Prolog in a Nutshell
Table of Contents
Introduction to Prolog
Prolog, invented in 1972 by Alain Colmerauer and Philippe Roussel, is a niche programming language that uses Horn clauses to represent relations in the form of facts and rules. These elements together form a knowledge base that can be used to run computations or queries.
Prolog is highly suitable for symbolic reasoning and artificial intelligence applications, and excels at tasks like natural language processing, expert systems, and theorem proving, as it can handle abstract relationships and patterns without needing specific step-by-step directives as in imperative programming languages, such as Java.
Facts in Prolog
Facts serve as the basic building blocks in constructing a knowledge base in Prolog. In the example below, we establish a knowledge base about a family, which consists of two parents, two children, and four grandparents:
% anne has two children parent(anne, axel). parent(anne, victor). % tim has two children parent(tim, axel). parent(tim, victor). % tim has two parents parent(bill, tim). parent(sally, tim). % anne has two parents parent(mark, anne). parent(jane, anne). % Define genders female(anne). female(sally). female(jane). male(tim). male(axel). male(victor). male(mark). male(bill).
These facts state that:
- Anne and Tim are parents to both Axel and Victor.
- Mark and Jane are Anne's parents.
- Bill and Sally are Tim's parents.
- Anne, Sally, and Jane are female.
- Tim, Axel, Victor, Mark, and Bill are male.
Note: In Prolog, facts (atoms or constants) and predicate names must start with a lowercase letter, while variables start with an uppercase letter.
Rules in Prolog
Using rules, we can build a relationships among facts.
% Define what a mother, father, and a grandparent is mother(M, C) :- parent(M, C), female(M). father(F, C) :- parent(F, C), male(F). grandparent(X, Z) :- parent(X, Y), parent(Y, Z).
The rule named mother defines the meaning of the mother relationship: If M is a parent of C, and if M is female, then M is a mother.
Similarly, the father rule defines the meaning of the father relationship: If F is a parent of C, and if F is male, then F is a father.
Lastly, the grandparent rule defines the meaning of the grandparent relationship: if X is a parent of Y, and Y is a parent of Z, then X is a grandparent of Z.
Queries in Prolog
Having defined the relationships (or Horn clauses) using facts and rules, we can execute queries or computations on our knowledge base.
To determine if Tim is the parent of Victor, input the following query in a Prolog interpreter:
% Ask if Tim is the parent of Victor parent(tim, victor).
The interpreter will return true, confirming Tim is indeed the parent of Victor.
Variables in Prolog
We can use variables to make our queries more general, allowing us to lookup parents and grandparents:
% Lookup parents ?- parent(Parent, victor). Parent = anne ; Parent = tim. % Lookup grandparents ?- grandparent(GrandParent, victor). GrandParent = bill ; GrandParent = sally ; GrandParent = mark ; GrandParent = jane.
These queries use variables named Parent and GrandParent to retrieve all matching results for the given conditions.
Note that in Prolog, variables, like GrandParent and Parent, must begin with an uppercase letter, while constants, such as anne and tim, must start with a lowercase letter.
Recursion in Prolog
Recursion in Prolog occurs when a rule refers to itself. Such recursive rules can be especially useful for capturing complex relationships and is commonly used to navigate hierarchical data, like family trees. We use recursion to find the ancestors of a person by adding the following two rules to our knowledge base:
% Use recursion to define the concept of ancestor ancestor(X, Z) :- parent(X, Z). ancestor(X, Z) :- parent(X, Y), ancestor(Y, Z).
With these rules in place, we can enumerate all the ancestors of Victor using the variable Ancestor:
?- ancestor(Ancestor, victor). Ancestor = anne ; Ancestor = tim ; Ancestor = bill ; Ancestor = sally ; Ancestor = mark ; Ancestor = jane ; false.
The false at the end of the result set is Prolog's way of saying, "I've given you all the solutions I can find, and there are no more left."
Data Types in Prolog
In Prolog, the data in our knowledge base is represented using terms. A term is the basic units of data and can be classified into four main categories: atoms, numbers, variables, and compound terms. Let's delve deeper into each of them:
- Atoms are a sequence of characters that are used to represent a specific piece of data.
- They start with a lowercase letter.
- They are analogous to constants, or identifiers, in other programming languages.
- Examples include anne, tim, victor, etc.
- Strings enclosed in single quotes are also considered atoms, like 'Hello World'.
- Variables are placeholders for any Prolog term and are useful in queries.
- They start with an uppercase letter, or an underscore.
- Examples: Parent, Child, GrandParent, Ancestor, _.
- Compound Terms:
- Compound terms are used to represent structures in Prolog.
- Rules, facts, and many queries in Prolog are expressed using compound terms.
- Example: in the compound term parent(anne, axel), parent is the functor, and its arity is 2 because it has two arguments: anne and axel.
- Example: in the rule ancestor(X, Y) :- parent(X, Y). The rule's head (ancestor(X, Y)) is a compound term. The body (parent(X, Y)) is also a compound term.
- Different functors can have different arities, which means they can take different numbers of arguments.
- The functor provides a name for the structure, and it's followed by its arguments, which are enclosed in parentheses.
- Examples: 123, -456, 3.14.
- Prolog supports both integer and floating-point numbers.
In Prolog, everything you work with – whether it's facts, rules, or queries – revolves around terms. Understanding terms is fundamental to grasping how Prolog represents and processes data.
Debugging and Tracing Prolog Programs
You can use the trace tool to debug and trace your queries and for understanding Prolog's backtracking mechanism.
The trace/0 predicate in Prolog activates the debugger. The /0 indicates that trace is a predicate with zero arguments. When activated, it allows you to step through the execution of your program to understand its flow or locate errors.
For example, using the following query and by using trace we see how Prolog figures out that Anne is an ancestor of Victor:
[trace] ?- trace, ancestor(anne, victor). Call: (11) ancestor(anne, victor) ? creep Call: (12) parent(anne, victor) ? creep Exit: (12) parent(anne, victor) ? creep Exit: (11) ancestor(anne, victor) ? creep true .
For each step, you see a marker (Call), a call number indicating the sequence and depth (11, etc), and then the goal (or predicate) being evaluated.
The markers you'll commonly encounter are:
- Call: This indicates that Prolog is attempting to satisfy the goal.
- Exit: This indicates that the goal has succeeded.
- Fail: This indicates that the goal has failed.
- Redo: This means Prolog is backtracking to find another solution.
For example, when asking if a grandparent is an ancestor, Prolog will use backtracking to find the answer:
``bash [trace] ?- trace, ancestor(bill, victor). Call: (11) ancestor(bill, victor) ? creep Call: (12) parent(bill, victor) ? creep Fail: (12) parent(bill, victor) ? creep Redo: (11) ancestor(bill, victor) ? creep Call: (12) parent(bill, _53830) ? creep Exit: (12) parent(bill, tim) ? creep Call: (12) ancestor(tim, victor) ? creep Call: (13) parent(tim, victor) ? creep Exit: (13) parent(tim, victor) ? creep Exit: (12) ancestor(tim, victor) ? creep Exit: (11) ancestor(bill, victor) ? creep true .
Here's a summary of a few key commands: - a: Abort the trace and return to the top level. - c or Space: Continue to the next step (creep mode). ## Writing Tests Let's write a unit test for the **ancestor** predicate. First, we need to make the ancestor code a **module** that we can import from an external file and from the tests that we're going to write. Add the following to the top of the family\.pl file: ```prolog :- module(family, [ancestor/2]).
This defines a module that we can import in our tests.
Save the following code in a file named family_test.pl:
:- use_module(library(plunit)). % The code we test is saved in 'family.pl' :- use_module(family). :- begin_tests(ancestor_tests). % Direct parent-child relationships test(anne_is_ancestor_of_axel) :- % We use once to get rid of this warning: % PL-Unit: Test anne_is_ancestor_of_axel: Test succeeded with choicepoint once(ancestor(anne, axel)). test(tim_is_ancestor_of_victor) :- once(ancestor(tim, victor)). % Grandparent relationships test(bill_is_ancestor_of_axel) :- once(ancestor(bill, axel)). test(jane_is_ancestor_of_victor) :- once(ancestor(jane, victor)). % Ancestor relationships with self should fail test(anne_is_not_ancestor_of_anne, [fail]) :- once(ancestor(anne, anne)). :- end_tests(ancestor_tests).
Run the tests from the CLI with the following command:
❯ swipl -s magic_square_test.pl -g run_tests -t halt
You should see the following output:
% PL-Unit: magic_square_tests Warning: magic_square_test.pl:7: PL-Unit: Test valid_magic_square: Test succeeded with choicepoint passed 0.021 sec % test passed
The key takeaways are that unit tests for the "ancestor" predicate in Prolog require module declaration and the once/1 predicate for removing the warning ”Test succeeded with choicepoint”.
In our exploration of Prolog, we began with Facts, the foundational building blocks of our knowledge base, and then advanced to Rules, which helped us establish relationships.
With Queries, we saw the unique power of Prolog in action, drawing inferences from established facts and rules. Variables added another layer of depth, allowing more generalized and complex queries.
The use of Recursion in Prolog allowed us to solve a complex problem through self-referencing techniques.
The Data Types section provided a deep dive into the various ways data is represented, offering a clearer picture of how Prolog views and processes information.
In Debugging and Tracing we saw the inner workings of Prolog, giving us the tools to troubleshoot and optimize our programs.
Prolog: A high-level programming language based on formal logic, known for its use in artificial intelligence and symbolic reasoning tasks.
Horn clauses: A specific kind of logical expression that forms the basis for defining relationships in Prolog. They can be facts, rules, or queries.
Knowledge Base: A database of facts and rules in a Prolog program. It's where the programmer defines all the known relationships and data.
Fact: A basic assertion in Prolog that something is true. For example, parent(anne, axel). states that Anne is a parent to Axel.
Rule: A logical formula in Prolog that defines a relationship in terms of other relationships and facts. It has a head and a body, e.g., grandparent(X, Z) :- parent(X, Y), parent(Y, Z). Here, the rule states that if X is a parent of Y and Y is a parent of Z, then X is a grandparent of Z.
Query: A question posed to the Prolog interpreter to retrieve information from the knowledge base. It can return true, false, or bind values to specific variables.
Backtracking: A cornerstone of Prolog's execution mechanism. If Prolog doesn't find a solution using the current path, it goes back (backtracks) to previous decision points and tries other possibilities.
Term: The basic unit of data in Prolog. Terms include atoms, numbers, variables, and compound terms.
Atom: A specific kind of term used to represent constant values. They start with a lowercase letter, e.g., anne.
Compound Term: A structure in Prolog representing a relation between different terms. It consists of a functor and its arguments, e.g., parent(anne, axel).
Functor: The name or label of a compound term, followed by its arguments enclosed in parentheses.
Predicate: A predicate is a combination of a functor and a specific arity (number of arguments). Predicates can represent facts, rules, or queries in the Prolog database.
The Power of Prolog: https://www.metalevel.at/prolog