Grouping inheritance arrow ends

Question

Hello,

Large inheritance trees would be more readable with all arrow ends merged. It seems that the samehead attribute of dot edges could be used to achieve this.

Is it possible to set something like "samehead=inheritance" on every inheritance link in the generated dot? This could possibly be triggered by an option like "group inheritance".

Additionally, the concentrate attribute of dot would merge even part of the path of inheritance arrows. This could be enabled by the same "group inheritance" option.

 

Thanks,

Fred

Answer 1

Let's use the same thread.

We've put another beta (well, work is still in progress, so you should save the previous beta if you try this one).
https://dl.dropboxusercontent.com/u/13064071/plantuml.jar

We decide to disable by default grouping, since it really change the drawings.

@startuml
skinparam groupInheritance 3

A1 <|-- B1

A2 <|-- B2
A2 <|-- C2

A3 <|-- B3
A3 <|-- C3
A3 <|-- D3

A4 <|-- B4
A4 <|-- C4
A4 <|-- D4
A4 <|-- E4
@enduml

We are doing something really tricky here, so even if it's working with the example, it does not work (yet) in general case.

And even with this simple example, the drawing is somehow strange, but I'm afraid that it's the best we can have. Is it still ok for you ?

Comments

Hello,

This version is indeed nicer. Thanks!

By the way, it seems that setting sametail to the same value for the whole diagram works. It does not seem necessary to set sametail to a different value per superclass as you currently do.

Although this is useless, I tried setting groupInheritance to 1, and it crashes. It should probably silently behave like if it was set to 2.


By playing with svek.dot, one can try different things. Unfortunately, simply using dot's splines=ortho does not produce nice results (notably because it seems incompatible with sametail). Using one invisible node per superclass + splines=ortho (see http://stackoverflow.com/questions/2901233/complicated-graphviz-tree-structure#12406246) results in a nicer output, but this is not perfect yet.


Best regards,

Fred

---

After trying on a more complex diagram, which worked with the previous beta, I encountered the following exception:

java.lang.NullPointerException
        at net.sourceforge.plantuml.cucadiagram.dot.Neighborhood.drawU(Neighborhood.java:78)
        at net.sourceforge.plantuml.svek.EntityImageProtected.drawUntranslated(EntityImageProtected.java:77)
        at net.sourceforge.plantuml.svek.SvekResult.drawU(SvekResult.java:81)
        at net.sourceforge.plantuml.ugraphic.ImageBuilder.writeImageTOBEMOVED(ImageBuilder.java:125)
        at net.sourceforge.plantuml.ugraphic.ImageBuilder.writeImageTOBEMOVED(ImageBuilder.java:119)
        at net.sourceforge.plantuml.svek.CucaDiagramFileMakerSvek.createFileInternal(CucaDiagramFileMakerSvek.java:150)
        at net.sourceforge.plantuml.svek.CucaDiagramFileMakerSvek.createFile(CucaDiagramFileMakerSvek.java:99)
        at net.sourceforge.plantuml.cucadiagram.CucaDiagram.exportDiagramInternal(CucaDiagram.java:314)
        at net.sourceforge.plantuml.classdiagram.ClassDiagram.exportDiagramInternal(ClassDiagram.java:203)
        at net.sourceforge.plantuml.UmlDiagram.exportDiagram(UmlDiagram.java:228)
        at net.sourceforge.plantuml.PSystemUtils.exportDiagramsCuca(PSystemUtils.java:202)
        at net.sourceforge.plantuml.PSystemUtils.exportDiagrams(PSystemUtils.java:65)
        at net.sourceforge.plantuml.SourceFileReader.getGeneratedImages(SourceFileReader.java:130)
        at net.sourceforge.plantuml.Run.manageFileInternal(Run.java:394)
        at net.sourceforge.plantuml.Run.processArgs(Run.java:316)
        at net.sourceforge.plantuml.Run.manageAllFiles(Run.java:291)
        at net.sourceforge.plantuml.Run.main(Run.java:135)

---

Could you send us by mail the source diagram ?
Thanks

---

Hey guys.

Is this only possible on Generalization?

Best regards,
Timo

---

Hi,
Sorry, I am not sure to understand your point. Which "Generalization" ?
Could you give an example ?

Thanks!

---

Hi.
In your example you showed this workaround:

A2 <|-- B2
A2 <|-- C2
A2 <|-- D2

But what I'm looking for if the same reaction on this:

A2 *-- "0..1" B2
A2 *-- "0..1" C2
A2 *-- "0..1" D2

Or that:

A2 -- "0..1" B2
A2 -- "0..1" C2
A2 -- "0..1" D2

And that:

A2 *--> "0..1" B2
A2 *--> "0..1" C2
A2 *--> "0..1" D2

And so on.

Hope these examples are enough.

Best regards,
Timo

Answer 2

Hi,

Thanks for the idea : we did not know about the "sametail" feature.

We have made a beta here: https://dl.dropboxusercontent.com/u/13064071/plantuml.jar

This beta starts to group inheritance when there are 3 or more inheritance links.
Merging path inheritance is possible, but this need some works...

Example:
@startuml
A1 <|-- B1

A2 <|-- B2
A2 <|-- C2

A3 <|-- B3
A3 <|-- C3
A3 <|-- D3

A4 <|-- B4
A4 <|-- C4
A4 <|-- D4
A4 <|-- E4
@enduml

We may add a skinparam parameter (like groupInheritance) that would allow to change this hardcoded limit of 3.
If you have ideas of a better syntax, please post here!

Thanks again for your suggestion!
 

Comments

Hi,

Thanks, it works as expected.

I had not thought about setting a limit, and making it changeable, but this seems like a good idea. "groupInheritance 0" could disable grouping (if enabled by default).

My large class diagrams are a bit more readable with this feature.

They should be even more readable with inheritance path merging. Do you want me to start a new thread about that other feature, to make it easier to keep track of it?


Thanks,

Frédéric

---

I tested today: 

• skinparam groupInheritance 0 or 1 disables the feature (it is disabled by default)
• skinparam groupInheritance 2 enables merging of all inheritance links as here: http://www.plantuml.com/plantuml/uml/SoWkIImgAStDuIhEpimhI2nAp5L8ByelBV3CoqWjoYn9p4jELJ3Yud8qLB2fqTLLS384iesWR2Cuqncan1WgR0ndEYEOBiP05IPG5IPmfZE2wO9WkffmEQJcfG2T1G00
• skinparam groupInheritance 3 enables merging of 3 or more links going to the same parent
• etc

---

Is there a way to make the triangle a circle to use for Subtype/supertype relationships in ER diagrams?  And to change that circle so it has details like if a super type can have a direct instance or must be part of a subtype, and if multiple subtypes can apply to one instance, etc.?