What about setting chebyshev coefficients directly

[Brian Gough]

=?ISO-8859-15?Q?Achim_G=E4dke?= writes:
 > I'd like to describe a current-distribution by a chebishev-series,
 > but like to manipulate the coeffitients directly, not by providing
 > a function to approximate.
 > 
 > Is there anything, that might cause trouble by including a simple
 > function for that (I'll provide it...)

Should be fine.  

The current interface is not ideal (cs->f[] is probably unnecessary
and could be replaced by an independent workspace, but better to leave
it now).