A tutorial on explicit/parametric
and
everything you did not dare to ask
about
curves and surfaces
in
gnuplot
Several types of curves and surface are supported in gnuplot. Of those
not every operation is supported for every curve or surface type and it
can be therefore useful to understand the different types, their advantages
and limitations.
Curves in gnuplot are almost always planar (with one exception which we
will deal with in the end) and are assumed to be in the XY plane.
Therefore only X and Y coordinates are needed for plotting curves.
The simplest curve is the `explicit function`. This curve is in fact a
function and for each given x, there is one and only one y value associated
with it. A gnuplot example for such type is `plot sin(x)` or
`plot "datafile" using 1". Note the later is using only a single column from
the data file which is assumed to be the y values.
Alternatively one can define a `parametric curve` form. In this case
x and y are both functions of a third free parameter t, while independent
of each other. A circle can be expressed parametrically as x = cos(t),
y = sin(t) and be plotted using gnuplot as
'set parametric; plot cos(t),sin(t)'.
This form is not a function since there can be unlimited number of y values
associated with same x. Furthermore the explicit form is a special case of
the parametric representation by letting x equal to t. The curve y = sin(x)
can be written in parametric form as y = sin(t), x = t.
We are used to think of the plane in cartesian coordinate system.
In practice, some coordinate systems may be easier to use then others
under some circumstances. The polar form uses a different basis
to span the XY plane. In this representation the cartesian x coordinate
is equal to r cos(t) and the cartesian y coordinate is equal to r sin(t).
To draw a unit circle using the polar coordinate system in gnuplot use the
following simple command: 'set polar; plot 1'. To better understand this
explicit form lets backup a little.
When we plot a regular explicit function like `y = sin(x)` we march in equal
steps in x, evaluate the provided function and plot a piecewise linear curve
between the sampled points approximating the real function. In the polar
explicit form we do exactly the same thing, but we march along the angular
direction - we turn around the origin, computing the length of the radius
at that angle. Since for the unit circle, this radius is a constant 1,
`plot 1` in polar form plots a circle (if t domain is from 0 or 2Pi).
Note the polar form is explicit in that for each angle there is only a
single radius.
Surprisingly (or maybe not so surprising) surfaces share the same
representations. Since surfaces are two dimensional entities, they
require two free parameters (like t for curves).
A surface explicit function uses x and y as the free parameters. For
each such pair it provides a single z value. An example for this form
can be `splot sin(sqrt(x**2+y**2))/sqrt(x**2+y**2)` for a three dimensional
sinc function or `splot 'datafile' using 1`. As for curves, the single column
used from the data file defines the function value or z in this case.
The order of the x and y function values is very strict in this form and
simply defines a rectangular grid in the XY plane. Fortunately this
strict form allows us to apply a very simplistic hidden line algorithm
called "the floating horizon". This hidden line algorithm exploits the
rectangular XY domain of the surface and therefore may be used for this
type of surfaces only. Since in gnuplot this is the only form of hidden
lines removing algorithm provided, only explicit surfaces may have their
hidden lines removed.
Parametric surfaces are the exact extension for explicit surfaces as in
the curves case. the x, y, and z are defined in terms of two new free
variables and are totally independent of each other as x(u, v), y(u, v),
and z(u, v). Again the explicit surface is a special case of the parametric
representation where x = u, and y = v. Examples for plotting parametric
surfaces in gnuplot can be `splot cos(u)*cos(v),cos(u)*sin(v),sin(u)` which
defines a sphere, or `splot "datafile" using 1:2:3`. Since these are
parametric surfaces, gnuplot must be informed to handle them by issuing
`set parametric`.
The curve polar form takes the obvious extensions in the surface world.
The first possible extension is spherical coordinate system, while the
second is the cylindrical one. These modes currently work for data files
only and both requires two parameters, theta and phi for mapping onto the
unit sphere, and theta and z form mapping on a unit radius cylinder as follow:
Spherical coord. Cylin. coord.
---------------- -------------
x = cos( theta ) * cos( phi ) x = cos( theta )
y = sin( theta ) * cos( phi ) y = sin( theta )
z = sin( phi ) z = z
This subject brings us back to non planar curves. When surfaces are displayed
under gnuplot, isocurves are actually getting plotted. An isocurve is a
curve on the surface in which one of the two free parameters of the
surface is fixed. For example the u isolines of a surface are drawn by
setting u to be fixed and varying v along the entire v domain. The v isolines
are similarly drawn by fixing v. When data files are specified they are
classified internally into two types. A surface is tagged to have grid
topology if all its specified isolines are of the same length. A data mesh
of five isolines, seven points each is an example. In such a case the
surface cross isolines are drawn as well. Seven isolines with five points
each will be automatically created and drawn for grid type data. If
however, isolines of different length are found in the data, it is
tagged as nongrid surface and in fact is nothing more than a collection
of three dimensional curves. Only the provided data is plotted in that
case (see world.dem for such an example).