Is there a way to use fdiff for functions, which allow only numerical input? It seems that fdiff is sending symbolics and that would not work in my case of interest (and I want to avoid to code such stuff, especially how to get the appropriate step size automatically ...). Related to Maple I only find this (general) source, but no according paper: M. Monagan, E. Cheb-Terrab A numerical differentiation routine for computing single, multiple and partial derivatives to arbitrary precision. This enables us to compute numerical values of derivatives of special functions for whic

May be a bit off that forum, but may be somebody has an answer or some reference: for the linear transforms w = 1 - z and w = 1/z there are exceptional cases given the formulae in Abramowitz & et al do involve infinite series in Psi (A&S 15.3.11 ff). A direct coding becomes slow and it is awesome to dig through Maple's code for that. Does there exist s.th. worth to look into it (please no pure source code like Macsyma or so ...) ?

Why there is no way (except I am stupidly missing something) to let
Maple convert exp(phi*I) to argument with piecewise discussing the
input (as it is explained quite nice in ?argument)?

Even assume(-Pi < phi, phi <= Pi) does not help and results in some
arctan(sin(phi),cos(phi)) and convert(%, piecewise, phi) does not
give the desired presentation.

I do not understand how to use this command (on strings), I want to convert
outputs from PARI like below - any hints for me doing that?

"5.2185454343674342011212095337 E444" ->
 5.2185454343674342011212095337E444
"0.E-66*I" -> 
 0.I

"0.14237172979226366716527232070 +
 3.1415926535897932384626433833*I" 
which would work with 'parse', but not if exponential notation is involved.

Where can I find a description for the various variants of numtheory:-cfrac, which goes more into details than the online help? For example one can guess how an approximating sum can be obtained from numtheory:-cfrac((a+x)^k, x, 7, 'diag', 'simregular', 'quotients'); but I prefer to cross-check against a confirmed formula instead of guessing. If somebody would be so kind to point me to some source to look at ... or even knows it already ...: how does one evaluate the above list to get a value? -- edited to add: I mean a way which can be compiled and without running through