How to layout Graphviz / dot layout and order question - graphviz

Question

I'm trying to use dot notation and set the order of the items. in the example below I would like FLIP/FLOPA to be on top and FLIP/FLOPB on the bottom. It would also be good to line everything up.
I'm a bit confused how invisible edges might work?
this is where im at:
digraph G {
graph [pad ="1", rankdir = LR, splines=ortho];
size = "16.66,8.33!"; // 1200x600 at 72px/in, "!" to force
ratio = "fill";
node[shape=record];
flipa[label="FLIPA", height=3];
flopa[label="FLOPA", height=3];
flipb[label="FLIPB", height=3];
flopb[label="FLOPB", height=3];
source1[shape=rarrow];
source2[shape=rarrow];
sink1[shape=rarrow];
sink2[shape=rarrow];
source1 -> flipa;
flipa -> flopa [label="red" color=red,penwidth=3.0];
flipa -> flopa [label="blue" color=blue,penwidth=3.0];
flopa -> sink1;
source2 -> flipb;
flipb -> flopb [label="red" color=red,penwidth=3.0];
flipb -> flopb [label="blue" color=blue,penwidth=3.0];
flopb -> sink2;
label="Graph";
labelloc=top;
labeljust=left;
}
Thanks in advance
Neil

Answer 1

The order in which you mention nodes in layout is important.
If you have horizontal layout (rankdir=LR), the nodes appear in your layout from bottom to top, look at this, for example:
digraph {
rankdir=LR
node1
node2
node3
}
This results in:
Now if we reverse the order:
digraph {
rankdir=LR
node3
node2
node1
}
We get this:
PS: I don't recommend using ortho splines, they may cause all sorts of artifacts, including very poor handling of edge labels. You problems with lining up may be caused by them. Even though when I compile your code on my machine, the nodes are lined up correctly.
To answer your question in the comments:
How to implement parallel edges without splines=ortho
But let me warn you, it is very ugly.
In graphviz you can use HTML-like syntax to define some structures, most important of them being tables. Any old enough frontender will tell you that you can design almost anything using tables. I use them a lot when work with graphviz.
What you can do in your situation: instead of your plain rectangular nodes place a table with three rows. Top row and bottom row are going to be empty. The middle one will contain your label: FLIPA or FLOPA.
Next you assign port attribute to all cells. This way you will be able to connect specific rows of tables together using the headport and tailport (or their synonims, the colon-syntax, which I used in example below).
Here's the example:
digraph G {
graph [pad ="1", rankdir = LR];
size = "16.66,8.33!"; // 1200x600 at 72px/in, "!" to force
ratio = "fill";
node[shape=record];
flipb[
shape=plain
label=<
<table border="1" cellspacing="0" cellborder="0">
<tr>
<td height="80" port="upper"> </td>
</tr>
<tr>
<td height="80" port="middle">FLIPB</td>
</tr>
<tr>
<td height="80" port="bottom"> </td>
</tr>
</table>
>
];
flopb[
shape=plain
label=<
<table border="1" cellspacing="0" cellborder="0">
<tr>
<td height="80" port="upper"> </td>
</tr>
<tr>
<td height="80" port="middle">FLOPB</td>
</tr>
<tr>
<td height="80" port="bottom"> </td>
</tr>
</table>
>
];
flipa[
shape=plain
label=<
<table border="1" cellspacing="0" cellborder="0">
<tr>
<td height="80" port="upper"> </td>
</tr>
<tr>
<td height="80" port="middle">FLIPA</td>
</tr>
<tr>
<td height="80" port="bottom"> </td>
</tr>
</table>
>
];
flopa[
shape=plain
label=<
<table border="1" cellspacing="0" cellborder="0">
<tr>
<td height="80" port="upper"> </td>
</tr>
<tr>
<td height="80" port="middle">FLOPA</td>
</tr>
<tr>
<td height="80" port="bottom"> </td>
</tr>
</table>
>
];
source1[shape=rarrow];
source2[shape=rarrow];
sink1[shape=rarrow];
sink2[shape=rarrow];
source1 -> flipa;
flipa:upper -> flopa:upper [label="red" color=red,penwidth=3.0];
flipa:bottom -> flopa:bottom [label="blue" color=blue,penwidth=3.0];
flopa -> sink1;
source2 -> flipb;
flipb:upper -> flopb:upper [label="red" color=red,penwidth=3.0];
flipb:bottom -> flopb:bottom [label="blue" color=blue,penwidth=3.0];
flopb -> sink2;
label="Graph";
labelloc=top;
labeljust=left;
}
Result: