Hacktoberfest 2026：メンテナが10月に向けて印を付けた、オープンで初心者向けの issue。 Hacktoberfest の issue を見る

[Bug]: graph_diff reports "no changes" when a call is reversed (skill --update summary)

クローズ
  • 何か問題がありますか？
#4,067 コメント 1 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

まだ誰も着手していません。

  • #4170 @Mpasha17 による — マージされずにクローズ

評価

難易度
4/5
見積もり時間
3〜5日
初心者へのやさしさ
68/100
issue の種類
バグ
明瞭さ
明確に書かれている
活発さ
活発
技術スタック
python
領域
devtools

調査の方向性

Start at analyze.graph_diff and run the two-file reproduction from the issue, then inspect tools/skillgen/fragments/references/shared/update.md and the aider and devin core fragments. Check how the skill loads G_old, including paths.load_node_link_graph, and use the mentioned regression test as the guide. Done means a reversed call reports one new and one removed edge with correct source and target while no-edit and line-shift cases remain unchanged.

索引モデルが issue の本文から書いたものです。

説明

Pre-flight checks
  • I have checked the Troubleshooting section in the README
What happened?

analyze.graph_diff reports no changes when a call reverses (a calls b becomes b calls a). The skill's --update flow uses it for the "what changed" summary (tools/skillgen/fragments/references/shared/update.md, plus the aider and devin core fragments). A reversed call is therefore invisible to the user.

There are two causes:

  1. graph_diff keys undirected edges by sorted endpoint pair, so a→b and b→a produce the same key.
  2. The skill loads G_old with a plain json_graph.node_link_graph() call, which loses the arc-order direction (see #4066). So even a direction-aware diff would have nothing to compare.

Expected: a reversed call shows as one removed and one new edge, with the correct source/target.

Measured through the skill's own code on a two-file project:

Edit Shown on 0.9.75 Expected
no edit no changes no changes
call reversed no changes 1 new, 1 removed edge
lines inserted above functions no changes no changes
new call added 1 new edge 1 new edge

Low severity: it only affects the update summary, not the graph.

Steps to reproduce
python - <<'EOF'
import networkx as nx
from graphify.analyze import graph_diff

def g(src, tgt):
    G = nx.Graph()
    G.add_node("normalize", label="normalize()")
    G.add_node("tokenize", label="tokenize()")
    G.add_edge(src, tgt, relation="calls", _src=src, _tgt=tgt)
    return G

print(graph_diff(g("tokenize", "normalize"), g("normalize", "tokenize")))
EOF
Error output or graph output
{'new_nodes': [], 'removed_nodes': [], 'new_edges': [], 'removed_edges': [], 'summary': 'no changes'}
Graphify version

0.9.75 (v8 @ 48d7c0e)

Operating System

Linux

Python Version

3.12

Installation Method

built from source (git clone)

Additional Environment Details

No provider environment variables set. Reproduced on a clean checkout.

Additional context

Proposed fix (I'm happy to send a PR after #4066):

  • Count an edge as reversed only when both graphs carry _src/_tgt and the markers differ. In that case it appears in both new_edges and removed_edges.
  • Report source/target from the markers.
  • Have the skill fragments load G_old through paths.load_node_link_graph and regenerate them with skillgen.

Gating on markers keeps today's output for callers that don't stamp them, which matters for the proposed public graph_diff in #164. The naive fix (key every edge by _src/_tgt) gives false positives today, because G_old comes back without markers.

Prototyped with a regression test that fails on v8 and passes with the fix. The no-edit and line-shift cases above still report no changes.

AI disclosure: I investigated this and wrote the reproduction and prototype with Claude Code (Claude Opus 5.5), and verified them myself.

主要言語
Python
スター
124k
フォーク
11.9k
30日以内にマージされた PR はありません

環境構築

このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

Graphify-Labs/graphify のほかの issue

Graphify-Labs/graphify の issue をすべて見る

似ている issue

Python の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。