Naming scheme for algorithm functions about tidygraph HOT 12 CLOSED

Proposed ontology WIP

Nodes

• centrality
  • degree (degree)
  • alpha (alpha_centrality)
  • authority (authority_score)
  • betweenness (betweenness)
  • power (power_centrality)
  • closeness (closeness)
  • eigen (eigen_centrality)
  • hub (hub_score)
  • rank (page_rank)
  • subgraph (subgraph_centrality)
• bfs/dfs (all derived from bfs()/dfs() output - should be memorised?)
  • order
  • parent
  • before
  • after
  • distance
• group (all clustering functions)
  • component (components)
  • edge_betweenness (cluster_edge_betweenness)
  • fast_greedy (cluster_fast_greedy)
  • infomap (cluster_infomap)
  • label_prop (cluster_label_prop)
  • leading_eigen(cluster_leading_eigen)
  • louvain (cluster_louvain)
  • optimal (cluster_optimal)
  • springlass (cluster_springlass)
  • walktrap (cluster_walktrap)
• local
  • size (ego_size)
  • members (ego)
  • triangles (count_triangles)
  • transitivity (transitivity)
  • ave_degree (knn)

eccentricity
articulation_points, constraint, coreness, diversity
dominator_tree, topo_sort
scan

Node pairs
edge_connectivity, shortest_paths, simple_paths, cocitation, similarity

Edges

• centrality
  • betweenness (edge_betweenness)

from tidygraph.

Going through the igraph docs I was quite surprised by the almost complete lack of algorithms calculating edge properties — are there simply not that many edge properties out there..?

Idea:
all node properties should be calculate-able for edges by automatically computing on the line-graph

Need a good way to deal with node pair computations — both in terms of how they are called and how their input is returned. Should they be vectorised like e.g. pmin so that the return is still a vector of the same length as the input nodes, or should they simply calculate on all pairs returning a list of vectors?

There are remaining properties of nodes that lack a main term. Some are related to neighbourhood, some are related to different tree orderings. scan is kind of like a neighbourhood version of invoke... Inputs are appreciated

from tidygraph.

maybe degree, strength, and all neighbour related functions should be grouped together under a neighbour term? degree would then be neighbour_count...

from tidygraph.

The centrality subset has been implemented in b7712cf

from tidygraph.

bfs and bfs has been implemented in 3d8a6fe

from tidygraph.

group functions has been added in 7fb7382

from tidygraph.

Local measures added in 99cecbe (neighbourhood, knn, transitivity, etc)

from tidygraph.

Also needs wrappers for everything computing a statistic on the full graph

from tidygraph.

eccentricity, articulation (cut), constraint, coreness, diversity and a lot of other types (leaf, root, sink, source, etc) added in 494d7f9

from tidygraph.

Remember: max_cardinality, min_cut, and maximum_flow

from tidygraph.

graph stats have been implemented

from tidygraph.

I think everything of interest from igraph has been wrapped now so closing

from tidygraph.