1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
|
\section{Extensibility\label{extend}}
Two kinds of extensibilities will be described in this section.
On one hand, regarding the \sdd\ design, the extensibility of the \RR\ graphics system will be discussed;
it provides the tools and features enabling the programmer to write complex
high level plot functions in a very general manner.
On the other hand we describe the extensibility of \sdd\ itself.
\subsection{Extensibility of \RR\ graphics\label{exR}}
\RR\ provides a huge collection of low level graphic functions like those
for adding elements to an existing plot or for computations related to
plotting.
%
These functions are used to build very general high level functions, at
least for the two dimensional case, and without them, the ``\RR\ code
only'' design of \sdd\ would be impossible.
A selection of these low level functions begins with the functions to obtain $x$, $y$ (and~$z$) coordinates for plotting,
namely {\tt xy.coords} and {\tt xyz.coords} (for the 3D case, cf.\ Section \ref{xyzcoords}).
Further on, the functions {\tt plot.new} and {\tt plot.window} can be used to set up the plotting region appropriately,
{\tt pretty} to calculate pretty axis tick marks,
{\tt segments} to draw line segments between pairs of points,
and functions like {\tt title}, {\tt axis}, {\tt points}, {\tt lines}, {\tt text} etc.~are self-explanatory.
A huge collection of graphical parameters for \RR\ is documented in the
help pages for {\tt par} and {\tt plot.default} (cf.\ R core \citeyear{r-ref}).
Almost all low level graphic functions make use of the argument `\code{...}'
which allows specifying most of these parameters in a very general manner.
If this argument, `\code{...}', is also used in a high level function,
arguments which are not \textsl{explicitly} introduced in the arguments list,
can be passed through to lower level graphic functions as well; this
is a powerful feature of the \textsf{S} language.
\enlargethispage{5mm}
Since the \RR\ graphics system is designed for two dimensional graphics,
it lacks of some features for the three dimensional case.
%%
Unfortunately, the {\tt axis} function works only for 2D graphics.
Consequently a large amount of code was required to
enable oblique axes for displaying the 3D scatter plot in an arbitrary
angle.
Locations in \RR\ graphics devices can be addressed with 2D coordinates,
Thus the information on the projection has to be calculated by the 3D graphic functions internally.
As described in Section \ref{design}, \sdd\ uses a parallel projection.
Since the \RR\ graphics device does not know anything about the projection,
without any appropriate additional tools it is not possible to add elements into an existing \sdd .
\subsection{Extensibility of \sdd\label{exsdd}}
In Sections \ref{introduction} and \ref{design} it was emphasized
that the \sdd\ design was intended to be as general as possible.
Some attempts to obtain this generality are described in Section \ref{design} and its subsections.
%%
Because of the missing projection information, the ability of adding
elements to an already existing \sdd\ would be restricted, if only the
already defined (and for the 2D case general) \RR\ functions could be used
(cf.\ Section \ref{exR}).
For this reason, \sdd\ (invisibly) returns a list of \textsl{function
closures} (cf.\ Section~\ref{structure}).
A \textsl{function closure} is a function together with an
\textsl{environment}, and an \textsl{environment} is a collection of
symbols and associated values (i.e. \RR\ variables). Thus these properties
of \RR's scoping rules, called \textsl{Lexical Scoping}
(\cite{gentleman}), are extensively used in \sdd.
%
Notice that \textsl{Lexical Scoping} is a feature of \RR, not defined as
such in the \textsf{S} language.
In other words, the values returned by \sdd\ are functions together with
the environment in which they (and the scatter plot) were created.
The benefit of returning function closures is, that the function somehow
``knows'' the values of variables (in the environment) that were assigned
to those variables at the time when the function was created. All in all,
we made those functions know details about the axis scaling and the
projection information that are required to add elements to an existing
plot appropriately.
The following functions are returned by \sdd , for details see the Appendix:\\ \vspace{-11mm}
\begin{description} \setlength\itemsep{0.5ex plus0.2ex minus0.3ex}
\item[{\tt xyz.convert}:] A function which converts 3D coordinates to the 2D parallel projection of the existing \sdd .
It is useful to add arbitrary elements into the plot.
\item[{\tt points3d}:] A function which draws points or lines into the existing plot.
\item[{\tt plane3d}:] A function which draws a plane into the existing plot:\\
\verb+plane3d(Intercept, x.coef=NULL, y.coef=NULL, lty="dashed", ...)+.
Instead of an intercept, a vector containing three elements or an \textsf{(g)lm} object can be specified.
\item[{\tt box3d}:] This function draws a box (or ``refreshes'' an existing one) around the plot.
\end{description}
{\tt xyz.convert} is the most important function, because it does the parallel projection
by converting the given 3D coordinates into the 2D coordinates needed for the \RR\ graphics devices.
Examples how to use the mentioned function closures are given in Section \ref{examples}.
|
