Prolog
Purpose
Prolog is declarative: a program is a theory (facts plus rules), and running it is asking queries. Logic, not control flow, is the programming unit.
The Problem It Solves
Problems that are naturally relational — genealogy, grammar parsing, expert rules, constraint satisfaction — become tiny in Prolog. You describe what is true; the engine searches for what follows. Backtracking explores alternatives automatically, and unification does pattern matching that other languages fake with libraries.
Where It Fits
Prolog anchors the logic-languages phase: Datalog (its data-focused relative), Clingo/ASP (its answer-set cousin), and MiniZinc (constraint modeling) all descend from or interoperate with this lineage. It remains the standard for symbolic AI work.
History
Prolog is a 1970s European creation with a remarkable staying power.
Origins
Alain Colmerauer and Philippe Roussel developed Prolog at the University of Aix-Marseille between 1970 and 1972, building on Robinson’s resolution theorem proving and Kowalski’s logic programming ideas. The name abbreviates PROgrammation en LOGique.
Milestones
- 1977 — the Edinburgh (DECsystem-10) syntax, created by David Warren and colleagues, becomes the standard dialect.
- 1980s — Warren’s Abstract Machine (WAM) makes compiled Prolog practical; Japan’s Fifth Generation project puts logic programming on the world stage.
- 1995 — ISO/IEC 13211-1 standardizes the language; Erlang, type systems, and Datalog inherit key ideas.
- Today — SWI-Prolog (maintained since 1987) is the community workhorse, >1 million downloads and counting.
Current Status
Mature, niche, and alive: ISO standard plus actively developed implementations (SWI-Prolog, SICStus, GNU Prolog, Trealla, Tau Prolog for the browser), with lively communities in computational linguistics, semantic web, and education.
Stage
Prolog is a mature standard with actively maintained, modern implementations.
Maturity
Fully mature: ISO/IEC 13211-1 defines the core, and SWI-Prolog (maintained since 1987) adds practical extensions — tabling, CLP(FD) constraint solving, RDF/semantic-web support, and web services.
Governance & Maintenance
Decentralized: SWI-Prolog is BSD-2-licensed with a dedicated maintainer; SICStus is commercial with academic licensing; GNU Prolog, Trealla, and Tau Prolog (browser) are open source. The ISO committee keeps the standard current.
Popularity & Usability
Prolog ranks low in popularity surveys and high in its niches.
Adoption
Standard tooling for computational linguistics (DCG grammars), semantic web/ontology work, expert and advisory systems, and university AI courses. SWI-Prolog reports over a million downloads.
Learning Curve
The paradigm is the challenge: you define relations and let the engine search, rather than write steps. Unification and backtracking take deliberate practice. Once clicked, the same skills transfer to Datalog, ASP, and type-inference machinery.
Tooling
SWI-Prolog ships an IDE (PceEmacs), tracer/debugger, tabling, CLP(FD/BD), web server (Pengines), and SWISH — a browser playground. Language servers and VS Code extensions exist for modern editing.
Use Cases
Prolog shines when the problem IS a set of facts and rules.
Primary Domains
- Natural-language grammars (DCGs) and parsing pipelines.
- Expert/advisory systems and rule engines for business and law.
- Constraint and scheduling problems via CLP(FD).
- Semantic web: RDF reasoning, ontology queries, SPARQL adjacent work.
- Education: the language that teaches search, unification, and logic.
Strengths
Declarative clarity for relational problems, automatic backtracking search, built-in pattern matching, and a lineage that produced Erlang and modern type systems.
Weak Spots
Performance ceilings on large state spaces, fewer libraries than mainstream languages, and a small hiring pool for production maintenance.
Performance
Prolog’s performance is about search efficiency, not raw speed.
Execution Model
Programs compile to Warren Abstract Machine (WAM) instructions; queries resolve via SLD resolution with chronological backtracking. SWI-Prolog’s tabling (SLG resolution) memoizes results and terminates more programs, at the cost of memory.
Published Claims
Modern Prolog systems execute millions of logical inferences per second — fine for search- and knowledge-based workloads. The honest guidance: Prolog is not for numeric loops; keep computation in constraints (CLP(FD)) or call out to C for hot paths.
Example
A family tree with rules: facts, a recursive rule, and a query Prolog answers by backtracking.
Facts, Rules, Query
% family.pl — facts and rules
parent(tom, ann). % fact: tom is a parent of ann
parent(ann, sam).
parent(ann, liz).
grandparent(X, Z) :- % rule: X is a grandparent of Z
parent(X, Y), % if X is a parent of someone Y,
parent(Y, Z). % and that Y is a parent of Z.The comma between goals means AND; the :- reads as “if”. When the query fails to find another solution, Prolog backtracks automatically and keeps searching.
How to Run
swipl -f family.pl -g "grandparent(tom, Who), write(Who), nl, fail." # prints sam, liz
# or interactively: swipl -f family.pl then: ?- grandparent(tom, Who).The fail goal forces backtracking so you see all solutions — the standard Prolog idiom for enumerating answers.
Learn More
Official sources and free materials; the full categorized catalog is on the References & Downloads page.