% Copyright 2019 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{Commands of the System Layer}

\makeatletter

\subsection{Beginning and Ending a Stream of System Commands}

A ``user'' of the \pgfname\ system layer (like the basic layer or a frontend)
will interface with the system layer by calling a stream of commands starting
with |\pgfsys@|. From the system layer's point of view, these commands form a
long stream. Between calls to the system layer, control goes back to the user.

The driver files implement system layer commands by inserting |\special|
commands that implement the desired operation. For example, |\pgfsys@stroke|
will be mapped to |\special{pdf: S}| by the driver file for |pdftex|.

For many drivers, when such a stream of specials starts, it is necessary to
install an appropriate transformation and perhaps perform some more
bureaucratic tasks. For this reason, every stream will start with a
|\pgfsys@beginpicture| and will end with a corresponding ending command.

\begin{command}{\pgfsys@beginpicture}
    Called at the beginning of a |{pgfpicture}|. This command should ``set up
    things''.

    Most drivers will need to implement this command.
\end{command}

\begin{command}{\pgfsys@endpicture}
  Called at the end of a |{pgfpicture}|.

  Most drivers will need to implement this command.
\end{command}

\begin{command}{\pgfsys@typesetpicturebox\marg{box}}
    Called \emph{after} a |{pgfpicture}| has been typeset. The picture will
    have been put in box \meta{box}. This command should insert the box into
    the normal text. The box \meta{box} will still be a ``raw'' box that
    contains only the |\special|'s that make up the description of the picture.
    The  job of this command is to resize and shift \meta{box} according to the
    baseline shift and the size of the box.

    This command has a default implementation and need not be implemented by a
    driver file.
\end{command}

\begin{command}{\pgfsys@beginpurepicture}
    This version of the |\pgfsys@beginpicture| picture command can be used for
    pictures that are guaranteed not to contain any escaped boxes (see below).
    In this case, a driver might provide a more compact version of the command.

    This command has a default implementation and need not be implemented by a
    driver file.
\end{command}

\begin{command}{\pgfsys@endpurepicture}
    Called at the end of a ``pure'' |{pgfpicture}|.

    This command has a default implementation and need not be implemented by a
    driver file.
\end{command}

Inside a stream it is sometimes necessary to ``escape'' back into normal
typesetting mode; for example to insert some normal text, but with all of the
current transformations and clippings being in force. For this escaping, the
following command is used:

\begin{command}{\pgfsys@hbox\marg{box number}}
    Called to insert a (horizontal) TeX box inside a |{pgfpicture}|.

    Most drivers will need to (re-)implement this command.
\end{command}

\begin{command}{\pgfsys@hboxsynced\marg{box number}}
    Called to insert a (horizontal) TeX box inside a |{pgfpicture}|, but with
    the current coordinate transformation matrix synced with the canvas
    transformation matrix.

    This command should do the same as if you used |\pgflowlevelsynccm|
    followed by |\pgfsys@hbox|. However, the default implementation of this
    command will use a ``TeX-translation'' for the translation part of the
    transformation matrix. This will ensure that hyperlinks ``survive'' at
    least translations. On the other hand, a driver may choose to revert to a
    simpler implementation. This is done, for example, for the \textsc{svg}
    implementation, where a \TeX-translation makes no sense.
\end{command}

\begin{command}{\pgfsys@pictureboxsynced\marg{box number}}
    Basically, this should do the same as doing a (scoped) low level sync
    followed by inserting the box \meta{box number} directly into the output
    stream. However, the default implementation uses |\pgfsys@hboxsynced| in
    conjunction with |\pgfsys@beginpicture| to ensure that, if possible,
    hyperlinks survive in \textsc{pdf}s. Drivers that are sensitive to
    picture-in-picture scopes should replace this implementation by
    %
\begin{codeexample}[code only]
\pgfsys@beginscope\pgflowlevelsynccm\box#1\pgfsys@endscope
\end{codeexample}
    %
\end{command}


\subsection{Scoping System Commands}

The scoping commands are used to keep changes of the graphics state local.

\begin{command}{\pgfsys@beginscope}
    Saves the current graphic state on a graphic state stack. All changes to
    the graphic state parameters mentioned for |\pgfsys@stroke| and
    |\pgfsys@fill| will be local to the current graphic state and the old
    values will be restored after |\pgfsys@endscope| is used.

    \emph{Warning:} \pdf\ and PostScript differ with respect to the question of
    whether the current path is part of the graphic state or not. For this
    reason, you should never use this command unless the path is currently
    empty. For example, it might be a good idea to use |\pgfsys@discardpath|
    prior to calling this command.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@endscope}
  Restores the last saved graphic state.

  This command is protocolled, see Section~\ref{section-protocols}.
\end{command}


\subsection{Path Construction System Commands}

\begin{command}{\pgfsys@moveto\marg{x}\marg{y}}
    This command is used to start a path at a specific point $(x,y)$ or to move
    the current point of the current path to  $(x,y)$ without drawing anything
    upon stroking (the current path is ``interrupted'').

    Both \meta{x} and \meta{y} are given as \TeX\ dimensions. It is the
    driver's job to transform these to the coordinate system of the backend.
    Typically, this means converting the \TeX\ dimension into a dimensionless
    multiple of $\frac{1}{72}\mathrm{in}$. The function |\pgf@sys@bp| helps
    with this conversion.

    \example Draw a line from $(10\mathrm{pt},10\mathrm{pt})$ to the origin of
    the picture.
    %
\begin{codeexample}[code only]
\pgfsys@moveto{10pt}{10pt}
\pgfsys@lineto{0pt}{0pt}
\pgfsys@stroke
\end{codeexample}

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@lineto\marg{x}\marg{y}}
    Continue the current path to $(x,y)$ with a straight line.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@curveto\marg{$x_1$}\marg{$y_1$}\marg{$x_2$}\marg{$y_2$}\marg{$x_3$}\marg{$y_3$}}
    Continue the current path to $(x_3,y_3)$ with a Bézier curve that has the
    two control points  $(x_1,y_1)$ and  $(x_2,y_2)$.

    \example Draw a good approximation of a quarter circle:
    %
\begin{codeexample}[code only]
\pgfsys@moveto{10pt}{0pt}
\pgfsys@curveto{10pt}{5.55pt}{5.55pt}{10pt}{0pt}{10pt}
\pgfsys@stroke
\end{codeexample}

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@rect\marg{x}\marg{y}\marg{width}\marg{height}}
    Append a rectangle to the current path whose lower left corner is at
    $(x,y)$ and whose width and height in big points are  given by \meta{width}
    and \meta{height}.

    This command can be ``mapped back'' to |\pgfsys@moveto| and
    |\pgfsys@lineto| commands, but it is included since \pdf\ has a special,
    quick version of this command.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@closepath}
    Close the current path. This results in joining the current point of the
    path with the point specified by the last |\pgfsys@moveto| operation.
    Typically, this is preferable over using |\pgfsys@lineto| to the last point
    specified by a |\pgfsys@moveto|, since the line starting at this point and
    the line ending at this point will be smoothly joined by
    |\pgfsys@closepath|.

    \example Consider
    %
\begin{codeexample}[code only]
\pgfsys@moveto{0pt}{0pt}
\pgfsys@lineto{10bp}{10bp}
\pgfsys@lineto{0bp}{10bp}
\pgfsys@closepath
\pgfsys@stroke
\end{codeexample}
    %
    and
    %
\begin{codeexample}[code only]
\pgfsys@moveto{0bp}{0bp}
\pgfsys@lineto{10bp}{10bp}
\pgfsys@lineto{0bp}{10bp}
\pgfsys@lineto{0bp}{0bp}
\pgfsys@stroke
\end{codeexample}

    The difference between the above will be that in the second triangle the
    corner at the origin will be wrong; it will just be the overlay of two
    lines going in different directions, not a sharp pointed corner.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}


\subsection{Canvas Transformation System Commands}

\begin{command}{\pgfsys@transformcm\marg{a}\marg{b}\marg{c}\marg{d}\marg{e}\marg{f}}
    Perform a concatenation of the canvas transformation matrix with the matrix
    given by the values \meta{a} to \meta{f}, see the \pdf\ or PostScript
    manual for details. The values \meta{a} to \meta{d} are dimensionless
    factors, \meta{e} and \meta{f} are \TeX\ dimensions

    \example |\pgfsys@transformcm{1}{0}{0}{1}{1cm}{1cm}|.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@transformshift\marg{x displacement}\marg{y displacement}}
    This command will change the origin of the canvas to $(x,y)$.

    This command has a default implementation and need not be implemented by a
    driver file.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@transformxyscale\marg{x scale}\marg{y scale}}
    This command will scale the canvas (and  everything that is drawn) by a
    factor of \meta{x scale} in the $x$-direction and \meta{y scale} in the
    $y$-direction. Note that this applies to everything, including  lines. So a
    scaled line will have a different width and may even have a different width
    when going along the $x$-axis and when going along the $y$-axis, if the
    scaling is different in these directions. Usually, you do not want this.

    This command has a default implementation and need not be implemented by a
    driver file.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@viewboxmeet\marg{$x_1$}\marg{$y_1$}\marg{$x_2$}\marg{$y_2$}%
    \marg{$x'_1$}\marg{$y'_1$}\marg{$x'_2$}\marg{$y'_2$}%
}
    Starts a ``view box'' scope, which must be ended using |\pgfsys@endviewbox|
    later on (with matching scopes).

    The effect of this command is as follows: Consider the rectangles $R$ with
    lower left corner $(x_1,y_1)$ and upper right corner $(x_2,y_2)$ and $R'$
    with corners $(x'_1,y'_1)$ and $(x'_2,y'_2)$. The command will install a
    canvas translation and uniform scaling such that $R'$ then has the same
    center as~$R$ and additionally, has maximum size such that it still fits
    inside~$R$. (Think of this as ``viewing'' $R'$ through $R$ such that the
    aspect ratio is kept.)

    This command has a default implementation. Its main purpose is to allow
    animations of the view box; for static drawings it is better to compute the
    necessary transformations directly.
\end{command}

\begin{command}{\pgfsys@viewboxslice\marg{$x_1$}\marg{$y_1$}\marg{$x_2$}\marg{$y_2$}%
    \marg{$x'_1$}\marg{$y'_1$}\marg{$x'_2$}\marg{$y'_2$}%
}
    Works like the previous command, but now $R'$ has minimal size such that it
    encompasses all of $R$.
\end{command}

\begin{command}{\pgfsys@endviewbox}
    Ends a viewbox previously started using |\pgfsys@viewboxmeet| or the
    |...slice| variant.
\end{command}


\subsection{Stroking, Filling, and Clipping System Commands}

\begin{command}{\pgfsys@stroke}
    Stroke the current path (as if it were drawn with a pen). A number of
    graphic state parameters influence this, which can be set using appropriate
    system commands described later.
    %
    \begin{description}
        \item[Line width] The ``thickness'' of the line. A width of 0 is the
            thinnest width renderable on the device. On a high-resolution
            printer this may become invisible and should be avoided. A good
            choice is 0.4pt, which is the default.
        \item[Stroke color] This special color is used for stroking. If it is
            not set, the current color is used.
        \item[Cap] The cap describes how the endings of lines are drawn. A
            round cap adds a little half circle to these endings. A butt cap
            ends the lines exactly at the end (or start) point without anything
            added. A rectangular cap ends the lines like the butt cap, but the
            lines protrude over the endpoint by the line thickness. (See also
            the \pdf\ manual.) If the path has been closed, no cap is drawn.
        \item[Join] This describes how a bend (a join) in a path is rendered. A
            round join draws bends using small arcs. A bevel join just draws
            the two lines and then fills the join minimally so that it becomes
            convex. A miter join extends the lines so that they form a single
            sharp corner, but only up to a certain miter limit. (See the \pdf\
            manual once more.)
        \item[Dash] The line may be dashed according to a dashing pattern.
        \item[Clipping area] If a clipping area is established, only those
            parts of the path that are inside the clipping area will be drawn.
    \end{description}

    In addition to stroking a path, the path may also be used for clipping
    after it has been stroked. This will happen if the |\pgfsys@clipnext| is
    used prior to this command, see there for details.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@closestroke}
    This command should have the same effect as first closing the path and then
    stroking it.

    This command has a default implementation and need not be implemented by a
    driver file.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}


\begin{command}{\pgfsys@fill}
    This command fills the area surrounded by the current path. If the path has
    not yet been closed, it is closed prior to filling. The path itself is not
    stroked. For self-intersecting paths or paths consisting of multiple parts,
    the nonzero winding number rule is used to determine whether a point is
    inside or outside the path, except if |\ifpgfsys@eorule| holds -- in which
    case the even-odd rule should be used. (See the \pdf\ or PostScript manual
    for details.)

    The following graphic state parameters influence the filling:
    %
    \begin{description}
        \item[Interior rule] If |\ifpgfsys@eorule| is set, the even-odd rule is
            used, otherwise the non-zero winding number rule.
        \item[Fill color] If the fill color is not especially set, the current
            color is used.
        \item[Clipping area] If a clipping area is established, only those
            parts of the filling area that are inside the clipping area will be
            drawn.
    \end{description}

    In addition to filling the path, the path will also be used for clipping if
    |\pgfsys@clipnext| is used prior to this command.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@fillstroke}
    First, the path is filled, then the path is stroked. If the fill and stroke
    colors are the same (or if they are not specified and the current color is
    used), this yields almost the same as a |\pgfsys@fill|. However, due to the
    line thickness of the stroked path, the fill-stroked area will be slightly
    larger.

    In addition to stroking and filling the path, the path will also be used
    for clipping if |\pgfsys@clipnext| is used prior to this command.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@discardpath}
    Normally, this command should ``throw away'' the current path. However,
    after |\pgfsys@clipnext| has been called, the current path should
    subsequently be used for clipping. See |\pgfsys@clipnext| for details.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@clipnext}
    This command should be issued after a path has been constructed, but before
    it has been stroked and/or filled or discarded. When the command is used,
    the next stroking/filling/discarding command will first be executed
    normally. Then, afterwards, the just-used path will be used for subsequent
    clipping. If there has already been a clipping region, this region is
    intersected with the new clipping path (the clipping cannot get bigger).
    The nonzero winding number rule is used to determine whether a point is
    inside or outside the clipping area or the even-odd rule, depending on
    whether |\ifpgfsys@eorule| holds.
\end{command}


\subsection{Graphic State Option System Commands}

\begin{command}{\pgfsys@setlinewidth\marg{width}}
    Sets the width of lines, when stroked, to \meta{width}, which must be a
    \TeX\ dimension.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@buttcap}
    Sets the cap to a butt cap. See |\pgfsys@stroke|.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@roundcap}
    Sets the cap to a round cap. See |\pgfsys@stroke|.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@rectcap}
    Sets the cap to a rectangular cap. See |\pgfsys@stroke|.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@miterjoin}
    Sets the join to a miter join. See |\pgfsys@stroke|.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@setmiterlimit\marg{factor}}
    Sets the miter limit of lines to \meta{factor}. See the \pdf\ or PostScript
    for details on what the miter limit is.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@roundjoin}
    Sets the join to a round join. See |\pgfsys@stroke|.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@beveljoin}
    Sets the join to a bevel join. See |\pgfsys@stroke|.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@setdash\marg{pattern}\marg{phase}}
    Sets the dashing patter. \meta{pattern} should be a list of \TeX\
    dimensions separated by commas. \meta{phase} should be a single dimension.

    \example |\pgfsys@setdash{3pt,3pt}{0pt}|

    The list of values in \meta{pattern} is used to determine the lengths of
    the ``on'' and ``off'' phases of the dashing. For example, if
    \meta{pattern} is |3bp,4bp|, then the dashing pattern is ``3bp on followed
    by 4bp off, followed by 3bp on, followed by 4bp off, and so on''. A pattern
    of |.5pt,4pt,3pt,1.5pt| means ``.5pt on, 4pt off, 3pt on, 1.5pt off, .5pt
    on, \dots'' If the number of entries is odd, the last one is used twice, so
    |3pt| means ``3pt on, 3pt off, 3pt on, 3pt off, \dots'' An empty list means
    ``always on''.

    The second argument determines the ``phase'' of the pattern. For example,
    for a pattern of |3bp,4bp| and a phase of |1bp|, the pattern would start:
    ``2bp on, 4bp off, 3bp on, 4bp off, 3bp on, 4bp off, \dots''

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

{\let\ifpgfsys@eorule=\relax
\begin{command}{\ifpgfsys@eorule}
    Determines whether the even odd rule is used for filling and clipping or
    not.
\end{command}
}


\subsection{Color System Commands}

The \pgfname\ system layer provides a number of system commands for setting
colors. These command coexist with commands from the |color| and |xcolor|
package, which perform similar functions. However, the |color| package does not
support having two different colors for stroking and filling, which is a useful
feature that is supported by \pgfname. For this reason, the \pgfname\ system
layer offers commands for setting these colors separately. Also, plain \TeX\
profits from the fact that \pgfname\ can set colors.

For \pdf, implementing these color commands is easy since \pdf\ supports
different stroking and filling colors directly. For PostScript, a more
complicated approach is needed in which the colors need to be stored in special
PostScript variables that are set whenever a stroking or a filling operation is
done.

\begin{command}{\pgfsys@color@rgb\marg{red}\marg{green}\marg{blue}}
    Sets the color used for stroking and filling operations to the given
    red/green/blue tuple (numbers between 0 and 1).

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@rgb@stroke\marg{red}\marg{green}\marg{blue}}
    Sets the color used for stroking operations to the given red/green/blue
    tuple (numbers between 0 and 1).

    \example Make stroked text dark red: |\pgfsys@color@rgb@stroke{0.5}{0}{0}|

    The special stroking color is only used if the stroking color has been set
    since the last |\color| or |\pgfsys@color@...| command. Thus, each |\color|
    command will reset both the stroking and filling colors by calling
    |\pgfsys@color@reset|.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@rgb@fill\marg{red}\marg{green}\marg{blue}}
    Sets the color used for filling operations to the given red/green/blue
    tuple (numbers between 0 and 1). This color may be different from the
    stroking color.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@cmyk\marg{cyan}\marg{magenta}\marg{yellow}\marg{black}}
    Sets the color used for stroking and filling operations to the given cmyk
    tuple (numbers between 0 and 1).

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@cmyk@stroke\marg{cyan}\marg{magenta}\marg{yellow}\marg{black}}
    Sets the color used for stroking operations to the given cmyk tuple
    (numbers between 0 and 1).

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@cmyk@fill\marg{cyan}\marg{magenta}\marg{yellow}\marg{black}}
    Sets the color used for filling operations to the given cmyk tuple (numbers
    between 0 and 1).

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@cmy\marg{cyan}\marg{magenta}\marg{yellow}}
    Sets the color used for stroking and filling operations to the given cmy
    tuple (numbers between 0 and 1).

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@cmy@stroke\marg{cyan}\marg{magenta}\marg{yellow}}
    Sets the color used for stroking operations to the given cmy tuple (numbers
    between 0 and 1).

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@cmy@fill\marg{cyan}\marg{magenta}\marg{yellow}}
    Sets the color used for filling operations to the given cmy tuple (numbers
    between 0 and 1).

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@gray\marg{black}}
    Sets the color used for stroking and filling operations to the given black
    value, where 0 means black and 1 means white.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@gray@stroke\marg{black}}
    Sets the color used for stroking operations to the given black value, where
    0 means black and 1 means white.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@gray@fill\marg{black}}
    Sets the color used for filling operations to the given black value, where
    0 means black and 1 means white.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@color@reset}
    This command will be called when the |\color| command is used. It should
    purge any internal settings of stroking and filling color. After this call,
    till the next use of a command like |\pgfsys@color@rgb@fill|, the current
    color installed by the |\color| command should be used.

    If the \TeX-if |\pgfsys@color@reset@inorder| is set to true, this command
    may ``assume'' that any call to a color command that sets the fill or
    stroke color came ``before'' the call to this command and may try to
    optimize the output accordingly.

    An example of an incorrect ``out of order'' call would be using
    |\pgfsys@color@reset| at the beginning of a box that is constructed using
    |\setbox|. Then, when the box is constructed, no special fill or stroke
    color might be in force. However, when the box is later on inserted at some
    point, a special fill color might already have been set. In this case, this
    command is not guaranteed to reset the color correctly.
\end{command}

\begin{command}{\pgfsys@color@reset@inordertrue}
    Sets the optimized ``in order'' version of the color resetting. This is the
    default.
\end{command}

\begin{command}{\pgfsys@color@reset@inorderfalse}
    Switches off the optimized color resetting.
\end{command}

\begin{command}{\pgfsys@color@unstacked\marg{\LaTeX\ color}}
    This slightly obscure command causes the color stack to be tricked. When
    called, this command should set the current color to \meta{\LaTeX\ color}
    without causing any change in the color stack.

    \example |\pgfsys@color@unstacked{red}|
\end{command}


\subsection{Pattern System Commands}

\begin{command}{\pgfsys@declarepattern
    \marg{name}\marg{$x_1$}\marg{$y_1$}\marg{$x_2$}\marg{$y_2$}%
    \marg{$x$ step}\marg{$y$ step}%
    \marg{$a$}\marg{$b$}\marg{$c$}\marg{$d$}\allowbreak\marg{$e$}\marg{$f$}%
    \marg{code}\marg{flag}%
}
    This command declares a new colored or uncolored pattern, depending on
    whether \meta{flag} is |0|, which means uncolored, or |1|, which means
    colored. Uncolored patterns have no inherent color, the color is provided
    when they are set. Colored patters have an inherent color.

    The \meta{name} is a name for later use when the pattern is to be shown.
    The pairs $(x_1,y_1)$ and $(x_2,y_2)$ must describe a bounding box of the
    pattern \meta{code}.

    The tiling step of the pattern is given by \meta{$x$ step} and \meta{$y$
    step}.

    The parameters \meta{$a$} to \meta{$f$} are entries of the transformation
    matrix that is applied to the pattern, see |\pgfsys@patternmatrix| for more
    information.

    \example
    %
\begin{codeexample}[code only]
\pgfsys@declarepattern
    {hori}{-.5pt}{0pt}{.5pt}{3pt}{3pt}{3pt}%
    {1.0}{0.0}{0.0}{1.0}{0.0pt}{0.0pt}%
    {\pgfsys@moveto{0pt}{0pt}\pgfsys@lineto{0pt}{3pt}\pgfsys@stroke}
    {0}
\end{codeexample}
    %
\end{command}


\begin{command}{\pgfsys@patternmatrix}
    For convenience \pgfname\ defines the transformation matrix that is applied
    to all patterns defined with |\pgfdeclarepatternformonly| and
    |\pgfdeclarepatterninherentlycolored| in a macro.  This can be used as an
    extension point for ad-hoc transformation of existing patterns.  The
    default definition is the
    identity matrix:
    %
\begin{codeexample}[code only]
\def\pgfsys@patternmatrix{{1.0}{0.0}{0.0}{1.0}{0.0pt}{0.0pt}}
\end{codeexample}
    %
    The entries of the enclosed array
    |{|\meta{$a$}|}{|\meta{$b$}|}{|\meta{$c$}|}{|\meta{$d$}|}{|\meta{$e$}|}{|\meta{$f$}|}|
    are entries in the transformation matrix, identified as in the following
    transformation prescription:
    \begin{equation*}
        \begin{pmatrix}
            x' \\
            y' \\
            1 \\
        \end{pmatrix}
        =
        \begin{pmatrix}
            a & c & e \\
            b & d & f \\
            0 & 0 & 1 \\
        \end{pmatrix}
        \begin{pmatrix}
            x \\
            y \\
            1 \\
        \end{pmatrix} .
    \end{equation*}
    Carrying out the matrix multiplication results in the following system of
    equations
    \begin{align*}
        x' &= a x + c y + e , \\
        y' &= b x + d y + f .
    \end{align*}
    Evidently, the parameters \marg{$a$} to \marg{$d$} have to be dimensionless
    because they are scaling factors, but the parameters \marg{$e$} and
    \marg{$f$} are offsets, therefore they have to carry a unit.
\end{command}

\begin{command}{\pgfsys@setpatternuncolored\marg{name}\marg{red}\marg{green}\marg{blue}}
    Sets the fill color to the pattern named \meta{name}. This pattern must
    previously have been declared with \meta{flag} set to |0|. The color of the
    pattern is given in the parameters \meta{red}, \meta{green}, and
    \meta{blue} in the usual way.

    The fill color ``pattern'' will persist till the next color command that
    modifies the fill color.
\end{command}

\begin{command}{\pgfsys@setpatterncolored\marg{name}}
    Sets the fill color to the pattern named \meta{name}. This pattern must
    have been declared with the |1| flag.
\end{command}


\subsection{Image System Commands}

The system layer provides some commands for image inclusion.

\begin{command}{\pgfsys@imagesuffixlist}
    This macro should expand to a list of suffixes, separated by `:', that will
    be tried when searching for an image.

    \example |\def\pgfsys@imagesuffixlist{eps:epsi:ps}|
\end{command}

\begin{command}{\pgfsys@defineimage}
    Called, when an image should be defined.

    This command does not take any parameters. Instead, certain macros will be
    preinstalled with appropriate values when this command is invoked. These
    are:
    %
    \begin{itemize}
        \item\declare{|\pgf@filename|} File name of the image to be defined.
        \item\declare{|\pgf@imagewidth|} Will be set to the desired (scaled)
            width of the image.
        \item\declare{|\pgf@imageheight|} Will be set to the desired (scaled)
            height of the image.

            If this macro and also the height macro are empty, the image should
            have its ``natural'' size.

            If only one of them is specified, the undefined value the image is
            scaled so that the aspect ratio is kept.

            If both are set, the image is scaled in both directions
            independently, possibly changing the aspect ratio.
    \end{itemize}

    The following macros presumable mostly make sense for drivers that
    can handle \pdf:
    %
    \begin{itemize}
        \item \declare{|\pgf@imagepage|} The desired page number to be
            extracted from a multi-page ``image''.
        \item\declare{|\pgf@imagemask|} If set, it will be set to
            |/SMask x 0 R| where |x| is the \pdf\ object number of a soft mask
            to be applied to the image.
        \item\declare{|\pgf@imageinterpolate|} If set, it will be set to
            |/Interpolate true| or |/Interpolate false|, indicating whether the
            image should be interpolated in \pdf.
    \end{itemize}

    The command should now set up the macro |\pgf@image| such that calling this
    macro will result in typesetting the image. Thus, |\pgf@image| is the
    ``return value'' of the command.

    This command has a default implementation and need not be implemented by a
    driver file.
\end{command}


\subsection{Shading System Commands}

\begin{command}{\pgfsys@horishading\marg{name}\marg{height}\marg{specification}}
    Declares a horizontal shading for later use. The effect of this command
    should be the definition of a macro called |\@pgfshading|\meta{name}|!| (or
    |\csname @pdfshading|\meta{name}|!\endcsname|, to be precise). When
    invoked, this new macro should insert a shading at the current position.

    \meta{name} is the name of the shading, which is also used in the output
    macro name. \meta{height} is the height of the shading and must be given as
    a TeX dimension like |2cm| or |10pt|. \meta{specification} is a shading
    color specification as specified in Section~\ref{section-shadings}. The
    shading specification implicitly fixes the width of the shading.

    When |\@pgfshading|\meta{name}|!| is invoked, it should insert a box of
    height \meta{height} and the width implicit in the shading declaration.
\end{command}

\begin{command}{\pgfsys@vertshading\marg{name}\marg{width}\marg{specification}}
    Like the horizontal version, only for vertical shadings. This time, the
    height of the shading is implicit in \meta{specification} and the width is
    given as \meta{width}.
\end{command}

\begin{command}{\pgfsys@radialshading\marg{name}\marg{starting point}\marg{specification}}
    Declares a radial shading. Like the previous macros, this command should
    set up the macro |\@pgfshading|\meta{name}|!|, which upon invocation should
    insert a radial shading whose size is implicit in \meta{specification}.

    The parameter \meta{starting point} is a \pgfname\ point specifying the
    inner starting point of the shading.
\end{command}

\begin{command}{\pgfsys@functionalshading\marg{name}\marg{lower left corner}\meta{upper right corner}\marg{type 4 function}}
    Declares a shading using a PostScript-like function that provides a color
    for each point. Like the previous macros, this command should set up the
    macro |\@pgfshading|\meta{name}|!| so that it will produce a box containing
    the desired shading.

    Parameter \meta{name} is the name of the shading. Parameter \meta{type 4
    function} is a Postscript-like function (type 4 function of the PDF
    specification) as described in Section~3.9.4 of the PDF specification
    version 1.7. Parameters \meta{lower left corner} and \meta{upper right
    corner} are \pgfname\ points that specifies the lower left and upper right
    corners of the shading, respectively.

    When \meta{type 4 function} is evaluated, the coordinate of the current
    point will be on the (virtual) PostScript stack in bp units. After the
    function has been evaluated, the stack should consist of three numbers (not
    integers! -- the Apple PDF renderer is broken in this regard, so add cvrs
    at the end if needed) that represent the red, green, and blue components of
    the color.

    A buggy function will result is \emph{totally unpredictable chaos} during
    rendering.
\end{command}


\subsection{Transparency System Commands}

\begin{command}{\pgfsys@opacity\marg{value}}
    Sets the opacity of all operations, treating stroking and filling as a
    transparency group. Some drivers support this operations, others do not and
    set the fill and stroke individually. This difference can only be seen when
    a path is stroked and filled at the same time: When the drawing and fill
    opacities are set individually, the effect of filling and drawing a path at
    the same time is the same as first filling the path and then drawing it. On
    the other, if the opacity is set using this command, the effect should
    rather be that same as first filling and then drawing the path without any
    opacity in an off-screen area and then copying the result to the target
    area with a homogeneous opacity of \meta{value}.

    Since \textsc{pdf} does not support this form of opacity, this command is
    only present on the system layer and not supported in the basic layer.
\end{command}

\begin{command}{\pgfsys@stroke@opacity\marg{value}}
    Sets the opacity of stroking operations.
\end{command}

\begin{command}{\pgfsys@fill@opacity\marg{value}}
    Sets the opacity of filling operations.
\end{command}

\begin{command}{\pgfsys@blend@mode\marg{value}}
    Sets the blend mode, see Section~7.2.4 of the \textsc{pdf} Specification,
    Version~1.7.
\end{command}

\begin{command}{\pgfsys@transparencygroupfrombox\marg{box}}
    This takes a \TeX\ box and converts it into a transparency group. This
    means that any transparency settings apply to the box as a whole. For
    instance, if a box contains two overlapping black circles and you draw the
    box and, thus, the two circles normally with 50\% transparency, then the
    overlap will be darker than the rest. By comparison, if the circles are
    part of a transparency group, the overlap will get the same color as the
    rest.
\end{command}

A transparency group can be \emph{isolated} and/or a \emph{knockout} group (see
Sections~7.3.4 and 7.3.5 of the \textsc{pdf} Specification Version~1.7). Which
of these is the case is dictated by the current settings of the following two
ifs, which must be set before the above command is called:

{\let\ifpgfsys@transparency@group@isolated=\relax
\begin{command}{\ifpgfsys@transparency@group@isolated}
    Determines whether a transparency group should be isolated.
\end{command}
}

{\let\ifpgfsys@transparency@group@knockout=\relax
\begin{command}{\ifpgfsys@transparency@group@knockout}
    Determines whether a transparency group is a knockout group or not.
\end{command}
}

\begin{command}{\pgfsys@fadingfrombox\marg{name}\marg{box}}
    Declares the fading \meta{name}. The \meta{box} is a \TeX-box. Its
    content's luminosity determines the opacity of the resulting fading. This
    means that the lighter a pixel inside the box, the more opaque the fading
    will be at this position.
\end{command}

\begin{command}{\pgfsys@usefading\meta{name}\marg{a}\marg{b}\marg{c}\marg{d}\marg{e}\marg{f}}
    Installs a previously declared fading \meta{name} in the current graphics
    state. Afterwards, all drawings will be masked by the fading. The fading
    should be centered on the origin and have its original size, except that
    the parameters \meta{a} to \meta{f} specify a transformation matrix that
    should be applied additionally to the fading before it is installed. The
    transformation should not apply to the following graphics, however.
\end{command}

\begin{command}{\pgfsys@clipfading}
    This command has a default implementation and need not be implemented by
    driver files other than |pgfsys-dvips.def|. The macro is called
    in |\pgfsetfadingforcurrentpath| and |\pgfsetfadingforcurrentpathstroked|
    of the basic layer, where it invokes the current path for clipping the
    shading just before installing it as an opacity mask for fading. The
    default implementation is actually a non-operation, but with |dvips| it
    is used to clip the fading as described.
\end{command}

\begin{command}{\pgfsys@definemask}
    This command declares a fading (known as a soft mask in this context) based
    on an image and for usage with images. It works similar to
    |\pgfsys@defineimage|: Certain macros are set when the command is called.
    The result should be to set the macro |\pgf@mask| to a pdf object count
    that can subsequently be used as a transparency mask. The following macros
    will be set when this command is invoked:
    %
    \begin{itemize}
        \item \declare{|\pgf@filename|}
            File name of the mask to be defined.
        \item \declare{|\pgf@maskmatte|}
            The so-called matte of the mask (see the \pdf\ documentation for
            details). The matte is a color specification consisting of 1, 3 or
            4 numbers between 0 and 1. The number of numbers depends on the
            number of color channels in the image (not in the mask!). It will
            be assumed that the image has been preblended with this color.
    \end{itemize}
\end{command}


\subsection{Animation Commands}

The animation system layer command (|\pgfsys@anim...|) are described in a
separate section, Section~\ref{section-pgfsys-anim}.


\subsection{Object Identification System Commands}
\label{section-sys-id}

The system layer provides commands for adding identification labels (ids) to
different objects in a graphic. These can be used for hyperlinking, which is
needed for instance in conjunction with animations.

The following ``objects'' can get an id assigned to them:
%
\begin{enumerate}
    \item Graphic scopes (namely when |\pgfsys@begin@idscope| is called),
    \item view boxes (namely when |\pgfsys@viewboxmeet| or
        |\pgfsys@viewboxslice| are called),
    \item paths (namely when |\pgfsys@fill|, |\pgfsys@stroke|, and so on are
        called),
    \item text boxes (namely when |\pgfsys@hbox| or |\pgfsys@hboxsynced| is
        called), and
    \item animations (namely when |\pgfsys@animate| is called).
\end{enumerate}

Creating and using ids is a two-step process. First, you create the id using
|\pgfsys@new@id|, which stores a fresh id in a macro. You can now pass this id
around and clone it. Then, at some point, you wish one of the above objects to
actually get this id. For this, you use |\pgfsys@use@id| just \emph{before} the
object since this command always influences the \emph{next} object.

The basic id management gets more powerful when you use \emph{id types}. The
idea is as follows: In reality, the objects from above do not get assigned only
an id, but rather a combination of an id and a type -- and you can set the type
independently of the id. This is used, for instance, to allow easy access to
the different parts of a node in animations: Each node has a single id, but
consists of several graphic objects (normally, at least a background path and a
text). Each of these uses the same underlying id of the node, but the path has
the type |path| (actually |background.path|) while the text has the type
|text|. The advantage is that for each node only one id must be stored instead
of a great number of the many different possible parts of a node.

\begin{command}{\pgfsys@new@id\marg{macro}}
    Creates a new id for later use and stores it in \meta{macro}. It is an
    internal text created by the driver and may not be changed or modified.
\end{command}

\begin{command}{\pgfsys@use@id\marg{id}}
    ``Uses'' an id previously created using |\pgfsys@new@id|. This causes the
    \emph{next} graphic object to get the \meta{id} (not the current one). Once
    used, the id-type-pair becomes \emph{invalid} and will not be attached to
    any other graphics objects. It is, however, not an error to try this. If
    \meta{id} is empty, no id-type-pair is attached to the next object.
\end{command}

\begin{command}{\pgfsys@use@type\marg{type}}
    Changes the type used with the next graphic object. As mentioned earlier,
    the id assigned to the next object is actually a pair consisting of the
    currently used id and the currently used type.
\end{command}

\begin{command}{\pgfsys@append@type\marg{text}}
    Appends the \meta{text} to the current type.
\end{command}

\begin{command}{\pgfsys@push@type}
    Pushes the current type on a global ``stack of types'' without opening a
    \TeX\ scope. The is useful when you temporarily wish to change the type
    (for instance, by appending something to it), but you cannot create a new
    scope.
\end{command}

\begin{command}{\pgfsys@pop@type}
    Restores the most recently pushed type.
\end{command}

\begin{command}{\pgfsys@begin@idscope}
    Starts a (graphics) scope whose sole purpose is to assign it an
    id-type-pair so that it can be referenced later. Note that this command
    does not always produce a graphics scope: If no id is currently in use or
    if the id-type-pair has already been used, a graphic scope may or may not
    be created as defined by the driver (but always a \TeX\ scope). This allows
    drivers to minimize the number of graphic scopes created.

    When an id scope is created, any code that has been ``attached'' to it
    using |\pgfsys@attach@to@id| gets executed, see that command.

    Note that |\pgfsys@beginscope| does not use the current id-type-pair. You
    need to call this command to attach an id to a group.
\end{command}

\begin{command}{\pgfsys@end@idscope}
    Ends the graphics id scope started by |\pgfsys@end@idscope|. It must nest
    correctly with other graphic scopes and \TeX\ scopes.
\end{command}

\begin{command}{\pgfsys@attach@to@id\marg{id}\marg{type}\marg{begin code}\marg{end code}\marg{setup code}}
    Attaches codes to the \meta{id}-\meta{type}-pair, where \meta{id} must have
    been created using |\pgfsys@new@id|. The effect is that just before the id
    scope for this pair is created, the \meta{setup code} is executed, then the
    scope is started, then the \meta{begin code} is executed at the beginning,
    and, finally, \meta{end code} gets executed just before the scope ends.
    Multiple calls of this macro accumulated.
\end{command}


\subsection{Resource Description Framework Annotations (RDFa)}
\label{section-sys-rdf}

With certain output formats (in particular, with \textsc{svg}) you can insert
annotations into the output file following the standard set by the
\emph{resource description framework} (\textsc{rdf}), please consult the
literature on \textsc{rdf} for an introduction to resource descriptions and
ontologies.

The support for \textsc{rdf} annotations works as follows in \pgfname: You use
the following commands before you create an id scope (using
|\pgfsys@begin@idscope|). Then the attributes set by the commands will be added
as an annotation to that object. Here is an example:
%
\begin{codeexample}[code only]
\pgfsys@rdf@resource{/fruits/apple}
\pgfsys@begin@idscope
  ...
\pgfsys@end@idscope
\end{codeexample}

If \textsc{svg} output is produced, this results in the following code in the
\textsc{svg} file:
%
\begin{codeexample}[code only]
<g resource="/fruits/apple">
  ...
</g>
\end{codeexample}

Note that a call to |\pgfsys@begin@idscope| adds all the set attributes, but
then clears the settings (globally). Thus, you should set all attributes more
or less right before the id scope is created. For most of these command, if you
call them multiple times before starting the id scope, the ``last call wins'',
that is, later values overwrite earlier ones. In contrast, for the commands
|\pgfsys@rdf@property|, |\pgfsys@rdf@rel|, |\pgfsys@rdf@rev|, as well as
|\pgfsys@rdf@typeof|, the calls accumulate, that is, the texts passed in each
call will all be added to the output, properly separated to form a list of
values. Consider for instance:
%
\begin{codeexample}[code only]
\pgfsys@rdf@resource{/fruits/apple}
\pgfsys@rdf@resource{/fruits/watermelon}
\pgfsys@rdf@property{http://foo.com/props/juicy}
\pgfsys@rdf@property{http://foo.com/props/green}
\pgfsys@begin@idscope
  ...
\pgfsys@end@idscope
\end{codeexample}

In the resulting id scope, we will have:
%
\begin{codeexample}[code only]
<g resource="/fruits/watermelon"
   property="http://foo.com/props/juicy http://foo.com/props/green">
  ...
</g>
\end{codeexample}

\begin{command}{\pgfsys@rdf@about\marg{text}}
    Adds the \textsc{rdf} attribute |about="|\meta{text}|"| to the next id
    scope (please see the \textsc{rdf}a specification for details on the
    semantics of |about| in the context of the resource description framework).
\end{command}

The following commands work the same way as the above command, except that the
set attribute is different. Please see the \textsc{rdf}a specification for
details on these attributes. Note that the |\pgfsys@rdf@inlist| command is the
only one that takes no argument.

\begin{command}{\pgfsys@rdf@content\marg{text}}
\end{command}

\begin{command}{\pgfsys@rdf@datatype\marg{text}}
\end{command}

\begin{command}{\pgfsys@rdf@href\marg{text}}
\end{command}

\begin{command}{\pgfsys@rdf@inlist}
\end{command}

\begin{command}{\pgfsys@rdf@prefix\marg{text}}
\end{command}

\begin{command}{\pgfsys@rdf@property\marg{text}}
\end{command}

\begin{command}{\pgfsys@rdf@rel\marg{text}}
\end{command}

\begin{command}{\pgfsys@rdf@resource\marg{text}}
\end{command}

\begin{command}{\pgfsys@rdf@rev\marg{text}}
\end{command}

\begin{command}{\pgfsys@rdf@src\marg{text}}
\end{command}

\begin{command}{\pgfsys@rdf@typeof\marg{text}}
\end{command}

\begin{command}{\pgfsys@rdf@vocab\marg{text}}
\end{command}


\subsection{Reusable Objects System Commands}

\begin{command}{\pgfsys@invoke\marg{literals}}
    This command gets protocolled literals and should insert them into the
    |.pdf| or |.dvi| file using an appropriate |\special|.
\end{command}

\begin{command}{\pgfsys@defobject\marg{name}\marg{lower left}\marg{upper right}\marg{code}}
    Declares an object for later use. The idea is that the object can be
    precached in some way and then be rendered more quickly when used several
    times. For example, an arrow head might be defined and prerendered in this
    way.

    The parameter \meta{name} is the name for later use. \meta{lower left} and
    \meta{upper right} are \pgfname\ points specifying a bounding box for the
    object. \meta{code} is the code for the object. The code should not be too
    fancy.

    This command has a default implementation and need not be implemented by a
    driver file.
\end{command}

\begin{command}{\pgfsys@useobject\marg{name}\marg{extra code}}
    Renders a previously declared object. The first parameter is the name of
    the object. The second parameter is extra code that should be executed
    right \emph{before} the object is rendered. Typically, this will be some
    transformation code.

    This command has a default implementation and need not be implemented by a
    driver file.
\end{command}

\begin{command}{\pgfsys@marker@declare\marg{macro}\marg{code}}
    Declares a \emph{marker} symbol for later use. The command is very similar
    to |\pgfsys@defobject|, but the use case is slightly different: The graphic
    object defined using the \meta{code} is stored in such a way that it can be
    used as an \emph{arrow tip marker symbol} in animations. The \meta{macro}
    is set to an identifier by which the marker can be referenced later on.

    This command has a default implementation and need not be implemented by a
    driver file.
\end{command}

\begin{command}{\pgfsys@marker@use\marg{macro}}
    Adds the marker object referenced by the \meta{macro} to the current
    output.

    This command has a default implementation and need not be implemented by a
    driver file.
\end{command}


\subsection{Invisibility System Commands}

All drawing or stroking or text rendering between calls of the following
commands should be suppressed. A similar effect can be achieved by clipping
against an empty region, but the following commands do not open a graphics
scope and can be opened and closed ``orthogonally'' to other scopes.

\begin{command}{\pgfsys@begininvisible}
    Between this command and the closing |\pgfsys@endinvisible| all output
    should be suppressed. Nothing should be drawn at all, which includes all
    paths, images and shadings. However, no groups (neither \TeX\ groups nor
    graphic state groups) should be opened by this command.

    This command has a default implementation and need not be implemented by a
    driver file.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}

\begin{command}{\pgfsys@endinvisible}
    Ends the invisibility section, unless invisibility blocks have been nested.
    In this case, only the ``last'' one restores visibility.

    This command has a default implementation and need not be implemented by a
    driver file.

    This command is protocolled, see Section~\ref{section-protocols}.
\end{command}


\subsection{Page Size Commands}

The following commands can be used to set the page size of a document in a
``portable'' way. Note, however, that many packages also (try to) set the page
size.

These commands are typically not given inside a |{pgfpicture}|, but on the
outer level of compilation.

\begin{command}{\pgfsys@papersize\marg{width}\marg{height}}
    Inserts the necessary |\special|s for the current driver into the output
    stream to ``locally'' change the page size. Whether such a ``local'' change
    is possible depends strongly on the driver. For instance, |dvips| will
    honor the first call to this command that is part of the shipped-out
    document and will ignore all other uses. In contrast, |pdftex| will use the
    current value of the paper size for each page and, additionally, setting
    the papersize is local to the current \TeX\ group.
\end{command}

\begin{command}{\pgfsys@global@papersize\marg{width}\marg{height}}
    Like the previous command, only for drivers where setting the paper size
    parameters is a \TeX-group-local operation, |\global| is prefixed to the
    setting of the page sizes.
\end{command}

\begin{command}{\pgfsys@thepageheight}
    This macro expands to the current page's height, provided \LaTeX\ is used,
    otherwise a best guess is returned (currently just |\the\vsize|).
\end{command}

\begin{command}{\pgfsys@thepagewidth}
    As above.
\end{command}


\subsection{Position Tracking Commands}

The following commands are used to determine the position of text on a page.
This is a rather complicated process in general since at the moment when the
text is read by \TeX, the final position cannot be determined, yet. For
example, the text might be put in a box which is later put in the headline or
perhaps in the footline or perhaps even on a different page.

For these reasons, position tracking is typically a two-stage process. In a
first stage you indicate that a certain position is of interest by
\emph{marking} it. This will (depending on the details of the backend driver)
cause page coordinates or this position to be written to an |.aux| file when
the page is shipped. Possibly, the position might also be determined at an even
later stage. Then, on a second run of \TeX, the position is read from the
|.aux| file and can be used.

\begin{command}{\pgfsys@markposition\marg{name}}
    Marks a position on the page. This command should be given while normal
    typesetting is done such as in
    %
\begin{codeexample}[code only]
The value of $x$ is \pgfsys@markposition{here}important.
\end{codeexample}
    %
    It causes the position |here| to be saved when the page is shipped out.
\end{command}

\begin{command}{\pgfsys@getposition\marg{name}\marg{macro}}
    This command retrieves a position that has been marked on an earlier run of
    \TeX\ on the current file. The \meta{macro} must be a macro name such as
    |\mymacro|. It will be redefined such that it is
    %
    \begin{itemize}
        \item either just |\relax| or
        \item a |\pgfpoint...| command.
    \end{itemize}
    %
    The first case will happen when the position has not been marked at all or
    when the file is typeset for the first time, when the coordinates are not
    yet available.

    In the second case, executing \meta{macro} yields the position on the page
    that is to be interpreted as follows: A coordinate like
    |\pgfpoint{2cm}{3cm}| means ``2cm to the right and 3cm up from the origin
    of the page''. The position of the origin of the page is not guaranteed to
    be at the lower left corner, it is only guaranteed that all pictures on a
    page use the same origin.

    To determine the lower left corner of a page, you can call
    |\pgfsys@getposition| with \meta{name} set to the special name
    |pgfpageorigin|. By shifting all positions by the amount returned by this
    call you can position things absolutely on a page.

    \example Referencing a point of the page:
    %
\begin{codeexample}[code only]
The value of $x$ is \pgfsys@markposition{here}important.

Lots of text.

\hbox{\pgfsys@markposition{myorigin}%
\begin{pgfpicture}
  % Switch of size protocol
  \pgfpathmoveto{\pgfpointorigin}
  \pgfusepath{use as bounding box}

  \pgfsys@getposition{here}{\hereposition}
  \pgfsys@getposition{myorigin}{\thispictureposition}

  \pgftransformshift{\pgfpointscale{-1}{\thispictureposition}}
  \pgftransformshift{\hereposition}

  \pgfpathcircle{\pgfpointorigin}{1cm}
  \pgfusepath{draw}
\end{pgfpicture}}
\end{codeexample}
    %
\end{command}


\subsection{Internal Conversion Commands}

The system commands take \TeX\ dimensions as input, but the dimensions that
have to be inserted into \pdf\ and PostScript files need to be dimensionless
values that are interpreted as multiples of $\frac{1}{72}\mathrm{in}$. For
example, the \TeX\ dimension $2bp$ should be inserted as |2| into a \pdf\ file
and the \TeX\ dimension $10\mathrm{pt}$ as |9.9626401|. To make this conversion
easier, the following command may be useful:

\begin{command}{\pgf@sys@bp\marg{dimension}}
    Inserts how many multiples of $\frac{1}{72}\mathrm{in}$ the
    \meta{dimension} is into the current protocol stream (buffered).

    \example |\pgf@sys@bp{\pgf@x}| or |\pgf@sys@bp{1cm}|.
\end{command}

Note that this command is \emph{not} a system command that can/needs to be
overwritten by a driver.


%%% Local Variables:
%%% mode: latex
%%% TeX-master: "pgfmanual"
%%% End: