scicloj.zulipdata.graph — org.scicloj/zulipdata 0.1.1

Graph-shaped views of the corpus, backed by JGraphT.

The two graphs we routinely care about:

• channel co-membership — undirected weighted; nodes are channel names, edges weighted by number of shared users.
• user co-presence — undirected weighted; nodes are anonymized user-keys, edges weighted by number of shared channels.

Plus utilities for community detection (Girvan-Newman, Label Propagation), betweenness centrality, and conversion to kind/cytoscape and kind/graphviz shapes.

->cytoscape-elements clj

(->cytoscape-elements g
                      &
                      {:keys [node-attrs edge-attrs]
                       :or {node-attrs (constantly {})
                            edge-attrs (constantly {})}})

Convert a JGraphT graph to a :elements map for kind/cytoscape. node-attrs and edge-attrs are optional fns of the node / [u v weight] returning a map of extra attributes (merged into :data).

->dot clj

(->dot g
       &
       {:keys [directed node-label edge-label name]
        :or
          {directed true node-label str edge-label (constantly nil) name "G"}})

Render a JGraphT graph as Graphviz DOT source. directed chooses between digraph/graph. node-label and edge-label are optional fns producing label strings.

betweenness clj

(betweenness g)

Map node → betweenness centrality score.

channel-comembership-graph clj

(channel-comembership-graph timeline & {:keys [min-shared] :or {min-shared 1}})

Undirected weighted graph: nodes are channels, edges weighted by shared user count. min-shared filters out edges with fewer than N shared users.

girvan-newman clj

(girvan-newman g k)

Vector of node-sets, one per cluster. k is the desired number of clusters.

label-propagation clj

(label-propagation g)

Vector of node-sets — communities found by label propagation (number of clusters chosen by the algorithm).

migration-graph clj

(migration-graph timeline from-set & {:keys [min-users] :or {min-users 3}})

Directed weighted graph: edge from from-channel to to-channel weighted by the number of users who posted in from-channel and later (after their last post in any from-set channel) posted in to-channel. Excludes self-loops and edges within from-set.

Only users with at least 5 posts in from-set are considered.

user-channel-sets clj

(user-channel-sets timeline)

(user-channel-sets timeline min-channels)

Map of user-key → set of channels they posted in. Drops users with fewer than min-channels channels (default 1).

user-copresence-graph clj

(user-copresence-graph timeline
                       &
                       {:keys [min-shared min-channels]
                        :or {min-shared 3 min-channels 3}})

Undirected weighted graph: nodes are users, edges weighted by shared channel count. min-shared filters edges; min-channels filters users (active in ≥ N channels).