It is not the purpose of a plotting program to provide
function definitions as Mathematica et al. do. Yet,
many functions cannot be easily represented through the
pre-defined ones. Plotting an exponential integral does
not represent a major problem as a data file can be
generated by means of other software. Fitting however
is a problem. Often it is possible to define a function
through a power series or through an iterative
algorithm within gnuplot, however, this is inefficient
and cumbersome. To address this I suggest the following
features, the first of which should be easy to implement.
The user creates an executable "foo" that accepts a
number of float input parameters and outputs the value
the function takes at these parameter values. Then a
f(x,y,...) = `./foo $1 $2 ...` (for foo in the current
defines the function which now can be used in fits and
Possibly a syntax more coherent with existing gnuplot
definitions can be found.
An additional enhancement would be the possibility to
define a function through interpolation of an existing
data file, possibly
with a syntax similar to the plot with line/with
(c)splines syntax. Again, the main purpose would be to
use this interpolated function within fits.