Closed Maxwell3e8 closed 5 years ago
Yes, there are things Rubi can't integrate, while Mathematica uses complicated algorithms that can return results in terms of hypergeometric functions. In your case, you can use:
expr = TrigToExp[Tan[Log[x]]]
Int[expr, x]
to get
-I x + 2 I x Hypergeometric2F1[-(I/2), 1, 1 - I/2, -x^(2 I)]
and show that
D[%, x] == expr // Simplify
(* True *)
Please see also this comment from Albert for a strongly related question.
Thanks, so it almost works! Would it make sense for Rubi to automatically try to apply some common transformations (using build-in Mathematica rules or its own) to try the integral in a few different forms?
@Maxwell3e8 That needs to be decided by @AlbertRich since he is the driving force behind Rubi's integration rules.
Thanks to your suggestion, I derived new rules for integrating expressions of the form (e x)^m Tan[d (a+b Log[c x^n])]^p. Since Tan[Log[x]] is a special case of that form, the next version of Rubi returns
-I*x + 2*I*x*Hypergeometric2F1[-I/2, 1, 1 - I/2, -x^(2*I)]
for its antiderivative. That is considerably simpler than the result returned by Mathematica 11.3 which involves two calls on Hypergeometric2F1.
Also the next release of Rubi will have analogous rules for integrating expressions of the form (e x)^m Cot[d (a+b Log[c x^n])]^p, (e x)^m Tanh[d (a+b Log[c x^n])]^p, and (e x)^m Coth[d (a+b Log[c x^n])]^p.
The currently available Rubi 4.16.1.0 is able to integrate Int[Tan[Log[x]],x]
and expressions of the above form.
Just downloaded Rubi to compare with Mathematica, after a couple of quick tests, Int[Tan[Log[x]],x] can't be evaluated, but works on Mathematica in terms of hypergeometric functions.