#!/usr/bin/env python3 """Syntax-facts tests for the single Evidence ledger parser (ADR-0025). The Evidence ledger is now parsed once in ``ledger_syntax.parse_ledger``; G2 and G6 project their existing policy from its result. These tests pin the information-preserving facts the ADR requires: row order, field occurrence order, duplicate values, raw observed text, and the derived leading artifact token (with the currently accepted trailing punctuation). They also pin the G6 ``ledger_observed`` projection so the migration off the old independent parser stays behavior-compatible. """ from __future__ import annotations import sys import unittest from pathlib import Path # One import seam (ADR-0022): package root on sys.path once, then absolute # design_playbook.* imports below. _PKG_ROOT = Path(__file__).resolve().parents[2] if str(_PKG_ROOT) not in sys.path: sys.path.insert(0, str(_PKG_ROOT)) from design_playbook.mcp.evidence.ledger_syntax import ( # noqa: E402 FieldOccurrence, LedgerFacts, LedgerRow, parse_ledger, ) from design_playbook.scripts.g6_records import ledger_observed # noqa: E402 def _row(text: str) -> LedgerRow: """Parse a single-block ledger and return its one row.""" rows = parse_ledger(text).rows assert len(rows) == 1, f"expected one row, got {len(rows)}: {rows!r}" return rows[0] class RowOrderTests(unittest.TestCase): def test_multiple_rows_preserve_input_order(self): text = ( "criterion: L6.1\nobserved: evidence/a.png\nresult: pass\n\n" "criterion: L6.2\nobserved: evidence/b.png\nresult: pass\n\n" "criterion: L6.3\nobserved: evidence/c.png\nresult: pass\n" ) rows = parse_ledger(text).rows self.assertEqual( [r.values("criterion")[0] for r in rows], ["L6.1", "L6.2", "L6.3"], ) def test_blocks_without_evidence_fields_are_skipped(self): text = ( "## Evidence ledger\n\n" "Some prose intro with no fields.\n\n" "criterion: L6.1\nobserved: evidence/a.png\nresult: pass\n" ) rows = parse_ledger(text).rows self.assertEqual(len(rows), 1) self.assertEqual(rows[0].values("criterion")[0], "L6.1") class FieldOccurrenceOrderTests(unittest.TestCase): def test_interleaved_fields_preserve_occurrence_order(self): # criterion, observed, criterion again, result - not in canonical order text = ( "criterion: L6.1\n" "observed: evidence/a.png\n" "criterion: L6.1\n" "result: pass\n" ) row = _row(text) self.assertEqual( [oc.name for oc in row.occurrences], ["criterion", "observed", "criterion", "result"], ) def test_field_names_are_canonical_lowercase(self): text = "Criterion: L6.1\nObserved: evidence/a.png\nResult: Pass\n" row = _row(text) self.assertEqual( [oc.name for oc in row.occurrences], ["criterion", "observed", "result"], ) class DuplicateValuesTests(unittest.TestCase): def test_repeated_field_keeps_every_value_in_order(self): text = "criterion: L6.1\ncriterion: L6.2\nobserved: evidence/a.png\n" row = _row(text) self.assertEqual(row.values("criterion"), ("L6.1", "L6.2")) def test_repeated_observed_keeps_every_value(self): text = ( "criterion: L6.1\n" "observed: evidence/a.png\n" "observed: evidence/b.png\n" ) row = _row(text) self.assertEqual(row.values("observed"), ("evidence/a.png", "evidence/b.png")) def test_missing_field_yields_empty_tuple(self): row = _row("criterion: L6.1\nobserved: evidence/a.png\n") self.assertEqual(row.values("required"), ()) self.assertEqual(row.values("result"), ()) class RawObservedTextTests(unittest.TestCase): def test_raw_observed_keeps_trailing_commentary(self): text = "criterion: L6.1\nobserved: evidence/a.png viewer state shot\n" row = _row(text) self.assertEqual(row.raw_observed, "evidence/a.png viewer state shot") def test_raw_observed_is_first_occurrence_when_duplicated(self): text = ( "criterion: L6.1\n" "observed: evidence/first.png\n" "observed: evidence/second.png\n" ) row = _row(text) self.assertEqual(row.raw_observed, "evidence/first.png") def test_no_observed_yields_empty_raw(self): row = _row("criterion: L6.1\nresult: pass\n") self.assertEqual(row.raw_observed, "") def test_empty_observed_value_yields_empty_raw(self): row = _row("criterion: L6.1\nobserved:\nresult: pass\n") self.assertEqual(row.raw_observed, "") class ArtifactTokenTests(unittest.TestCase): def test_token_equals_raw_when_no_separator(self): row = _row("criterion: L6.1\nobserved: evidence/a.png\n") self.assertEqual(row.artifact_token, "evidence/a.png") def test_token_stops_at_whitespace(self): row = _row("criterion: L6.1\nobserved: evidence/a.png viewer shot\n") self.assertEqual(row.artifact_token, "evidence/a.png") def test_token_stops_at_half_width_paren(self): row = _row("criterion: L6.1\nobserved: evidence/a.png (shot)\n") self.assertEqual(row.artifact_token, "evidence/a.png") def test_token_stops_at_full_width_paren(self): row = _row("criterion: L6.1\nobserved: evidence/a.png（截图）\n") self.assertEqual(row.artifact_token, "evidence/a.png") def test_token_stops_at_half_width_comma(self): row = _row("criterion: L6.1\nobserved: evidence/a.png, shot\n") self.assertEqual(row.artifact_token, "evidence/a.png") def test_token_stops_at_full_width_comma(self): row = _row("criterion: L6.1\nobserved: evidence/a.png，截图\n") self.assertEqual(row.artifact_token, "evidence/a.png") def test_token_stops_at_half_width_colon(self): row = _row("criterion: L6.1\nobserved: evidence/a.png: shot\n") self.assertEqual(row.artifact_token, "evidence/a.png") def test_token_stops_at_full_width_colon(self): row = _row("criterion: L6.1\nobserved: evidence/a.png：截图\n") self.assertEqual(row.artifact_token, "evidence/a.png") def test_token_empty_when_no_observed(self): row = _row("criterion: L6.1\nresult: pass\n") self.assertEqual(row.artifact_token, "") def test_token_empty_when_observed_empty(self): row = _row("criterion: L6.1\nobserved:\nresult: pass\n") self.assertEqual(row.artifact_token, "") def test_observed_starting_with_separator_keeps_raw_as_token(self): # Matches the historical G6 derivation: when the raw value starts with # a tolerated separator, the leading-token match fails and the whole # raw value is the token (which then fails the evidence/ prefix check # at the consumer, so it is treated as free text). row = _row("criterion: L6.1\nobserved: (free text comment)\n") self.assertEqual(row.artifact_token, "(free text comment)") class LedgerFactsShapeTests(unittest.TestCase): def test_empty_text_yields_no_rows(self): self.assertEqual(parse_ledger("").rows, ()) def test_result_is_immutable_tuple(self): rows = parse_ledger("criterion: L6.1\nobserved: evidence/a.png\n").rows self.assertIsInstance(rows, tuple) self.assertIsInstance(rows[0].occurrences, tuple) def test_field_occurrence_is_frozen(self): oc = FieldOccurrence("criterion", "L6.1") with self.assertRaises(Exception): oc.name = "result" # type: ignore[misc] class G6ProjectionTests(unittest.TestCase): """``ledger_observed`` is the G6 projection over the new module. Pins the (criterion, artifact_token) pairs G6 consumes so the migration off the independent parser stays behavior-compatible, including the tolerated trailing punctuation from the commentary fixture. """ def test_pairs_preserve_row_order_and_tokens(self): text = ( "criterion: L6.1\nobserved: evidence/a.png\nresult: pass\n\n" "criterion: L6.2\nobserved: evidence/b.png, shot\nresult: pass\n" ) self.assertEqual( ledger_observed(text), [("L6.1", "evidence/a.png"), ("L6.2", "evidence/b.png")], ) def test_commentary_separators_strip_to_token(self): text = ( "criterion: L6.3\n" "observed: evidence/L6.3-error.png（截图：错误态）\n" "result: pass\n\n" "criterion: L6.4\n" "observed: evidence/L6.4-viewer.png: viewer state screenshot\n" "result: pass\n\n" "criterion: L6.5\n" "observed: evidence/L6.5-retry.png, retry state screenshot\n" "result: pass\n" ) self.assertEqual( ledger_observed(text), [ ("L6.3", "evidence/L6.3-error.png"), ("L6.4", "evidence/L6.4-viewer.png"), ("L6.5", "evidence/L6.5-retry.png"), ], ) def test_row_without_observed_is_skipped(self): text = "criterion: L6.1\nresult: pass\n" self.assertEqual(ledger_observed(text), []) def test_row_with_empty_observed_is_skipped(self): text = "criterion: L6.1\nobserved:\nresult: pass\n" self.assertEqual(ledger_observed(text), []) def test_row_without_criterion_is_skipped(self): text = "observed: evidence/a.png\nresult: pass\n" self.assertEqual(ledger_observed(text), []) if __name__ == "__main__": unittest.main(verbosity=2)