#4398 laplace (delta (t - a) * sin(b*t), t, s) gives zero

[Eric Majzoub]

(%i1) laplace (delta (t - a) * sin(b*t), t, s);
(%o1) 0

[Robert Dodier]

git bisect says that commit [ 4c9f030a ] is the place where the laplace result changed from nonzero to zero. Looking at the commit, I see it changed the way laplace handles its arguments, but I don't see how that leads to the observed result. More investigation needed.

[Barton Willis]

My theory is that the time-var and parm are swapped somewhere. In the function lapsin, I see (laptimes rest parm time-var), but I think that should be
(laptimes rest time-var param . Did commit [ 4c9f030a ] involve this code?

Making the swap, I get

(%i13)  laplace (delta (t - a) * sin(b*t), t, s);
Is a positive, negative or zero?

p;
(%o13) -(((%i*%e^(2*%i*a*b)-%i)*%e^(-(a*s)-%i*a*b))/2)
(%i14) trigrat(%);

(%o14) sin(a*b)*%e^-(a*s)

And %o14 is correct, I think. Specifically, I swapped:

(defun lapsin (fun rest trigswitch time-var parm)
  (let ((ab (islinear (cadr fun) time-var)))
    (cond (ab
       (cond (rest
          (compose (laptimes rest parm time-var) ;; Looks suspicious!! 
               parm
               trigswitch
               (car ab)
               (cdr ab)))

[Robert Dodier]

For the record, 4c9f030a does make several changes to LAPTIMES. Here is a link to the commit: https://sourceforge.net/p/maxima/code/ci/4c9f030a (I thought SF would automatically generate a link but I guess it didn't.)

[Raymond Toy]

Oops. I guess I messed things up a bit here. After looking at [4c9f03], I think the error is in line 355 where I changed (laptimes rest parm) to (laptimes rest parm var). I think it should var should come before parm in the call to laptimes. I think there are few more places where I accidentally swapped var and parm.

SF does generate links but you need to provide exactly 6 digits. If git annotate is any suggestion, we need to provide probably maybe 10 digits to identify a git commit.

[Raymond Toy]

Fixing the two places where laptimes had the args in the wrong order now results in:

(%i1) laplace (delta (t - a) * sin(b*t), t, s);
Is a positive, negative or zero?

pos;
                            2 %i a b         - a s - %i a b
                      (%i %e         - %i) %e
(%o1)               - ─────────────────────────────────────
                                        2
(%i2) 

Don't know if that's right or not, but at least it's not zero anymore. Wolfram Alpha seems to agree with this if a, b, and s are positive.

[Barton Willis]

Thanks--that commit didn't change lapsin (where I think the bug is located), but maybe it fixed something triggers the putative bug in lapsin. I'll try to fix this bug.

[Raymond Toy]

Fixed in commit [9c52b5]