Node-and-edge diagrams

The diagram package draws node-and-edge diagrams — block diagrams, state machines, flowcharts, dependency graphs — over the vector-graphics layer. A node is a named, measured box of some shape; an edge is an arrow between two named nodes. Three thin presets build on it: flowchart gives the shapes their program-flowchart meanings, automaton gives them finite-state-machine meanings, and er draws entity-relationship diagrams.

\use{diagram}

Everything is declared inside a \picture: open one, declare nodes and edges into it, and the whole diagram is a single box that flows in the text.

\picture width:3.4in height:2.6in {
  \node{start}{at 1.2in 2.2in}{Begin}
  \node[diamond]{check}{below start}{ok?}
  \node{done}{below check 40}{Finish}
  \edge{start}{check}
  \edge[yes]{check}{done}
}

The model

A node records its centre, half-width and half-height (sized to its label), and publishes its centre as a named picture coordinate — so (start) works in raw picture code too. Every edge reads that geometry back and attaches to the node’s true boundary, along the line toward the other node, so an arrow always meets the outline cleanly whatever the shape: straight, diagonal, elbowed and curved edges all land on the edge of the box, the diamond, the circle or the parallelogram, never inside it and never short of it.

There is no diagram-specific engine primitive — placement, the shapes and the arrow routing are all document language over the picture primitives.

Nodes

\node [shape] {name} {placement} {label}

The optional [shape] defaults to box. The recognised shapes:

Placement

The {placement} argument is a small word language. Relative placement measures from the reference node’s boundary, then steps in by the gap and this node’s own half-size, so neighbours sit a true gap apart whatever their sizes.

\node{a}{at 1in 4in}{First}
\node{b}{below a}{Second}        % default gap below a
\node{c}{right b 50}{Beside}     % 50pt to the right of b

Edges

Every edge inks in the configured colour and caps with the configured arrowhead. An optional [label] rides beside the edge.

\edgehv and \edgevh are the natural way to send a decision’s side branch across and then down, or to route a feedback line. When the two nodes happen to be aligned on the elbow’s axis they reduce to a straight edge automatically.

\cedge‘s bend is positive to bow left of the from→to direction; give the two directions of a back-and-forth pair the same bend and they separate into two arcs instead of drawing one arrow over another.

\picture width:4.6in height:1in {
  \node[circle]{sum}{at 0.5in 0.5in}{$\Sigma$}
  \node{ctrl}{right sum 40}{Controller}
  \node{plant}{right ctrl 40}{Plant}
  \edge[$e$]{sum}{ctrl}
  \edge[$u$]{ctrl}{plant}
}

Customising the look

Set any of these with \set before drawing. Sizes are in points unless they carry a unit.

Flowcharts

The flowchart package names the diagram shapes by their roles and reads the edges as flow. It draws nothing of its own — each command maps to a diagram node or edge.

\use{flowchart}

Each role takes the same arguments as \node — a name, a placement and a label.

\picture width:3in height:4in {
  \start{begin}{at 1.2in 3.6in}{Start}
  \io{read}{below begin}{Read $n$}
  \decision{check}{below read}{$n > 0$?}
  \process{print}{below check}{Print $n$}
  \stop{done}{below print}{End}
  \flow{begin}{read}
  \flow{read}{check}
  \flow[yes]{check}{print}
  \flow{print}{done}
  \branch[no]{check}{done}
}

Automata

The automaton package gives the circle and the edges their finite-state-machine meanings.

\use{automaton}

\picture width:5in height:1.7in {
  \state{q0}{at 0.7in 0.7in}{$q_0$}
  \state{q1}{right q0 78}{$q_1$}
  \accepting{q2}{right q1 78}{$q_2$}
  \initial{q0}
  \trans[$a$]{q0}{q1}
  \trans[$b$]{q1}{q2}
  \loopabove{q0}{$b$}
  \loopabove{q2}{$a, b$}
}

The presets carry no geometry of their own, so anything diagram exposes — the other shapes, the orthogonal and curved edges, and the full configuration — is available alongside them.

Entity-relationship diagrams

The er package draws entity-relationship diagrams in Chen notation: an entity is a rectangle, a relationship a diamond, an attribute an ellipse, joined by undirected lines (diagram‘s \link, no arrowhead) that carry cardinalities.

\use{er}

\picture width:4.4in height:2in {
  \keyattribute{ssn}{at 0.5in 1.7in}{ssn}
  \entity{emp}{at 1.1in 1in}{Employee}
  \relationship{wf}{right emp 60}{Works For}
  \entity{dept}{right wf 60}{Department}
  \connect{emp}{ssn}
  \connect[N]{emp}{wf}
  \connect[1]{wf}{dept}
}

The weak forms and the double ellipse add a second outline (like an automaton’s accepting state); the dashed ellipse and the underlined key attribute are an ordinary ellipse with a line-style or an \underlined label. Everything else is a diagram node or \link.

A note on spacing

These packages only load diagram, not the document format. A bare-engine document is vertically justified by default, which on a sparse page stretches the small inter-paragraph glue and spreads the paragraphs apart. Set \raggedbottom (what the document format does for you) to keep natural spacing:

\set raggedbottom {1}

To centre a figure, an \hbox with \hfil on each side does it:

\def centerline c {\hbox to:\hsize {\hfil\c\hfil}}
\centerline{\picture width:3in height:2in { … }}