This repository is archived. See https://github.com/GavinMendelGleason/mavis/ for the current maintainer.
:- use_module(library(mavis)).
%% even(+X:integer) is semidet.
even(X) :-
0 is X mod 2.
The =mavis= module (because she helps with typing ;-) allows one to use optional type declarations in Prolog code. During development, these declarations throw informative exceptions when values don't match types. A typical development environment converts this into a helpful stack track which assists in locating the error.
In production, the declarations are completely removed by macros
and do nothing. Production time
is defined as any time when optimization is enabled:
current_prolog_flag(optimise, true)
.
Type declarations can be give manually by calling the/2. mavis
also inserts
type declarations for you based on your PlDoc structured comments. For
example, during development, the definition of even
above becomes
even(A) :-
the(integer, A),
0 is A mod 2.
We love dynamic types. That's one reason we love Prolog. But sometimes more precise types are a helpful tool. They can:
- offer documentation to those reading our code
- help find errors during development
- structure our thinking during development
- provide data for static analysis tools
Mavis types are defined using error:has_type/2. We might define an
even_integer
type with
error:has_type(even_integer, X) :-
0 is X mod 2.
We can use the definition manually:
frobnify(A, B) :-
the(integer, A),
the(even_integer, B),
B is 2*A.
or simply add it to our PlDoc comments:
%% frobnify(+A:integer, -B:even_integer)
frobnify(A, B) :-
B is 2*A.
We can declare types for bound variables, like A
, and
not-yet-bound variables, like B
. The type constraints are implemented
with when/2 so they apply as soon as a variable is ground.
To disable type checking in production, start Prolog with the
-O
command line argument. A macro eliminates calls to the/2 so they
have no runtime overhead.
- Fix packaging error
Using SWI-Prolog 6.3.16 or later:
$ swipl
1 ?- pack_install(mavis).
Source code available and pull requests accepted on GitHub: https://github.com/mndrix/mavis