🚀 go-pugleaf

	Question of style for finished applications in Prolog.  What is
the prevailing opinion/practice for cuts?

     It seems to me that applications which do procedural work need a lot more
cuts than the books recommend.  I understand why it's a good thing to get rid
of cuts.  But most procedural predicates should have guaranteed deterministic
behavior.  Cut is the way to guarantee that.  For instance,

  program :-
     getInput(SomeInput),
     proveStuffAboutInput(SomeInput, SomeOutput),
     writeOutput(SomeOutput).

     The non-logical procedures getInput/1 and writeOutput/1, once they are
first satisfied, should never be backtracked into.  Assuming each has some
subgoals for which the same thing is true, 2/3 of the program should be
protected by cuts.

     Some predicates can be written so that unwanted backtracking will never
match the arguments.  The trivial cases of predicates with a single clause,
and the last predicate satisfied, never need cuts.  But I believe these cases
are in the minority for procedural code which models lots of if-type choices. 

	I measured a real application I wrote, with 809 clauses and 491 cuts. 
There's probably some room for improvement there, but I think that's in the
ballpark for what this type of application really needs.  Am I missing
something, or do Prolog programmers at large find this to be true?

	Looking forward to your comments,
-- 
Glenn Holliday | holliday@csgrad.cs.vt.edu