path: root/graphics/pgf/base/doc/text-en/pgfmanual-en-tikz-pics.tex

% Copyright 2013 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{Pics: Small Pictures on Paths}
\label{section-pics}

\subsection{Overview}

A ``pic'' is a ``short picture'' (hence the short name\dots) that can be
inserted anywhere in \tikzname\ picture where you could also insert a node.
Similarly to nodes, pics have a ``shape'' (called \emph{type} to avoid
confusion) that someone has defined. Each time a pic of a specified type is
used, the type's code is executed, resulting in some drawings to be added to
the current picture. The syntax for adding nodes and adding pics to a picture
are also very similar. The core difference is that pics are typically more
complex than nodes and may consist of a whole bunch of nodes themselves
together with complex paths joining them.

As a very simple example, suppose we want to define a pic type |seagull| that
just draw ``two bumps''. The code for this definition is quite easy:
%
\begin{codeexample}[code only]
\tikzset{
  seagull/.pic={
    % Code for a "seagull". Do you see it?...
    \draw (-3mm,0) to [bend left] (0,0) to [bend left] (3mm,0);
  }
}
\end{codeexample}

The first line just tells \TeX\ that you set some \tikzname\ options for the
current scope (which is the whole document); you could put |seagull/.pic=...|
anywhere else where \tikzname\ options are allowed (which is just about
anywhere). We have now defined a |seagull| pic type and can use it as follows:
%
\tikzset{
  seagull/.pic={
    % Code for a "seagull". Do you see it?...
    \draw (-3mm,0) to [bend left] (0,0) to [bend left] (3mm,0);
  }
}
\begin{codeexample}[]
\tikz \fill [fill=blue!20]
     (1,1)
  -- (2,2) pic             {seagull}
  -- (3,2) pic             {seagull}
  -- (3,1) pic [rotate=30] {seagull}
  -- (2,1) pic [red]       {seagull};
\end{codeexample}

As can be see, defining new types of pics is much easier than defining new
shapes for nodes; but see Section~\ref{section-new-pic-types} for the fine
details.

Since defining new pics types is easier than defining new node shapes and since
using pics is as easy as using nodes, why should you use nodes at all? There
are chiefly two reasons:
%
\begin{enumerate}
    \item Unlike nodes, pics cannot be referenced later on. You \emph{can}
        reference nodes that are inside a pic, but not ``the pic itself''. In
        particular, you cannot draw lines between pics the way you can draw
        them between nodes. In general, whenever it makes sense that some
        drawing could conceivably be connected to other node-like-things, then
        a node is better than a pic.
    \item If pics are used to emulate the full power of a node (which is
        possible, in principle), they will be slower to construct and take up
        more memory than a node achieving the same effect.
\end{enumerate}

Despite these drawbacks, pics are an excellent choice for creating highly
configurable reusable pieces of drawings that can be inserted into larger
contexts.


\subsection{The Pic Syntax}

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

The syntax for adding a pic to a picture is very similar to the syntax used for
nodes (indeed, internally the same parser code is used). The main difference is
that instead of a node contents you provide the picture's type between the
braces:

\begin{pathoperation}{pic}{
    \opt{\meta{foreach statements}}
    \opt{|[|\meta{options}|]|}
    \opt{|(|\meta{prefix}|)|}
    \opt{|at(|\meta{coordinate}|)|}
    \opt{|:|\meta{animation attribute}|=|\marg{options}}
    \opt{\marg{pic type}}%
}
    Adds a pic to the current \tikzname\ picture of the specified \meta{pic
    type}. The effect is, basically, that some code associated with the
    \meta{pic type} is executed (how this works, exactly, is explained later).
    This code can consist of arbitrary \tikzname\ code. As for nodes, the
    current path will not be modified by this path command, all drawings
    produced by the code are ``external'' to the path the same way neither a
    node nor its border are part of the path on which they are specified.

    Just like the |node| command, this path operation is somewhat complex and
    we go over it step by step.

    \medskip
    \textbf{Order of the parts of the specification.}
    Just like for nodes, everything between ``|pic|'' and the opening brace of
    the \meta{pic type} is optional and can be given in any order. If there are
    \meta{foreach statements}, they must come first, directly following
    ``|pic|''. As for nodes, the ``end'' of the pic specification is normally
    detected by the presence of the opening brace. You can, however, use the
    |pic type| option to specify the pic type as an option.

    \begin{key}{/tikz/pic type=\meta{pic type}}
        This key sets the pic type of the current~|pic|. When this option is
        used inside an option block of a |pic|, the parsing of the |pic| ends
        immediately and no pic type in braces is expected. (In other words,
        this option behaves exactly like the |node contents| option and,
        indeed, the two are interchangeable.)
        %
\begin{codeexample}[]
\tikz {
  \path (0,0) pic [pic type = seagull]
        (1,0) pic                      {seagull};
}
\end{codeexample}
    \end{key}

    \medskip
    \textbf{The location of a pic.}
    Just like nodes, pics are placed at the last position mentioned on the path
    or, when |at| is used, at a specified position. ``Placing'' a pic somewhere
    actually means that the coordinate system is translated (shifted) to this
    last position. This means that inside of the pic type's code any mentioning
    of the origin refers to the last position used on the path or to the
    specified |at|.
    %
\begin{codeexample}[]
\tikz { % different ways of placing pics
  \draw [help lines] (0,0) grid (3,2);
  \pic  at (1,0)     {seagull};
  \path (2,1)    pic {seagull};
  \pic  [at={(3,2)}] {seagull};
}
\end{codeexample}

    As for nodes, except for the described shifting, the coordinate system of a
    pic is reset prior to executing the pic type's code. This can be changed
    using the |transform shape| option, which has the same effect as for nodes:
    The ``outer'' transformation gets applied to the node:
    %
\begin{codeexample}[]
\tikz [scale=2] {
  \pic at (0,0)                   {seagull};
  \pic at (1,0) [transform shape] {seagull};
}
\end{codeexample}

    When the \meta{options} contain transformation commands like |scale| or
    |rotate|, these transformations always apply to the pic:
    %
\begin{codeexample}[]
\tikz [rotate=30] {
  \pic at (0,0)             {seagull};
  \pic at (1,0) [rotate=90] {seagull};
}
\end{codeexample}

    Just like nodes, pics can also be positioned implicitly and, somewhat
    unsurprisingly, the same rules concerning positioning and sloping apply:
    %
\begin{codeexample}[]
\tikz \draw
  (0,0) to [bend left]
           pic [near start]       {seagull}
           pic                    {seagull}
           pic [sloped, near end] {seagull} (4,0);
\end{codeexample}

    \medskip
    \textbf{The options of a node.}
    As always, any given \meta{options} apply only to the pic and have no
    effect outside. As for nodes, most ``outside'' options also apply to the
    pics, but not the ``action'' options like |draw| or |fill|. These must be
    given in the \meta{options} of the pic.

    \medskip
    \textbf{The code of a pic.}
    As stated earlier, the main job of a pic is to execute some code in a scope
    that is shifted according to the last point on the path or to the |at|
    position specified in the pic. It was also claimed that this code is
    specified by the \meta{pic type}. However, this specification is somewhat
    indirect. What really happens is the following: When a |pic| is
    encountered, the current path is suspended and a new internal scope is
    started. The \meta{options} are executed and also the \meta{pic type} (as
    explained in a moment). After all this is done, the code stored in the
    following key gets executed:

    \begin{key}{/tikz/pics/code=\meta{code}}
        This key stores the \meta{code} that should be drawn in the current
        pic. Normally, setting this key is done by the \meta{pic type}, but you
        can also set it in the \meta{options} and leave the \meta{pic type}
        empty:
        %
\begin{codeexample}[]
\tikz \pic [pics/code={\draw (-3mm,0) to[bend left] (0,0)
                                      to[bend left] (3mm,0);}]
      {}; % no pic type specified
\end{codeexample}
    \end{key}

    Now, how does the \meta{pic type} set |pics/code|? It turns out that the
    \meta{pic type} is actually just a list of keys that are executed with the
    prefix |/tikz/pics/|. In the above examples, this ``list of keys'' just
    consisted of the single key ``|seagull|'' that did not take any arguments,
    but, in principle, you could provide any arbitrary text understood by
    |\pgfkeys| here. This means that when we write |pic{seagull}|, \tikzname\
    will execute the key |/tikz/pics/seagull|. It turns out, see
    Section~\ref{section-new-pic-types}, that this key is just a style set to
    |code={\draw(-3mm,0)...;}|. Thus, |pic{seagull}| will cause the |pics/code|
    key to be set to the text needed to draw the
    seagull.

    Indeed, you can also use the \meta{pic type} simply to set the |code| of
    the pic. This is useful for cases when you have some code that you ``just
    want to execute, but do not want to define a new pic type''. Here is a
    typical example where we use pics to add some markings to a path:
    %
\begin{codeexample}[]
\tikz \draw (0,0) .. controls(1,0) and (2,1) .. (3,1)
  foreach \t in {0, 0.1, ..., 1} {
    pic [pos=\t] {code={\draw circle [radius=2pt];}}
  };
\end{codeexample}

    In our seagull example, we always explicitly used |\draw| to draw the
    seagull. This implies that when a user writes something
    |pic[fill]{seagull}| in the hope of having a ``filled'' seagull, nothing
    special will happen: The |\draw| inside the pic explicitly states that the
    path should be drawn, not filled, and the fact that in the surrounding
    scope the |fill| option is set has no effect. The following key can be used
    to change this:
    %
    \begin{key}{/tikz/pic actions}
        This key is a style that can be used (only) inside the code of a pic.
        There, it will set the ``action'' keys set inside the \meta{options} of
        the pic (``actions'' are drawing, filling, shading, and clipping or any
        combination thereof).

        To see how this key works, let us define the following pic:
        %
\begin{codeexample}[code only]
\tikzset{
  my pic/.pic = {
    \path [pic actions] (0,0) circle[radius=3mm];
    \draw (-3mm,-3mm) rectangle (3mm,3mm);
  }
}
\end{codeexample}
        %
        In the code, whether or not the circle gets drawn/filled/shaded
        depends on which options where given to the |pic| command when it
        is used. In contrast, the rectangle will always (just) be drawn.
        %
\tikzset{
  my pic/.pic = {
    \path [pic actions] (0,0) circle[radius=3mm];
    \draw (-3mm,-3mm) rectangle (3mm,3mm);
  }
}
\begin{codeexample}[width=6cm]
\tikz \pic                      {my pic}; \space
\tikz \pic [red]                {my pic}; \space
\tikz \pic [draw]               {my pic}; \space
\tikz \pic [draw=red]           {my pic}; \space
\tikz \pic [draw, shading=ball] {my pic}; \space
\tikz \pic [fill=red!50]        {my pic};
\end{codeexample}
    \end{key}

    \medskip
    \textbf{Code executed behind or in front of the path.}
    As for nodes, a pic can be ``behind'' the current path or ``in front of
    it'' and, just as for nodes, the two options |behind path| and
    |in front of path| are used to specify which is meant. In detail, if |node|
    and |pic| are both used repeatedly on a path, in the resulting picture we
    first see all nodes and pics with the |behind path| option set in the order
    they appear on the path (nodes and pics are interchangeable in this
    regard), then comes the path, and then come all nodes and pics that are in
    front of the path in the order they appeared.
    %
\begin{codeexample}[]
\tikz \fill [fill=blue!20]
     (1,1)
  -- (2,2) pic [behind path]      {seagull}
  -- (3,2) pic                    {seagull}
  -- (3,1) pic [rotate=30]        {seagull}
  -- (2,1) pic [red, behind path] {seagull};
\end{codeexample}
    %
    In contrast to nodes, a pic need not only be completely behind the path or
    in front of the path as specified by the user. Instead, a pic type may also
    specify that a certain part of the drawing should always be behind the path
    and it may specify that a certain other part should always be before the
    path. For this, the values of the following keys are relevant:

    \begin{key}{/tikz/pics/foreground code=\meta{code}}
        This key stores \meta{code} that will always be drawn in front of the
        current path, even when |behind path| is used. If |behind path| is not
        used and |code| is (also) set, the code of |code| is drawn first,
        following by the foreground \meta{code}.
    \end{key}

    \begin{key}{/tikz/pics/background code=\meta{code}}
        Like |foreground code|, only that the \meta{code} is always put behind
        the path, except when the |behind path| option is applied to the pic,
        then the background code is drawn in front of the ``behind path'' code.
    \end{key}

    \medskip
    \textbf{The foreach statement for pics.}
    As for nodes, a pic specification may start with |foreach|. The effect and
    semantics are the same as for nodes.
    %
\begin{codeexample}[]
\tikz \pic foreach \x in {1,2,3} at (\x,0) {seagull};
\end{codeexample}

    \medskip
    \textbf{Styles for pics.}
    The following styles influence how nodes are rendered:
    %
    \begin{stylekey}{/tikz/every pic (initially \normalfont empty)}
        This style is installed at the beginning of every pic.
        %
\begin{codeexample}[]
\begin{tikzpicture}[every node/.style={draw}]
  \draw (0,0) node {A} -- (1,1) node {B};
\end{tikzpicture}
\end{codeexample}
    \end{stylekey}

    \medskip
    \textbf{Name scopes.}
    You can specify a \meta{name} for a pic using the key |name=|\meta{name} or
    by giving the name in parenthesis inside the pic's specification. The
    effect of this is, for once, quite different from what happens for nodes:
    All that happens is that |name prefix| is set to \meta{name} at the
    beginning of the pic.

    The |name prefix| key was already introduced in the description of the
    |node| command: It allows you to set some text that is prefixed to all
    nodes in a scope. For pics this makes particular sense: All nodes defined
    by a pic's code can be referenced from outside the pic with the prefix
    provided.

    To see how this works, let us add some nodes to the code of the seagull:
    %
\begin{codeexample}[code only]
\tikzset{
  seagull/.pic={
    % Code for a "seagull". Do you see it?...
    \coordinate (-left wing) at (-3mm,0);
    \coordinate (-head)      at (0,0);
    \coordinate (-right wing) at (3mm,0);

    \draw (-left wing) to [bend left] (0,0) (-head) to [bend left] (-right wing);
  }
}
\end{codeexample}
    %
\tikzset{
  seagull/.pic={
    % Code for a "seagull". Do you see it?...
    \coordinate (-left wing) at (-3mm,0);
    \coordinate (-head)      at (0,0);
    \coordinate (-right wing) at (3mm,0);

    \draw (-left wing) to [bend left] (0,0) (-head) to [bend left] (-right wing);
  }
}

    Now, we can use it as follows:
    %
\begin{codeexample}[code only]
\tikz {
  \pic (Emma)               {seagull};
  \pic (Alexandra) at (0,1) {seagull};

  \draw (Emma-left wing) -- (Alexandra-right wing);
}
\end{codeexample}

    Sometimes, you may also wish your pic to access nodes outside the pic
    (typically, because they are given as parameters). In this case, the name
    prefix gets in the way since the nodes outside the picture do not have this
    prefix. The trick is to locally reset the name prefix to the value it had
    outside the picture, which is achieved using the following style:

    \begin{key}{/tikz/name prefix ..}
        This key is available only inside the code of a pic. There, it
        (locally) changes the name prefix to the value it had outside the pic.
        This allows you to access nodes outside the current pic.
    \end{key}

    \medskip
    \textbf{Animations for pics.}
    Just as for nodes, you can use the attribute--colon syntax to add an
    animation to a pic:
    %
\begin{codeexample}[animation list={0.5,1,1.5,2}]
\tikz {
  \pic :rotate={0s="0", 20s="90"} {seagull};
  \pic at (1.5,1.5)               {seagull};
}
\end{codeexample}
    %
    Naturally, you can also use animations in the code of a picture:
    %
\begin{codeexample}[animation list={0.5,1,1.5,2},width=3cm]
\begin{tikzpicture} [flapping seagull/.pic={
    \draw (0,0) :path={
      0s= {"{(180:3mm) to [bend left] (0,0) to [bend left] (0:3mm)}"=base},
      1s=  "{(160:3mm) to [bend left] (0,0) to [bend left] (20:3mm)}",
      2s=  "{(180:3mm) to [bend left] (0,0) to [bend left] (0:3mm)}",
      repeats };
  }]
  \pic :rotate={0s="0", 20s="90"} {flapping seagull};
  \pic at (1.5,1.5)               {flapping seagull};
\end{tikzpicture}
\end{codeexample}
    %
\end{pathoperation}

There are two general purpose keys that pics may find useful:

\begin{key}{/tikz/pic text=\meta{text}}
    This macro stores the \meta{text} in the macro |\tikzpictext|, which is
    |\let| to |\relax| by default. Setting the |pic text| to some value is the
    ``preferred'' way of communicating a (single) piece of text that should
    become part of a pic (typically of a node). In particular, the |quotes|
    library maps quoted parameters to this key.
\end{key}

\begin{key}{/tikz/pic text options=\meta{options}}
    This macro stores the \meta{options} in the macro |\tikzpictextoptions|,
    which is |\let| to the empty string by default. The |quotes| library maps
    options for quoted parameters to this key.
\end{key}


\subsubsection{The Quotes Syntax}
\label{section-pic-quotes}

When you load the |quotes| library, you can use the ``quotes syntax'' inside
the options of a pic. Recall that for nodes this syntax is used to add a label
to a node. For pics, the quotes syntax is used to set the |pic text| key.
Whether or not the pic type's code takes this key into consideration is,
however, up to the key.

In detail, when the |quotes| library is loaded, each time a key--value pair in
a list of options passed to an |pic| starts with |"|, the key--value pair must
actually be a string of the following form:
%
\begin{quote}
    |"|\meta{text}|"|\opt{|'|}\opt{\meta{options}}
\end{quote}
%
This string is transformed into the following:
%
\begin{quote}
    |every pic quotes/.try,pic text=|\meta{text}|, pic text options={|\meta{options}|}|
\end{quote}

As example of a pic type that takes these values into account is the |angle|
pic type:
%
\begin{codeexample}[]
\tikz \draw (3,0) coordinate (A)
         -- (0,1) coordinate (B)
         -- (1,2) coordinate (C)
            pic [draw, "$\alpha$"] {angle};
\end{codeexample}

As described in Section~\ref{section-label-quotes}, the apostrophe becomes part
of the \meta{options}, when present. As can be seen above, the following style
is executed:
%
\begin{stylekey}{/tikz/every pic quotes (initially \normalfont empty)}
\end{stylekey}


\subsection{Defining New Pic Types}
\label{section-new-pic-types}

As explained in the description of the |pic| command, in order to define a new
pic type you need to
%
\begin{enumerate}
    \item define a key with the path prefix |/tikz/pics| that
    \item sets the key |/tikz/pics/code| to the code of the pic.
\end{enumerate}

It turns out that this is easy enough to achieve using styles:

\begin{codeexample}[code only]
\tikzset{
  pics/seagull/.style ={
    % Ok, this is the key that should, when
    % executed, set the code key:
    code = { %
      \draw (...) ... ;
    }
  }
}
\end{codeexample}

Even though the above pattern is easy enough, there is a special so-called key
handler that allows us to write even simpler code, namely:
%
\begin{codeexample}[code only]
\tikzset{
  seagull/.pic = {
    \draw (...) ... ;
  }
}
\end{codeexample}

\begin{handler}{{.pic}|=|\meta{some code}}
    This handler can only be used with a key with the prefix |/tikz/|, so just
    should normally use it only as an option to a \tikzname\ command or to the
    |\tikzset| command. It takes the \meta{key}'s path and, inside that path,
    it replaces |/tikz/| by |/tikz/pics/| (so, basically, it adds the
    ``missing'' |pics| part of the path). Then, it sets up things so that the
    resulting name to key is a style that executes |code=some code|.
\end{handler}

In almost all cases, the |.pic| key handler will suffice to setup keys.
However, there are cases where you really need to use the first version using
|.style| and |code=|:
%
\begin{itemize}
    \item Whenever your pic type needs to set the foreground or the background
        code.
    \item In case of complicated arguments given to the keys.
\end{itemize}

As an example, let us define a simple pic that draws a filled circle behind the
path. Furthermore, we make the circle's radius a parameter of the pic:
%
\begin{codeexample}[]
\tikzset{
  pics/my circle/.style = {
    background code = { \fill circle [radius=#1]; }
  }
}
\tikz [fill=blue!30]
  \draw (0,0) pic {my circle=2mm} -- (1,1) pic {my circle=5mm};
\end{codeexample}