13/07/16 18:24, Paul Houle wrote:
In the following docs:
https://jena.apache.org/documentation/inference/
The arithmetic builtins are described as such:
"Note that these do not run backwards, if in sum a and c are bound and b
is unbound then the test will fail rather than bind b to (c-a). This
could be fixed."
This is repeated in the source code for said functions.
https://github.com/apache/jena/blob/master/jena-core/src/main/java/org/apache/jena/reasoner/rulesys/builtins/Sum.java
in that case at line 78.
My question is: is the only reason this does not work because the
Builtin doesn't handle it, or because the engine wouldn't be able to
work correctly if the Builtin was fixed?
If I remember correctly the engine requires a deterministic answer from
the builtin - it doesn't support builtins that expand the search space
(doesn't support return of a set of legal answers). So specific cases
where there's only one unbound to an arithmetic builtin and only one
result to bind could be supported. However, in general once you allow
arithmetic expressions to run backwards you are getting into equational
constraint solve territory.
Dave