[Networkit] Graph Drawing & Dropping the "viz" Module
meyerhenke at kit.edu
Wed May 20 17:17:21 CEST 2015
Strong general disagreement about dropping the graph drawing /
visualization code from my end. I agree that some refactoring is
necessary. We can talk about this later in more detail.
To avoid confusion:
- The Postscript-based visualization writes a given layout to a file
(can be convenient; if we have better options, *also from the C++ part*,
this part can be dropped)
- The actual drawing part (i.e. computing coordinates in the first
place) is what needs to stay.
Both things (visualization vs drawing) need to be distinguished.
Am 20.05.15 um 16:56 schrieb Christian Staudt:
> Hi NetworKit developers,
> as agreed, we are going to clean up the code base for the next release to focus more on a strong core of functionality.
> At this point graph drawing is not a strong selling point for NetworKit. We have already dropped the “viztools” module, a student project which contained some code for layout and drawing in Python. I also propose that we drop the code under networkit/cpp/viz from future releases. Some reasons:
> - the classes are not pythonized
> - the code is not release-ready in my opinion, e.g. it contains things like hard-coded paths for the output files
> - it implements postscript-based graphics from scratch, relatively much low-level code for relatively little functionality
> - it is one of 4 (!) rudimentary options in NetworKit for drawing graphs on the screen, which is probably confusing for users. We should get down to 2 proper ones (see below).
> Does any of that code implement a competitive method for graph drawing that other tools do not provide (better)? Are there other reasons for keeping the layout code in a release? If yes, it should be definitely refactored: Layout algorithms should return simply 2D-coordinates for nodes. Drawing the actual graphics is not NetworKit’s job and can be delegated completely to other tools:
> - NetworkX for small graph drawings within the IPython Notebook
> - Gephi for anything larger
> I recommend that we focus on a good interface to Gephi to support visual inspection of graphs.
Karlsruhe Institute of Technology (KIT)
Institute of Theoretical Informatics (ITI)
Juniorprof. Dr. Henning Meyerhenke
Theoret. Informatics / Parallel Computing
KIT - University of the State of Baden-Wuerttemberg and
National Research Center of the Helmholtz Association
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 5316 bytes
Desc: S/MIME Cryptographic Signature
More information about the NetworKit