🚧 Add sqlite extraction

tests/integration/test_run_cycle.py

@@ -169,6 +169,132 @@ class TestRunCycle:
         assert "Station PS101/räumliche Datei.xml" in files
 
 
+class TestExtraction:
+    """Section 45: derived artifacts flow through the full collection."""
+
+    def _make_sqlite(self, git_repo: Path) -> Path:
+        import sqlite3
+
+        history = git_repo / "history"
+        history.mkdir(exist_ok=True)
+        db = history / "alarms.sqlite"
+        connection = sqlite3.connect(db)
+        connection.execute("CREATE TABLE alarms (id INTEGER PRIMARY KEY, text TEXT)")
+        connection.execute("INSERT INTO alarms VALUES (1, 'alpha')")
+        connection.commit()
+        connection.close()
+        return db
+
+    def _config_with_extraction(self, tracker_config: dict) -> dict:
+        tracker_config["extraction"] = {
+            "enabled": True,
+            "output_root": "extracted",
+            "format": "jsonl",
+            "sources": [{"path": "history/alarms.sqlite"}],
+        }
+        tracker_config["files"]["exclude"].append("**/*.sqlite")
+        return tracker_config
+
+    def test_first_run_commits_artifacts(
+        self, git_repo: Path, tracker_config: dict, tmp_path: Path
+    ) -> None:
+        self._make_sqlite(git_repo)
+        config = self._config_with_extraction(tracker_config)
+        outcome = _run(config, tmp_path)
+        # artifacts only: raw DB excluded, no other changes
+        assert outcome.change_count == 1
+        assert outcome.commit_hash is not None
+        body = _git_out("log", "-1", "--format=%b", cwd=git_repo)
+        assert "extracted/alarms/alarms.jsonl" in body
+        tracked = _git_out("ls-files", "-z", cwd=git_repo)
+        assert "alarms.sqlite" not in tracked  # raw DB excluded
+        content = (git_repo / "extracted" / "alarms" / "alarms.jsonl").read_text(
+            encoding="utf-8"
+        )
+        assert '{"id":1,"text":"alpha"}' in content
+
+    def test_unchanged_db_no_second_commit(
+        self, git_repo: Path, tracker_config: dict, tmp_path: Path
+    ) -> None:
+        self._make_sqlite(git_repo)
+        config = self._config_with_extraction(tracker_config)
+        _run(config, tmp_path)
+        outcome = _run(config, tmp_path)
+        assert outcome.change_count == 0  # FR-031: byte-identical artifacts
+        assert outcome.commit_hash is None
+
+    def test_changed_row_produces_artifact_commit(
+        self, git_repo: Path, tracker_config: dict, tmp_path: Path
+    ) -> None:
+        import sqlite3
+
+        db = self._make_sqlite(git_repo)
+        config = self._config_with_extraction(tracker_config)
+        _run(config, tmp_path)
+
+        connection = sqlite3.connect(db)
+        connection.execute("INSERT INTO alarms VALUES (2, 'beta')")
+        connection.commit()
+        connection.close()
+
+        outcome = _run(config, tmp_path)
+        assert outcome.change_count == 1
+        body = _git_out("log", "-1", "--format=%b", cwd=git_repo)
+        assert "M ; extracted/alarms/alarms.jsonl" in body
+
+    def test_failed_source_skipped_with_warning_others_extracted(
+        self, git_repo: Path, tracker_config: dict, tmp_path: Path
+    ) -> None:
+        """Section 45.7: per-source isolation - one bad file, others proceed."""
+        self._make_sqlite(git_repo)
+        # A second, non-database source alongside the valid one
+        (git_repo / "history" / "impostor.db").write_text(
+            "definitely not a database", encoding="utf-8"
+        )
+        config = self._config_with_extraction(tracker_config)
+        config["extraction"]["sources"].append({"path": "history/impostor.db"})
+        config["files"]["exclude"].append("history/impostor.db")  # raw file not tracked
+
+        outcome = _run(config, tmp_path)
+
+        # The valid source was still extracted and committed
+        assert outcome.change_count == 1
+        assert outcome.commit_hash is not None
+        body = _git_out("log", "-1", "--format=%b", cwd=git_repo)
+        assert "extracted/alarms/alarms.jsonl" in body
+        # Nothing was extracted for the bad source
+        assert not (git_repo / "extracted" / "impostor").exists()
+
+    def test_non_sqlite_file_not_committed_as_artifact(
+        self, git_repo: Path, tracker_config: dict, tmp_path: Path
+    ) -> None:
+        """A non-database source is skipped, never dumped or committed."""
+        history = git_repo / "history"
+        history.mkdir(exist_ok=True)
+        (git_repo / "history" / "notadb.sqlite").write_text(
+            "plain text wearing a database name", encoding="utf-8"
+        )
+        config = self._config_with_extraction(tracker_config)
+        config["extraction"]["sources"] = [{"path": "history/notadb.sqlite"}]
+
+        outcome = _run(config, tmp_path)
+        assert outcome.change_count == 0  # nothing extracted, nothing committed
+        assert not (git_repo / "extracted" / "notadb").exists()
+
+    def test_dry_run_skips_refresh(
+        self, git_repo: Path, tracker_config: dict, tmp_path: Path, caplog
+    ) -> None:
+        self._make_sqlite(git_repo)
+        config = self._config_with_extraction(tracker_config)
+        before = _git_out("status", "--porcelain", "-z", cwd=git_repo)
+        outcome = _run(config, tmp_path, dry_run=True)
+        assert outcome.dry_run
+        # Artifacts never created: the refresh is the step that writes them
+        assert (git_repo / "extracted").exists() is False  # tree untouched
+        after = _git_out("status", "--porcelain", "-z", cwd=git_repo)
+        assert before == after
+
+
 class TestSafetyAborts:
     def test_pre_existing_staged_changes_abort(
         self, git_repo: Path, tracker_config: dict, tmp_path: Path