Provide Gemini with the list of IDs in correction request

tests/test_correction_ids.py

@@ -0,0 +1,69 @@
+import json
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import Mock, patch
+
+from copienator import EvaluationWorkspace, atomic_write_json
+from copienator.commands import correction
+
+
+class CorrectionIdTests(unittest.TestCase):
+    def test_mismatched_ids_are_rejected_before_any_side_effects(self):
+        for ids, diagnostic in (
+            (["37", "03", "05"], "unexpected IDs: ['03']"),
+            (["37"], "missing IDs: ['05']"),
+            (["37", "05", "05"], "duplicate IDs: ['05']"),
+        ):
+            with self.subTest(ids=ids), tempfile.TemporaryDirectory() as directory:
+                root = Path(directory)
+                group = root / "Par label" / "Ex 1" / "Group_1.jpg"
+                group.parent.mkdir(parents=True)
+                group.touch()
+                atomic_write_json(group.with_suffix(".json"), [
+                    ["37", 0, 500, 1], ["05", 550, 1000, 1],
+                ])
+                pdf = root / "Copies" / "Copie37" / "Ex 1.pdf"
+                pdf.parent.mkdir(parents=True)
+                pdf.write_bytes(b"original")
+                workspace = EvaluationWorkspace(root)
+                prior = {"Other": [[{"id": "22", "result": {"score": 4}}]]}
+                atomic_write_json(workspace.correction_file, prior)
+                before = workspace.correction_file.read_bytes()
+                args = correction.build_parser().parse_args([str(root)])
+                correction.configure_runtime(workspace, [(str(group), "Ex 1")], args, api_client=Mock())
+                response = [{"id": pid, "result": {
+                    "score": 0, "error": "empty-answer", "feedback": [],
+                }} for pid in ids]
+                with patch.object(correction.prompting, "generate_request", return_value=([], {})), patch.object(
+                    correction, "handle_label_errors"
+                ) as handle:
+                    correction.process_single_task((str(group), "Ex 1"), json.dumps(response))
+                handle.assert_not_called()
+                self.assertEqual(pdf.read_bytes(), b"original")
+                self.assertFalse(pdf.with_name("Ex 1_old.pdf").exists())
+                self.assertEqual(workspace.correction_file.read_bytes(), before)
+                self.assertEqual(correction.completed_tasks, [])
+                self.assertEqual(correction.results, prior)
+                self.assertIn(diagnostic, correction.errors_summary[0][0])
+
+    def test_valid_ids_can_be_returned_in_different_order(self):
+        with tempfile.TemporaryDirectory() as directory:
+            root = Path(directory)
+            group = root / "Par label" / "Ex 1" / "Group_1.jpg"
+            group.parent.mkdir(parents=True)
+            group.touch()
+            atomic_write_json(group.with_suffix(".json"), [
+                ["37", 0, 500, 1], ["05", 550, 1000, 1],
+            ])
+            workspace = EvaluationWorkspace(root)
+            args = correction.build_parser().parse_args([str(root)])
+            correction.configure_runtime(workspace, [(str(group), "Ex 1")], args, api_client=Mock())
+            response = [{"id": pid, "result": {
+                "score": 4, "error": "", "feedback": [],
+            }} for pid in ["05", "37"]]
+            with patch.object(correction.prompting, "generate_request", return_value=([], {})):
+                correction.process_single_task((str(group), "Ex 1"), json.dumps(response))
+            self.assertEqual(correction.errors_summary, [])
+            self.assertEqual(correction.results["Ex 1"], [response])
+            self.assertEqual(correction.completed_tasks, [(str(group), "Ex 1")])