ggnetwork about ggraph HOT 6 CLOSED

Competition is a good thing :-) I like how we both realised the need for a modular approach to building up network graphics instead of defining a single geom_network function with thousands of of different parameters. The main difference is really that I patched ggplot2 so I didn't need to flatten the network into a single data.frame which is really a bad abstraction for a network (especially big ones)

In your view what is the usage distribution between the igraph/graph/network packages?

from ggraph.

@sctyner and @heike have done a nice job at coding up a geom that gets the network plotted in a single call: https://github.com/sctyner/geomnet

… and there is the other approach that we both followed.

Patching ggplot2 is certainly preferable to my own "flattening" technique, although that technique appeared to be quite robust on the kind-of-large networks that I used during testing. I'm talking tens of thousands, but not millions, of nodes and edges.

In your view what is the usage distribution between the igraph/graph/network packages?

In my view, it's

1. igraph -- because the larger public seems to be using that package primarily, and
2. network -- because the statnet people running graph models are using it.

The two packages are complementary in view, despite the incompatibilities that can occur if you load the sna and igraph packages together (some functions have similar names).

I don't know graph despite having read quite a few blog posts and other things on manipulating networks in R, so I would remove it from the ranking entirely--it must be far, far away from the two others in terms of usage.

from ggraph.

Forgot to say that we have written a paper with the two aforementioned authors on plotting networks with R: https://github.com/sctyner/ggnet-paper

Also, my ggnetwork package was written as a proof-of-concept: unless I need to submit it for CRAN to get the paper published, I won't submit it and just leave it on GitHub, so feel free to take everything you need from it for your own package, if there is anything useful in there for you!

from ggraph.

I did not think about robustness too much when discussing a tabular format for networks, more the fact that you obscure a lot of information with it - I wouldn't want to derive paths for hierarchical edge bundles based on that format for instance:-)

graph is the base package for graph data in Bioconductor so unless you're working within computational biology (which I am) you've probably not come across it...

from ggraph.

Sorry, forgot closing the issue myself.

May I bug you with something else?

I have compiled that list of network analysis resources, but know close to nothing about Bioconductor and computational biology. If you have any suggestions for the R packages subsection of the list, or any other part of it really, please feel free to send links or to submit a PR.

(Naturally, ggraph is already in the list.)

from ggraph.

I'll have a look at it once time permits - skimmed over the list some time ago and made a mental note to come back once I had time to actually look into all the resources

from ggraph.