Most probably because of AI::Prolog

I have the impression - I might be wrong - that these clauses can be represented as entries in a 3x3x3 matrix = (name x profession x fruit) in the example given.

Entries would be undef, 0 and 1 and each 3x3 sub-square must have exactly one 1 entry and all others 0 for a solution.

A solver could be built by filling that matrix and constantly checking all 9 sub layers for deduced entries. (see example here)

Giving a set of random clauses it should be possible to pick n til the solver proves it's solvable (rejecting contradictory clauses)

This reminds me of a graphic technique we learned at uni for finding boolean rules and logical circuits from a given truth table, but I can't remember the name.

(Update: this looks right http://en.wikipedia.org/wiki/Karnaugh_map )

You are probably right about the matrix. I have a website which contains the puzzles and there are some diagrams generated to fill them:

http://www.pawelbiernacki.net/4kids/logicalpuzzle/4x4/puzzle_4x4_en_US_0.jsp.

I think the diagram reflects the matrix you have mentioned.

The version on my website is available in different languages but I was not sure how to apply gettext on a Perl project by ExtUtils::MakeMaker, so I simply removed the translations.

Pawel Biernacki

Actually I'm not sure how complicated your clauses can become, I'm a bit too busy to dive in now.

But the example given in the OP should be solvable by filling 3 zeros for each clause and applying simple deduction rules on the 3x3 cuts.

I think the strength of prolog comes to play if you apply more complicated rules.

Please don't take this as a critic of your work, just theoretical observations. :)

update

Applying the rules given on a 3x3x3 matrix displayed by 3 cuts by name:

• like The one who likes apples is not a blacksmith => 3 zeroes at (apple&blacksmith) x names.
• and so on:

FRUITS   PROFESSIONS NAMES

            W B F   
pineapples  
apples        0 0    John
pears       0 0 0

pineapples  0
apples      0 0 0    Patrick
pears       0

pineapples  0 0 0
apples      0 0 0    James
pears           0
[download]

Deductions:

• The apple cut has only one free field
• => John is a witcher who likes apples
• All free fields in the witcher and John cut must be filled with zeroes.

FRUITS   PROFESSIONS NAMES

            W B F   
pineapples  0 0 0
apples      1 0 0    John
pears       0 0 0

pineapples  0
apples      0 0 0    Patrick
pears       0

pineapples  0 0 0
apples      0 0 0    James
pears       0   0
[download]

Deductions:

• The James cut has only one field free
• => James is a blacksmith who likes pears.
• the remaining fields in blacksmith and pears cut are filled with zeroes.

FRUITS   PROFESSIONS NAMES

            W B F   
pineapples  0 0 0
apples      1 0 0    John
pears       0 0 0

pineapples  0 0
apples      0 0 0    Patrick
pears       0 0 0

pineapples  0 0 0
apples      0 0 0    James
pears       0 1 0
[download]

• Only one field free
• => Patrick is a fisherman who like pineapples.

I did it manually, hope everything is right and I made the technique of a solver clearer.

I think you need more complicated rules to challenge people and take advantage of the power of prolog. :)

Cheers Rolf
_{(addicted to the Perl Programming Language and ☆☆☆☆ :)
Wikisyntax for the Monastery}

• it is slow, it is my algorithm + AI::Prolog, you can add a parameter debug=>1 to see what happens
• you are right
• sorry about that, I will fix the POD
• thank you, I will fix it
• it is already on github: https://github.com/pawelbiernacki/App-ForKids-LogicalPuzzleGenerator

You can add the github details in your Makefile.PL under the META_MERGE key. Then it will appear in the sidebar for your module's page on metacpan.

The details are in the ExtUtils::MakeMaker docs (search for META_MERGE) but it is perhaps easier to see from an existing implementation such as List::Util: https://metacpan.org/source/PEVANS/Scalar-List-Utils-1.50/Makefile.PL.