Information Technology Reference
In-Depth Information
(a)
Layout with layer-based methods by Schulze et al.
(b)
Layout with the CoDaFlow algorithm presented here.
Fig. 5.
Two layouts of the same Ptolemy diagram. While two distinct networks are
interleaved in (a), they are clearly separated and the two crossings are avoided in (b).
networks are interleaved. A
global
approach would solve this issue, positioning
all compound nodes along with their children at the same time.
Even though we focus our attention on a global approach in what follows, our
methods are flexible in that we may choose between a bottom-up and a global
strategy in each stage of our pipeline.
A compound graph
G
is transformed into
G
as above, which is used to con-
struct a
flat
graph
G
=(
A, E
)where
A
V
is the set of atomic nodes and
ↆ
their port nodes, and
E
=
U
∪
H
with
U
=
{
(
ʴ
(
p
1
)
,ʴ
(
p
n
)) :
ˀ
(
p
1
)
,ˀ
(
p
n
)
∈ A}
H
=
{
(
ʴ
(
p
1
)
,ʴ
(
p
n
)) :
∃
(
p
1
,p
2
)
,
(
p
2
,p
3
)
,...,
(
p
n−
2
,p
n−
1
)
,
(
p
n−
1
,p
n
)
∈ E
:
ˀ
(
p
1
)
,ˀ
(
p
n
)
∈
A
∧
ˀ
(
p
2
)
,...,ˀ
(
p
n−
1
)
∈
V
\
A
}
Intuitively, compound nodes are neglected along with their ports and only
atomic nodes are retained. Sequences of edges that span hierarchy boundaries,
e. g., the three edges between
in Fig. 5b, are replaced
by a single edge that directly connectsthetwoatomicnodes.Notethatfor
hyperedges multiple edges have to be created.
Cluster constraints
[3] guarantee
that children of compound nodes are kept close together and are not interleaved
with any other nodes. For instance, the
Sampler2
and
Controller2
CompositeActor
in Fig. 5b yields a cluster
containing
Controller1
and
Controller2
.
