OSSプロジェクトの健康診断

from pydriller import Repository
import pandas as pd

url = "https://github.com/ishepard/pydriller"

# コミットを1件ずつたどり、変更ファイルごとに1行を作る
rows = []
for commit in Repository(url).traverse_commits():
    for mod in commit.modified_files:
        path = mod.new_path or mod.old_path
        if path is None:
            continue
        rows.append({
            "commit": commit.hash[:8],
            "author": commit.author.email,
            "date": commit.author_date,
            "file": path,
            "added": mod.added_lines,
            "deleted": mod.deleted_lines,
        })

df = pd.DataFrame(rows)
print(df.shape)
print(df.head())

import networkx as nx
from itertools import combinations

# 同じコミットで一緒に変わったファイルを辺で結ぶ
G = nx.Graph()
for commit, grp in df.groupby("commit"):
    files = sorted(grp["file"].unique())
    for a, b in combinations(files, 2):
        if G.has_edge(a, b):
            G[a][b]["weight"] += 1
        else:
            G.add_edge(a, b, weight=1)

# 次数中心性と媒介中心性
degree = nx.degree_centrality(G)
between = nx.betweenness_centrality(G)

cent = pd.DataFrame({
    "file": list(degree.keys()),
    "degree": list(degree.values()),
    "betweenness": [between[f] for f in degree],
})
print("nodes:", G.number_of_nodes(), "edges:", G.number_of_edges())
print(cent.sort_values("betweenness", ascending=False).head())

import matplotlib.pyplot as plt

fig, ax = plt.subplots(1, 2, figsize=(11, 4))

# 左: ファイルごとの変更回数の分布（ロングテール）
ax[0].hist(feat["changes"], bins=30, color="#0E6F5C")
ax[0].set_xlabel("changes per file")
ax[0].set_ylabel("number of files")
ax[0].set_title("Change frequency")

# 右: 共変更ネットワークの中心部分（次数の高い上位30ノード）
top_nodes = [n for n, _ in sorted(G.degree,
             key=lambda x: x[1], reverse=True)[:30]]
H = G.subgraph(top_nodes)
pos = nx.spring_layout(H, seed=42)
nx.draw_networkx_nodes(H, pos, ax=ax[1], node_size=120, node_color="#0E6F5C")
nx.draw_networkx_edges(H, pos, ax=ax[1], edge_color="#B7C0BB")
ax[1].set_title("Co-change network (top 30)")
ax[1].axis("off")

plt.tight_layout()
plt.show()