Posterizing Diagrams with Top-Down Layout

Maximilian Kasperowski

and Reinhard von Hanxleden

Department of Computer Science, Kiel University, Kiel, Germany

{mka, rvh}@informatik.uni-kiel.de

Keywords:

Automatic Layout, Layered Graph Drawing, Nested Graph Drawing.

Abstract:

Diagrams serve as a communication tool and automatic graph drawing enables the creation of diagramming

tools that allow seamless graph modeling and visualization. These tools typically have the goal of optimiz-

ing the produced diagrams for computer screens, which support zooming in and out of diagrams and further

interaction techniques. In this paper, we consider what automatic graph drawing needs to provide to produce

diagrams suitable for printing instead, where we can no longer zoom and are restricted to a ﬁxed area. We iden-

tify key differences in the requirements between printed static diagrams and interactive on-screen diagrams,

including layout speed and layout area. Our goal is to automatically draw a complex diagram to be printed

on a large poster, with maximal readability. We propose a collection of reﬁnements to the well-established

Sugiyama algorithm and illustrate our approach with a large SCChart that implements a railway controller.

1 INTRODUCTION

Automatic graph drawing describes the automatic

creation of node-link diagrams from a graph. In prac-

tice this graph has typically been extracted from some

kind of model. Modeling tools frequently use visu-

alizations of models, and automatic graph drawing

technology allows the creation of these visualizations

on-the-ﬂy during the modeling process (Lohstroh

et al., 2021; Glaser and Bork, 2021; Kasperowski

et al., 2024). While such tooling typically focuses

on creating diagrams that work well on a zoomable

computer screen, another use case are static diagrams.

In industry, it is not uncommon to create “diagram

walls” that spread out over multiple square meters.

These walls typically consist of numerous printouts

of system components represented by large nested

graphs. For example, calibration engineers use such

diagram walls to explore complex engine control soft-

ware. Creating such diagram walls manually is a

rather time-consuming process, which motivates the

work presented here.

We investigate the requirements printed diagrams

have of automatic graph drawing techniques and how

they differ from diagrams primarily displayed on

computer screens. We perform this analysis by exem-

plarily creating a large printable diagram using auto-

matic graph drawing techniques. We utilize existing

https://orcid.org/0000-0002-7509-1678

https://orcid.org/0000-0001-5691-1215

techniques to optimize space usage as well as apply

new techniques that are not as important or applicable

for computer screens.

The diagrams included in this paper were pro-

duced using the KIELER tooling (Kasperowski et al.,

2024), which uses the Eclipse Layout Kernel (Domr

et al., 2023b) to perform automatic layout. The ex-

ample used throughout this paper is an excerpt of a

larger poster

created to demonstrate the techniques

discussed in this paper.

A common method to draw large nested graphs

is to use a divide-and-conquer approach and work

bottom-up, i. e., one ﬁrst creates layouts for the

bottom-most elements of the graph hierarchy and then

combines those to determine size requirements of the

parents, which are then laid out in turn. E. g., in

Fig. 1a the layout of the subgraph in Giraffe deter-

mines the ﬁnal size of the node. This is a recursive

procedure terminating at the graph root.

Such bottom-up layouts produce drawings where

low-level elements and labels are drawn at a uniform

scale, however, for large graphs this method leads to

large and hard-to-read graphs with no good overview

over any given hierarchy level. This is why we in-

stead propose to use a top-down procedure where

node sizes are restricted and ﬁxed by higher-level

nodes (Kasperowski and von Hanxleden, 2023). Sub-

layouts are scaled down to ﬁt within their parents. A

https://rtsys.informatik.uni-kiel.de/

∼

biblio/

downloads/papers/ivapp25-showpiece.pdf

930

Kasperowski, M. and von Hanxleden, R.

Posterizing Diagrams with Top-Down Layout.

DOI: 10.5220/0013299200003912

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 20th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2025) - Volume 1: GRAPP, HUCAPP

and IVAPP, pages 930-937

ISBN: 978-989-758-728-3; ISSN: 2184-4321

Cat

Elephant

Chameleon

Giraffe

Dog

(a) Bottom-up layout.

Cat

Elephant

Chameleon

Giraffe

Dog

(b) Top-down layout.

Figure 1: This simple example highlights the difference between bottom-up and top-down layout. In the bottom-up case, the

children deﬁne the sizes of the parents, while in the top-down case the parents restrict the space available to their children.

comparison of the bottom-up and top-down procedure

is shown in Fig. 1.

Top-down layout itself, while creating helpful

overviews and structure in the drawings, also adds

whitespace through its scaling. This can be seen in

the nodes in Fig. 1b. This additional whitespace must

be kept to a minimum for top-down layout to produce

useful results. Essentially this translates to the re-

quirement that layouts must use their provided space

effectively.

This can be rephrased as stating that we want to

create a good layout for a prescribed drawing area,

which is itself a topic that has previously been studied

extensively (R

uegg, 2018). We compare the require-

ments of printed diagrams with those of diagrams

used on computer screens with respect to layout speed

and layout area. The general layout techniques we

cover in this paper are graph wrapping, layer split-

ting, and label management.

1.1 Contributions and Outline

Sec. 2 presents related work. In Sec. 3 we discuss how

we apply graph drawing techniques to the problem of

producing printable diagrams using existing methods

as well as the following new approaches:

• We introduce the concept of layer splitting as a

pendant to graph wrapping for producing layouts

closer to a desired aspect ratio.

• We extend semantic soft wrapping with fuzzi-

ness to produce aesthetically pleasing and space-

efﬁcient labels.

• We propose adaptive font sizes to improve read-

ability in large drawings containing many small

elements.

We discuss our results and how the techniques cov-

ered in this paper help to produce a readable diagram

in Sec. 4. Finally, Sec. 5 concludes this paper.

2 RELATED WORK

At ﬁrst glance the task of creating a single diagram

for a poster poses the question: Do we need automatic

graph drawing or is a manual layout potentially bet-

ter? We argue that the main advantage of a manual

layout, control, is greatly outweighed by the many ad-

vantages of automatic layout. While users often have

a very strong desire for control, manually positioning

and repositioning many graph elements becomes very

tedious (Petre, 1995). Automatic layout on the other

provides us with consistency, we can easily adjust pa-

rameters that affect the entire diagram, and rapid pro-

totyping is possible. Furthermore, using model order

to control the automatic layout actually provides us

with a large degree of control over the positioning of

elements in the diagram (Domr

os et al., 2023a).

To achieve minimize whitespace we apply the

concept of prescribed drawing areas or area-adaptive

graph drawings. In our case we are interested in tech-

niques for layered graph drawings (R

uegg, 2018), but

the problem has also been studied for other types of

graph drawings such as tree drawings (Misue, 2024).

When creating large diagrams that are viewed on

computer screens there are dynamic focus and context

visualization techniques that can be used to aid with

viewing large diagrams. Fisheye techniques encom-

pass methods of distorting the view or the graph to

highlight certain portions of the diagram (Sarkar and

Brown, 1992; Formella and Keller, 1996; Tominski

et al., 2006). This effect is restricted to one area of

interest that is brought into focus. We could apply

such an effect selectively to portions of our diagram

poster, but we want to create a static view of the entire

diagram to let observers explore it.

Another technique commonly used to make large

diagrams readable when zoomed out is semantic

zooming, which replaces details of the diagram with

readable abstract views (Frisch et al., 2008; De Carlo

et al., 2022). The detailed views are shown as the

Posterizing Diagrams with Top-Down Layout

931

view is zoomed in. We cannot dynamically switch

elements of a printed diagram, but we do want to en-

able observers to zoom in by standing closer to the

poster. Therefore, we also need graphic elements that

are readable from a distance and detailed elements

that only become readable upon stepping closer.

3 POSTERIZING DIAGRAMS

Diagrams as a communication medium are often

found printed on posters of varying sizes. For this pa-

per we considered the challenges of automatic graph

drawing speciﬁcally for the use case of producing an

aesthetic and readable diagram for print. For this

application we used SCCharts (von Hanxleden et al.,

2014), which is a statechart (Harel, 1987) dialect

and have an automatic visualization provided by the

KIELER tool. Because SCCharts can be hierarchical,

we can utilize top-down layout (Kasperowski and von

Hanxleden, 2023) to produce drawings that provide

good overviews, i. e., views of the diagram where

graph elements that have relations to one another are

readable and their relationships clear.

As an example for this problem we looked at a

part of a model railway controller created as a student

project. We chose to use only an excerpt of this rather

large controller to create something that is interesting

and can still be reasonably printed within a paper. The

result can be seen in Fig. 2. Fig. 3 shows the same

SCChart but laid out using a bottom-up procedure.

3.1 Requirements of Printed Diagrams

In this paper we are concerned with the problem of

creating a diagram suitable for printing on a large

poster. This use case differs from the viewing of di-

agrams on computer screens, and we therefore need

to consider what the requirements are. The following

are the core goals of large printed diagrams.

The most important elements must be clearly visi-

ble and recognizable by an observer standing in front

of the poster. Further details may be smaller such that

they can be read by stepping closer to the poster. Any-

thing smaller than that can no longer be read.

A poster provides us with a larger base area than

a computer screen. This means we can draw larger

boxes with relatively smaller labels and still have

them readable. We cannot, however, zoom in as we

could on a computer. Therefore, it is important that

we ﬁll the space that we have available effectively and

try to reduce the amount of elements and labels that

are printed so small that they can no longer be dis-

cerned easily. With a poster we have a bit of leeway

since we can move away from the poster and move

closer to it to examine details, but the overall range of

viable scale levels is limited.

A poster is also a communication medium. The

diagram that we draw on it should be helpful and un-

derstandable. We want to clearly show “interesting”

elements and relations and ﬁlter out or reduce un-

wanted clutter or redundant information. This will

further help us in highlighting interesting features.

Note that these interesting elements are domain-

dependent and therefore vary between modeling lan-

guages and models.

3.2 Top-Down Layout Conﬁguration

Using top-down layout requires setting a desired dia-

gram size as a constraint. In our scenario these con-

straints are provided by the poster size.

The constraint we provide for a layout of a node is

the ﬁxed size of the node. This precisely deﬁnes how

much space is available to the children of the node,

i. e., the to-be-drawn graph, e. g., the size available

to the subgraph in the Giraffe node in Fig. 1b. If the

provided space is exceeded by the graph drawing pro-

cedure, a scaling is applied to ﬁt the drawing into the

prescribed area. This way top-down layout provides

us with a simple mechanism to control the size ele-

ments should be drawn at.

Dynamically estimating suitable sizes is in itself

a challenging problem and remains a topic for future

work, however, we can manually adjust the size con-

trol to achieve an aesthetic result. For the example

in Fig. 2 we set the size of the main state to roughly

match the page dimensions and all other nodes were

set to a size of approximately 650 px by 400 px.

There are many ways to conﬁgure layout options

that affect the resulting aspect ratio of the drawing in

hard-to-predict ways. A key importance in using au-

tomatic graph drawing for generating interactive di-

agrams is that the algorithms must be fast. In the

printed case the speed of the layout computation is

of secondary importance. For a given diagram to be

printed it can be worthwhile to explore different lay-

out options to determine the optimal set of layout op-

tions for a given input graph. For this we determined

a small set of layout options that would have a strong

effect on the area of the drawing and generated draw-

ings for all permutations of these options.

The layout options we searched optimal settings

for were the layout direction (down or right) and the

layer splitting3 (see Sec. 3.3) strategy (no split or 2-

layer-split). These two options have a large impact

on the resulting aspect ratio of a drawing, but their

effect is difﬁcult to predict. Finally, we use the op-

IVAPP 2025 - 16th International Conference on Information Visualization Theory and Applications

932

Figure 2: The drive SCChart laid out using top-down layout to neatly ﬁt on a page or more suitably on a poster. This example

demonstrates all the layout tools that are explained in the following sections. The model itself stems from a student project and

is an excerpt from a larger model that controls a model railway. This drive state controls the behaviour of an individual train.

There are several highlighted excerpts that are shown and discussed in later sections. Excerpt A highlights graph wrapping

and is shown in Fig. 4. Excerpt B highlights layer splitting and is shown in Fig. 5. Excerpt C highlights fuzzy soft wrapping

and is shown in Fig. 6. Excerpt D highlights adaptive fonts and is shown in Fig. 7.

Posterizing Diagrams with Top-Down Layout

933

Figure 3: The drive SCChart laid out using bottom-up layout. The adaptive fonts feature is also enabled in this version. Despite

this the diagram remains wholly unreadable as it would require a space of nearly fourteen meters by ﬁve meters to print fonts

at their default size of 11 pt. At the current scale none of the labels are readable. Some used tools such as graph wrapping and

adaptive fonts do technically help a bit, but it should be clear that this is a diagram that is too large to draw in the bottom-up

manner and still produce good results. We also lose the ability to exercise direct control over the size of the root node, which

in this case leads to a wide diagram that cannot effectively use the available page area.

Figure 4: This state is an excerpt of Fig. 2 showing how

wrapping ﬁts long graphs into restricted spaces.

tion settings that result in an aspect ratio closest to

the prescribed area’s aspect ratio as it leads to mini-

mal scaling in top-down layout. The scale measure

introduced by R

uegg expresses this property (R

uegg,

2018).

3.3 Whitespace Management

In this section we discuss techniques to reduce the

whitespace in a layered graph drawing. The topic of

prescribed drawing areas for layered graph drawings

has previously been studied by R

uegg and in partic-

ular the concept of graph wrapping is a helpful ap-

proach to control and restrict the aspect ratio required

by graphs with many layers (R

uegg and von Hanxle-

den, 2018; R

uegg, 2018).

We utilize graph wrapping to effectively ﬁll the

space provided by our ﬁxed sizes in situations where

the drawing would normally become very long and

thin. Fig. 4 demonstrates how wrapping was used for

our example. One wrapped graph is contained in the

travelKHKH state, and one wrapped graph is contained

in the right child state of the travelKHIC state.

We propose layer splitting as a post-processor af-

ter the layer assignment step of the Sugiyama Algo-

rithm (Sugiyama et al., 1981). Layer splitting ad-

dresses a problem that is orthogonal to the graph

wrapping approach. While graph wrapping helps to

reduce the effect of long layouts, layer splitting im-

proves bad aspect ratios of high (in a left-to-right

layout direction) layouts. The rotated case (top-to-

bottom layout direction) is demonstrated in Fig. 5.

Here, the travel. . . states are partitioned into rows.

Instead of having many thin layers as in graph

wrapping, we have few layers that are very fat, i. e.,

they have many nodes next to each other. To decrease

the width of the layer we can split up the nodes of a

layer into two or more layers. In our situation with

top-down layout the nodes are all of similar sizes and

therefore simply staggering the nodes in a brick-like

layout is sufﬁcient to achieve a compact result.

3.4 Control and Model Order

Our use case of creating a large printed poster of a

speciﬁc diagram means that we want to be able to ex-

ercise ﬁne control over the layout of certain portions

of the diagram. The diagram is merely a visualization

of an underlying textual model (Kasperowski et al.,

2024). The textual order in this model, i. e., the order

the states are deﬁned in, is used to infer the model or-

der and to guide decisions during the automatic layout

of the graph (Domr

os et al., 2023a).

This allows us to easily adjust and control what

the diagram looks like by modifying the textual model

rather than being forced to ﬁddle with the layout al-

IVAPP 2025 - 16th International Conference on Information Visualization Theory and Applications

934

Figure 5: This state is an excerpt of Fig. 2 where layer splitting was used to decrease the width of the ﬁnal layer in the main

layout. This ensures that the diagram does not become too wide for the restricted page space.

gorithms themselves. This ﬁne-granular control over

the layout is invaluable when tweaking a diagram to

achieve the desired appearance. Moreover, model or-

der shows the underlying structure of a textual model

even though the diagram may be cluttered. It also

helps us retain stability, i. e., the general structure of

diagram does not change between versions.

3.5 Label Management

In this section we discuss label management tech-

niques that create aesthetic and space-efﬁcient labels.

Fuzzy Label Wrapping. In realistic models there

are typically many labels that also tend to be long. La-

bels in general and edge labels in particular can have a

large effect on the overall size requirements of a graph

drawing. This effect compounds in diagrams of large

models and good label management is required to re-

duce the amount of whitespace that is introduced.

Schulze has previously studied how to perform la-

bel management for layered graph drawings (Schulze,

2019). In particular the method of semantic soft wrap-

ping is of interest to us because it allows us to re-

duce the width and use the available vertical space.

In soft wrapping long labels are split into multiple

lines based on a maximum line length. The line break

is performed on space between words if the cutoff

length has been reached. Semantic soft wrapping ex-

tends this by adding additional linebreaks at semanti-

cally suitable positions. For SCCharts these are for ex-

ample the slashes in the transition labels as they serve

as visual separator between the trigger and action of

the transition.

There are situations in which the hard linewrap

cutoff creates awkward empty spaces in parts of the

wrapped label. We improve the label wrapping by

adding a fuzzy cutoff that permits a line to exceed its

line length limit if the next line would be underﬁlled.

A line is underﬁlled if it only covers a deﬁned percent-

railNextStation(trainNum) == 1

/ targetTrack =

railNextTrack(trainNum)

railNextStation(trainNum) == 2

/ targetTrack =

railNextTrack(trainNum)

railNextStation(trainNum) == 3

/ targetTrack =

railNextTrack(trainNum)

3 h dAb

5: homeStation == 4

/ targetTrack =

homeTrack

1: cleanup &

!t i T lli

Figure 6: The edge labels in this excerpt of Fig. 2 demon-

strate the result of semantic fuzzy soft wrapping.

age of the available line space. This enhancement fur-

ther compacts labels and also produces more aesthetic

results. Long and mostly empty lines are greatly re-

duced when adding a fuzzy cutoff. In Fig. 2 a fuzzy

cutoff of 30 percent is used.

Fig. 6 shows this in action. Soft wrapping would

usually result in the segments of the form: == x, where

x is some short string or number, to be placed in a

new line. We instead keep these short label fragments

on the previous line, which reduces the overall area

occupied by the label.

Adaptive Font Sizes. When using top-down lay-

out there can be signiﬁcant scale differences between

neighbouring nodes due to the contained subgraphs

differing strongly in their complexity, as shown in

Fig. 2. This makes larger layouts with more nodes and

therefore stronger downscaling harder to read. Top-

down layout is intended to maintain the overview. Sit-

uations such as this undermine the efﬁcacy at provid-

ing readable overviews. To alleviate this problem, we

introduce adaptive font sizes for labels, as shown in

Fig. 7. The font sizes of the state labels have been in-

creased to make them more readable, while the edge

labels remain unchanged. This is a decision based on

the notion that the states themselves act as landmarks

and are important to obtain an overview, whereas the

edge labels are details that are more relevant upon

closer inspection. It is therefore a decision based on

Posterizing Diagrams with Top-Down Layout

935

init

KHr

travelKHrKHr

req_KH_in

KH_LN_7_KH_LN_5

KH_LN_5

Travel

Cleanup

KH_LN_1_KH_LN_0

KH_LN_0

Entry

Travel

Permissions

1: perm_in_L

KH_ST_0_req[trainNum] = false;

KH_LN_8_req[trainNum] = false

railContact(KH_LN_5,1)

railContact(KH_LN_0,1)

2: homeStation == 2

/ realTrack =

homeTrack

3: homeStation == 3

/ realTrack =

homeTrack

railNextStation(trainNum) == 2

/ targetTrack =

railNextTrack(trainNum)

railNextStation(trainNum) == 3

/ targetTrack =

railNextTrack(trainNum)

railNextStation(trainNum) == 4

/ targetTrack =

railNextTrack(trainNum)

railNextStation(trainNum) == 0

2: schedAbortedStation == 2

Figure 7: This excerpt of Fig. 2 demonstrates adaptive fonts.

domain knowledge for SCCharts models.

Adaptive font sizes are set automatically depend-

ing on the number of nodes in a layout. The most

important elements to retain an overview over a local

subdiagram are the labels of the elements. In large

layouts (due to many nodes being present) the labels

become small and more difﬁcult to read. We increase

the font sizes of these labels by a factor that scales

linearly in the number of nodes of the layout.

In SCCharts we differentiate between simple states

and macro states. Macro states contain inner be-

haviour and therefore these nodes often have children.

Because these nodes are larger and more important,

we scale their fonts up at twice the rate compared to

the simple states. In Fig. 7 the init state is a simple

state and its font is slightly smaller than the fonts of

the two macro states in the image.

4 DISCUSSION

We have highlighted the key goal of readability and

in summary our tools achieve this using two things.

Firstly, we purposefully use size and scale to make

graph elements more visible if they are important.

Secondly, we exercise control over the layout to min-

imize whitespace. This interacts with the scaling

as additional whitespace in a ﬁxed space decreases

the available space for displaying interesting things.

Whitespace minimization leads to efﬁcient utilization

of the prescribed drawing area, which in turn leads

less downscaling being necessary, and therefore in-

creasing the readability of the diagram.

We identiﬁed labels in graphs to be key elements

as they are readable features that provide context and

detailed information about the graph elements they

are attached to. Understanding a labelled graph re-

quires being able to read the labels. Labels also have

a large impact on the layout of a graph. We hence put

a lot of focus on the readability of labels and on the

control and display of labels.

It is evident that drawing large labelled graphs

readably in their entirety is a difﬁcult undertaking.

Performing a bottom-up layout of a large graph, as

seen in Fig. 3, imposes an upper limit on the graph

size after which the diagrams inevitably become un-

readable. Dynamic visualizations and ﬁltering out

portions of the diagrams are always viable techniques,

but they are compromises as they hide or distort in-

formation. Top-down layout on the other hand aims

to show the entire graph structure in a readable form.

Additionally, when viewing a top-down layout on

a computer screen, zooming allows the total explo-

ration of a graph in what we would argue is an eas-

ier manner than panning and zooming in a bottom-up

layout.

So far, we have only looked at node-link diagrams

that are drawn using a layered approach. Some ideas

transfer well to other types of graph drawing prob-

lems or other types of diagrams and others are more

restricted in their general applicability. Layered graph

wrapping and layer splitting are very speciﬁc strate-

gies for optimizing the area occupied by layered lay-

outs, but in general they are techniques to ﬁll pre-

scribed areas, which reduces whitespace. Whitespace

minimization is a general goal in many graph drawing

applications. The notion of model order to control the

topology of a diagram is also more broadly applicable

as long as there is an inherent ordering in the underly-

ing model and a graph drawing algorithm that is able

to leverage that ordering. Both fuzzy label wrapping

and adaptive font sizes are label management tools

that can be applied to any type of labelled graph.

We focused on an application where the graph

comes from a model made by a developer, and we had

the goal of producing a diagram that can be printed on

a poster in a readable way. There are of course many

other types of diagrams and graphs from a wide range

of domains. In particular, graphs from the area of data

visualization may come with unique challenges and

requirements that are not captured by our example,

because they are not manually modelled. We think

the ideas presented in this paper are nonetheless help-

ful and can be partially applied to other applications.

5 CONCLUSION AND OUTLOOK

We showed how top-down layout can be used to pro-

duce diagrams of large models that are easier to read

at a high-level than corresponding diagrams drawn

using a bottom-up approach. In particular, bottom-

up layouts do not allow us to easily exercise control

IVAPP 2025 - 16th International Conference on Information Visualization Theory and Applications

936

over the total area required by a layout, and for large

diagrams we obtain large and hard-to-read drawings

where details and high-level elements blend together.

This makes it quite unsuitable for producing large

printable diagrams, whereas top-down layout can pro-

duce promising aesthetic results. We applied layered

graph wrapping and semantic soft wrapping for la-

bels for drawing graphs in prescribed areas. In ad-

dition, we introduced layer splitting, fuzzy soft wrap-

ping, and adaptive font sizes as new techniques to fur-

ther improve the results, and we leveraged the abil-

ity to sacriﬁce layout speed for better layout results.

Searching for optimal option settings increased our

layout times by a factor of four.

While this work introduces the main concepts of

posterizing diagrams, we plan to evaluate the effect

of the new techniques on the whitespace quantity and

readability of diagrams. We also plan to develop

better techniques for automatically determining good

size constraints to set on nodes for top-down layout

and to evaluate the usability of top-down diagrams.

Especially the question of how humans who work

with diagrams use them and whether top-down lay-

outs beneﬁt their use cases is an open question that

requires further investigation.

REFERENCES

De Carlo, G., Langer, P., and Bork, D. (2022). Advanced vi-

sualization and interaction in GLSP-based web mod-

eling: Realizing semantic zoom and off-screen ele-

ments. In Proceedings of the 25th International Con-

ference on Model Driven Engineering Languages and

Systems, MODELS ’22, pages 221–231, New York,

NY, USA. Association for Computing Machinery.

Domr

os, S., Riepe, M., and von Hanxleden, R. (2023a).

Model order in Sugiyama layouts. In Proceedings of

the 18th International Joint Conference on Computer

Vision, Imaging and Computer Graphics Theory and

Applications (VISIGRAPP 2023) - Volume 3: IVAPP,

pages 77–88. INSTICC, SciTePress.

Domr

os, S., von Hanxleden, R., Sp

onemann, M., R

uegg,

U., and Schulze, C. D. (2023b). The Eclipse Layout

Kernel. https://doi.org/10.48550/arXiv.2311.00533.

Formella, A. and Keller, J. (1996). Generalized ﬁsh-

eye views of graphs. In Brandenburg, F. J., editor,

Graph Drawing, pages 242–253, Berlin, Heidelberg.

Springer Berlin Heidelberg.

Frisch, M., Dachselt, R., and Br

uckmann, T. (2008).

Towards seamless semantic zooming techniques for

UML diagrams. In Proceedings of the 4th ACM Sym-

posium on Software Visualization, pages 207–208.

Glaser, P.-L. and Bork, D. (2021). The bigER tool - hy-

brid textual and graphical modeling of entity relation-

ships in VS Code. In 2021 IEEE 25th International

Enterprise Distributed Object Computing Workshop

(EDOCW), pages 337–340. IEEE.

Harel, D. (1987). Statecharts: A visual formalism for com-

plex systems. Science of Computer Programming,

8(3):231–274.

Kasperowski, M., Rentz, N., Domr

os, S., and von Hanxle-

den, R. (2024). KIELER: A text-ﬁrst framework

for automatic diagramming of complex systems. In

Proceedings of the 14th International Conference on

the Theory and Application of Diagrams, DIAGRAMS

’24, volume 14981 of LNCS, pages 402–418. Springer

Nature Switzerland”.

Kasperowski, M. and von Hanxleden, R. (2023).

Top-down drawings of compound graphs.

https://doi.org/10.48550/arXiv.2312.07319.

Lohstroh, M., Menard, C., Bateni, S., and Lee, E. A. (2021).

Toward a Lingua Franca for deterministic concurrent

systems. ACM Transactions on Embedded Computing

Systems (TECS), 20(4):Article 36.

Misue, K. (2024). Area-adaptive drawing of rooted trees.

In 2024 IEEE 17th Paciﬁc Visualization Conference

(PaciﬁcVis), pages 152–161.

Petre, M. (1995). Why looking isn’t always seeing: Read-

ership skills and graphical programming. Communi-

cations of the ACM, 38(6):33–44.

uegg, U. (2018). Sugiyama layouts for prescribed draw-

ing areas. Number 2018/1 in Kiel Computer Sci-

ence Series. Department of Computer Science. Disser-

tation, Faculty of Engineering, Christian-Albrechts-

Universit

at zu Kiel.

uegg, U. and von Hanxleden, R. (2018). Wrapping layered

graphs. In Proceedings of the 10th International Con-

ference on the Theory and Application of Diagrams

(DIAGRAMS ’18), pages 743–747. Springer.

Sarkar, M. and Brown, M. H. (1992). Graphical ﬁsh-

eye views of graphs. In Proceedings of the SIGCHI

Conference on Human Factors in Computing Systems,

pages 83–91. ACM.

Schulze, C. D. (2019). Text in Diagrams: Challenges to and

Opportunities of Automatic Layout. Number 2019/4 in

Kiel Computer Science Series. Department of Com-

puter Science, CAU Kiel. Dissertation, Faculty of En-

gineering, Kiel University.

Sugiyama, K., Tagawa, S., and Toda, M. (1981). Methods

for visual understanding of hierarchical system struc-

tures. IEEE Transactions on Systems, Man and Cy-

bernetics, 11(2):109–125.

Tominski, C., Abello, J., van Ham, F., and Schumann, H.

(2006). Fisheye tree views and lenses for graph visu-

alization. In Tenth International Conference on Infor-

mation Visualisation (IV’06), pages 17–24.

von Hanxleden, R., Duderstadt, B., Motika, C., Smyth, S.,

Mendler, M., Aguado, J., Mercer, S., and O’Brien,

O. (2014). SCCharts: Sequentially Constructive

Statecharts for safety-critical applications. In Proc.

ACM SIGPLAN Conference on Programming Lan-

guage Design and Implementation (PLDI ’14), pages

372–383, Edinburgh, UK. ACM.

Posterizing Diagrams with Top-Down Layout

937