[Termtools] semantic issues with the LP examples

[Peter Schneider-Kamp]

Dear all,

there are also a number of semantic issues with the current set
of LP examples.

Consider, for instance, the following two examples:

tpdb-4.0/talp/apt: dc_mod.pl
tpdb-4.0/talp/apt: gt_mod.pl

These do not contain any clause definitions at all. My proposal would
be to remove these from the tpdb.

There is also the issue of "pure" logic programs vs. Prolog.
There are 43 examples that contain non-trivial built-in predicates
(like numbervars, findall, etc.).

This is a problem as basically all tools handle unknown
predicates differently. For instance, consider TALP and AProVE.
TALP just removes atoms containing unknown predicates, thereby
assuming that they always terminate and succeed with the trivial
answer substitution. AProVE handles a few built-ins by translation,
others by removal. If we regard other termination tools that could
take part in the competition, the situation gets even messier:
cTI, for eample assumes finite failure for all unknown predicates,
TerminWeb throws an error when encountering unknown predicates, ...

To eliminate the influence of the translation from Prolog to logic
programs, I would suggest to remove all these examples from the
LP collections that use non-trivial built-in predicates or that are not
definite logic programs (maybe after ignoring the cut operator, which is
safe and done by all tools as far as I know). Removing these examples
would give a more fair view of the relative power of the tools.
The deleted examples could also be moved to a category Prolog.

What do you think?

Best regards,
Peter