|
|
|
|
MAX - PLANCK - INSTITUT - FÜR AERONOMIE
D-37191 Katlenburg-Lindau, Federal Republic of Germany
Graphics and Colour with 
PATRICK W.
DALY
1998 June 4
[This document is also available in PostScript and PDF form.] 1 Introduction
2 Importing external graphics
2.1 Importing with the graphics package
2.2 Scaling
2.3 Reflection
2.4 Rotation
2.5 Importing, scaling, rotating with
the graphicx package
3 Additional graphics packages
3.1 The epsfig package
3.2 The lscape package
3.3 The rotating package
4 Troubleshooting
4.1 How importation works
4.2 Problems with importation
5 Configuring graphics importation
5.1 Default driver
5.2 Search path for graphics files
5.3 Default extensions
5.4 Graphics types
5.5 Compressing graphics files
6 Colours
7 The named colour model
The advent of PostScript drivers for TeX has offered the opportunity to expand the capabilities
of to include a number of very useful features, such as the importation, scaling, and
rotation of external graphics files, and colour. However, since each driver tended to have its own
packages and user interface commands, the resulting document files threatened to
become tied to one particular driver.
With 2e, a standardized set of commands is made available for all drivers by means
of the packages graphics and color. The specialized coding to translate the user commands
into instructions for the specific drivers is placed in certain .def files which are loaded by
options to these packages. Thus one only needs to change the option to switch to another
driver; the main text remains unchanged.
The commands defined by these packages are the building blocks for other packages that
either emulate the older driver-specific ones or provide a more comfortable syntax for these
functions. As long as these other packages are based on graphics and color, they should be
equally compatible with all the supported drivers.
The problem is to have a graphics file produced by some other program included in the
document, most likely as a figure. It may be necessary to scale it or to rotate it by 90o. One wants
to do by computer what used to be done with scissors and glue.
There are two packages available for importing and manipulating external graphics files: the
more basic graphics and the more extended graphicx. They both have identical functionality,
differing only in syntax. In addition to the driver names, there are some other options that may be invoked when the
packages graphics or graphicx are loaded:
-
draft - does not import but places a framed box where the graphic would appear, with
the file name printed inside; this speeds up the processing considerably when one
is only working on the text;
-
final
- counteracts
draft; needed when the draft option has been issued globally in
\documentclass; -
hidescale
- leaves blank space where scaled text should be;
-
hiderotate
- leaves blank space where rotated text should be; this and the option
hidescale
are useful if the previewer cannot handle scaling or rotation;
-
hiresbb
- look for bounding box values in
%%HiResBoundingBox instead of the the normal
%%BoundingBox line.
The basic importing command with the graphics package is
\includegraphics[llx,lly][urx,ury]{file_name}
where llx, lly are the coordinates of the lower left corner, and urx, ury those of the upper right
corner of the bounding box containing the part of the picture that is to be included. In other
words, they say where the scissors are to be applied. Units may be specified (like [3cm, 2in])
but if they are omitted, big points (bp, 72 per inch, 28.3464. . . per cm) are assumed. If only one
optional argument is given, it is the upper right corner, and the lower left
is assumed to be [0,0].
 If
no bounding box coordinates are given, the driver will obtain them some other way, depending
on the type of graphics file. For example, for the very common encapsulated PostScript files
with extension .eps, the bounding box information is extracted from the graphics file
itself. The figure at the left is stored in such a file and is included simply with the
command
\includegraphics{clock.eps}
With \includegraphics*, the figure is clipped, so that any drawing outside the
specified bounding box is suppressed. This is useful if only part of a figure is to be
reproduced. It is also vital with some perverse figures that paint the whole page
white!
The graphics file in the above commands is transferred to the document in its original size. In
order to rescale it, there are two commands available:
\scalebox{h_scale}[v_scale]{text}
which applies horizontal and vertical scale factors to the contents text; if v_scale is omitted, it is
the same as h_scale;
\resizebox{h_length}{v_length}{text}
adjusts the figure to fit into the specified horizontal and vertical sizes; if either length is given
as !, the one scale factor is used for both dimensions. A *-form allows v_length to refer to
the height + depth of the box, rather than just to the height. In both cases, the
contents text may be an \includegraphics command, but it may also be any arbitrary
text.
The contents of a box may be reflected horizontally with
\reflectbox{text}
Rotation of a box about the left-hand end of its baseline is done with
\rotatebox{angle}{text}
where angle is in degrees, and the rotation is counterclockwise.
To
illustrate this, we have scaled the previous clock figure to a height of
2 cm and then rotated it by 30o, using the commands
\rotatebox{30}{\resizebox{!}{2cm}{% \includegraphics{clock.eps}}}
For demonstration purposes, we have added framed boxes around the figure before
and after the rotation. It is the inner (tilted) box that is
2 cm high, while the overall outer box is somewhat higher and broader due to
its inclined contents. Without these frames the figure seems to have extraneous
space above and to the left, something that can be very puzzling even to
experienced users. (The frames would not normally appear with the above commands.)
If one selects the graphicx rather than the graphics package, a different interface is available
for both importing and rotation, one making use of keys and values:
\includegraphics[key=value,. . . ]{file_name}
The keys are of two types: those that take a numerical value, and those that are flags with the
values true or false. Simply giving the name of a flag without a value is equivalent to setting
it to true. Possible keys and their values are:
scale= - number; enters the number by which the figure size should be magnified over its
natural size;
width= - length; specifies the width to which the figure should be scaled to; if
height not
given, it is scaled with the same factor as the width;
height= - length; specifies the height to which the figure should be scaled to; if
width is
not given, it is scaled with the same factor as the height;
totalheight= - length; like
height but specifies the height plus depth of the figure; should
always be used in place of height if the figure has been rotated;
keepaspectratio - (
= true/false); if both height and width are specified, this flag
ensures that the original height/width ratio remains unchanged; the figure will not
exceed either of the given dimensions;
angle= - number; the angle by which the figure is to be rotated counterclockwise, in
degrees; any
height or width specifications coming before this key are also rotated,
so that the height becomes the width, while the width becomes either the height
(positive angle) or depth (negative angle);
origin= - loc; determines the point about which the rotation occurs; default is
bl for
bottom left corner; also possible are c for center, t for top, r for right, and B for
baseline; any sensible combination, such as tr, is allowed;
draft - (
= true/false); like the draft package option but applied to the one graphics
file; the figure is not imported, but rather a framed box of the correct size is printed
containing the name of the file;
clip - (
= true/false); suppresses the printing of any graphic outside the bounding box;
bb= - llx lly urx ury; enters the coordinates of the bounding box manually, if they
are missing or incorrect in the graphics file, or to be deliberately altered; the
specifications are four lengths separated by blanks; units may be given, but if
omitted, big points (
bp) are assumed;
viewport= - llx lly urx ury; specifies the bounding box but relative to the lower left corner
of the existing one; useful for correcting the bounding box, or (with
clip) to select
only a portion of the whole figure;
trim= - dllx dlly durx dury; reduces the existing bounding box by the amounts specified;
hiresbb - (
= true/false); like the hiresbb package option but applied to the one graphics
file; reads bounding box information from the %%HiResBoundingBox line in the
graphics file.
The keys are all optional; they are included as needed. Their order is not important other than
that angle can change previous height and width meanings. The sets of key/values are
separated from each other by commas.
With
the key/value syntax, the tilted, scaled graphic on page 4 is produced with
\includegraphics[height=2cm,angle=30]{clock.eps}
For compatibility with the graphics package, there also exists an \includegraphics* version
that clips the imported figure; this is equivalent to including the key clip.
With the graphicx package, the \rotatebox command is similarly redefined to accept the
optional key origin.
Sebastian Rahtz has provided a package epsfig that not only updates the earlier (2.09)
version, but also re-implements Rokicki's epsf package by means of the graphics
commands. This is helpful for users who are accustomed to those syntaxes. For epsf, this
is
\epsfysize=y_size or \epsfxsize=x_size \epsf[llx lly urx ury]{file_name}
The epsfig package also defines an importing command that makes use of the regular keys and
values to enter its parameters:
\epsfig{file=file_name,key=value,. . . }
For compatibility with some older versions, there is a \psfig command which is synonymous
with the above.
The epsfig package is included in the bundle of graphics packages and drivers.
Another extra package in the graphics bundle is lscape, by David Carlisle. This defines a
landscape environment that prints its contents rotated 90o on
a page for itself. Head and footlines remain as normal. This is intended primarily
for inserting figures that are in landscape mode, that is, wider than they are
high.
The rotating package by Sebastian Rahtz and Leonor Barroca tries to make the interface for
rotation somewhat simpler. It defines
\begin{sideways} text \end{sideways} \begin{turn}{angle} text \end{turn} \begin{rotate}{angle} text \end{rotate} \turnbox{angle}{text}
where sideways rotates text by 90o, turn by an arbitrary angle. The environment rotate and
command turnbox are equivalent: they rotate but in a box of zero size, so that the contents
overlap the surroundings.
This package is not part of the graphics bundle and must be obtained separately.
In order to understand better what can go wrong when
a graphics file is imported, and to know what to do about, it is important to
realize how the interplay between and the driver
program functions.
has no idea what
is in the graphics file; for it, the figure is simply a box of a given height,
width, and depth, as indeed are all the characters that processes. The
information on the graphic's natural size is somehow obtained, by reading %%BoundingBox line
in the graphics file itself, or in some other file, or through the optional entries
in the
\includegraphics command or equivalent. After scaling and rotating, knows the final
size that it must reserve in the output text for the figure.
What is then written to the .dvi file is the name of the graphics file and information on
how it should be transformed. Just how this information is coded depends on the
graphics driver selected. When the printer driver program processes the .dvi file, it interprets these
special instructions, reads in the specified graphics file, performs the transformations,
and places the result where has said it should
go. The end result is that the area inside the designated bounding box coincides
with the box that has reserved for
it. If the bounding box information is incorrect, the figure is obviously going
to be misplaced. The most common type of graphics file to be imported is encapsulated PostScript. Files
adhering to this standard are intended to be included within other PostScript files, and so must
not contain certain PostScript commands that reset the whole graphics page. Most
importantly, it must contain a comment line (ignored by PostScript itself) of the
form:
%%BoundingBox: llx lly urx ury
giving the coordinates of the lower left and upper right corners of the bounding box. The units
are never specified, always being big points.
Having pointed out how the importation takes place, we can now discuss what can go wrong
along this chain of processes.
The most common problems encountered when importing encapsulated PostScript files are
listed here.
-
No bounding box.
- If the bounding box information is totally missing
from the graphics file,
issues the error
message
!LaTeX Error: Cannot determine size of graphic
in...(no
BoundingBox)
- The solution is to determine the bounding box coordinates somehow (see
next point) and to include them, either in the
\includegraphics command, or by editing the graphics file itself. However,
if there really is no bounding box information in the file, it is unlikely
to conform to the encapsulated standard and will cause other problems.
-
-
The placement is incorrect.
- Both the and driver process
without error messages, but the figure is either displaced from the expected
position, or is far too small.
Most likely the bounding box information is incorrect. Many applications that produce
PostScript files are too lazy to calculate the true bounding box, or they think they are
generating a whole page with a figure somewhere in the middle. In either case, the
bounding box corresponds to the full page even though the printed figure occupies only a
portion of it. Find the true bounding box by one of the following methods:
- Print the figure, mark the lower left and upper right corners of the box
containing the figure, and measure their distances from the left and bottom
edges. Enter these distances in the
\includegraphics command, or edit the
PostScript file. In the latter case, convert to big points.
Difficulties with this are that some encapsulated PostScript files cannot be
printed on their own, and that the left and bottom edges of the paper need
not be the exact lines from which the printer really measures.
- Include the figure in a short file such as
\setlength{\fboxsep}{-\fboxrule} \fbox{\includegraphics{test.eps}}
and
print the output. The apparent bounding box appears as a framed
box. Measure the true bounding box relative to the left and bottom
edges of the apparent one, and enter the values in the \includegraphics command with the viewport= key. If necessary, scale the figure down to fit on the page,
but then remember to increase the measurements by the same scale
factor.
- Use the GhostView program to fix up the bounding
box, either automatically or manually. This is the most convenient
method if you have this utility for viewing and manipulating PostScript
files.
-
-
Immovable graphic.
- It does not shift, nor scale nor rotate, no matter what is specified. In
this case, it violates the encapsulated PostScript rules and contains some global plotting
commands. Graphics produced by word processing programs are notorious for this. Often
the offending command is
setpagedevice.
There is little that can be done to correct this, other than trying to regenerate the
graphics file with an option for encapsulated PostScript. Judicious editing can
remove the troublesome lines, but this could result in the file becoming totally
unreadable.
5 Configuring graphics importation
Although the graphics syntax has been standardized, and most of the driver-specific coding
hidden in the .def files, there are still a number of items that must be set up for any particular
installation and operating system. These are most conveniently placed in the local configuration
file graphics.cfg which is read in if it is present.
The choice of driver option must always be specified, but a local default can be established by
placing the line
\ExecuteOptions{driver}
into the configuration file, where driver is one of the allowed driver options. This may be
overridden by any explicit driver option specified in the document file itself.
The rest of the configuring commands in this section can be issued either in the
graphics.cfg file, or in the document.
One can specify the directories where is to look for
graphics files with
\graphicspath{dir_list}
where dir_list is a list of directory names, separated by commas. The syntax
of the local operating system must be used. Without this command, searches for graphics
files in the same directories as for all other TeX files.
A list of default extensions for the graphics files can be defined with
\DeclareGraphicsExtensions{ext_list}
This means that only the root name of the file must
be given and will attempt to
find it by attaching all the possible extensions. For PostScript drivers,
the ext_list is usually set to
.eps,.ps. At our installation, we also include the non-standard extension .psc. Note that
the above command does not add to the list of extensions but rewrites it anew;
if you wish to add to the list, you must include all the allowed extensions in
the one declaration.
Defining the extensions is only part of the task: one must also associate each extension with
a graphics type so the ðknows how to process
it. PostScript recognizes only one type, eps, encapsulated PostScript, but there do exist other types such as bmp
and pcx for other drivers. For the non-standard .psc extension above, we must also
give
\DeclareGraphicsRule{.psc}{eps}{}{}
to inform that this extension
belongs to type eps. The other two (empty) arguments
specify that the bounding box information is to be read from the file itself,
and that no other program needs to be applied to the file.
Since PostScript files are often extremely large,
it makes sense to try to compress them with either the zip or gzip programs. In such a case, the .eps file is replaced by a file with
extension .zip, or .eps.gz, or .eps-gz. Two problems now arise: first, cannot read
such files to obtain the bounding box information, and secondly, the driver
needs to unpack such a file to include it in the final output. This can
be accomplished with, for example,
\DeclareGraphicsRule{.eps.gz}{eps}{.eps.bb}{`gnzip -c #1}
which establishes the graphics type as eps, with the bounding box information in the file of the
same name and extension .eps.bb, and that the operating system command gunzip -c must
be applied to the file (represented as #1). The single quote ` is required to indicate a system
command. The %%BoundingBox line of the original file must be copied and stored in the
.eps.bb file.
Such decompression rules are system dependent and thus need to be configured for the local
installation. For example, under the VMS operating system, the gzip program produces files
with extension .eps-gz and decompression is performed with gzip -d rather than with
gunzip. The corresponding rule becomes
\DeclareGraphicsRule{.eps-gz}{eps}{.bb}{`gzip -d -c #1}
The color package recognizes the options:
-
monochrome
- to convert all colour commands to black and white, for previewers that cannot
handle colour;
-
dvipsnames
- makes the
named colour model of dvips (Section 7) available to other drivers; -
nodvipsnames
- disables the
named model for dvips, to save memory;
-
usenames
- loads all the
named colours as defined ones; again, see Section 7 for details.
A local configuration file color.cfg can be set up in the same way as for the
graphics package. The default driver option is specified in exactly the
same way as in Section 5.
Colours are specified either by a defined name, or by the form
[model]{specs}
where model is one of rgb (red, green, blue), cmyk (cyan, magenta, yellow, black), gray, or
named. The specs is a list of numbers from 0 to 1 giving the strengths of the components
in the model. Thus [rgb]{1,0,0} defines red, [cmyk]{0,0,1,0} yellow. The gray model takes only
one number. The named model accesses colours by internal names that were originally built
into the dvips driver, but which may now be used by some other drivers too. This
model is described in Section 7.
A colour can be defined with
\definecolor{name}{model}{specs}
and then the name may be used in all the following colour commands. Certain colours are
automatically predefined for all drivers: red, green, blue, yellow, cyan, magenta, black,
white.
In the following colour commands, col_spec is either the name of a defined colour, like
{blue}, or [model]{spec}, like [rgb]{0,1,0}.
\pagecolor col_spec - sets the background colour for the current and following pages;
\color col_spec - is a declaration to switch to setting text in the given colour;
\textcolor col_spec{text} - sets the text of its argument in the given colour;
\colorbox col_spec{text} - sets its argument in a box with the given colour as background;
\fcolorbox col_spec1col_spec2{text} - is like
\colorbox, with a frame of col_spec1 around
a box of background colour col_spec2; the two specifications must either both be
defined ones, or both use the same model, which is given only once; for example,
\fcolorbox{red}{green}{Text} sets `Text' in the current text colour on a green
background with a red frame;
\normalcolor - switches to the colour that was active at the end of the preamble. Thus
placing a
\color command in the preamble can change the standard colour for the
whole document. This is the equivalent to \normalfont for font selection.
Normally one would try to define all the colours needed as names for the col_spec entries. This
simplifies changing the colour definition everywhere should fine tuning be required after the
initial printed results are seen. The same colour definition can produce quite different effects on
different printers. Even the display on the monitor is no reliable guide as to how the output will
appear on paper.
One very useful colour model is called named and is based on the 68 predefined internal colours
of the dvips PostScript driver. Sample names are BurntOrange or DarkOrchid. This model can
be activated for other drivers with the option dvipsnames, in which case one can define colours
as, for example
\definecolor{titlecol}[named]{DarkOrchid}
The colour titlecol can then be used as col_spec in the various colour commands.
The named colours can be defined with their own names if one invokes the option
usenames, which effectively declares
\definecolor{BurntOrange}[named]{BurntOrange}
and so on, for all 68 colours.
It is possible to generate a palette of the named
colours by processing the following short
file and sending
the output to the desired printer.
\documentclass[12pt,a4paper]{article}
\usepackage[dvips]{color}
\usepackage{multicol}
\pagestyle{empty}
\setlength{\oddsidemargin}{0pt}
\setlength{\textwidth}{16cm}
\setlength{\textheight}{22cm}
\setlength{\parindent}{0pt}
\setlength{\parskip}{0pt}
\begin{document}
\renewcommand*{\DefineNamedColor}[4]{%
\textcolor[named]{#2}{\rule{7mm}{7mm}}\quad
\texttt{#2}\strut\\}
\begin{center}\Large Named colours in \texttt{dvipsnam.def}
\end{center}
\begin{multicols}{3}
\input{dvipsnam.def}
\end{multicols}
\end{document}
Remember, each printer can reproduce the colours differently, so it is important to test this
table with every colour printer that might be used. The output of this job is shown next page |
|
