Problem with use_module(library(option)) and :-debug(topic)

[hnbeck]

Hi all,

at the moment I'm porting SWI Prolog code to a Logtalk object. Backend is SWI Prolog 8.3.22
In the source file in Logtalk, the line
:- use_module(library(option)).
results in error message

! Domain error: value meta_options(1,:,-) is not in domain meta_argument_specifier
!       in clause for predicate meta_options/3
!       in auxiliary clause generated for file c:/dev/workspace/ectrafficmodeller/logic/xmpp/xmpp_config.lgt
!     

It is only with library(option), other libraries work as always.

Independently from this, the line
:- debug(xmpp).
results in

!     Domain error: value debug/1 is not in domain directive
!       in directive debug/1
!       in file c:/dev/workspace/ectrafficmodeller/logic/xmpp/xmpp_util.lgt at or above line 8

For the first case, I've read somethin in the SWI Group that there may be a problem with use_module/1 in some cases, but I've not understand the discussion
For the second case, I've no idea.
Can someone help me?

(BTW: there is nearly no code, I've just at the step setting up the file structure, containg of a loader file, a Power shell script for compliation and the source file itself. :) )

[pmoura]

Note that Logtalk doesn't support a use_module/1 directive, only its own use_module/2 directive (which takes as first argument a module identifier instead of a file path). For details, see:

https://logtalk.org/manuals/refman/directives/use_module_2.html

True, there's code in some of the Prolog adapter files, including the one for SWI-Prolog, that try to expand Prolog use_module/1 directives into Logtalk use_module/2 directives. But that code doesn't work in all scenarios as is mainly intended to help in the compilation of modules as objects. One of the main issues is that the : meta-argument indicator is ambiguous, hence the error you got. Logtalk is not based on a predicate prefixing mechanism as found on modules systems. Thus, the compiler is unable to parse any argument declared as : as it can mean a goal, a closure, or neither. One possible solution is to override the problematic module meta_predicate/1 directive as explained at:

https://logtalk.org/manuals/userman/predicates.html#calling-prolog-module-meta-predicates

An alternative solution would to use Logtalk's own options library:

https://logtalk.org/manuals/libraries/options.html

As this library is fully portable, that would be the best solution.

For debugging support, the Logtalk debug_messages tool provides similar functionality and is also fully portable:

https://logtalk.org/manuals/devtools/debug_messages.html

You get an error in that debug/1 directive due to Logtalk not supporting (by design) the problematic use of arbitrary goals as directives.

[hnbeck]

Hi Paulo,

Thanks again, all is solved. To be honest, I still don't understand every detail, term expansions and meta predicats are a little bit blurry to me :). But because I'm working now on a much bigger application now (Key words : Metaverse ready logical traffic simulation +, AR/VR visualization over an external engine) it is now time to read the Logtalk handbook again and in more detail. It is fascinating how much work you do and have done for compatibility and development. Therefore I'm now doing my XMPP stuff in Logtalk, and my goal is to get much more value out of Logtalk and its libraries. So any hint like yours above is warmly welcomed :)

Cheers + thanks

Hans

[pmoura]

Good to know that all issues are sorted out. Let's do a Q&A Logtalk webinar in January on those topics.

[hnbeck]

That would be a good idea. Here a unstructured collection of points which would interest me from my knowledge today:

• difference (or better) consequence of this and self
• where goes the memory? -> what happens at instantiaton of objects, "persistence" of data (in the sense of assert)
• examples ("Patterns") for using categories and protocols
• help for the developers -> tools for the developers provided by Logtalk
• Logtalk way of building knowlegde heavy systems (-> "patterns" or best practice)
• Using the multitask system and event library for building responsible servers (-> "patterns")

You may know, I'm coding in Prolog now for 4 years. I was also Smalltalker for many years. Now my overall question is to use Logtalk at the best, because it is a great system. But I'm struggling a little bit to "harmonize" my Smalltalk knowledge with my Prolog knowledge :)
My usage of Logtalk is limited up to now using simple objects with no explicit instantiation, now I want more :)

Cheers

Hans

[pmoura]

Those are all interesting topics to discuss. I will lookup and suggest a date for the webinar in January. Meanwhile, some tips on some of those topics:

• on this and self, see (if not already) the self_vs_this programming example; also their glossary entries
• "[object ] instantiation" is a relation between objects that doesn't necessarily imply any sort of side effects (both the object playing the role of the class and the object playing the role of the instance can and often are static)
• for a gentle introduction to protocols and categories, see the planets programming example, notably the comments in the planets.lgt source file.
• if you're familiar with design patterns, there's also a design_patterns programming example covering the most common ones

Given your knowledge of Smalltalk, your feedback on the Smalltalk nomenclature section on the Handbook would be much appreciated.

[hnbeck]

Thx for the pointers, and I'm looking forward to a new year full of Logtalk :)
So Happy New Year to all :)

Hans