#425 simplifer returns %i*%i


((-%i)^(1/2)*%i)*((-%i)^(1/2)) => -%i*%i

Resimplifying -- expand(%,0,0) -- correctly returns 1.

Maxima 5.9.0 gcl 2.5.0 mingw32 W2k Athlon


  • Stavros Macrakis

    Logged In: YES

    This may be related to the inconsistent simplification of simple
    expressions involving %i:

    Mathematically, %i^(1/4) = (-%i)^(-1/4), but the first
    simplifies to (-1)^(1/8) and the second to (-%i)^(1/4) .

    Mathematically, (-1)^(-1/4) = %i^(-1/2) = (-%i)^(1/2), but
    the first two simplify to 1/(-1)^(1/4), while the third simplifies
    to sqrt(-%i).

    There are other similar cases. This is also reminiscent of the
    the non-normalization of 1/sqrt(2) -- bug # 721575.

  • Robert Dodier

    Robert Dodier - 2006-07-11

    Logged In: YES

    Observed in 5.9.3cvs.

  • Robert Dodier

    Robert Dodier - 2006-07-11
    • labels: --> 460522
  • Robert Dodier

    Robert Dodier - 2006-09-09
    • labels: 460522 --> Lisp Core - Simplification
  • Dieter Kaiser

    Dieter Kaiser - 2009-05-25
    • milestone: --> Includes_proposed_fix
  • Dieter Kaiser

    Dieter Kaiser - 2009-05-25

    I had a look at this bug. Here are some more general examples related to this report:

    (%i3) a*sqrt(a*b)*sqrt(a*b);
    (%o3) a*(a*b)

    (%i4) a*b*sqrt(a*b)*sqrt(a*b);
    (%o4) a*b*(a*b)

    Now with another exponent:

    (%i5) a*b*(a*b)^(1/4)*(a*b)^(3/4);
    (%o5) a*b*(a*b)

    Here is the example from this bug report:

    (%i6) %i*sqrt(-%i)*sqrt(-%i);
    (%o6) -%i*%i

    This simplification error can be located to occur in the routine timesin in simp.lisp. This is the small piece of code which is responsible:

    ((onep1 w)
    (return (rplaca (cdr fm) (car x))))

    When two factors are equal and the exonentents of the factors sum to 1, e. g. for sqrt(a*b)*sqrt(a*b) the factor is replaced by the base, that is for this example (a*b). Unfortunately, this is a mtimes expression. The simplifier generates a result with an unsimplified and nested mtimes expression in it.

    I have tried to find a better implementation and I have found the following code which works:

    ((onep1 w)
    (cond ((mtimesp (car x))
    ;; A base which is a mtimes expression.
    ;; Remove the factor from the lists of products.
    (rplacd fm (cddr fm))
    ;; We multiply the complete base with the list
    ;; of all remaining factors.
    (setq rulesw t)
    (return (muln (nconc y (cdar x)) t)))
    ;; The old code for the other cases.
    (t (return (rplaca (cdr fm) (car x))))))

    Again the above examples with the modified code:

    (%i3) a*b*sqrt(a*b)*sqrt(a*b);
    (%o3) a^2*b^2

    (%i4) a*sqrt(a*b)*sqrt(a*b);
    (%o4) a^2*b

    (%i5) a*b*sqrt(a*b)*sqrt(a*b);
    (%o5) a^2*b^2

    (%i6) a*b*(a*b)^(1/4)*(a*b)^(3/4);
    (%o6) a^2*b^2

    That is the example of this bug report:

    (%i7) %i*sqrt(-%i)*sqrt(-%i);
    (%o7) 1

    The testsuite has no problems with the new code. Because this is a fundamental simplification error I supposed to get more problems with the testsuite. Perhaps a lot of examples do some extra expansions to get the desired simple result.

    The other inconsistencies reported in the bug report are not caused by this problem. I am not sure if the code above is the best way to remove this problem, but it works.

    Any comments?

    Dieter Kaiser

  • Dieter Kaiser

    Dieter Kaiser - 2009-05-28
    • status: open --> closed
  • Dieter Kaiser

    Dieter Kaiser - 2009-05-28

    Suggested test for a base which is a mtimes expressions in expressions like (a*b)^q*(a*b)^q, where q+r=1 has been checked out.

    The example of this bug report now simplifies:

    (%i6) sqrt(-%i)*sqrt(-%i)*%i;
    (%o6) 1

    Other examples are:

    (%i7) sqrt(a*b)*sqrt(a*b)*a*b;
    (%o7) a^2*b^2

    (%i8) (a*b*x)^(1/5)*(a*b*x)^(4/5)*x;
    (%o8) a*b*x^2

    The initial example simplifies correctly. Closing this bug report as fixed.

    Dieter Kaiser


Log in to post a comment.

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:

JavaScript is required for this form.

No, thanks