#1408 opproperties

[Andrej Vodopivec]

This is a bug from share_testsuite:

(%i1) load(multiadditive)$
(%i2) declare(myabs, idempotent, myabs, multiplicative)$
(%i3) myabs(xp * myabs(z));
(%o3) myabs(xp)*myabs(myabs(z))

(%o3) should be myabs(xp)*myabs(z)

Andrej

[Robert Dodier]

Logged In: YES
user_id=501686
Originator: NO

Looks like the result is not maximally simplified --- reevaluating %o3 => myabs(xp)*myabs(z) as expected.

[Barton Willis]

Logged In: YES
user_id=895922
Originator: NO

...with GCL it's OK:

(%i1) load(multiadditive);
(%o1) C:/PROGRA~1/MAXIMA~2.0/share/maxima/5.15.0/share/contrib/multiadditive.lisp
(%i2) declare(myabs, idempotent, myabs, multiplicative);
(%o2) done
(%i3) myabs(xp)*myabs(myabs(z));
(%o3) myabs(xp)*myabs(z)
(%i4) build_info();
Maxima version: 5.15.0
Maxima build date: 17:36 4/20/2008
host type: i686-pc-mingw32
lisp-implementation-type: GNU Common Lisp (GCL)
lisp-implementation-version: GCL 2.6.8

Is this a noun / verb problem?

[Barton Willis]

Logged In: YES
user_id=895922
Originator: NO

Try this experiment with non-GCL Maxima

(%i7) :lisp(trace eq idempotent);
Warning:
EQ is being redefined.
Warning:
EQ is being redefined.
(EQ IDEMPOTENT)

myabs(xp)*myabs(myabs(z));

<junk>

1> (IDEMPOTENT (($MYABS) $XP) NIL)
<1 (IDEMPOTENT (($MYABS SIMP) $XP))
1> (IDEMPOTENT (($MYABS) $Z) NIL)
<1 (IDEMPOTENT (($MYABS SIMP) $Z))
1> (IDEMPOTENT (($MYABS) (($MYABS SIMP) $Z)) NIL)
2> (EQ $MYABS $MYABS)
<2 (EQ T)
<1 (IDEMPOTENT (($MYABS SIMP) $Z))

[Robert Dodier]

Logged In: YES
user_id=501686
Originator: NO

Barton, I think you have tried the wrong example. You have myabs(xp)*myabs(myabs(z)) as the test case, but the original report has myabs(xp * myabs(z)). I see the bug when I try the original test case with GCL (on Windows). How about you?

[Barton Willis]

Logged In: YES
user_id=895922
Originator: NO

Yes, I did try the wrong example.

Changing the order of *opers-list fixes this problem. Specifically,
in contrib/multiadditive.lisp, change

(setq opers (cons '$idempotent opers)
*opers-list (cons '($idempotent . idempotent) *opers-list))

to

(setq opers (cons '$idempotent opers)
*opers-list `(,@*opers-list ($idempotent . idempotent)))

Then

(%i10) declare(myabs, multiplicative, myabs,idempotent)$
(%i11) myabs(xp * myabs(z));
(%o11) myabs(xp)*myabs(z)

After testing and collecting comments, I'll commit this fix. It's unfortunate
that the *opers-list scheme is order dependent. But that's the way it is
intended to work, I think.

[Barton Willis]

Logged In: YES
user_id=895922
Originator: NO

There are other problems with the ordering of *opers-list:

(%i1) declare(f, multiplicative, f, additive)$

Not OK:

(%i2) f(a*b+c);
(%o2) f(c)+f(a*b)
(%i3) expand(%,0,0);
(%o3) f(c)+f(a)*f(b)

Change the order of *opers-list --- no more bug:

(%i4) :lisp(setq *opers-list (reverse *opers-list));

(%i4) f(a*b+c);
(%o4) f(c)+f(a)*f(b)

[Barton Willis]

Logged In: YES
user_id=895922
Originator: NO

fixed by contrib/multiadditive CVS version 1.8. As noted below, there are other
problems with the ordering of *opers-list.

[Dieter Kaiser]

The reported problems of this bug report are no longer present:

(%i2) load(multiadditive)$
(%i3) declare(myabs,idempotent,myabs,multiplicative)$

(%i4) myabs(xp*myabs(z));
(%o4) myabs(xp)*myabs(z)

And:

(%i5) declare(f, multiplicative, f, additive)$

(%i6) f(a*b+c);
(%o6) f(c)+f(a)*f(b)

This has been fixed in asum.lisp revision 1.37.
Closing this bug report as fixed.
Dieter Kaiser