Graph::Easy - Manual

Syntax

If you haven't done so, please read the Overview first.

This chapter describes the syntax of the text format that is parsed by Graph::Easy::Parser into a Graph::Easy object.

General rules

Input and Encoding

Graph::Easy, and especially the Parser, expect the input to be in utf-8. This means that the input must be valid Unicode.
But it also means you can embed arbitrary characters, like Japanese, box art drawing, Umlaute etc. into the labels without the need for cumbersome quoting.

Comments

Comments start with a # character (a hatch), and and run til the end of the line.

To make sure that you comments are parsed correctly put a space as the first character after the hatch (entire lines consisting only of the '#' character are ok, though). In addition, inside attributes you should escape the '#' character with a backslash.
For historical reasons, certain things are recognized even without escaping, like color values like #ff00aa inside attributes, where the unescaped '#' will not start a comment:

##############################################################
# This is a good comment.

##############################################################
#This is bad.

node { label: \#5; }	  # Note the "\"!
edge { color: #aabbcc; }  # the color value will be all right

However, to be sure that the parser does not get confused, always follow these two simple rules:

add a space after the '#' for comments
escape '#' inside attribute values with a backslash: '\#'

Whitespace

Whitespace does generally not matter, e.g. multiple spaces are collapsed together to one, and linebreaks are ignored. Whitespace at the front and end of label texts is removed, likewise. Thus the following two are equivalent:

[A]->[B][C]->[D]

[ A ] -> [ B ]
[ C ] -> [ D ]

When writing graph text, you are encouraged to use whitespace and linebreaks for clarity, like in the second example above.

Linebreaks and wrapping

To insert a line break into a node or edge label, use literally \n. Note that the graph description becomes more readable if you insert linebreaks after every literal \n:

[ My\n long\n node\n name ] 
  -- A\n
     longer\n
     label --> [ B ]

+------+          +---+
| My   |  A       |   |
| long |  long    | B |
| node |  label   |   |
| name | -------> |   |
+------+          +---+

You can also replace the \n by \r, \c or \l to align the next line right, left or center, respectively:

[ Long Node Label\l left\r right\c center ]
 -- A\r long\n edge label --> [ B ]

+-----------------+               +---+
| Long Node Label |  A            |   |
| left            |        long   | B |
|           right |  edge label   |   |
|     center      | ------------> |   |
+-----------------+               +---+

Another method to create more readable labels is to use the textwrap-attribute, either setting it for all types of labels in the graph at once (by putting it on the graph), or only for certain classes:

graph { textwrap: auto; }

[ Long Node Label left right center ] 
  -- A long edge label --> { text-wrap: 10; }
[ B ]

+--------+           +---+
|  Long  |           |   |
|  Node  |           |   |
| Label  |  A long   | B |
|  left  |  edge     |   |
| right  |  label    |   |
| center | --------> |   |
+--------+           +---+

Nodes

Nodes are written (or "quoted", if you wish) with enclosing square brackets:

[ Single node ]
[ Node A ] --> [ Node B ]

You can also have a list of nodes by separating them with a comma:

[ A ], [ B ], [ C ] --> [ D ]

+---+     +---+     +---+
| A | --> | D | <-- | C |
+---+     +---+     +---+
            ^
            |
            |
          +---+
          | B |
          +---+

It also works for lists on the right side of an edge:

[ A ] -> [ B ], [ C ], [ D ]

+---+     +---+     +---+
| B | <-- | A | --> | C |
+---+     +---+     +---+
            |
            |
            v
          +---+
          | D |
          +---+

In addition, you can chain nodes together like this:

[ A ] -> [ B ] -> [ C ]
 -> [ D ]
 -> [ E ]

You can also create invisible, anonymous nodes with [ ]. These nodes are called anonymous, because you do not know their name and thus cannot refer to them again:

[ ] -> [ Karlsruhe ] -> [ ] -> [ Plauen ]

That would be rendered like:

        +-----------+             +--------+
    --> | Karlsruhe | -->     --> | Plauen |
        +-----------+             +--------+

If you merely want an invisible node, use either shape: invisible; (for a node with a certain minimum size) or shape: point; point-style: invisible; for a very small and invisible node:

[ $sys$Node ] { shape: invisible; title: You don't see me! }
  -> [ Bischofswerda ]
  -> [ Borna ] { shape: point; point-style: invisible; }
  -> [ Bremen ]
  -> [ $sys$Node ]


  +------------------------------------+
  v                                    |
          +---------------+          +--------+
      --> | Bischofswerda | -->  --> | Bremen |
          +---------------+          +--------+

Attributes

Attributes are enclosed in { }, are in the format attributename: attributevalue;.

Attributes names must not be quoted, attribute values can be quoted, and any double quotes (when they appear at start and end) will be removed. Note that the quotes are only removed from attribute values, and not from node or group names:

[ "Monitor Size" ] --> { label: 21"; } [ Big ] { label: "Huge"; }

This will produce:

+----------------+  21"   +------+
| "Monitor Size" | -----> | Huge |
+----------------+        +------+

If you want to embed a ; (semicolon), then either quote the string or escape the semicolon with a backslash:

[ Baz;Bar ] --> { label: "Baz;Bar"; } [ Bonn ] { label: \;\;; }

+---------+  Baz;Bar   +----+
| Baz;Bar | ---------> | ;; |
+---------+            +----+

If you want to embed a semicolon and have quotes at the start and end of the string value, simple use backslashes to escape the quotes and semicolon:

[ A ] { label: \"Hello\;\"; } --> [ Berlin ]

+----------+     +--------+
| "Hello;" | --> | Berlin |
+----------+     +--------+

If you want to embed a # (hashmark), then you must escape the hashmark, or the parser will see it as the start of a comment:

[ Baz\#Bar ] --> { label: "Baz\#Bar"; } [ Bonn ] { label: \#\#; }

+---------+  Baz#Bar   +----+
| Baz#Bar | ---------> | ## |
+---------+            +----+

Note that the need to quote or escape only applies to attribute values, and not to node or group names, since these are always used as-they-are.

Attribute Scope

Attributes follow immidiately the object(s) for that they apply.
Please note that attributes following a node in a node list apply to all nodes in the list that came to this point:

graph { background: white; }	# for the graph itself
node { background: white; }	# for all nodes
edge { style: bold; }		# for all edges
node.city { background: red; }	# for all nodes with class "city"

[ Bonn ] { class: city; }	# for the node "Bonn"

[ Bonn ] --> { style: dotted; }	# for the edge "Bonn" to "Berlin"
[ Berlin ] { color: green; }	# for the node "Berlin"

[ ABC ] { border: bold; color: white; }

[ DEF ] 			# DEF has not yet attributes
  {
  fill: #ff8060;		# applies only to DEF
  } ,
[ GHI ]
  {				# these apply to DEF and GHI!
    border: bold; 
    color: white; 
  } ,
[ JKL ]
  {
    shape: circle;		# applies to DEF, GHI and JKL
  }

[ ] { fill: brown; }		# for the anonymous node only

Multiple Attributes

As a special case for autosplit nodes, you can separate attribute values with "|" (vertical bar) for each part of the auto-split node:

[ Bonn | Berlin | Frankfurt ] { fill: red|yellow|blue; }

In the example above, "Bonn" would be red, "Berlin" yellow and "Frankfurt" blue. Missing values will NOT change the attribute on the node as in the following example, where "Berlin" and "Ulm" will get the default background (white):

node { fill: white; }

[ Bonn | Berlin | Hahn | Ulm ] { fill: red||blue; }

The || here means that the attribute value is missing, please do not confuse that with || inside the autosplit node itself, where this introduces a "linebreak".

You can also set an attribute for all parts, or only for specific parts:

[ A|B ] { class: |legend; } will put B into class 'legend' and leave the class of A alone
[ A|B ] { class: legend|; } will put A into class 'legend' and leave the class of B alone
[ A|B ] { class: legend; } will put A and B into class 'legend'

Here are the rules from above in an example showing their effect:

node.1 { border: dotted; }
node.2 { border: dashed; }
node.3 { border: dot-dash; }
node { border: double; }

[ AAAA|BBBB ] { class: |1; }
[ CCCC|DDDD ] { class: 2|; }
[ EEEE|FFFF ] { class: 3; }

#======#.......
H AAAA H BBBB :
#======#......:
+ - - -+======#
' CCCC ' DDDD H
+ - - -+======#
+-.-.-.+.-.-.-+
! EEEE ! FFFF !
+-.-.-.+.-.-.-+

For a complete listing of possible attributes see the appropriate chapter.

Edges

The edges between the nodes can have the following styles:

        ->              solid
        =>              double
        .>              dotted
        ~>              wave

        - >             dashed
        .->             dot-dash
        ..->            dot-dot-dash
        = >             double-dash

In addition to these, the following styles exist:

bold
bold-dash
broad
wide
invisible

Unlike the normal edge styles, these can only be set via the (optional) edge attributes:

        [ A ] --> { style: bold; }      [ B ]
              --> { style: broad; }     [ B ]
              --> { style: bold-dash; } [ C ]
              --> { style: invisible; } [ D ]

            ####
            v  #
+---+     +------+     +---+     +---+
| A | ##> |  B   | # > | C |     | D |
+---+     +------+     +---+     +---+

You can repeat each of the style-patterns as much as you like:

        --->
        ==>
        =>
        ~~~~~>
        ..-..-..->

Note that in patterns longer than one character, the entire pattern must be repeated e.g. all characters of the pattern must be present. Thus:

        ..-..-..->      # valid dot-dot-dash
        ..-..-..>       # invalid!

        .-.-.->         # valid dot-dash
        .-.->           # invalid!

In additon to the styles, the following directions are possible:

         --             edge without arrow heads
         -->            arrow at target node (end point)
        <-->            arrow on both the source and target node
                        (end and start point)

Of course you can combine all directions with all styles. However, note that edges without arrows must use at least two repetitions of the full pattern:

        ---             # valid
        .-.-            # valid
        .-              # invalid!
        -               # invalid!
        ~               # invalid!

You can also give edges a label, either by inlining it into the style, or by setting it via the attributes:

        [ AB ] --> { style: bold; label: foo; } [ ABC ]

        -- foo -->
        ... baz ...>

        -- solid -->
        == double ==>
        .. dotted ..>
        ~~ wave ~~>

        -  dashed - >
        =  double-dash = >
        .- dot-dash .->
        ..- dot-dot-dash ..->

Note that the two patterns on the left and right of the label must be the same, and that there is a mandatory space between the label and the patterns on the left and right side.

You may use inline labels only with edges that have at least one arrow. Thus:

        <-- label -->   # valid
        -- label -->    # valid

        -- label --     # invalid!

To use a label with an edge without arrow heads, use the attributes:

        [ AB ] -- { label: edgelabel; } [ CD ]

For a complete listing of possible edge attributes see the appropriate chapter.

Groups

You can group nodes together by using parantheses:

( German Cities
  [ Berlin ] -> [ Potsdam ]
) { 
  background: lightbrown;
  }

Putting nodes into a group gives the layouter the hint that these nodes are related and must be laid out closely together.

If the name of the group is ommitted, an anonymous group is created. These groups do not have a border and background, and since the group label defaults to the group name you will get an invisible label as a side-effect:

( [ Bremen ] -> [ Bremerhaven ] )

+--------+     +-------------+
| Bremen | --> | Bremerhaven |
+--------+     +-------------+

Please see the section about grouping for further details and examples.

Advanced Layouts

Joints

Sometimes you want on edge join another, or have a pair of edges start at a common point, and then split up. Or you even want two edges being connected by a third edge. Graph::Easy allows all these things by a feature called shared ports.

For detailed information and examples please see the chapter about joints.

Classes

Each type of object in a graph is a (primary) class and you can set attributes on them on a per-class basis:

graph { color: red; }
edge { color: blue; }
node { color: green; }
group { color: brown; }

All objects in a graph automatically belong to their primary class, e.g. nodes are in the class "node". Except graph, all primary classes can have subclasses:

edge { color: blue; }
edge.train { color: darkblue; }
node { color: green; }
node.cities { color: darkgreen; }
group { color: brown; }
group.cities { color: darkbrown; }

See also the section about class names for reference and more examples, including class selectors.

Relative placement (via auto-split)

You can cluster nodes together by placing them relatively to each other.
Perhaps the easiest way to achive the placement is to use the auto-split feature:

a | (vertical bar) in the node name will split the node into two parts, and place them next to each other, horizontally
likewise, || (two vertical bars) in the node name will split the node into two parts, but place the second part at the start of a new row
If a part between two | consists of exactly one space, an invisible cell will be generated, e.g. one without borders and background
If a part between two | consists of more than one space, an empty cell (e.g. with borders and background) will be generated
Trailing empty parts will be treated just like when they apeared in the middle, e.g. both [ |... and [|..- create an invisible leading part, while [ |... (two spaces) creates an empty, but visible leading part. Likewise for trailing parts.

Please see the section in Hinting for examples and explanations.

Please continue with the chapter about the attributes.