Rust implementation of prolog (originally) based on miniprolog: http://andrej.com/plzoo/html/miniprolog.html
I translated miniprolog from the above SML sources to Rust for two primary reasons:
- Firstly, I wanted a project to explore programming in Rust
- Secondly, prolog is interesting but I never felt like I properly learned it
Starting from an existing implementation was nice because it shortened the time to a working version. The current version in this repository is very much a work in progress and overly simplistic. I plan to improve it to better learn Rust and I plan to improve it as I learn Rust :)
In other words, this is really a playground for me to play with both Rust and prolog.
Tested on Linux (Fedora), OSX, and Windows.
Should be as simple as: cargo run
$ cargo run
Welcome to rust-prolog!
This prolog interpreter is based on the ML code at the PLZoo:
http://andrej.com/plzoo/html/miniprolog.html
Input syntax:
?- query. Make a query.
a(t1, ..., tn). Assert an atomic proposition.
A :- B1, ..., Bn. Assert an inference rule.
$quit Exit interpreter.
$use "filename" Execute commands from a file.
Prolog> $use "src/likes.pl"
Prolog> ?- likes(X,Y).
Y = mary
X = john
more? (y/n) [y]
I haven't figured out exactly which directions I want to take this in, but some ideas I've been kicking around include:
- Add a type system
- Add constraints and constraint checker.
- Add a bottom-up search procedure (perhaps even replace the existing top-down search procedure entirely).
- Add program transformations to generate more efficient queries (for instance, magic sets and entailment elimination).
- Change the syntax to one that supports mixfix declarations.
Completed:
- Change the search procedure to a complete search. This is now done. It uses iterative deepening with an admissable heuristic that does a lower bound estimate on the number of unifications required to solve the current goal. It also generates contrapositives (instead of using restarts).
- Add more primitive types, lists, numbers, etc and make a small standard library.
- Improve the garbage collection. Currently cycles are allowed but due to the hash consing, nothing is ever collected. A simple scheme that should allow some collection is to throw away terms generated by unification and terms that are part of a query after the query has terminated. This should lend itself to a simplistic form of regions.
Old goals that I may not pursue after all:
- Compile to the Warren Abstract Machine. The reason I've reconsidered this one is because my search procedure is slowing moving towards a more general constraint logic search procedure. The WAM is still rather elegant and I may follow the development at some point and see if I can adapt it to fit the general CPL algorithm. Perhaps someone else has already done this work.