Working group for API for networks

[gaborcsardi]

Maybe tidy networks, in the spirit of tidytext. It would be great to manipulate networks using some concise data-frame-like API, whatever that means.

To be honest I am not even sure how I would map a network with metadata to a data frame. Or maybe two data frames? But it would be great to discuss this with people working with network data.

[gaborcsardi]

For now I am putting this issue here to see if others are interested.

[ghost]

This definitely seems interesting. Would this be something of an R equivalent for Networkx?

[gaborcsardi]

Maybe, I don't know networkx well enough to tell.

[mdsumner]

I see this as a superset requirement for more general tools for geo-spatial. #12

I'm very interested to find ways of working with multiple tables, and generalizing away from specific type definitions of attributes on vertices. The topology of a data structure that is possible with networks is easily adapted to geo-spatial data, and I think it's a key topic. I'm very keen to have discussions about leveraging network tools for better handling of primitives-based topological spatial data, and I think the hierarchies of structure in GIS that aren't usually dealt with topologically are a useful use-case.

As an example I think there's value in more purely network approach to the tables of mesh data used here for polygons, where sets of parts using a shared vertex pool also have a table-based "object" front end:

http://rpubs.com/cyclemumner/spatial-normal-forms (especially under "Constrained polygon triangulation starts with line primitives")

This is also hosted here, but the WebGL plots have not survived the trip to GH static docs
https://r-gris.github.io/rangl/articles/spatial-normal-forms.html#constrained-polygon-triangulation-starts-with-line-primitives