tidygraphtool 0.1

A tidy API for network manipulation with Graph-tool inspired by tidygraph.

tidygraphtool

A tidy API for network manipulation with Graph-tool inspired by tidygraph.

Usage

from tidygraphtool.all import *
g = play_sbm(n_k=500)
(
  g >>
    activate("nodes") >>
    add_property("degree", centrality_degree, mode="total") >>
    filter_on("degree == 2") >>
    summary()
)

Features

• tidy and functional API for graph wranling with the graph-tool library.

Setting up conda environment

Graph-tool is best installed in a fresh environment.

conda create --name graph_tool_env &&
conda activate graph_tool_env &&
conda install -c conda-forge graph-tool &&
conda install -c conda-forge ipython jupyter pandas pandas-flavor &&
pip install matplotlib ipykernel 

# optional on jupyter notebook
# python -m ipykernel install --user --name=graph_tool_env

INSPIRATION

• graph_tool: tidygraph is thin wrapper of graph_tool (https://graph-tool.skewed.de/).
• networkx: we also draw inspiration from networkx.
• tidygraph: the tidy API that cast graph analysis as two dataframe (https://github.com/thomasp85/tidygraph).
• dplyr: verbs like API from which tidygraph draw inspiration .
• pyjanitor: general organisation of python data wranling with pandas (https://github.com/ericmjl/pyjanitor).
• thinc: functional API for deep learning (especially chaining and operator overloading; https://github.com/explosion/thinc).

MODELS TO INTEGRATE

• hSBM_Topicmodel: topic modeling based on graph_tool (https://github.com/martingerlach/hSBM_Topicmodel)
• bipartiteSBM: bipartite community detection based on graph_tool (https://github.com/junipertcy/bipartiteSBM)

PROTOYPE FUNCTIONAL INTERFACE

#### Current
g = gt_graph(nodes=nodes, edges=edges)
g = add_property(g, "node_coreness", node_coreness)
g = add_property(g, "pr", centrality_pagerank)
g = filter_on(g, "node_coreness > 3")
g = activate(g, "edges")
g = add_property(g, "edge_bet", centrality_edge_betweenness)

#### With #pipey, we can currently do
g = gt_graph(nodes=nodes, edges=edges)
g = (g >>
       add_property("node_coreness", node_coreness(g)) >>
       add_property("pr", centrality_pagerank(g)) >>
       filter_on("node_coreness > 3 & pr > 10") >>
       activate("edges") >>
       add_property("edge_bet", centrality_edge_betweenness(g)))