@Section
    @Tag { dia_defi }
    @Title { Expert usage:  defining new shapes }
@Begin
@PP
@@Diag permits you to create your own node outlines and link paths, by
diagrams. @RawIndex { diagrams }
diagrams.definitions @SubIndex { definitions }
definitions. @RawIndex { definitions }
definitions.use.with.diagrams @SubIndex { use with diagrams }
giving non-standard values to the @Code outline and @Code path
options.  This section shows how to do this for very simple shapes
only; the following section introduces the large repertoire of geometrical
symbols that @@Diag offers for helping you create complex shapes.
@PP
As explained earlier, a node outline is drawn over its {@I base}, which
is a rectangle containing the following object plus margins.  The base
defines a coordinate system with the point (0, 0) at the bottom left
corner, and @M { (xsize, ysize) } at the top right:
@CD @OneRow @Diag {
@Box
    nodelabelmargin { 0.3f }
    blabel { @M { ysize } }
    blabelprox { E }
    clabel { @M { 0 } }
    clabelprox { E }
    dlabel { @M { xsize } }
    dlabelprox { N }
    alabel { @M { 0 } }
    alabelpos { SW }
    alabelprox { N }
    paint { lightgrey }
    outlinestyle { noline }
    margin { 0c }
{ 3c @Wide 2c @High }
//0.5c
}
The value of the @Code outline option is a sequence of points defined in
this coordinate system:
@ID {
@Code @Verbatim {
@Node
    outline {
        0 0
        xsize 0
        0 ysize
        0 0
    }
}
||7ct
@Diag {
@Box
    margin { 0c }
    outlinestyle { noline }
    paint { lightgrey }
@Node
    outline {
        0 0
        xsize 0
        0 ysize
        0 0
    }
    margin { 0c }
{ 3c @Wide 2c @High }
}
}
As shown, the resulting outline is created by joining each point to the
next with a straight line.  It is conventional to proceed anticlockwise
around the outline, but you may start anywhere.
@PP
The {@Code paint}, {@Code texture}, {@Code outlinestyle},
{@Code outlinedashlength}, and {@Code outlinewidth} options of
@Code "@Node" work for user-defined outlines exactly as they do
for the standard ones:
@ID {
@Code @Verbatim {
@Node
    outline {
        0 0
        xsize 0
        0 ysize
        0 0
    }
    paint { lightgrey }
    outlinestyle { solid dashed }
}
||7ct
@Diag {
@Node
    outline {
        0 0
        xsize 0
        0 ysize
        0 0
    }
    paint { lightgrey }
    outlinestyle { solid dashed }
    margin { 0c }
{ 3c @Wide 2c @High }
}
}
Each line in the outline is one segment for {@Code outlinestyle}.
@PP
If two points in an outline are separated by {@Code "[]"}, no line is
drawn between them, and the outline is treated as two separate,
disconnected regions when painting.
@PP
Two points may also be separated by {@Code "["}{@I point}{@Code "]"},
where @I point stands for any point.  This causes the two points to be
joined by an arc whose centre is at the given point:
@ID {
@Code @Verbatim {
@Node
    outline {
        0 0
        ysize 0
        [ 0 0 ]
        0 ysize
        0 0
    }
}
||7ct
@Diag {
@Box
    margin { 0c }
    outlinestyle { noline }
    paint { lightgrey }
@Node
    outline {
        0 0
        ysize 0
        [ 0 0 ]
        0 ysize
        0 0
    }
    margin { 0c }
{ 3c @Wide 2c @High }
}
}
The arc will be circular if possible, otherwise it will be part of
elliptical. @Index { elliptical arcs }
an ellipse whose axes are oriented horizontally and vertically.  The
arc goes anticlockwise; to get a clockwise arc, use
{@Code "["}{@I point}{@Code " clockwise]"}.
@PP
Two points may be separated by
@M { [x sub 1 ``` y sub 1 ```  x sub 2 ``` y sub 2 ] }, which requests
that a Bezier curve be drawn between them with control points
bezier.curve @Index { Bezier curve }
@M { (x sub 1 , y sub 1 ) } and
@M { (x sub 2 , y sub 2 ) }:
@CD @Diag {
@Node
    outline {
        A:: { xsize*0.2 ysize*0.5 }
        B:: { xsize*0.4 ysize*0.9 }
        C:: { xsize*0.9 ysize*0.4 }
        D:: { xsize*0.3 ysize*0.1 }
        A B C D A
    }
    alabelpos { A }
    blabelpos { B }
    clabelpos { C }
    dlabelpos { D }
    alabelprox { SE }
    blabelprox { SW }
    clabelprox { SW }
    dlabelprox { NW }
    outlinestyle { cdashed cdashed cdashed noline }
    alabel { @M { ( x sub 0 , y sub 0 ) } }
    blabel { @M { ( x sub 1 , y sub 1 ) } }
    clabel { @M { ( x sub 2 , y sub 2 ) } }
    dlabel { @M { ( x sub 3 , y sub 3 ) } }
{ 6c @Wide 2c @High }
//
@Link
    path { A [B C] D }
}
The curve is attracted toward the control points, without reaching
them; it is tangent to the straight line from the start point to the
first control point, and from the second control point to the finishing
point, and it lies wholly inside the quadrilateral formed by the four
points.  Owing to the author's laziness, dashes and dots do not fit as
neatly onto Bezier curves as they do onto lines and arcs.
@PP
Tags (Section {@NumberOf dia_tags}) may be assigned to points within
the outline option, like this:
@ID {
@Code @Verbatim {
@Node
    outline {
        LR:: { xsize 0 }
        UL:: { 0 ysize }
        0 0  LR  UL  0 0
    }
}
||7ct
@Diag {
//0.5f
@ShowTags @Node
    outline {
        LR:: { xsize 0 }
        UL:: { 0 ysize }
        0 0  LR  UL  0 0
    }
    { 2c @High 3c @Wide }
}
}
The tagged point does not have to lie on the outline, and it
is not automatically added to the outline.  Once defined, a
tag stands for a point in the usual way; it may be used later in the
outline, as was done above, relabelled, and so on, exactly like the tags
of the standard nodes.
@PP
Once a point has been tagged, a @I direction may be associated
with it, to inform @@Diag which way the outline or
link path is going at that point.  The standard outlines have directions:
@ID {
@Code {
"@Ellipse { 3c @Wide 1c @High }"
}
||7ct
@Diag {
//0.5f
@ShowTags @ShowDirections @Ellipse { 3c @Wide 1c @High }
}
}
@Code CTR has no direction.  If available, direction information
is used when placing labels, in the proximity step (by {@Code above}, for
example) and in the angle step if the label is aligned, perpendicular,
parallel, or antiparallel.  A direction is given using the
@Code ":<" symbol within an outline:
@ID {
@Code @Verbatim {
@Node
    outline {
        LR:: { xsize 0 }
        LR:< 0d
        UL:: { 0 ysize }
        UL:< 270d
        0 0 LR UL 0 0
    }
}
||7ct
@Diag {
//0.5f
@ShowTags @ShowDirections @Node
    outline {
        LR:: { xsize 0 }
        LR:< 0d
        UL:: { 0 ysize }
        UL:< 270d
        0 0 LR UL 0 0
    }
    { 2c @High 3c @Wide }
}
}
It is often helpful when creating outlines to check where the tagged
points and directions really are, by printing them out as is done
above.  For this there is a @Code "@ShowTags" symbol whose result is
the following (arbitrary) object with its tagged points visible, and
a @Code "@ShowDirections" symbol which works similarly and shows the
directions.  The diagram above was printed using
{@Code "@ShowTags @ShowDirections @Node ..."}.  There is also a
@Code "@ShowPoints" symbol which is like @Code "@ShowTags" except
that it omits the tags, just placing circles on the points.
@PP
Link paths are similar to node outlines, created
using the @Code path option of @Code "@Link" instead of the
@Code outline option of {@Code "@Node"}.  The major difference is that
links have no base, so @Code xsize and @Code ysize cannot be
used.  Indeed, even @Code "0 0" does not have any useful
meaning inside a link path.
@PP
Within a link path, the symbols @Code from and @Code to denote the
values of the link's @Code from and @Code to options, and these
form the basis of constructing the link path:
@ID {
@Code @Verbatim {
@Link
    path {
        FROM:: from
        TO:: to
        FROM TO
    }
}
||7ct
{
//1.0c
@VContract @Diag {
3c @Wide 1c @High
//
@ShowTags @Link
    path {
        FROM:: from
        TO:: to
        FROM TO
    }
    from { 0,1 }
    to { 1,0 }
}
}
}
This simple example creates two tagged points and joins them with
a straight line.  If you want a link that can carry arrowheads, it is
best to ensure that it creates @Code FROM and @Code TO tags, with
directions pointing along the link from @Code FROM to @Code TO at
both points, since then the default values of the various arrow
options will do the rest.  Similarly, if you want labels you need to
define {@Code LFROM}, {@Code LMID}, and {@Code LTO} labels, ideally
also with directions.
@PP
Once the outline or path is complete, unless it is really a one-off
production the best thing to do with it is to add it to your
extend. @Index { @Code extend keyword }
@Code "mydefs" file in the following form:
@ID @OneRow @Code @Verbatim {
extend @DiagSetup @Diag
macro @MyNode {
    @Node
        outline {
            LR:: { xsize 0 }
            LR:< 0d
            UL:: { 0 ysize }
            UL:< 270d
            0 0 LR UL 0 0
        }
}
}
This says that we are `extending' the @@Diag symbol by adding a new
symbol, {@Code "@MyNode"}, which stands for what follows it between
braces.  @Code "@MyNode" will then behave exactly like @Code "@Circle"
and the other standard node symbols.  The same pattern works for links:
@ID @OneRow @Code @Verbatim {
extend @DiagSetup @Diag
macro @MyLink {
    @Link
        path {
            FROM:: from
            TO:: to
            FROM TO
        }
}
}
If it is worth the effort to construct a new outline or link path, it
is worth packaging it like this and thinking up a good name for it,
for then it will be available, easily, forever.
@PP
This same approach is also useful to define common combinations of
options, even when there is no new outline or path:
@ID @OneRow @Code @Verbatim {
extend @DiagSetup @Diag
macro @BigOctagon {
    @Polygon
        sides { 8 }
        hsize { 5c }
        vsize { 5c }
        font { Bold }
}
}
Such definitions are very useful if the combinations occur
frequently.  Any options not mentioned have their usual default values,
and may be set in the usual way:
@ID @Code "@BigOctagon outlinestyle { dashed } ..."
Attempts to reset an already set option will elicit a warning message.
@End @Section