any solutions by searching).

I was wondering if there was a way to intercept the <return> key pressed by the user, i.e. wait for input, and if it's <return>, do something with it. I know I can't use a blank character of some sort because SWI returns the 'Unexpected end of clause' error. Any suggestions?

Secondly, my program works off a menu based system, and I have options 1 through 7 doing something, while entering anything other than these at the prompt will output an error message and loop back to the menu. If a user were to close SWI using the windows 'X' (top right) while the prompt was waiting for input, the menu loop goes on infinitely and SWI doesn't close. Any suggestions that don't involve removing the invalid input catcher?

Thanks a ton!

Re: Critical bug in zlib library

2009-11-25T11:37:50Z

On Tue, 2009-11-24 at 18:39 +0100, Roberto Bagnara wrote:
> We've found a critical bug in zlib library.
>
> The predicate test/0 above can be used to trigger the bug:

<snip>

> pl is blocked in this infinite loop present in function zread (zlib4pl.c):
>
> while( (rc=gz_skip_footer(ctx)) == -2 )
> { /* TBD: read more */
> }
>
> The "To Be Done" marker seems to imply that the problem is known, but
> we wonder if its importance has been underestimated.
>
> Note that we have found this misbehaviour in a real application that
> hangs unexpectedly reading gzipped files of unlucky size.

Thanks. Fixed. The patch has been merged into both pl-devel and pl.

Cheers --- Jan

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

CHR: optimization takes forever

2009-11-25T09:58:57Z

Hello!

Here's the test program that compiles ~ 13 seconds on my machine:

:- use_module(library(chr)).
:- chr_option(debug, off).
:- chr_option(optimize, full).

:- chr_constraint baz/3, bar/2, foo/2, quux/2.

foo(R, RH), foo(A1, AH1), foo(A2, AH2),
baz(_, GR, RH), bar(GR, PR),
baz(_, GA1, AH1), bar(GA1, PA1),
baz(_, GA2, AH2), bar(GA2, PA2),
quux(R, A1) \ quux(R, A2) <=> PR == PA1, PR \== PA2 | true.

If I duplicate the only rule in the program, compilation takes forever.
Thus the question: what is time complexity of CHR optimizer? Is the test
program illegal in any way?

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

Re: compiling a file into its variants at once on the fly

2009-11-25T09:15:05Z

Hi Jan,

> Note that modules are loaded only once.

Thank you for this clear answer. Now I understand it.

> You can do such tricks using :- include/1.

Replacing use_module/1 in my examples with include/1 has resulted
exactly what I want. It is very simple, and works beautifully.
Many thanks.

(a modified version for using inlcude/1)

%%% file: b.pl %%%%%%
:-if(nb_getval(flag, true)).
d(0).
:-else.
d(1).
:- endif.
%%% end of file %%%%%%

%%% module file: a1.pl %%%%%%
:-module(a1,[]).
:-include(b).
%%% end of file %%%%%%

%%% module file: a2.pl %%%%%%
:-module(a2,[]).
:-include(b).
%%% end of file %%%%%%

%%% main.pl %%%%%%
:-nb_setval(flag, true), use_module(a1).
:-nb_setval(flag, false), use_module(a2).
%%% end of file %%%%%%

?- [main].
?- % a1 compiled into a1 0.00 sec, 1,584 bytes
% a2 compiled into a2 0.00 sec, 1,224 bytes
% /Users/cantor/tmp/main.pl compiled 0.00 sec, 4,400 bytes
true.

?- a1:d(X).
X = 0.

?- a2:d(X).
X = 1.

Kuniaki

?-

On Nov 26, 2009, at 1:26 AM, Jan Wielemaker wrote:

> Hi Kuniaki,
>
> Note that modules are loaded only once. So, for the two references to
> reexport(b, []), only the first does something. The second is a simple
> no-op. You force the module to be reloaded using load_files/2, but you
> will always get the same result for a1:d(X) and a2:d(X) because they
> both export from the same module and there is always only one instance
> of that module. I.e. a reload will wipe b and then load the new
> definitions into it.
>
> You can do such tricks using :- include/1. That loads the text verbatim
> into the module in which the directive appears, clausing all predicates
> and definitions that appear in them to exist multiple times.
>
> Cheers --- Jan
>
> On Wednesday 25 November 2009 04:59:19 pm Kuniaki Mukai wrote:
>> Hi,
>>
>> Preparing four tiny files below, I expected
>> the following results.
>>
>> ?- [main].
>>
>> ?- a1:d(X).
>> X = 0.
>>
>> ?- a2:d(X).
>> X = 1.
>>
>> However, as log below shows, it is not the case.
>> Although I read the help on module at least twice,
>> I could not point out myself what is wrong with the code below.
>>
>> What I want to do actually is to compile a file
>> into its (two) variants at once by switching parameters
>> on the fly.
>>
>> Thank you in advance for any hint.
>>
>> %%% module file: b.pl %%%%%%
>>
>> :- module(b,[]).
>> :-if(nb_getval(flag, true)).
>>
>> d(0).
>>
>> :-else.
>>
>> d(1).
>>
>> :- endif.
>>
>> %%% end of file %%%%%%
>>
>> %%% module file: a1.pl %%%%%%
>>
>> :-module(a1,[]).
>> :-reexport(b, []).
>>
>> %%% end of file %%%%%%
>>
>> %%% module file: a2.pl %%%%%%
>>
>> :-module(a2,[]).
>> :-reexport(b, []).
>>
>> %%% end of file %%%%%%
>>
>> %%% main.pl %%%%%%
>>
>> :-nb_setval(flag, true), use_module(a1).
>> :-nb_setval(flag, false), use_module(a2).
>>
>> %%% end of file %%%%%%
>>
>> %%%
>> Welcome to SWI-Prolog (Multi-threaded, 64 bits, Version 5.8.1-4-ge648422)
>> Copyright (c) 1990-2009 University of Amsterdam.
>> SWI-Prolog comes with ABSOLUTELY NO WARRANTY. This is free software,
>> and you are welcome to redistribute it under certain conditions.
>> Please visit http://www.swi-prolog.org for details.
>>
>> For help, use ?- help(Topic). or ?- apropos(Word).
>>
>> ?- [main].
>> % b compiled into b 0.00 sec, 968 bytes
>> % a1 compiled into a1 0.00 sec, 2,144 bytes
>> % a2 compiled into a2 0.00 sec, 832 bytes
>> % main compiled 0.00 sec, 4,728 bytes
>> true.
>>
>> ?- a1:d(X).
>> Correct to: "b:d(X)"? y
>> X = 0.
>>
>> ?- a2:d(X).
>> Correct to: "b:d(X)"? y
>> X = 0.
>>
>> ?- listing.
>> true.
>>
>> ?- current_module(X).
>> X = user ;
>> X = error ;
>> X = lists ;
>> X = system ;
>> X = a1 ;
>> X = toplevel_variables ;
>> X = license ;
>> X = prolog_listing ;
>> X = prolog ;
>> X = a2 ;
>> X = b ;
>> false.
>>
>> ?- module(b).
>> true.
>>
>> b: ?- listing.
>>
>> d(0).
>> true.
>>
>> b: ?- module(a1).
>> true.
>>
>> a1: ?- listing.
>> true.
>>
>> a1: ?- module(a2).
>> true.
>>
>> a2: ?- listing.
>> true.
>>
>> <
>> Kuniaki Mukai
>>
>> _______________________________________________
>> 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: compiling a file into its variants at once on the fly

2009-11-25T08:26:26Z

Hi Kuniaki,

Note that modules are loaded only once. So, for the two references to
reexport(b, []), only the first does something. The second is a simple
no-op. You force the module to be reloaded using load_files/2, but you
will always get the same result for a1:d(X) and a2:d(X) because they
both export from the same module and there is always only one instance
of that module. I.e. a reload will wipe b and then load the new
definitions into it.

You can do such tricks using :- include/1. That loads the text verbatim
into the module in which the directive appears, clausing all predicates
and definitions that appear in them to exist multiple times.

Cheers --- Jan

On Wednesday 25 November 2009 04:59:19 pm Kuniaki Mukai wrote:

> Hi,
>
> Preparing four tiny files below, I expected
> the following results.
>
> ?- [main].
>
> ?- a1:d(X).
> X = 0.
>
> ?- a2:d(X).
> X = 1.
>
> However, as log below shows, it is not the case.
> Although I read the help on module at least twice,
> I could not point out myself what is wrong with the code below.
>
> What I want to do actually is to compile a file
> into its (two) variants at once by switching parameters
> on the fly.
>
> Thank you in advance for any hint.
>
> %%% module file: b.pl %%%%%%
>
> :- module(b,[]).
> :-if(nb_getval(flag, true)).
>
> d(0).
>
> :-else.
>
> d(1).
>
> :- endif.
>
> %%% end of file %%%%%%
>
> %%% module file: a1.pl %%%%%%
>
> :-module(a1,[]).
> :-reexport(b, []).
>
> %%% end of file %%%%%%
>
> %%% module file: a2.pl %%%%%%
>
> :-module(a2,[]).
> :-reexport(b, []).
>
> %%% end of file %%%%%%
>
> %%% main.pl %%%%%%
>
> :-nb_setval(flag, true), use_module(a1).
> :-nb_setval(flag, false), use_module(a2).
>
> %%% end of file %%%%%%
>
> %%%
> Welcome to SWI-Prolog (Multi-threaded, 64 bits, Version 5.8.1-4-ge648422)
> Copyright (c) 1990-2009 University of Amsterdam.
> SWI-Prolog comes with ABSOLUTELY NO WARRANTY. This is free software,
> and you are welcome to redistribute it under certain conditions.
> Please visit http://www.swi-prolog.org for details.
>
> For help, use ?- help(Topic). or ?- apropos(Word).
>
> ?- [main].
> % b compiled into b 0.00 sec, 968 bytes
> % a1 compiled into a1 0.00 sec, 2,144 bytes
> % a2 compiled into a2 0.00 sec, 832 bytes
> % main compiled 0.00 sec, 4,728 bytes
> true.
>
> ?- a1:d(X).
> Correct to: "b:d(X)"? y
> X = 0.
>
> ?- a2:d(X).
> Correct to: "b:d(X)"? y
> X = 0.
>
> ?- listing.
> true.
>
> ?- current_module(X).
> X = user ;
> X = error ;
> X = lists ;
> X = system ;
> X = a1 ;
> X = toplevel_variables ;
> X = license ;
> X = prolog_listing ;
> X = prolog ;
> X = a2 ;
> X = b ;
> false.
>
> ?- module(b).
> true.
>
> b: ?- listing.
>
> d(0).
> true.
>
> b: ?- module(a1).
> true.
>
> a1: ?- listing.
> true.
>
> a1: ?- module(a2).
> true.
>
> a2: ?- listing.
> true.
>
> <
> Kuniaki Mukai
>
> _______________________________________________
> 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

compiling a file into its variants at once on the fly

2009-11-25T07:59:19Z

Hi,

Preparing four tiny files below, I expected
the following results.

?- [main].

?- a1:d(X).
X = 0.

?- a2:d(X).
X = 1.

However, as log below shows, it is not the case.
Although I read the help on module at least twice,
I could not point out myself what is wrong with the code below.

What I want to do actually is to compile a file
into its (two) variants at once by switching parameters
on the fly.

Thank you in advance for any hint.

%%% module file: b.pl %%%%%%
:- module(b,[]).
:-if(nb_getval(flag, true)).
d(0).
:-else.
d(1).
:- endif.
%%% end of file %%%%%%

%%% module file: a1.pl %%%%%%
:-module(a1,[]).
:-reexport(b, []).
%%% end of file %%%%%%

%%% module file: a2.pl %%%%%%
:-module(a2,[]).
:-reexport(b, []).
%%% end of file %%%%%%

%%% main.pl %%%%%%
:-nb_setval(flag, true), use_module(a1).
:-nb_setval(flag, false), use_module(a2).
%%% end of file %%%%%%

%%%
Welcome to SWI-Prolog (Multi-threaded, 64 bits, Version 5.8.1-4-ge648422)
Copyright (c) 1990-2009 University of Amsterdam.
SWI-Prolog comes with ABSOLUTELY NO WARRANTY. This is free software,
and you are welcome to redistribute it under certain conditions.
Please visit http://www.swi-prolog.org for details.

For help, use ?- help(Topic). or ?- apropos(Word).

?- [main].
% b compiled into b 0.00 sec, 968 bytes
% a1 compiled into a1 0.00 sec, 2,144 bytes
% a2 compiled into a2 0.00 sec, 832 bytes
% main compiled 0.00 sec, 4,728 bytes
true.

?- a1:d(X).
Correct to: "b:d(X)"? y
X = 0.

?- a2:d(X).
Correct to: "b:d(X)"? y
X = 0.

?- listing.
true.

?- current_module(X).
X = user ;
X = error ;
X = lists ;
X = system ;
X = a1 ;
X = toplevel_variables ;
X = license ;
X = prolog_listing ;
X = prolog ;
X = a2 ;
X = b ;
false.

?- module(b).
true.

b: ?- listing.

d(0).
true.

b: ?- module(a1).
true.

a1: ?- listing.
true.

a1: ?- module(a2).
true.

a2: ?- listing.
true.

<
Kuniaki Mukai

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

Re: Bug in zlib

2009-11-24T23:25:56Z

I apologize for multiple posting: I got confused by mailing list manager
bounce backs.

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

Re: New to Prolog.A question.

2009-11-24T13:59:50Z

This function already exists in the library, but for the sake of the
exercise, here goes:

> substring([],[]).
> substring(X,X).
> substring(X,[H|T]):-substring(X,H).
> substring(X,[H|T]):-substring(X,T).

What you are trying to do is match any part of the second parameter to
all of the first parameter. It is therefore important to distinguish 2
different cases: you find a partial match, or you don't find a partial
match. First we'll rename it, because while you're applying this to a
string, it'll work on any list where the order matters.

Second, lets give the corrected predicate:

sublist([], _).
sublist([M|MTail], [M|STail]) :- sublist_(MTail, STail).
sublist(Pattern, [H|T]) :- sublist(Pattern, T).

sublist_([], _).
sublist_([M|MTail], [M|STail]) :- sublist_(MTail, STail).

This can probably be written shorter and more optimized, but this way
it's clear what's going on. The base case is in case your pattern is
empty, in that case it should always succeed.
The other two cases are the important bits: if the head of your
pattern matches the head of your list, then you want to check that the
rest of your pattern matches exactly with the next part of the list.
That is what the sublist_/2 predicate does. If that succeeds the
predicate succeeds and you've found a match.

In any other case, the predicate should recurse over the list, using
the complete pattern.

Cheers,
Andrew

PS. I don't think there are any errors, but I don't have prolog
installed on this computer and haven't checked it.

Quoting Raul Reinaldo <momentofpeace1@...>:

>
> Hi everyone,
> I've been trying to create function "substring"(X,Y),that returns "true" if
> Y contains X string(f.e. if X is [5,1] Y must have "5,1" in order for the
> function to be true).
> I've tried the following:
>

>
>
> But this doesn't work correctly(f.e. substring([5,1],[5,1,7]) will be
> false). What am I doing wrong?
> --
> View this message in context:
> http://old.nabble.com/New-to-Prolog.A-question.-tp26499935p26499935.html
> Sent from the SWI Prolog mailing list archive at Nabble.com.
>
> _______________________________________________
> SWI-Prolog mailing list
> SWI-Prolog@...
> https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog
>

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

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

New to Prolog.A question.

2009-11-24T11:16:04Z

Hi everyone,
I've been trying to create function "substring"(X,Y),that returns "true" if Y contains X string(f.e. if X is [5,1] Y must have "5,1" in order for the function to be true).
I've tried the following:

substring([],[]).
substring(X,X).
substring(X,[H|T]):-substring(X,H).
substring(X,[H|T]):-substring(X,T).

But this doesn't work correctly(f.e. substring([5,1],[5,1,7]) will be false). What am I doing wrong?

Re: Translation problem with Prolog

2009-11-24T10:40:04Z

BasHV:
>With these lists I know how to translate from NE to EN and from EN to FA,
>but I would like to know if it is possible to use these 2 lists and
>translate from NE to FA.
>
>I have defined the first two Prolog clauses like this:
>
>trans_ne([], []).
>trans_ne([NE|Ta], [EN|Tb]) :- ne(NE,EN), trans_ne(Ta,Tb).

Consider:

?- maplist(ne, NEs, ENs).

?- maplist(\NE^EN^ne(NE,EN), NEs, ENs).

?- maplist(\NE^FR^(ne(NE,EN),ef(EN,FR)), NEs, FRs).

Using library(apply) and library(lambda)

http://www.complang.tuwien.ac.at/ulrich/Prolog-inedit/lambda.pl
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

RE: Translation problem with Prolog

2009-11-24T09:56:48Z

Dear <Sorry, I can’t read your name because the characters are in Chinese and there is no signature block>,

You are right about there being no need for an accumulator. The translation works without them. I stand corrected.

Regards,
Mike

Michael L. Carroll
Software Integrator
GPS SE&I Program
SAIC
310-416-8357
BB: 424-207-7221

From: swi-prolog-admin@... [mailto:swi-prolog-admin@...] On Behalf Of ???
Sent: Tuesday, November 24, 2009 9:05 AM
To: SWI-Prolog maillinglist
Subject: Re: [SWIPL] Translation problem with Prolog

Bas,

if you have two relations trans_ne and trans_ef, and then you must have a transitive relation. Try to query:

?- trans_ne( [een,one], X), trans_ef(X, Y).

You will get:

X = [one, two],

Y = [un, deux].

Thus you may have a trans_nf/2:

trans_nf(NEs, FAs) :- trans_ne(NEs, Tmp), trans_ef(Tmp, FAs).

This is the general transitive law that if you have relations A->B and B->C, you must have A->C.

-------------

Besides, Carroll, I'm not really understand what you said about "rather than just check if they can be accomplished." For me, the "rather-than" method works in many cases.

----------------

And back to the list-translating problem, what if some of all terms in the list cannot be translated? For example, when you query:

?- trans_ne( [eenn, twee], ENs).

false.

Some error occurring in a list means the whole list false. Is it a meaningful answer?

On Tue, Nov 24, 2009 at 8:40 PM, BasHV <bashovestad@...> wrote:

ne(een,one).
ne(twee,two).

ef(one,un).
ef(two,deux).

With these lists I know how to translate from NE to EN and from EN to FA,
but I would like to know if it is possible to use these 2 lists and
translate from NE to FA.

I have defined the first two Prolog clauses like this:

trans_ne([], []).
trans_ne([NE|Ta], [EN|Tb]) :- ne(NE,EN), trans_ne(Ta,Tb).

trans_ef([], []).
trans_ef([EN|Ta], [EN|Tb]) :- ef(EN,FA), trans_ef(Ta,Tb).

Thanks in advance,

Bas
--
View this message in context: http://old.nabble.com/Translation-problem-with-Prolog-tp26495179p26495179.html
Sent from the SWI Prolog mailing list archive at Nabble.com.

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

Critical bug in zlib library

2009-11-24T09:39:10Z

We've found a critical bug in zlib library.

The predicate test/0 above can be used to trigger the bug:

random_codes(N, N, []) :-
!.
random_codes(Count, N, [Code|Tail]) :-
Code is random(256),
Count1 is Count+1,
random_codes(Count1, N, Tail).

test(N, File) :-
format('Writing ~a bytes~n', [N]),
gzopen(File, write, Write),
random_codes(0, N, Codes),
atom_codes(Atom, Codes),
write(Write, Atom),
close(Write),
format('Reading ~a bytes~n', [N]),
gzopen(File, read, Read),
read_stream_to_codes(Read, Codes),
close(Read).

test(N, N, _File) :-
!.
test(Count, N, File) :-
test(Count, File),
Count1 is Count+1,
test(Count1, N, File).

test :-
set_random(seed(0)),
test(3300, 5000, 'out.gz').

From our analysis the bug is triggered when the gzipped file is a
little bigger than SIO_BUFSIZE (4096 bytes).

pl is blocked in this infinite loop present in function zread (zlib4pl.c):

while( (rc=gz_skip_footer(ctx)) == -2 )
{ /* TBD: read more */
}

The "To Be Done" marker seems to imply that the problem is known, but
we wonder if its importance has been underestimated.

Note that we have found this misbehaviour in a real application that
hangs unexpectedly reading gzipped files of unlucky size.

Abramo and Roberto Bagnara

--
Prof. Roberto Bagnara
Applied Formal Methods Laboratory
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara@...
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Translation problem with Prolog

2009-11-24T09:04:47Z

Bas,

if you have two relations trans_ne and trans_ef, and then you must have a transitive relation. Try to query:

?- trans_ne( [een,one], X), trans_ef(X, Y).

You will get:

X = [one, two],

Y = [un, deux].

Thus you may have a trans_nf/2:

trans_nf(NEs, FAs) :- trans_ne(NEs, Tmp), trans_ef(Tmp, FAs).

This is the general transitive law that if you have relations A->B and B->C, you must have A->C.

-------------

Besides, Carroll, I'm not really understand what you said about "rather than just check if they can be accomplished." For me, the "rather-than" method works in many cases.

----------------

And back to the list-translating problem, what if some of all terms in the list cannot be translated? For example, when you query:

?- trans_ne( [eenn, twee], ENs).

false.

Some error occurring in a list means the whole list false. Is it a meaningful answer?

On Tue, Nov 24, 2009 at 8:40 PM, BasHV <bashovestad@...> wrote:

ne(een,one).
ne(twee,two).

ef(one,un).
ef(two,deux).

With these lists I know how to translate from NE to EN and from EN to FA,
but I would like to know if it is possible to use these 2 lists and
translate from NE to FA.

I have defined the first two Prolog clauses like this:

trans_ne([], []).
trans_ne([NE|Ta], [EN|Tb]) :- ne(NE,EN), trans_ne(Ta,Tb).

trans_ef([], []).
trans_ef([EN|Ta], [EN|Tb]) :- ef(EN,FA), trans_ef(Ta,Tb).

Thanks in advance,

Bas
--
View this message in context: http://old.nabble.com/Translation-problem-with-Prolog-tp26495179p26495179.html
Sent from the SWI Prolog mailing list archive at Nabble.com.

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

Bug in zlib

2009-11-24T08:33:02Z

Bug in zlib

2009-11-24T08:20:50Z

RE: Translation problem with Prolog

2009-11-24T07:38:31Z

Bas,

Ditto what Graham said. In addition, you might want to add an
accumulator so that your list predicates actually produce the
translations rather than just check if they can be accomplished.
Something along these lines:

trans_ne(NEs,ENs) :- trans_ne(NEs,[],Tmp),reverse(Tmp,ENs).
trans_ne([], ENs,ENs).
trans_ne([NE|Ta], Acc1, ENs) :-
ne(NE,EN),
Acc2=[EN|Acc1],
trans_ne(Ta,Acc2, ENs).

Regards,
Mike

Michael L. Carroll
Software Integrator
GPS SE&I Program
SAIC
310-416-8357
BB: 424-207-7221

-----Original Message-----
From: swi-prolog-admin@...
[mailto:swi-prolog-admin@...] On Behalf Of BasHV
Sent: Tuesday, November 24, 2009 4:40 AM
To: swi-prolog@...
Subject: [SWIPL] Translation problem with Prolog

Hi everyone,

I'm struggling with a problem and I hope someone can help me. I'm
working on
a small translation system in order to extend my knowledge of Prolog,
however there is a thing I don't get.

The situation is as follows:

I have for example 2 lists like this:

ne(een,one).
ne(twee,two).
ne(drie,three).
ne(vier,four).
ne(vijf,five).
ne(zes,six).
ne(zeven,seven).
ne(acht,eight).
ne(negen,nine).
ne(tien,ten).

ef(one,un).
ef(two,deux).
ef(three,trois).
ef(four,quatre).
ef(five,cinq).
ef(six,six).
ef(seven,sept).
ef(eight,huit).
ef(nine,neuf).
ef(ten,dix).

With these lists I know how to translate from NE to EN and from EN to
FA,
but I would like to know if it is possible to use these 2 lists and
translate from NE to FA.

I have defined the first two Prolog clauses like this:

trans_ne([], []).
trans_ne([NE|Ta], [EN|Tb]) :- ne(NE,EN), trans_ne(Ta,Tb).

trans_ef([], []).
trans_ef([EN|Ta], [EN|Tb]) :- ef(EN,FA), trans_ef(Ta,Tb).

Thanks in advance,

Bas
--
View this message in context:
http://old.nabble.com/Translation-problem-with-Prolog-tp26495179p2649517
9.html
Sent from the SWI Prolog mailing list archive at Nabble.com.

_______________________________________________
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: Translation problem with Prolog

2009-11-24T06:41:24Z

>
> The situation is as follows:
>
> I have for example 2 lists like this:
>
> ne(een,one).
> ne(twee,two).
> ne(drie,three).
> ne(vier,four).
> ne(vijf,five).
> ne(zes,six).
> ne(zeven,seven).
> ne(acht,eight).
> ne(negen,nine).
> ne(tien,ten).
>
> ef(one,un).
> ef(two,deux).
> ef(three,trois).
> ef(four,quatre).
> ef(five,cinq).
> ef(six,six).
> ef(seven,sept).
> ef(eight,huit).
> ef(nine,neuf).
> ef(ten,dix).
>
> With these lists I know how to translate from NE to EN and
> from EN to FA, but I would like to know if it is possible to
> use these 2 lists and translate from NE to FA.
>
> I have defined the first two Prolog clauses like this:
>
> trans_ne([], []).
> trans_ne([NE|Ta], [EN|Tb]) :- ne(NE,EN), trans_ne(Ta,Tb).
>
> trans_ef([], []).
> trans_ef([EN|Ta], [EN|Tb]) :- ef(EN,FA), trans_ef(Ta,Tb).
>

The second clause for trans_ef/2 is wrong - it will translate EN to EN,
not EN to EF.

NB you should have got a singleton variable warning for FA when you
compiled this clause. This is often a sign of something amiss and always
worth looking for!

To achieve your other aim you could either translate a complete list NE
to EN, and then translate this intermediate result EN to FA.
Alternatively, you could translate each element of the list NE to EN
then EN to FA.

HTH

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

Translation problem with Prolog

2009-11-24T04:40:19Z

Hi everyone,

I'm struggling with a problem and I hope someone can help me. I'm working on a small translation system in order to extend my knowledge of Prolog, however there is a thing I don't get.

The situation is as follows:

I have for example 2 lists like this:

ne(een,one).
ne(twee,two).
ne(drie,three).
ne(vier,four).
ne(vijf,five).
ne(zes,six).
ne(zeven,seven).
ne(acht,eight).
ne(negen,nine).
ne(tien,ten).

ef(one,un).
ef(two,deux).
ef(three,trois).
ef(four,quatre).
ef(five,cinq).
ef(six,six).
ef(seven,sept).
ef(eight,huit).
ef(nine,neuf).
ef(ten,dix).

With these lists I know how to translate from NE to EN and from EN to FA, but I would like to know if it is possible to use these 2 lists and translate from NE to FA.

I have defined the first two Prolog clauses like this:

trans_ne([], []).
trans_ne([NE|Ta], [EN|Tb]) :- ne(NE,EN), trans_ne(Ta,Tb).

trans_ef([], []).
trans_ef([EN|Ta], [EN|Tb]) :- ef(EN,FA), trans_ef(Ta,Tb).

Thanks in advance,

Bas

Re: help with swi prolog program

2009-11-23T19:38:10Z

Richard O'Keefe:
>setof/3 and bagof/3 were originally designed as more principled
>"relational calculus"-like operations to do database-y things.
>Very little thought was given to what they should do with non-ground
>results.
...
>Right up to the (very much delayed) time that the Prolog standard
>came out, nobody (except Lee Naish) had given any thought to what
>setof/3 and bagof/3 should do in the presence of *constrained*
>variables.
>
>That was a mistake. By the time the ISO Prolog standard came out,
>it was already obsolete.

A standard is not a substitute for research. At that time, systems
like Quintus did not have constraints - not even freeze/2. No current
practice - no standard for it.

Even today the situation is not that rosy. Look at all the various
approaches to clpfd!

What the standard did quite nicely w.r.t constraints was to very
clearly distinguish between instantiation errors and type errors. It
even reserved a place to integrate constraints via arithmetic
predicates. But that idea never caught on.

>Adding constraints to SWI Prolog made a lot of things rather puzzling.
>Subsumption checks are actually at the top of my list.

Please do not miss my current version (V5) for subsumes_term/2!

http://www.complang.tuwien.ac.at/ulrich/Prolog-inedit/subsumes_term.pl

It is certainly not the optimum, but it is difficult to proceed.
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Re: Coding Guidelines for Prolog

2009-11-23T18:29:39Z

Tom Schrijvers:
>Since you mention skeletons, have you looked at the logical loops library
>by Joachim Schimpf? ...

Logical loops, while a quite expressive formalism, have some
undesirable properties that seem to be deeply connected to the precise
formalism. In the implementation, the actual source of these problems
are the red cuts used.

As a consequence, programs that would be pure monotonic programs with
easy to grasp termination properties, now become very difficult to
model. Some examples:

Omission of solutions:

?- freeze(Xs, (Xs=[];Xs=[a])), ( foreach(a, Xs) do true ).

Finds only the first one. Expected:

?- freeze(Xs, (Xs=[];Xs=[a])), maplist(=(a),Xs).
Xs = [] ;
Xs = "a".

?- ( foreach(a,Xs) do true ), Xs = [_|_].

fails, in constrast to

?- maplist(=(a),Xs), Xs = [_|_].
Xs = "a" ;
Xs = "aa" ;

Erratic nontermination properties:

?- ( for(_,1,2), for(_,1,3) do true ).

?- ( for(I, 1,1), foreach(I, Is) do true ).

terminates, but adding Is = [_,_|_] in front loops.

These problems are very reminiscent of the many nonsteadfast programs
in the 1980s.

Another problem of loops is nesting. On each level you have to
"redeclare" variables with global scope. Try to define the following
using logical loops. Here it is defined with the traditional higher
order library(maplist) plus library(lambda).

?- Xss = [[_,_]],maplist(maplist(Y+\X^Z^(X+Y#=Z)), Xss, Zss).
Xss = [[_G416, _G419]],
Zss = [[_G449, _G452]],
_G416+Y#=_G449,
_G419+Y#=_G452.

There are nevertheless many good points of logical loops: One is that
frequently the variables that belong to each other are textually
closer and thus much easier to read. Also "adding another argument"
is often very easy to express. Further, the default variable scoping
is local, and thus anonymous variables are local.

There is still room for improvement. And somewhat I am convinced that
the traditional call/N approach will be part of it. Higher order
programming based on call/N still has not found too widespread
acceptance. In fact, so far I have seen three different definitions
of call/2 as built-in predicate in three systems that are incompatible
with call/N. Try call(functor(F,c), 0).

Library lambda:

http://www.complang.tuwien.ac.at/ulrich/Prolog-inedit/lambda.pl
_______________________________________________
SWI-Prolog mailing list
SWI-Prolog@...
https://mailbox.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog

Windows VC9 runtime version issue

2009-11-23T07:49:17Z

Hi,

I've got a bug-report for the Windows version that one needs the VC8
runtime libraries. See
http://gollem.science.uva.nl/bugzilla/show_bug.cgi?id=432

Does anyone know how to deal with this in Microsoft land? In particular,
can anyone propose a change to the NullSoft installerscript that takes
an action that would make Windows users happy, such as giving the
appropriate message or better, install the runtime for the poor soul.

You find the current installer script at

http://www.swi-prolog.org/git/pl.git?a=blob;f=src/win32/installer/pl.nsi

(click pl.nsi there to get the raw file).

Thanks --- Jan

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

RE: help with swi prolog program

2009-11-23T07:37:39Z

"People think a programming paradigm is intuitive if that's the one the learned first"

I can't speak for others, but I learned imperative languages first (BASIC -> Assembler -> Ada -> C -> Glide ... -> Prolog -> ...). At present, Prolog comes closest to my own "internal" way of expressing a problem. I do not find imperative languages more intuitive, in fact I find it frustrating to have to drop to a "lower" level of algorithm specification when I have to provide a step-by-step procedural specification.

I hope one day to find a language based on first-order logic that is even more declarative (clause-order independence, classical negation). Perhaps when answer-set programming matures it'll provide such a programming language.

Parker

Subject: RE: [SWIPL] help with swi prolog program
From: gearoid.oneill@...

RE: [SWIPL] help with swi prolog program
The first programming language I encountered was Algol 60 but I feel that Prolog is more 'intuitive', in so far as one could use such a term in relation to a programming language. However, i feel that the programming languages - rather like 'natural languages' - are not easily mastered and once one is learned, it would, in practical terms, tend to be be a person's preferred language.

The hypothesis "People think a programming paradigm is intuitive if
that's the one the learned first" seems plausible, and I would be
interested to hear of any evidence against it.

What I've heard from the Erlang people is that it's easier to teach
Erlang (with concurrency and no assignment statements) to telecoms
engineers who aren't programmers than it is to teach concurrency to
people who _are_ (imperative or OO) programmers.

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

New Windows 7: Find the right PC for you. Learn more.

Re: help with swi prolog program

2009-11-23T03:27:26Z

Hello Richard,

"Richard O'Keefe" <ok@...> wrote:

> > I can't imagine such a language, neither as a programming nor a
> > native language. For the native languages the last centuries of
> > human evolution
> > show that it's at least not quite easy to get one -- if I'm not told
> > nonsense, most inuit languages are not very suited for describing
> > the life of bush people in central africa, while central african
> > languages don't very precise fit for the life of the Inuit. :D
>
> You were told nonsense.
> The *vocabulary* of Inuit languages may not suit Central Africa,
> but that's a trivial matter, easily amended in a generation.
> The *grammar* of Inuit languages is well up to the job.

OK, I wasn't very precise here. But usually I learned (or tried to;
maybe the number of tries is too small yet) another language, I had to
pick up a few words first (vocabulary) and then had to start with
grammar while continuously picking up more words, means: *practically*
your very exact distinction is less useful.

Moreover, have you ever tried to automatically translate texts into a
different language and back again (try a much different language like
chinese or the like) or even transcribe english words/sentences to
japanese script and let somebody read that? The result may not very
well sell as english ;)

What I was trying to tell for programming language wasn't that any
language is *useless* to implement a solution for a given problem (what
indeed may be the case for special purpose languages ore special
problems), but that a given language may be of *less use* for that on
a general level because it will limit the set of generally possible
solutions to a subset.

> English has picked up a lot of words from other languages.
> New Zealanders are expected to know "iwi", "whanua", "wai", "kai",
> "mana", "kaumatua", "runanga", "tane", "wahine", "kuia", "matua",
> "tamariki", "reo", "kohanga", "maunga", "awa", "ringaringa",
> "whare", oh, the list goes on, and I haven't covered plant names
> ("kauri", "matai", "manuka", ...) or bird names ("kiwi", "weka",
> "pukeko", ...) or fish names ... It's something that happens
> fairly quickly.

AFAIK what is sold under the name "english" in any part of the world
can easily broken down into at least a lot of dialects and a few
languages (this because of it's not only a matter of *vocabulary*,
but also *grammar*).

If it were much different for programming languages, why then write
style guides? Or what about predicate names in looksLikeJavaCamelCase?

> In programming languages, we are in fact seeing a Grand Convergence.
> The designers of Ada originally rejected function pointers and
> dot syntax for method calls; now it has them. Eiffel picked up
> closures (called "agents" in Eiffel), C++ is getting them, Java is
> getting them, C# has them. C# picked up the idea of higher order
> functions on collections from Smalltalk and list comprehensions from
> Haskell and crossed them with SQL to make LINQ, which is being picked
> up by other languages.

That's how evolution works -- I was referring to a (any) given version
of a (programming) language. In contrast to native languages most of
them are at least semi-standardized (maybe only by single
implementation). My theory is, that simply no such standard will fir
for a really all-purpose programming language.

Cheers,
Eckard

signature.asc (205 bytes) Download Attachment

Re: Line numbers in HTML parsing

2009-11-22T22:45:07Z

Thanks for the advice, Jan,

I think that there might be another way based on leaving the whitespaces
in the parsed tree and matching them. However, I couldn't find how to
pattern-match a newline character... Does anyone knows that?

(The good solution is of course to hack the source, but that's going to
take a while...)

Thank again,

Guy Wiener.

Jan Wielemaker wrote:

> On Sun, 2009-11-22 at 13:35 +0200, Guy Wiener wrote:
>
>> Hello everyone,
>>
>> I want to use the SWI library HTML parser (that is, the SGML/XML
>> parser), but I also need to get the line number for elements. For
>> example, instead of just getting
>>
>> "element(foo, [...], [...])"
>>
>> I prefer
>>
>> "element(foo, LineNumber, [...], [...])"
>>
>>
>> Is there a way to get the line numbers of elements back from the library
>> parser?
>>
>
> Only in call-back mode. In normal mode you'd have to hack sgml2pl.c.
> If you do so, my first option would be to add something that allows
> adding the line-number as extra attributes to the attribute-list.
> It is not really hard because all the information is available
> AFAIK.
>
> 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: Coding Guidelines for Prolog

2009-11-22T21:23:48Z

On Nov 21, 2009, at 12:31 PM, <rahelzer@...>
<rahelzer@...> wrote:
> ins_bst(leaf,K, branch(leaf,K,leaf)).
...

One other thing. Since this uses a BST to represent
a set, I'd expect it to follow the naming convention
for sets, which is
<implementation tag>_add_element(+Set0, +Elt, -Set)
so I'd expect it to be called bstset_add_element/3 and to
live in a module devoted to such sets.

The Craft of Prolog mentions that the definition of a predicate
often looks like the definition of its main argument type, but
doesn't talk about "skeletons", mainly because that's not how I
think about code.

A skeletons approach could be very helpful IF you had a
programming tool (not entirely unlike the transformation
based program Martin Feather once wrote for either HOPE or
NPL -- can't recall which) where you could state your
starting point and the transformation steps and automatically
replay the transformation process when one or the other of
them changed.

There's been some relevant work in the Haskell community, but
there they start from a higher order specification, which can
commonly be related somehow to category theory, and just hit
it with lots of optimisations, so that the code the programmer
sees *is* the high level version, not the low level transformed
version.

A lot of things can be done as folds and unfolds (in the
functional programming sense, not the transformation sense).
I'm going to ask about this in the Haskell Café, because I
don't see how to do it myself.

For what it's worth, the Haskell code would be
data BST k = Empty | NonEmpty (BST k) k (BST k)

insert :: Ord k => BST k -> k -> BST k
insert Empty x = BST Empty k Empty
insert (NonEmpty l m r) x =
case compare x m of
LT -> BST (insert l x) m r
GT -> BST l m (insert r x)
EQ -> BST l m r

where the type of the function could be inferred by the compiler,
and the data type would be shared by all the functions. So we're
really talking about 6 lines of code.

I have a Prolog preprocessor that provides functional syntax,
so this could be written as
:- data leaf, bst(Less,Key,Greater).

insert(fork(L,K,G), X) | compare(C, X, K) ->
( C == (<) -> bst(insert(L,X),K,H)
; C == (>) -> bst(L,K,insert(H,X))
; C == (=) -> bst(L,K,G)
).
insert(leaf, X) -> bst(leaf,X,leaf).

from which the code I presented before results by
(a) desugaring and (b) outlining the case as a predicate.

What, other than better variable names, would make code
like this clearer?

One of the reasons I detest syntax colouring is that it usurps
the use of colour for other purposes. For example, you could
use
black for the type case-analysis skeleton,
blue for the ordering case-analysis skeleton,
red for the result construction code.

I've actually seen something like this done on a large scale
in the listing of a compiler. Three or four logically separate
passes had been folded into one, but the lines were coloured
according to which pass they belonged to.

writing Prolog code using

>

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

Re: help with swi prolog program

2009-11-22T20:36:30Z

On Nov 23, 2009, at 1:37 PM, Marius Filip wrote:

> I agree regarding the role of paradigms in lifting many burdens off
> the
> programmer's shoulders. But for reliable dealing with concurency a NEW
> paradigm is needed, orthogonal to what currently exists (namely
> procedural,
> functional, object-oriented, etc). Concurrency is a "new" problem
> which with
> the advent of multi-cores puts enough pressure upon the industry
> calling for
> a solution - in terms of paradigm - and the existing languages with
> their
> multi-threading patchwork are not enough anymore.

I was using a concurrent programming language back in the 1970s.
It's NOT a new problem in any interesting sense. (In fact, by
1979 I had programmed in PL/I, Burroughs Algol, Simula 67, and
Algol 68, all of which had some form of concurrency. My Burroughs
Algol programs _could_ have run on a multihead machine, although
we happened not to have one. Note the "67" and "68" in the names.)

I have no idea what a paradigm orthogonal to imperative, declarative,
&c might look like. If you do, more power to you! But it sounds very
much like "it shouldn't make any difference to the imperative part".
Well, we _have_ one of those. It's called LINDA. There's more than
one Linda implementation for Java.

It's hard to call Ada a "patchwork"; concurrency was designed into Ada
from the outset. (In the 1970s, in fact. This is NOT a new problem.)
>
> Concerning Prolog, my opinion is that it's not recoverable.

Fortunately, we have Mercury and Mozart/Oz, and there is prior art
(some of it at Quintus) for an Erlang-like Prolog. Indeed, there
is a Prolog interpreter written in Erlang.

> A completely new
> language blending the logical paradigm with the safe concurrent
> access a la
> Erlang (or some other approach), yes. But a NEW language, not Prolog
> (not
> surprisingly, Erlang is deterministic, no such thing as
> backtracking). That
> language might look and feel like Prolog, just as C++ still looks
> and feels
> like "C with classes" - but it would be quite different in nature.

The C++ I'm seeing these days does not look like C with classes.
(Except for one of my colleagues, who insists on using it so,
because he's red hot keen on writing programs that just work,
rather than using the latest and greatest of anything.)
C++ these days, thanks to things like BOOST, looks like macro
processing on steroids, and doesn't really feel like C at all.
A very nice example of how adding one feature can change the
character of a language dramatically.

>

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

RE: help with swi prolog program

2009-11-22T16:05:53Z

RE: [SWIPL] help with swi prolog program

The first programming language I encountered was Algol 60 but I feel that Prolog is more 'intuitive', in so far as one could use such a term in relation to a programming language. However, i feel that the programming languages - rather like 'natural languages' - are not easily mastered and once one is learned, it would, in practical terms, tend to be be a person's preferred language.

The hypothesis "People think a programming paradigm is intuitive if
that's the one the learned first" seems plausible, and I would be
interested to hear of any evidence against it.

What I've heard from the Erlang people is that it's easier to teach
Erlang (with concurrency and no assignment statements) to telecoms
engineers who aren't programmers than it is to teach concurrency to
people who _are_ (imperative or OO) programmers.

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

Re: Coding Guidelines for Prolog

2009-11-22T15:45:01Z

On Nov 21, 2009, at 12:31 PM, <rahelzer@...>
<rahelzer@...> wrote:

> ins_bst(leaf,K, branch(leaf,K,leaf)).
> ins_bst(branch(L1,K1,R1),K, T) :-
> ( (K @< K1) ->
> ins_bst(L1,K,L2),
> T=branch(L2,K1,R1)
> ; ( (K == K1) ->
> T = branch(L1,K1,R1)
> ; ins_bst(R1,K,R2),
> T = branch(L1,K1,R2)
> )
> ).

(A) it's pretty obviously insertion into a binary search tree
(B) it's rather ugly (those extra (...) around comparisons, the
indentation that puts things at the same conceptual level
at differnt physical levels).

Let's rewrite.

%- type bst(K) ---> leaf | branch(bst(K),K,bst(K)).
%- pred ins_bst(+bst(K), +K, ?bst(K)).

% ins_bst(+BST0, +Key, -BST1)
% is true when BST0 represents some set S0 as a
% binary search tree and BST1 is the result of
% inserting Key into S0 without rebalancing.
% BST1 represents S0 U {Key}.

ins_bst(leaf, Key, branch(leaf,Key,leaf).
ins_bst(branch(L,K,R), Key, BST) :-
compare(Order, Key, K),
ins_bst(Order, Key, BST, L, K, R).

ins_bst(<, Key, branch(L1,K,R), L0, K, R) :-
ins_bst(L0, Key, L1).
ins_bst(>, Key, branch(L,K,R1), L, K, R0) :-
ins_bst(R0, Key, R1).
ins_bst(=, _, branch(L,K,R), L, K, R).

>
> program which used it), I had no idea what it meant or whether it had
> a bug in it or not.

That's funny, because when I looked at it, it was _obvious_ what
it meant.
> Well, how much plainer could it be? Its only 11 lines of code,
> and the indentation style is sane.

No, the indentation style is not good.
The code above, using an auxiliary predicate, is 9 lines.

Don't forget, two of the recommendations are
- use brief but helpful variable names
- provide header comments.

> %% insert_binary_search_tree(*T0,*K,*T1)

No, "*" is wrong here. *T0, *K, maybe, although
+K would be more typical. But -T1.

>
> % inserts the prolog term K into a binary binary

It's pointless to say that K is a Prolog term.
Every Prolog term is a Prolog term.

>
> % search tree T0, yielding T1
> insert_binary_search_tree(leaf,K, branch(leaf,K,leaf)).
> insert_binary_search_tree(branch(L1,K1,R1),K, T) :-
> ( (K @< K1) ->
> insert_binary_search_tree(L1,K,L2),
> T=branch(L2,K1,R1)
> ; ( (K == K1) ->
> T = branch(L1,K1,R1)
> ; insert_binary_search_tree(R1,K,R2),
> T = branch(L1,K1,R2)
> )
> ).

This leaves the variable names and the wonky indentation
and the asymmetric presentation unchanged.\

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

Re: help with swi prolog program

2009-11-22T15:18:56Z

On Nov 21, 2009, at 4:46 AM, Eckard Brauer wrote:
> I can't imagine such a language, neither as a programming nor a native
> language. For the native languages the last centuries of human
> evolution
> show that it's at least not quite easy to get one -- if I'm not told
> nonsense, most inuit languages are not very suited for describing the
> life of bush people in central africa, while central african languages
> don't very precise fit for the life of the Inuit. :D

You were told nonsense.
The *vocabulary* of Inuit languages may not suit Central Africa,
but that's a trivial matter, easily amended in a generation.
The *grammar* of Inuit languages is well up to the job.

English has picked up a lot of words from other languages.
New Zealanders are expected to know "iwi", "whanua", "wai", "kai",
"mana", "kaumatua", "runanga", "tane", "wahine", "kuia", "matua",
"tamariki", "reo", "kohanga", "maunga", "awa", "ringaringa",
"whare", oh, the list goes on, and I haven't covered plant names
("kauri", "matai", "manuka", ...) or bird names ("kiwi", "weka",
"pukeko", ...) or fish names ... It's something that happens
fairly quickly.

In programming languages, we are in fact seeing a Grand Convergence.
The designers of Ada originally rejected function pointers and
dot syntax for method calls; now it has them. Eiffel picked up
closures (called "agents" in Eiffel), C++ is getting them, Java is
getting them, C# has them. C# picked up the idea of higher order
functions on collections from Smalltalk and list comprehensions from
Haskell and crossed them with SQL to make LINQ, which is being picked
up by other languages.

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