"""Tests for `graphify extract` CLI dispatch path in graphify.__main__.""" from __future__ import annotations import pytest import graphify.__main__ as mainmod def _make_corpus(tmp_path): """Minimal corpus: one Go code file + one Markdown doc. Both file types are needed so semantic extraction is requested (docs path triggers the LLM step we want to assert against). """ (tmp_path / "main.go").write_text("package main\nfunc main() {}\n") (tmp_path / "README.md").write_text("# Notes\nThe main function entry point.\n") return tmp_path def test_extract_exits_nonzero_when_all_semantic_chunks_fail( monkeypatch, tmp_path, capsys ): """When every semantic chunk errors (e.g. backend SDK not installed), the CLI must exit non-zero instead of silently writing an AST-only graph. The bug this guards: `pip install graphifyy` doesn't pull in `anthropic`, so `graphify extract --backend claude` would print per-chunk errors and still exit 0 with a graph.json. Callers checking exit status saw success. """ corpus = _make_corpus(tmp_path) out_dir = tmp_path / "out" # Stub the API-key check so the backend gate doesn't reject before we # reach the semantic-extraction step. monkeypatch.setenv("ANTHROPIC_API_KEY", "sk-test-fake-key") # Patch extract_corpus_parallel to simulate "all chunks failed": # return an empty merged accumulator without ever invoking on_chunk_done. # This matches the real behavior of extract_corpus_parallel when every # chunk raises (the per-chunk failures print to stderr and the loop # continues without calling the success callback). def _all_chunks_failed(paths, **kwargs): return { "nodes": [], "edges": [], "hyperedges": [], "input_tokens": 0, "output_tokens": 0, } monkeypatch.setattr( "graphify.llm.extract_corpus_parallel", _all_chunks_failed ) monkeypatch.setattr(mainmod, "_check_skill_version", lambda _: None) monkeypatch.setattr( mainmod.sys, "argv", ["graphify", "extract", str(corpus), "--backend", "claude", "--out", str(out_dir)], ) with pytest.raises(SystemExit) as exc_info: mainmod.main() assert exc_info.value.code == 1, ( f"expected exit code 1 when all semantic chunks fail, " f"got {exc_info.value.code}" ) stderr = capsys.readouterr().err assert "all semantic chunks failed" in stderr assert "claude" in stderr # No graph.json should have been written - the failure must abort before # the merge/cluster/write phase, not after. assert not (out_dir / "graphify-out" / "graph.json").exists(), ( "graph.json must not be written when semantic extraction fails" ) def test_extract_succeeds_when_at_least_one_chunk_completes( monkeypatch, tmp_path ): """Sanity counter-test: a successful chunk run keeps exit 0. Confirms the new guard only fires on the all-failed path, not on every extract.""" corpus = _make_corpus(tmp_path) out_dir = tmp_path / "out" monkeypatch.setenv("ANTHROPIC_API_KEY", "sk-test-fake-key") def _one_chunk_succeeded(paths, **kwargs): on_chunk = kwargs.get("on_chunk_done") if on_chunk: on_chunk(0, 1, {"nodes": [], "edges": [], "hyperedges": []}) return { "nodes": [], "edges": [], "hyperedges": [], "input_tokens": 100, "output_tokens": 50, } monkeypatch.setattr( "graphify.llm.extract_corpus_parallel", _one_chunk_succeeded ) monkeypatch.setattr(mainmod, "_check_skill_version", lambda _: None) monkeypatch.setattr( mainmod.sys, "argv", ["graphify", "extract", str(corpus), "--backend", "claude", "--out", str(out_dir)], ) # extract may still raise SystemExit at the end (clean exit code 0) # depending on platform; accept either no exception or SystemExit(0). try: mainmod.main() except SystemExit as exc: assert exc.code in (None, 0), f"unexpected exit code {exc.code}" # graph.json should exist on the happy path assert (out_dir / "graphify-out" / "graph.json").exists(), ( "graph.json must be written on the happy path" )