path: root/Master/texmf-dist/doc/generic/pgf/text-en/pgfmanual-en-tikz-actions.tex

% Copyright 2006 by Till Tantau
%
% This file may be distributed and/or modified
%
% 1. under the LaTeX Project Public License and/or
% 2. under the GNU Free Documentation License.
%
% See the file doc/generic/pgf/licenses/LICENSE for more details.


\section{Actions on Paths}

\subsection{Overview}

Once a path has been constructed, different things can be done with it. It can
be drawn (or stroked) with a ``pen'', it can be filled with a color or shading,
it can be used for clipping subsequent drawing, it can be used to specify the
extend of the picture -- or any combination of these actions at the same time.

To decide what is to be done with a path, two methods can be used. First, you
can use a special-purpose command like |\draw| to indicate that the path should
be drawn. However, commands like |\draw| and |\fill| are just abbreviations for
special cases of the more general method: Here, the |\path| command is used to
specify the path. Then, options encountered on the path indicate what should be
done with the path.

For example, |\path (0,0) circle (1cm);| means: ``This is a path consisting of
a circle around the origin. Do not do anything with it (throw it away).''
However, if the option |draw| is encountered anywhere on the path, the circle
will be drawn. ``Anywhere'' is any point on the path where an option can be
given, which is everywhere where a path command like |circle (1cm)| or
|rectangle (1,1)| or even just |(0,0)| would also be allowed. Thus, the
following commands all draw the same circle:
%
\begin{codeexample}[code only]
\path [draw] (0,0) circle (1cm);
\path (0,0) [draw] circle (1cm);
\path (0,0) circle (1cm) [draw];
\end{codeexample}
%
Finally, |\draw (0,0) circle (1cm);| also draws a path, because |\draw| is an
abbreviation for |\path [draw]| and thus the command expands to the first line
of the above example.

Similarly, |\fill| is an abbreviation for |\path[fill]| and |\filldraw| is an
abbreviation for the command |\path[fill,draw]|. Since options accumulate, the
following commands all have the same effect:
%
\begin{codeexample}[code only]
\path [draw,fill]   (0,0) circle (1cm);
\path [draw] [fill] (0,0) circle (1cm);
\path [fill] (0,0) circle (1cm) [draw];
\draw [fill] (0,0) circle (1cm);
\fill (0,0) [draw] circle (1cm);
\filldraw (0,0) circle (1cm);
\end{codeexample}

In the following subsection the different actions that can be performed on a
path are explained. The following commands are abbreviations for certain sets
of actions, but for many useful combinations there are no abbreviations:

\begin{command}{\draw}
    Inside |{tikzpicture}| this is an abbreviation for |\path[draw]|.
\end{command}

\begin{command}{\fill}
    Inside |{tikzpicture}| this is an abbreviation for |\path[fill]|.
\end{command}

\begin{command}{\filldraw}
    Inside |{tikzpicture}| this is an abbreviation for |\path[fill,draw]|.
\end{command}

\begin{command}{\pattern}
    Inside |{tikzpicture}| this is an abbreviation for |\path[pattern]|.
\end{command}

\begin{command}{\shade}
    Inside |{tikzpicture}| this is an abbreviation for |\path[shade]|.
\end{command}

\begin{command}{\shadedraw}
    Inside |{tikzpicture}| this is an abbreviation for |\path[shade,draw]|.
\end{command}

\begin{command}{\clip}
    Inside |{tikzpicture}| this is an abbreviation for |\path[clip]|.
\end{command}

\begin{command}{\useasboundingbox}
    Inside |{tikzpicture}| this is an abbreviation for |\path[use as bounding box]|.
\end{command}


\subsection{Specifying a Color}

The most unspecific option for setting colors is the following:
%
\begin{key}{/tikz/color=\meta{color name}}
        \indexoption{color option}%
    This option sets the color that is used for fill, drawing, and text inside
    the current scope. Any special settings for filling colors or drawing
    colors are immediately ``overruled'' by this option.

    The \meta{color name} is the name of a previously defined color. For
    \LaTeX\ users, this is just a normal ``\LaTeX-color'' and the |xcolor|
    extensions are allowed. Here is an example:
    %
\begin{codeexample}[]
\tikz \fill[color=red!20] (0,0) circle (1ex);
\end{codeexample}

    It is possible to ``leave out'' the |color=| part and you can also write:
    %
\begin{codeexample}[]
\tikz \fill[red!20] (0,0) circle (1ex);
\end{codeexample}
    %
    What happens is that every option that \tikzname\ does not know, like
    |red!20|, gets a ``second chance'' as a color name.

    For plain \TeX\ users, it is not so easy to specify colors since plain
    \TeX\ has no ``standardized'' color naming mechanism. Because of this,
    \pgfname\ emulates the |xcolor| package, though the emulation is
    \emph{extremely basic} (more precisely, what I could hack together in two
    hours or so). The emulation allows you to do the following:
    %
    \begin{itemize}
        \item Specify a new color using |\definecolor|. Only the two color
            models |gray| and |rgb| are supported\footnote{Con\TeX t users
            should be aware that \texttt{\textbackslash definecolor} has a
            different meaning in Con\TeX t. There is a low-level equivalent
            named \texttt{\textbackslash pgfutil@definecolor} which can be
            used instead.}.
            %
            \example |\definecolor{orange}{rgb}{1,0.5,0}|
        \item Use |\colorlet| to define a new color based on an old one.
            Here, the |!| mechanism is supported, though only ``once'' (use
            multiple |\colorlet| for more fancy colors).
            %
            \example |\colorlet{lightgray}{black!25}|
        \item Use |\color|\marg{color name} to set the color in the current
            \TeX\ group. |\aftergroup|-hackery is used to restore the color
            after the group.
    \end{itemize}
\end{key}

As pointed out above, the |color=| option applies to ``everything'' (except to
shadings), which is not always what you want. Because of this, there are
several more specialized color options. For example, the |draw=| option sets
the color used for drawing, but does not modify the color used for filling.
These color options are documented where the path action they influence is
described.


\subsection{Drawing a Path}

You can draw a path using the following option:
%
\begin{key}{/tikz/draw=\meta{color} (default \normalfont is scope's color setting)}
    Causes the path to be drawn. ``Drawing'' (also known as ``stroking'') can
    be thought of as picking up a pen and moving it along the path, thereby
    leaving ``ink'' on the canvas.

    There are numerous parameters that influence how a line is drawn, like the
    thickness or the dash pattern. These options are explained below.

    If the optional \meta{color} argument is given, drawing is done using the
    given \meta{color}. This color can be different from the current filling
    color, which allows you to draw and fill a path with different colors. If
    no \meta{color} argument is given, the last usage of the |color=| option is
    used.

    If the special color name |none| is given, this option causes drawing to be
    ``switched off''. This is useful if a style has previously switched on
    drawing and you locally wish to undo this effect.

    Although this option is normally used on paths to indicate that the path
    should be drawn, it also makes sense to use the option with a |{scope}| or
    |{tikzpicture}| environment. However, this will \emph{not} cause all paths
    to be drawn. Instead, this just sets the \meta{color} to be used for
    drawing paths inside the environment.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \path[draw=red] (0,0) -- (1,1) -- (2,1) circle (10pt);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

The following subsections list the different options that influence how a path
is drawn. All of these options only have an effect if the |draw| option is
given (directly or indirectly).


\subsubsection{Graphic Parameters: Line Width, Line Cap, and Line Join}
\label{section-cap-joins}

\begin{key}{/tikz/line width=\meta{dimension} (initially 0.4pt)}
    Specifies the line width. Note the space.
    %
\begin{codeexample}[]
  \tikz \draw[line width=5pt] (0,0) -- (1cm,1.5ex);
\end{codeexample}
    %
\end{key}

There are a number of predefined styles that provide more ``natural'' ways of
setting the line width. You can also redefine these styles.

\begin{stylekey}{/tikz/ultra thin}
    Sets the line width to 0.1pt.
    %
\begin{codeexample}[]
  \tikz \draw[ultra thin] (0,0) -- (1cm,1.5ex);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/very thin}
    Sets the line width to 0.2pt.
    %
\begin{codeexample}[]
  \tikz \draw[very thin] (0,0) -- (1cm,1.5ex);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/thin}
    Sets the line width to 0.4pt.
    %
\begin{codeexample}[]
  \tikz \draw[thin] (0,0) -- (1cm,1.5ex);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/semithick}
    Sets the line width to 0.6pt.
    %
\begin{codeexample}[]
  \tikz \draw[semithick] (0,0) -- (1cm,1.5ex);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/thick}
    Sets the line width to 0.8pt.
    %
\begin{codeexample}[]
  \tikz \draw[thick] (0,0) -- (1cm,1.5ex);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/very thick}
    Sets the line width to 1.2pt.
    %
\begin{codeexample}[]
  \tikz \draw[very thick] (0,0) -- (1cm,1.5ex);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/ultra thick}
    Sets the line width to 1.6pt.
    %
\begin{codeexample}[]
  \tikz \draw[ultra thick] (0,0) -- (1cm,1.5ex);
\end{codeexample}
    %
\end{stylekey}

    \label{section-line-cap}
\begin{key}{/tikz/line cap=\meta{type} (initially butt)}
    Specifies how lines ``end''. Permissible \meta{type} are |round|, |rect|,
    and |butt|. They have the following effects:
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \begin{scope}[line width=10pt]
    \draw[line cap=round] (0,1 ) -- +(1,0);
    \draw[line cap=butt]  (0,.5) -- +(1,0);
    \draw[line cap=rect]  (0,0 ) -- +(1,0);
  \end{scope}
  \draw[white,line width=1pt]
    (0,0 ) -- +(1,0) (0,.5) -- +(1,0) (0,1 ) -- +(1,0);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

\begin{key}{/tikz/line join=\meta{type} (initially miter)}
    Specifies how lines ``join''. Permissible \meta{type} are |round|, |bevel|,
    and |miter|. They have the following effects:
    %
\begin{codeexample}[]
\begin{tikzpicture}[line width=10pt]
  \draw[line join=round] (0,0) -- ++(.5,1) -- ++(.5,-1);
  \draw[line join=bevel] (1.25,0) -- ++(.5,1) -- ++(.5,-1);
  \draw[line join=miter] (2.5,0) -- ++(.5,1) -- ++(.5,-1);
  \useasboundingbox (0,1.5); % enlarge bounding box
\end{tikzpicture}
\end{codeexample}

    \begin{key}{/tikz/miter limit=\meta{factor} (initially 10)}
        When you use the miter join and there is a very sharp corner (a small
        angle), the miter join may protrude very far over the actual joining
        point. In this case, if it were to protrude by more than \meta{factor}
        times the line width, the miter join is replaced by a bevel join.
        %
\begin{codeexample}[]
\begin{tikzpicture}[line width=5pt]
  \draw                 (0,0) -- ++(5,.5) -- ++(-5,.5);
  \draw[miter limit=25] (6,0) -- ++(5,.5) -- ++(-5,.5);
  \useasboundingbox (14,0); % make bounding box bigger
\end{tikzpicture}
\end{codeexample}
    \end{key}
\end{key}


\subsubsection{Graphic Parameters: Dash Pattern}

\begin{key}{/tikz/dash pattern=\meta{dash pattern}}
    Sets the dashing pattern. The syntax is the same as in \textsc{metafont}.
    For example following pattern |on 2pt off 3pt on 4pt off 4pt| means ``draw
    2pt, then leave out 3pt, then draw 4pt once more, then leave out 4pt again,
    repeat''.
    %
\begin{codeexample}[]
\begin{tikzpicture}[dash pattern=on 2pt off 3pt on 4pt off 4pt]
  \draw (0pt,0pt) -- (3.5cm,0pt);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

\begin{key}{/tikz/dash phase=\meta{dash phase} (initially 0pt)}
    Shifts the start of the dash pattern by \meta{phase}.
    %
\begin{codeexample}[]
\begin{tikzpicture}[dash pattern=on 20pt off 10pt]
  \draw[dash phase=0pt] (0pt,3pt) -- (3.5cm,3pt);
  \draw[dash phase=10pt] (0pt,0pt) -- (3.5cm,0pt);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

\begin{key}{/tikz/dash=\meta{dash pattern}|phase|\meta{dash phase}}
    Sets the dashing pattern and phase at the same time.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw [dash=on 20pt off 10pt phase  0pt] (0pt,3pt) -- (3.5cm,3pt);
  \draw [dash=on 20pt off 10pt phase 10pt] (0pt,0pt) -- (3.5cm,0pt);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

As for the line thickness, some predefined styles allow you to set the dashing
conveniently.

\begin{stylekey}{/tikz/solid}
    Shorthand for setting a solid line as ``dash pattern''. This is the default.
    %
\begin{codeexample}[]
\tikz \draw[solid] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/dotted}
    Shorthand for setting a dotted dash pattern.
    %
\begin{codeexample}[]
\tikz \draw[dotted] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/densely dotted}
    Shorthand for setting a densely dotted dash pattern.
    %
\begin{codeexample}[]
\tikz \draw[densely dotted] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/loosely dotted}
    Shorthand for setting a loosely dotted dash pattern.
    %
\begin{codeexample}[]
\tikz \draw[loosely dotted] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/dashed}
    Shorthand for setting a dashed dash pattern.
    %
\begin{codeexample}[]
\tikz \draw[dashed] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/densely dashed}
    Shorthand for setting a densely dashed dash pattern.
    %
\begin{codeexample}[]
\tikz \draw[densely dashed] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/loosely dashed}
    Shorthand for setting a loosely dashed dash pattern.
    %
\begin{codeexample}[]
\tikz \draw[loosely dashed] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/dash dot}
    Shorthand for setting a dashed and dotted dash pattern.
    %
\begin{codeexample}[]
\tikz \draw[dash dot] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/densely dash dot}
    Shorthand for setting a densely dashed and dotted dash pattern.
    %
\begin{codeexample}[]
\tikz \draw[densely dash dot] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/loosely dash dot}
    Shorthand for setting a loosely dashed and dotted dash pattern.
    %
\begin{codeexample}[]
\tikz \draw[loosely dash dot] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/dash dot dot}
    Shorthand for setting a dashed and dotted dash pattern with more dots.
    %
\begin{codeexample}[]
\tikz \draw[dash dot dot] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/densely dash dot dot}
    Shorthand for setting a densely dashed and dotted dash pattern with more dots.
    %
\begin{codeexample}[]
\tikz \draw[densely dash dot dot] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}

\begin{stylekey}{/tikz/loosely dash dot dot}
    Shorthand for setting a loosely dashed and dotted dash pattern with more dots.
    %
\begin{codeexample}[]
\tikz \draw[loosely dash dot dot] (0pt,0pt) -- (50pt,0pt);
\end{codeexample}
    %
\end{stylekey}


\subsubsection{Graphic Parameters: Draw Opacity}

When a line is drawn, it will normally ``obscure'' everything behind it as if
you had used perfectly opaque ink. It is also possible to ask \tikzname\ to use
an ink that is a little bit (or a big bit) transparent using the |draw opacity|
option. This is explained in Section~\ref{section-tikz-transparency} on
transparency in more detail.


\subsubsection{Graphic Parameters: Double Lines and Bordered Lines}

\begin{key}{/tikz/double=\meta{core color} (default white)}
    This option causes ``two'' lines to be drawn instead of a single one.
    However, this is not what really happens. In reality, the path is drawn
    twice. First, with the normal drawing color, secondly with the \meta{core
    color}, which is normally |white|. Upon the second drawing, the line width
    is reduced. The net effect is that it appears as if two lines had been
    drawn and this works well even with complicated, curved paths:
    %
\begin{codeexample}[]
\tikz \draw[double]
  plot[smooth cycle] coordinates{(0,0) (1,1) (1,0) (0,1)};
\end{codeexample}

    You can also use the doubling option to create an effect in which a line
    seems to have a certain ``border'':
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw (0,0) -- (1,1);
  \draw[draw=white,double=red,very thick] (0,1) -- (1,0);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

\begin{key}{/tikz/double distance=\meta{dimension} (initially 0.6pt)}
    Sets the distance the ``two'' lines are spaced apart. In reality, this is
    the thickness of the line that is used to draw the path for the second
    time. The thickness of the \emph{first} time the path is drawn is twice the
    normal line width plus the given \meta{dimension}. As a side-effect, this
    option ``selects'' the |double| option.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw[very thick,double]              (0,0) arc (180:90:1cm);
  \draw[very thick,double distance=2pt] (1,0) arc (180:90:1cm);
  \draw[thin,double distance=2pt]       (2,0) arc (180:90:1cm);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

\begin{key}{/tikz/double distance between line centers=\meta{dimension}}
    This option works like |double distance|, only the distance is not the
    distance between (inner) borders of the two main lines, but between their
    centers. Thus, the thickness the \emph{first} time the path is drawn is the
    normal line width plus the given \meta{dimension}, while the line width of
    the \emph{second} line that is drawn is \meta{dimension} minus the normal
    line width. As a side-effect, this option ``selects'' the |double| option.
    %
\begin{codeexample}[]
\begin{tikzpicture}[double distance between line centers=3pt]
  \foreach \lw in {0.5,1,1.5,2,2.5}
    \draw[line width=\lw pt,double] (\lw,0) -- ++(4mm,0);
\end{tikzpicture}
\end{codeexample}
    %
\begin{codeexample}[]
\begin{tikzpicture}[double distance=3pt]
  \foreach \lw in {0.5,1,1.5,2,2.5}
    \draw[line width=\lw pt,double] (\lw,0) -- ++(4mm,0);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

\begin{stylekey}{/tikz/double equal sign distance}
    This style selects a double line distance such that it corresponds to the
    distance of the two lines in an equal sign.
    %
\begin{codeexample}[]
\Huge $=\implies$\tikz[baseline,double equal sign distance]
                    \draw[double,thick,-implies](0,0.55ex) --++(3ex,0);
\end{codeexample}
    %
\begin{codeexample}[]
\normalsize $=\implies$\tikz[baseline,double equal sign distance]
                          \draw[double,-implies](0,0.6ex) --++(3ex,0);
\end{codeexample}
    %
\begin{codeexample}[]
\tiny $=\implies$\tikz[baseline,double equal sign distance]
                   \draw[double,very thin,-implies](0,0.5ex) -- ++(3ex,0);
\end{codeexample}
    %
\end{stylekey}


\subsection{Adding Arrow Tips to a Path}
\label{section-arrow-tip-action}

In different situations, \tikzname\ will add arrow tips to the end of a path.
For this to happen, a number of different things need to be specified:
%
\begin{enumerate}
    \item You must have used the |arrows| key, explained in detail in
        Section~\ref{section-tikz-arrows}, to setup which kinds of arrow tips
        you would like.
    \item The path may not be closed (like a circle or a rectangle) and, if
        it consists of several subpath, further restrictions apply as
        explained in Section~\ref{section-tikz-arrows}.
    \item The |tips| key must be set to an appropriate value, see
        Section~\ref{section-tikz-arrows} once more.
\end{enumerate}

For the current section on paths, it is only important that when you add the
|tips| option to a path that is not drawn, arrow tips will still be added at
the beginning and at the end of the current path. This is true even when
``only'' arrow tips get drawn for a path without drawing the path itself. Here
is an example:
%
\begin{codeexample}[width=2cm]
\tikz \path[tips, -{Latex[open,length=10pt,bend]}] (0,0) to[bend left] (1,0);
\end{codeexample}
%
\begin{codeexample}[width=2cm]
\tikz \draw[tips, -{Latex[open,length=10pt,bend]}] (0,0) to[bend left] (1,0);
\end{codeexample}


\subsection{Filling a Path}
\label{section-rules}

To fill a path, use the following option:
%
\begin{key}{/tikz/fill=\meta{color} (default \normalfont is scope's color setting)}
    This option causes the path to be filled. All unclosed parts of the path
    are first closed, if necessary. Then, the area enclosed by the path is
    filled with the current filling color, which is either the last color set
    using the general |color=| option or the optional color \meta{color}. For
    self-intersection paths and for paths consisting of several closed areas,
    the ``enclosed area'' is somewhat complicated to define and two different
    definitions exist, namely the nonzero winding number rule and the even odd
    rule, see the explanation of these options, below.

    Just as for the |draw| option, setting \meta{color} to |none| disables
    filling locally.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \fill (0,0) -- (1,1) -- (2,1);
  \fill (4,0) circle (.5cm)  (4.5,0) circle (.5cm);
  \fill[even odd rule] (6,0) circle (.5cm)  (6.5,0) circle (.5cm);
  \fill (8,0) -- (9,1) -- (10,0) circle (.5cm);
\end{tikzpicture}
\end{codeexample}

    If the |fill| option is used together with the |draw| option (either
    because both are given as options or because a |\filldraw| command is
    used), the path is filled \emph{first}, then the path is drawn
    \emph{second}. This is especially useful if different colors are selected
    for drawing and for filling. Even if the same color is used, there is a
    difference between this command and a plain |fill|: A ``filldrawn'' area
    will be slightly larger than a filled area because of the thickness of the
    ``pen''.
    %
\begin{codeexample}[]
\begin{tikzpicture}[fill=yellow!80!black,line width=5pt]
  \filldraw (0,0) -- (1,1) -- (2,1);
  \filldraw (4,0) circle (.5cm)  (4.5,0) circle (.5cm);
  \filldraw[even odd rule] (6,0) circle (.5cm)  (6.5,0) circle (.5cm);
  \filldraw (8,0) -- (9,1) -- (10,0) circle (.5cm);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}


\subsubsection{Graphic Parameters: Fill Pattern}
\label{section-fill-pattern}

Instead of filling a path with a single solid color, it is also possible to
fill it with a \emph{tiling pattern}. Imagine a small tile that contains a
simple picture like a star. Then these tiles are (conceptually) repeated
infinitely in all directions, but clipped against the path.

Tiling patterns come in two variants: \emph{inherently colored patterns} and
\emph{form-only patterns}. An inherently colored pattern is, say, a red star
with a black border and will always look like this. A form-only pattern may
have a different color each time it is used, only the form of the pattern will
stay the same. As such, form-only patterns do not have any colors of their own,
but when it is used the current \emph{pattern color} is used as its color.

Patterns are not overly flexible. In particular, it is not possible to change
the size or orientation of a pattern without declaring a new pattern. For
complicated cases, it may be easier to use two nested |\foreach| statements to
simulate a pattern, but patterns are rendered \emph{much} more quickly than
simulated ones.

\begin{key}{/tikz/pattern=\meta{name} (default \normalfont is scope's pattern)}
    This option causes the path to be filled with a pattern. If the \meta{name}
    is given, this pattern is used, otherwise the pattern set in the enclosing
    scope is used. As for the |draw| and |fill| options, setting \meta{name} to
    |none| disables filling locally.

    The pattern works like a fill color. In particular, setting a new fill
    color will fill the path with a solid color once more.

    Strangely, no \meta{name}s are permissible by default. You need to load for
    instance the |patterns| library, see
    Section~\ref{section-library-patterns}, to install predefined patterns.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw[pattern=dots] (0,0) circle (1cm);
  \draw[pattern=fivepointed stars] (0,0) rectangle (3,1);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

\begin{key}{/tikz/pattern color=\meta{color}}
    This option is used to set the color to be used for form-only patterns.
    This option has no effect on inherently colored patterns.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw[pattern color=red,pattern=fivepointed stars]  (0,0) circle (1cm);
  \draw[pattern color=blue,pattern=fivepointed stars] (0,0) rectangle (3,1);
\end{tikzpicture}
\end{codeexample}

\begin{codeexample}[]
\begin{tikzpicture}
  \def\mypath{(0,0) -- +(0,1) arc (180:0:1.5cm) -- +(0,-1)}
  \fill   [red]                                \mypath;
  \pattern[pattern color=white,pattern=bricks] \mypath;
\end{tikzpicture}
\end{codeexample}
    %
\end{key}


\subsubsection{Graphic Parameters: Interior Rules}

The following two options can be used to decide how interior points should be
determined:
%
\begin{key}{/tikz/nonzero rule}
    If this rule is used (which is the default), the following method is used
    to determine whether a given point is ``inside'' the path: From the point,
    shoot a ray in some direction towards infinity (the direction is chosen
    such that no strange borderline cases occur). Then the ray may hit the
    path. Whenever it hits the path, we increase or decrease a counter, which
    is initially zero. If the ray hits the path as the path goes ``from left to
    right'' (relative to the ray), the counter is increased, otherwise it is
    decreased. Then, at the end, we check whether the counter is nonzero (hence
    the name). If so, the point is deemed to lie ``inside'', otherwise it is
    ``outside''. Sounds complicated? It is.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \filldraw[fill=yellow!80!black]
  % Clockwise rectangle
  (0,0) -- (0,1) -- (1,1) -- (1,0) -- cycle
  % Counter-clockwise rectangle
  (0.25,0.25) -- (0.75,0.25) -- (0.75,0.75) -- (0.25,0.75) -- cycle;

  \draw[->] (0,1) -- (.4,1);
  \draw[->] (0.75,0.75) -- (0.3,.75);

  \draw[->] (0.5,0.5) -- +(0,1) node[above] {crossings: $-1+1 = 0$};

  \begin{scope}[yshift=-3cm]
    \filldraw[fill=yellow!80!black]
    % Clockwise rectangle
    (0,0) -- (0,1) -- (1,1) -- (1,0) -- cycle
    % Clockwise rectangle
    (0.25,0.25) -- (0.25,0.75) -- (0.75,0.75) -- (0.75,0.25) -- cycle;

    \draw[->] (0,1) -- (.4,1);
    \draw[->] (0.25,0.75) -- (0.4,.75);

    \draw[->] (0.5,0.5) -- +(0,1) node[above] {crossings: $1+1 = 2$};
  \end{scope}
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

\begin{key}{/tikz/even odd rule}
    This option causes a different method to be used for determining the inside
    and outside of paths. While it is less flexible, it turns out to be more
    intuitive.

    With this method, we also shoot rays from the point for which we wish to
    determine whether it is inside or outside the filling area. However, this
    time we only count how often we ``hit'' the path and declare the point to
    be ``inside'' if the number of hits is odd.

    Using the even-odd rule, it is easy to ``drill holes'' into a path.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \filldraw[fill=yellow!80!black,even odd rule]
    (0,0) rectangle (1,1) (0.5,0.5) circle (0.4cm);
  \draw[->] (0.5,0.5) -- +(0,1) [above] node{crossings: $1+1 = 2$};
\end{tikzpicture}
\end{codeexample}
    %
\end{key}


\subsubsection{Graphic Parameters: Fill Opacity}
\label{section-fill-opacity}

Analogously to the |draw opacity|, you can also set the fill opacity. Please
see Section~\ref{section-tikz-transparency} for more details.


\subsection{Generalized Filling: Using Arbitrary Pictures to Fill a Path}

Sometimes you wish to ``fill'' a path with something even more complicated than
a pattern, let alone a single color. For instance, you might wish to use an
image to fill the path or some other, complicated drawing. In principle, this
effect can be achieved by first using the path for clipping and then,
subsequently, drawing the desired image or picture. However, there is an option
that makes this process much easier:

\begin{key}{/tikz/path picture=\meta{code}}
    When this option is given on a path and when the \meta{code} is not empty,
    the following happens: After all other ``filling'' operations are done with
    the path, which are caused by the options |fill|, |pattern| and  |shade|, a
    local scope is opened and the path is temporarily installed as a clipping
    path. Then, the \meta{code} is executed, which can now draw something.
    Then, the local scope ends and, possibly, the path is stroked, provided the
    |draw| option has been given.

    As with other keys like |fill| or |draw| this option needs to be given on a
    path, setting the |path picture| outside a path has no effect (the path
    picture is cleared at the beginning of each path).

    The \meta{code} can be any normal \tikzname\ code like |\draw ...| or
    |\node ...|. As always, when you include an external graphic, you need to
    put it inside a |\node|.

    Note that no special actions are taken to transform the origin in any way.
    This means that the coordinate |(0,0)| is still where is was when the path
    was being constructed and not -- as one might expect -- at the lower left
    corner of the path. However, you can use the following special node to
    access the size of the path:
    %
    \begin{predefinednode}{path picture bounding box}
        This node is of shape |rectangle|. Its size and position are those of
        |current path bounding box| just before the \meta{code} of the path
        picture started to be executed. The \meta{code} can construct its own
        paths, so accessing the |current path bounding box| inside the
        \meta{code} yields the bounding box of any path that is currently being
        constructed inside the \meta{code}.
    \end{predefinednode}
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw [help lines] (0,0) grid (3,2);
  \filldraw [fill=blue!10,draw=blue,thick] (1.5,1) circle (1)
    [path picture={
      \node at (path picture bounding box.center) {
        This is a long text.
      };}
    ];
\end{tikzpicture}
\end{codeexample}

\begin{codeexample}[]
\begin{tikzpicture}[cross/.style={path picture={
      \draw[black]
            (path picture bounding box.south east) --
            (path picture bounding box.north west)
            (path picture bounding box.south west) --
            (path picture bounding box.north east);
    }}]
  \draw [help lines] (0,0) grid (3,2);
  \filldraw [cross,fill=blue!10,draw=blue,thick] (1,1) circle (1);
  \path     [cross,top color=red,draw=red,thick] (2,0) -- (3,2) -- (3,0);
\end{tikzpicture}
\end{codeexample}

\begin{codeexample}[]
  \begin{tikzpicture}[path image/.style={
      path picture={
        \node at (path picture bounding box.center) {
          \includegraphics[height=3cm]{#1}
        };}}]
  \draw     [help lines] (0,0) grid (3,2);

  \draw [path image=brave-gnu-world-logo,draw=blue,thick]
          (0,1) circle (1);
  \draw [path image=brave-gnu-world-logo,draw=red,very thick,->]
          (1,0) parabola[parabola height=2cm] (3,0);

\end{tikzpicture}
\end{codeexample}
    %
\end{key}


\subsection{Shading a Path}

You can shade a path using the |shade| option. A shading is like a filling,
only the shading changes its color smoothly from one color to another.

\begin{key}{/tikz/shade}
    Causes the path to be shaded using the currently selected shading (more on
    this later). If this option is used together with the |draw| option, then
    the path is first shaded, then drawn.

    It is not an error to use this option together with the |fill| option, but
    it makes no sense.
    %
\begin{codeexample}[]
\tikz \shade (0,0) circle (1ex);
\end{codeexample}

\begin{codeexample}[]
\tikz \shadedraw (0,0) circle (1ex);
\end{codeexample}
    %
\end{key}

For some shadings it is not really clear how they can ``fill'' the path. For
example, the |ball| shading normally looks like this:
\tikz \shade[shading=ball] (0,0) circle (0.75ex);. How is this supposed to
shade a rectangle? Or a triangle?

To solve this problem, the predefined shadings like |ball| or |axis| fill a
large rectangle completely in a sensible way. Then, when the shading is used to
``shade'' a path, what actually happens is that the path is temporarily used
for clipping and then the rectangular shading is drawn, scaled and shifted such
that all parts of the path are filled.

The default shading is a smooth transition from gray to white and from top to
bottom. However, other shadings are also possible, for example a shading that
will sweep a color from the center to the corners outward. To choose the
shading, you can use the |shading=| option, which will also automatically
invoke the |shade| option. Note that this does \emph{not} change the shading
color, only the way the colors sweep. For changing the colors, other options
are needed, which are explained below.

\begin{key}{/tikz/shading=\meta{name}}
    This selects a shading named \meta{name}. The following shadings are
    predefined: |axis|, |radial|, and |ball|.
    %
\begin{codeexample}[]
\tikz \shadedraw [shading=axis] (0,0) rectangle (1,1);
\tikz \shadedraw [shading=radial] (0,0) rectangle (1,1);
\tikz \shadedraw [shading=ball] (0,0) circle (.5cm);
\end{codeexample}

    The shadings as well as additional shadings are described in more detail in
    Section~\ref{section-library-shadings}.

    To change the color of a shading, special options are needed like
    |left color|, which sets the color of an axis shading from left to right.
    These options implicitly also select the correct shading type, see the
    following example
    %
\begin{codeexample}[]
\tikz \shadedraw [left color=red,right color=blue]
    (0,0) rectangle (1,1);
\end{codeexample}

    For a complete list of the possible options see
    Section~\ref{section-library-shadings} once more.

    \begin{key}{/tikz/shading angle=\meta{degrees} (initially 0)}
        This option rotates the shading (not the path!) by the given angle. For
        example, we can turn a top-to-bottom axis shading into a left-to-right
        shading by rotating it by $90^\circ$.
        %
\begin{codeexample}[]
\tikz \shadedraw [shading=axis,shading angle=90] (0,0) rectangle (1,1);
\end{codeexample}
    \end{key}
\end{key}

You can also define new shading types yourself. However, for this, you need to
use the basic layer directly, which is, well, more basic and harder to use.
Details on how to create a shading appropriate for filling paths are given in
Section~\ref{section-shading-a-path}.


\subsection{Establishing a Bounding Box}

\pgfname\ is reasonably good at keeping track of the size of your picture and
reserving just the right amount of space for it in the main document. However,
in some cases you may want to say things like ``do not count this for the
picture size'' or ``the picture is actually a little large''. For this you can
use the option |use as bounding box| or the command |\useasboundingbox|, which
is just a shorthand for |\path[use as bounding box]|.

\begin{key}{/tikz/use as bounding box}
    Normally, when this option is given on a path, the bounding box of the
    present path is used to determine the size of the picture and the size of
    all \emph{subsequent} paths are ignored. However, if there were previous
    path operations that have already established a larger bounding box, it
    will not be made smaller by this operation (consider the
    |\pgfresetboundingbox| command to reset the previous bounding box).

    In a sense, |use as bounding box| has the same effect as clipping all
    subsequent drawing against the current path -- without actually doing the
    clipping, only making \pgfname\ treat everything as if it were clipped.

    The first application of this option is to have a |{tikzpicture}| overlap
    with the main text:
    %
\begin{codeexample}[]
Left of picture\begin{tikzpicture}
  \draw[use as bounding box] (2,0) rectangle (3,1);
  \draw (1,0) -- (4,.75);
\end{tikzpicture}right of picture.
\end{codeexample}

    In a second application this option can be used to get better control over
    the white space around the picture:
    %
\begin{codeexample}[]
Left of picture
\begin{tikzpicture}
  \useasboundingbox (0,0) rectangle (3,1);
  \fill (.75,.25) circle (.5cm);
\end{tikzpicture}
right of picture.
\end{codeexample}

    Note: If this option is used on a path inside a \TeX\ group (scope), the
    effect ``lasts'' only until the end of the scope. Again, this behavior is
    the same as for clipping.

    Consider using |\useasboundingbox| together with |\pgfresetboundingbox| in
    order to replace the bounding box with a new one.
\end{key}

There is a node that allows you to get the size of the current bounding box.
The |current bounding box| node has the |rectangle| shape and its size is
always the size of the current bounding box.

Similarly, the |current path bounding box| node has the |rectangle| shape and
the size of the bounding box of the current path.
%
\begin{codeexample}[]
\begin{tikzpicture}
  \draw[red] (0,0) circle (2pt);
  \draw[red] (2,1) circle (3pt);

  \draw (current bounding box.south west) rectangle
        (current bounding box.north east);

  \draw[red] (3,-1) circle (4pt);

  \draw[thick] (current bounding box.south west) rectangle
               (current bounding box.north east);
\end{tikzpicture}
\end{codeexample}

Occasionally, you may want to align multiple |tikzpicture| environments
horizontally and/or vertically at some prescribed position. The vertical
alignment can be realized by means of the |baseline| option since \TeX\
supports the concept of box depth natively. For horizontal alignment, things
are slightly more involved. The following approach is realized by means of
negative |\hspace|s before and/or after the picture, thereby removing parts of
the picture. However, the actual amount of negative horizontal space is
provided by means of image coordinates using the |trim left| and |trim right|
keys:

\begin{key}{/tikz/trim left=\meta{dimension or coordinate or \texttt{default}} (default 0pt)}
    The |trim left| key tells \pgfname\space to discard everything which is
    left of the provided \meta{dimension or coordinate}. Here, \meta{dimension}
    is a single $x$ coordinate of the picture and \meta{coordinate} is a point
    with $x$ and $y$ coordinates (but only its $x$ coordinate will be used).
    The effect is the same as if you issue |\hspace{-s}| where |s| is the
    difference of the picture's bounding box lower left $x$ coordinate and the
    $x$ coordinate specified as \meta{dimension or coordinate}:
    %
\begin{codeexample}[]
Text before image.%
    \begin{tikzpicture}[trim left]
        \draw (-1,-1) grid (3,2);
        \fill (0,0) circle (5pt);
    \end{tikzpicture}%
Text after image.
\end{codeexample}
    %
    Since |trim left| uses the default |trim left=0pt|, everything left of
    $x=0$ is removed from the bounding box.

    The following example has once the relative long label $-1$ and once the
    shorter label $1$. Horizontal alignment is established with |trim left|:
    %
\begin{codeexample}[pre={\vbox\bgroup\hsize=5cm},post=\egroup,width=8cm]
\begin{tikzpicture}
    \draw (0,1) -- (0,0) -- (1,1) -- cycle;
    \fill (0,0) circle (2pt);
    \node[left] at (0,0) {$-1$};
\end{tikzpicture}
\par
\begin{tikzpicture}
    \draw (0,1) -- (0,0) -- (1,1) -- cycle;
    \fill (0,0) circle (2pt);
    \node[left] at (0,0) {$1$};
\end{tikzpicture}
\par
\begin{tikzpicture}[trim left]
    \draw (0,1) -- (0,0) -- (1,1) -- cycle;
    \fill (0,0) circle (2pt);
    \node[left] at (0,0) {$-1$};
\end{tikzpicture}
\par
\begin{tikzpicture}[trim left]
    \draw (0,1) -- (0,0) -- (1,1) -- cycle;
    \fill (0,0) circle (2pt);
    \node[left] at (0,0) {$1$};
\end{tikzpicture}
\end{codeexample}

    Use |trim left=default| to reset the value.
\end{key}

\begin{key}{/tikz/trim right=\meta{dimension or coordinate or \texttt{default}}}
    This key is similar to |trim left|: it discards everything which is right
    of the provided \meta{dimension or coordinate}. As for |trim left|,
    \meta{dimension} denotes a single $x$ coordinate of the picture and
    \meta{coordinate} a coordinate with $x$ and $y$ value (although only its
    $x$ component will be used).

    We use the same example from above and add |trim right|:
    %
\begin{codeexample}[]
Text before image.%
    \begin{tikzpicture}[trim left, trim right=2cm, baseline]
        \draw (-1,-1) grid (3,2);
        \fill (0,0) circle (5pt);
    \end{tikzpicture}%
Text after image.
\end{codeexample}
    %
    In addition to |trim left=0pt|, we also discard everything which is right
    of $x$|=2cm|. Furthermore, the |baseline| key supports vertical alignment
    as well (using the $y$|=0cm| baseline).

    Use |trim right=default| to reset the value.
\end{key}

Note that |baseline|, |trim left| and |trim right| are currently the
\emph{only} supported way of truncated bounding boxes which are compatible with
image externalization (see the |external| library for details).

\begin{key}{/pgf/trim lowlevel=\mchoice{true,false} (initially false)}
    This affects only the basic level image externalization: the initial
    configuration |trim lowlevel=false| stores the normal image, without
    trimming, and the trimming into a separate file. This allows reduced
    bounding boxes without clipping the rest away. The |trim lowlevel=true|
    information causes the image externalization to store the trimmed image,
    possibly resulting in clipping.
\end{key}


\subsection{Clipping and Fading (Soft Clipping)}

\emph{Clipping path} means that all painting on the page is restricted to a
certain area. This area need not be rectangular, rather an arbitrary path can
be used to specify this area. The |clip| option, explained below, is used to
specify the region that is to be used for clipping.

A \emph{fading} (a term that I propose, fadings are commonly known as soft
masks, transparency masks, opacity masks or soft clips) is similar to clipping,
but a fading allows parts of the picture to be only ``half clipped''. This
means that a fading can specify that newly painted pixels should be partly
transparent. The specification and handling of fadings is a bit complex and it
is detailed in Section~\ref{section-tikz-transparency}, which is devoted to
transparency in general.

\begin{key}{/tikz/clip}
    This option causes all subsequent drawings to be clipped against the
    current path and the size of subsequent paths will not be important for the
    picture size. If you clip against a self-intersecting path, the even-odd
    rule or the nonzero winding number rule is used to determine whether a
    point is inside or outside the clipping region.

    The clipping path is a graphic state parameter, so it will be reset at the
    end of the current scope. Multiple clippings accumulate, that is, clipping
    is always done against the intersection of all clipping areas that have
    been specified inside the current scopes. The only way of enlarging the
    clipping area is to end a |{scope}|.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw[clip] (0,0) circle (1cm);
  \fill[red] (1,0) circle (1cm);
\end{tikzpicture}
\end{codeexample}

    It  is usually a \emph{very} good idea to apply the |clip| option only to
    the first path command in a scope.

    If you ``only wish to clip'' and do not wish to draw anything, you can use
    the |\clip| command, which is a shorthand for |\path[clip]|.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \clip (0,0) circle (1cm);
  \fill[red] (1,0) circle (1cm);
\end{tikzpicture}
\end{codeexample}

    To keep clipping local, use |{scope}| environments as in the following
    example:
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw (0,0) -- ( 0:1cm);
  \draw (0,0) -- (10:1cm);
  \draw (0,0) -- (20:1cm);
  \draw (0,0) -- (30:1cm);
  \begin{scope}[fill=red]
    \fill[clip] (0.2,0.2) rectangle (0.5,0.5);

    \draw (0,0) -- (40:1cm);
    \draw (0,0) -- (50:1cm);
    \draw (0,0) -- (60:1cm);
  \end{scope}
  \draw (0,0) -- (70:1cm);
  \draw (0,0) -- (80:1cm);
  \draw (0,0) -- (90:1cm);
\end{tikzpicture}
\end{codeexample}

    There is a slightly annoying catch: You cannot specify certain graphic
    options for the command used for clipping. For example, in the above code
    we could not have moved the |fill=red| to the |\fill| command. The reasons
    for this have to do with the internals of the \pdf\ specification. You do
    not want to know the details. It is best simply not to specify any options
    for these commands.
\end{key}


\subsection{Doing Multiple Actions on a Path}

If more than one of the basic actions like drawing, clipping and filling are
requested, they are automatically applied in a sensible order: First, a path is
filled, then drawn, and then clipped (although it took Apple two mayor
revisions of their operating system to get this right\dots). Sometimes,
however, you need finer control over what is done with a path. For instance,
you might wish to first fill a path with a color, then repaint the path with a
pattern and then repaint it with yet another pattern. In such cases you can use
the following two options:

\begin{key}{/tikz/preaction=\meta{options}}
    This option can be given to a |\path| command (or to derived commands like
    |\draw| which internally call |\path|). Similarly to options like |draw|,
    this option only has an effect when given to a |\path| or as part of the
    options of a |node|; as an option to a |{scope}| it has no effect.

    When this option is used on a |\path|, the effect is the following: When
    the path has been completely constructed and is about to be used, a scope
    is created. Inside this scope, the path is used but not with the original
    path options, but with \meta{options} instead. Then, the path is used in
    the usual manner. In other words, the path is used twice: Once with
    \meta{options} in force and then again with the normal path options in
    force.

    Here is an example in which the path consists of a rectangle. The main
    action is to draw this path in red (which is why we see a red rectangle).
    However, the preaction is to draw the path in blue, which is why we see a
    blue rectangle behind the red rectangle.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw[help lines] (0,0) grid (3,2);

  \draw
    [preaction={draw,line width=4mm,blue}]
    [line width=2mm,red] (0,0) rectangle (2,2);
\end{tikzpicture}
\end{codeexample}

    Note that when the preactions are preformed, then the path is already
    ``finished''. In particular, applying a coordinate transformation to the
    path has no effect. By comparison, applying a canvas transformation does
    have an effect. Let us use this to add a ``shadow'' to a path. For this, we
    use the preaction to fill the path in gray, shifted a bit to the right and
    down:
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw[help lines] (0,0) grid (3,2);
  \draw
    [preaction={fill=black,opacity=.5,
                transform canvas={xshift=1mm,yshift=-1mm}}]
    [fill=red] (0,0) rectangle (1,2)
               (1,2) circle (5mm);
\end{tikzpicture}
\end{codeexample}

    Naturally, you would normally create a style |shadow| that contains the
    above code. The shadow library, see Section~\ref{section-libs-shadows},
    contains predefined shadows of this kind.

    It is possible to use the |preaction| option multiple times. In this case,
    for each use of the |preaction| option, the path is used again (thus, the
    \meta{options} do not accumulate in a single usage of the path). The path
    is used in the order of |preaction| options given.

    In the following example, we use one |preaction| to add a shadow and
    another to provide a shading, while the main action is to use a pattern.
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw[help lines] (0,0) grid (3,2);
  \draw [pattern=fivepointed stars]
    [preaction={fill=black,opacity=.5,
                transform canvas={xshift=1mm,yshift=-1mm}}]
    [preaction={top color=blue,bottom color=white}]
               (0,0) rectangle (1,2)
               (1,2) circle (5mm);
\end{tikzpicture}
\end{codeexample}

    A complicated application is shown in the following example, where the path
    is used several times with different fadings and shadings to create a
    special visual effect:
    %
\begin{codeexample}[]
\begin{tikzpicture}
  [
    % Define an interesting style
    button/.style={
      % First preaction: Fuzzy shadow
      preaction={fill=black,path fading=circle with fuzzy edge 20 percent,
                 opacity=.5,transform canvas={xshift=1mm,yshift=-1mm}},
      % Second preaction: Background pattern
      preaction={pattern=#1,
                 path fading=circle with fuzzy edge 15 percent},
      % Third preaction: Make background shiny
      preaction={top color=white,
                 bottom color=black!50,
                 shading angle=45,
                 path fading=circle with fuzzy edge 15 percent,
                 opacity=0.2},
      % Fourth preaction: Make edge especially shiny
      preaction={path fading=fuzzy ring 15 percent,
                 top color=black!5,
                 bottom color=black!80,
                 shading angle=45},
      inner sep=2ex
    },
    button/.default=horizontal lines light blue,
    circle
  ]

  \draw [help lines] (0,0) grid (4,3);

  \node [button] at (2.2,1) {\Huge Big};
  \node [button=crosshatch dots light steel blue,
         text=white] at (1,1.5) {Small};
\end{tikzpicture}
\end{codeexample}
    %
\end{key}

\begin{key}{/tikz/postaction=\meta{options}}
    The postactions work in the same way as the preactions, only they are
    applied \emph{after} the main action has been taken. Like preactions,
    multiple |postaction| options may be given to a |\path| command, in which
    case the path is reused several times, each time with a different set of
    options in force.

    If both pre- and postactions are specified, then the preactions are taken
    first, then the main action, and then the post actions.

    In the first example, we use a postaction to draw the path, after it has
    already been drawn:
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw[help lines] (0,0) grid (3,2);

  \draw
    [postaction={draw,line width=2mm,blue}]
    [line width=4mm,red,fill=white] (0,0) rectangle (2,2);
\end{tikzpicture}
\end{codeexample}

    In another example, we use a postaction to ``colorize'' a path:
    %
\begin{codeexample}[]
\begin{tikzpicture}
  \draw[help lines] (0,0) grid (3,2);
  \draw
    [postaction={path fading=south,fill=white}]
    [postaction={path fading=south,fading angle=45,fill=blue,opacity=.5}]
    [left color=black,right color=red,draw=white,line width=2mm]
               (0,0) rectangle (1,2)
               (1,2) circle (5mm);
\end{tikzpicture}
\end{codeexample}
    %
\end{key}


\subsection{Decorating and Morphing a Path}

Before a path is used, it is possible to first ``decorate'' and/or ``morph''
it. Morphing means that the path is replaced by another path that is slightly
varied. Such morphings are a special case of the more general ``decorations''
described in detail in Section~\ref{section-tikz-decorations}. For instance, in
the following example the path is drawn twice: Once normally and then in a
morphed (=decorated) manner.
%
\begin{codeexample}[]
\begin{tikzpicture}
  \draw (0,0) rectangle (3,2);
  \draw [red, decorate, decoration=zigzag]
        (0,0) rectangle (3,2);
\end{tikzpicture}
\end{codeexample}

Naturally, we could have combined this into a single command using pre- or
postaction. It is also possible to deform shapes:
%
\begin{codeexample}[]
\begin{tikzpicture}
  \node [circular drop shadow={shadow scale=1.05},minimum size=3.13cm,
         decorate, decoration=zigzag,
         fill=blue!20,draw,thick,circle] {Hello!};
\end{tikzpicture}
\end{codeexample}