besselk does not implement derivative and does not raise an error when called with derivative

[electroflow]

Thank you for your very useful software package!

for funcname in ["besseli", "besselj", "besselk", "bessely"]:
    print(funcname, getattr(mpmath, funcname)(1, 1, derivative = 0),  getattr(mpmath, funcname)(1, 1, derivative = 1))

outputs

besseli 0.565159103992485 0.700906773759523
besselj 0.440050585744934 0.325147100813033
besselk 0.601907230197235 0.601907230197235
bessely -0.781212821300289 0.869469785515966

No warning or error message is raised that the derivative is not implemented for besselk. I think at least an error should be raised that the derivative is not implemented, this silent defaulting to derivative = 0 cost me a significant amount of debugging time.

[WarrenWeckesser]

It looks like all those functions ignore unknown keyword parameters (and for some reason besselk doesn't implement the derivative parameter). Ideally, the following call should raise a TypeError (something like TypeError: 'foo' is an invalid keyword argument for besselj()):

In [15]: mp.besselj(3, 1.5, foo=999)
Out[15]: mpf('0.060963951141139631')

[skirpichev]

Many functions, esp. using functions from hypergeometric.py have **kwargs. And they don't check keyword parameter names. Proper fix could be using explicit keyword arguments. But that's not an easy and simple change, see e.g. sympy/sympy#14030.

Meanwhile, I'll mark this as a bug. As a partial solution, we could raise NotImplementedError.

[skirpichev]

Ok, now we have NotImplementedError