Any one know why risch.lisp uses $%e/_to/_numlog in risch.lisp?
AFAICT, it is only used in risch.lisp and it doesn't actually do
anything because in the places used, it's merely bound to either NIL
or T.
I am guessing that it is really a typo for $%e_to_numlog, which
controls how exp(r*log(x)) is simplified when r is rational.
I think $%e/_to/_numlog is really $%e_to_numlog, and unless someone
objects, I'm going to make this change. It does affect one test case
in the testsuite:
Running tests in rtest15:
********************** Problem 168 ***************
Input:
2 2 4
1 - x + sqrt(1 - 6 x + x )
integrate(----------------------------, x)
2 x
Result:
/ 2 2 2
[ sqrt(x - 2 x - 1) sqrt(x + 2 x - 1) x
I ------------------------------------- dx + log(x) - --
] x 2
/
--------------------------------------------------------
2
This differed from the expected result:
2 2
log(- 1 - 2 x + x ) log(- 1 + 2 x + x )
------------------- + -------------------
2 / 2 2
- x [ %e
---- + log(x) + I ------------------------------------------- dx
2 ] x
/
----------------------------------------------------------------
2
Ray