[Bioc-devel] Classes to be supported by ggraph
Kasper Daniel Hansen
kasperdanielhansen at gmail.com
Sun Feb 7 03:57:41 CET 2016
>From my perspective Rgraphviz now has a quite good (cross-platform) access
to the Graphviz layout routines. But how to actually use these routines
from R is certainly something I think could be vastly improved upon. Right
now, Rgraphviz basically has two different interfaces to graph layout and
graph rendering - the "really, really old" interface and the "only old"
interface. As you may sense I don't have strong feelings about the R
interface to Rgraphviz, except that I would be very open to someone
improving on it. I don't have the time to do that myself. Basically what
I do right now is maintain the C interface, since I have spent so much time
hacking on it. But the R interface needs a lot of love. I think it would
be a shame to not continue to have access to the Graphviz library from R
given that it now works and it is extremely widely used.
For the ggraph package I think it would be worthwhile to also be able to
use the layout routines in (R)Graphviz, and I don't see why we would not be
able to plug into it.
How to display a general graph can get pretty burdensome wrt. arguments.
One of the original ideas in Rgraphviz was that a use would be able to put
a full R plot (say a scatterplot or histogram) inside each node.
On Sat, Feb 6, 2016 at 4:16 PM, Wolfgang Huber <whuber at embl.de> wrote:
> Dear Thomas
> thank you for asking about this.
> The main information in a graphNEL is, of course, the vector of nodes
> (character) and a list of edges. That information can for sure go into an
> Then there is the slot `renderInfo`, which afaIu is (or can be) used to
> control how the graph should be rendered when plotted e.g. on a 2D screen
> with Rgraphivz. The last big refactoring seems (according to svn blame) due
> to Seth Falcon, Nishant Gopalakrishnan and Florian Hahne in 2010/11. (The
> package has been worked on by a variety of of people over the years; e.g.
> my own involvement dates back to two bursts in 2004 and 07).
> It is hard to tell how much of the renderInfo functionality is used out
> there in the wild, but the fact that its code hasn’t been touched for 5
> years is consistent with assuming that interest in it is limited (otherwise
> there tend to be bug reports or requests for more functionality).
> Kapser and Florian are probably lurking on this list and waiting to make
> additional comments.
> In summary: I think your proposal is great, and you should go ahead. I
> expect the downsides, if any, to be miniscule.
> Hope this helps
> > On Feb 6, 2016, at 12:20 GMT+1, Thomas Lin Pedersen <thomasp85 at gmail.com>
> > I’m in the process of developing ggraph (
> https://github.com/thomasp85/ggraph) - An extensive plotting framework
> for graph/tree/network data based on ggplot2. Currently the class support
> is focused on igraph and dendrogram, but the idea is to extend it with
> support for additional graph representations in R. The easy way to do this
> is through conversion to one of the two currently supported classes, but I
> would like to hear from the developers and users of especially the graph
> package, whether there are any downsides to this approach, i.e. are there
> features/information captured by the graphNEL class that would disappear
> with conversion to an igraph object.
> > Also if someone is working with, or developing other graph
> representations and would like to see support in ggraph, please let me know…
> > best
> > Thomas
> > [[alternative HTML version deleted]]
> > _______________________________________________
> > Bioc-devel at r-project.org mailing list
> > https://stat.ethz.ch/mailman/listinfo/bioc-devel
> Wolfgang Huber
> Principal Investigator, EMBL Senior Scientist
> Genome Biology Unit
> European Molecular Biology Laboratory (EMBL)
> Heidelberg, Germany
> wolfgang.huber at embl.de
> Bioc-devel at r-project.org mailing list
[[alternative HTML version deleted]]
More information about the Bioc-devel