🚀 go-pugleaf

/* Written  2:18 pm  Feb  6, 1989 by gerdeman@clio.las.uiuc.edu in clio:comp.lang.prolog */
On this same topic, I have noticed some peculiar features
of 'findall' in TI-Prolog. I don't know whether these are
standard or peculiar to Texas Instrument.  
 
First: Why does 'findall not require existential quantification
at all? The manual doesn't say anything at all about this.

Second: why does 'findall' change variable names? For example consider
the following perverse little procedure:

?-findall(X,member(X,[Y,Z],S).

 X = _54
 Y = _98
 Z = _111
 S = [209,201]

At first I thought this was just an annoying quirk but then I 
discovered that this feature is useful for copying in the
following peculiar program:

copy(X,Y):- findall(Z,Z=X,[Y]).

This procedure turns out to be much faster than the copy
program given in Sterling and Shapiro (which uses assert
and retract). I used this copy procedure in my own implementation
of a PATR style parser and it sped things up by an order of 
magnitude. Then I discovered that this explicit copying is not 
necessary since any destructive procedure embedded inside
of 'findall' will fail to have destructive consequences.
For example 'graph_unify/3' (for DAG unification) normally
binds variables in its input arguments but 

non_destructive_unify(A,B,A_u_B):-
  findall(X,graph_unify(A,B,X),A_u_B).

Does anyone out there know why this works?

                                 ---- Dale Gerdemann
                                      University of Ill
                                      Dept Linguistics




/* End of text from clio:comp.lang.prolog */