1 files changed, 210 insertions, 0 deletions
diff --git a/tests/test_mml_conformance.py b/tests/test_mml_conformance.py
new file mode 100644
index 0000000..0e38945
--- /dev/null
+++ b/tests/test_mml_conformance.py
@@ -0,0 +1,210 @@
+"""BNF conformance tests spanning lexer, parser, semantics, and MCP validate."""
+
+import pytest
+from hypothesis import given, seed, settings
+from hypothesis import strategies as st
+
+from mml_composemusic_mcp.ast_nodes import NoteStmt, TransposeStmt
+from mml_composemusic_mcp.ir import ErrorCode, NoteEvent
+from mml_composemusic_mcp.lexer import TokenType, tokenize
+from mml_composemusic_mcp.parser_ppmck import parse_ppmck
+from mml_composemusic_mcp.parser_pyxel import parse_pyxel
+from mml_composemusic_mcp.semantic_ppmck import analyze_ppmck
+from mml_composemusic_mcp.semantic_pyxel import analyze_pyxel
+from mml_composemusic_mcp.server import _parse_mml, compose_mml
+
+
+def parse_pipeline(source: str, mode: str):
+    tokens = tokenize(source, mode)
+    parser = parse_ppmck if mode == "ppmck" else parse_pyxel
+    analyzer = analyze_ppmck if mode == "ppmck" else analyze_pyxel
+    program, errors = parser(source, tokens)
+    if any(error.severity == "error" for error in errors):
+        return tokens, program, None, errors
+    sequence, semantic_errors = analyzer(source, program)
+    return tokens, program, sequence, errors + semantic_errors
+
+
+def error_codes(errors):
+    return {error.code for error in errors}
+
+
+@pytest.mark.parametrize(
+    ("mode", "source", "accidental", "length", "dots"),
+    [
+        ("ppmck", "A c+4.", 1, 4, 1),
+        ("ppmck", "A d#8", 1, 8, 0),
+        ("ppmck", "A e-16..", -1, 16, 2),
+        ("pyxel", "0: C+4.", 1, 4, 1),
+        ("pyxel", "0: D#8", 1, 8, 0),
+        ("pyxel", "0: E-16..", -1, 16, 2),
+    ],
+)
+def test_note_follows_bnf_accidental_before_length(
+    mode, source, accidental, length, dots
+):
+    _, program, _, errors = parse_pipeline(source, mode)
+    assert not any(error.severity == "error" for error in errors)
+    note = next(
+        stmt for stmt in program.tracks[0].statements if isinstance(stmt, NoteStmt)
+    )
+    assert (note.accidental, note.length, note.dots) == (
+        accidental,
+        length,
+        dots,
+    )
+
+
+@pytest.mark.parametrize(
+    ("mode", "source"),
+    [
+        ("ppmck", "A o l v q t c"),
+        ("pyxel", "0: O L V Q @ T K Y C"),
+        ("pyxel", "0: K+ Y- C"),
+    ],
+)
+def test_commands_requiring_numbers_reject_omission(mode, source):
+    _, _, _, errors = parse_pipeline(source, mode)
+    assert ErrorCode.SYNTAX_INVALID_NUMBER in error_codes(errors)
+
+
+@pytest.mark.parametrize(
+    ("mode", "source"), [("ppmck", "A c++4"), ("pyxel", "0: C##4")]
+)
+def test_multiple_accidentals_are_rejected(mode, source):
+    _, _, _, errors = parse_pipeline(source, mode)
+    assert ErrorCode.SYNTAX_UNEXPECTED_TOKEN in error_codes(errors)
+
+
+@pytest.mark.parametrize(("command", "value"), [("K-12", -12), ("K+7", 7)])
+def test_pyxel_signed_transpose_is_preserved(command, value):
+    _, program, sequence, errors = parse_pipeline(f"0: O4 {command} C", "pyxel")
+    assert not any(error.severity == "error" for error in errors)
+    transpose = next(
+        stmt for stmt in program.tracks[0].statements if isinstance(stmt, TransposeStmt)
+    )
+    note = next(
+        event
+        for event in sequence.channels["Pulse1"].events
+        if isinstance(event, NoteEvent)
+    )
+    assert transpose.value == value
+    assert note.note_number == 60 + value
+
+
+def test_pyxel_signed_detune_reaches_ir():
+    _, _, sequence, errors = parse_pipeline("0: Y-25 C", "pyxel")
+    assert not any(error.severity == "error" for error in errors)
+    note = next(
+        event
+        for event in sequence.channels["Pulse1"].events
+        if isinstance(event, NoteEvent)
+    )
+    assert note.detune_cents == -25.0
+
+
+@pytest.mark.parametrize("source", ["L c\nL d", "A c\nA d"])
+def test_ppmck_all_track_ids_reject_duplicates(source):
+    _, _, _, errors = parse_pipeline(source, "ppmck")
+    assert ErrorCode.SYNTAX_DUPLICATE_TRACK in error_codes(errors)
+
+
+@pytest.mark.parametrize(
+    "source",
+    ['#UNKNOWN "x"\nA c', "#TITLE unquoted\nA c", 'A c\n#TITLE "late"'],
+)
+def test_ppmck_header_key_and_placement_are_enforced(source):
+    _, _, _, errors = parse_pipeline(source, "ppmck")
+    assert ErrorCode.SYNTAX_UNEXPECTED_TOKEN in error_codes(errors)
+
+
+def test_lexer_reports_precise_token_position_and_eof():
+    tokens = tokenize("0: C\n  ?", "pyxel")
+    invalid = next(token for token in tokens if token.type == TokenType.INVALID)
+    assert (invalid.line, invalid.column, invalid.raw) == (2, 3, "?")
+    assert tokens[-1].type == TokenType.EOF
+
+
+@pytest.mark.parametrize("mode", ["ppmck", "pyxel"])
+def test_empty_and_unicode_inputs_do_not_crash_pipeline(mode):
+    sequence, errors = _parse_mml("", mode)
+    assert sequence is not None
+    assert isinstance(errors, list)
+    sequence, errors = _parse_mml("♬", mode)
+    assert sequence is None
+    assert ErrorCode.SYNTAX_INVALID_TOKEN in error_codes(errors)
+
+
+def test_validate_full_pyxel_path_preserves_warning_and_summary():
+    result = compose_mml(
+        action="validate",
+        mode="pyxel",
+        mml="0: T120 L4 O4 V100 Q80 @1 K-1 Y+5 @ENV1 100 10 C+4.",
+    )
+    assert result["valid"] is True
+    assert result["channel_summary"]
+    assert any(
+        warning["code"] == ErrorCode.SEMANTIC_UNSUPPORTED_FEATURE.value
+        for warning in result["warnings"]
+    )
+
+
+note_names = st.sampled_from(tuple("cdefgab"))
+accidentals = st.sampled_from(("", "+", "#", "-"))
+lengths = st.one_of(st.just(""), st.integers(1, 192).map(str))
+dots = st.integers(0, 3).map(lambda count: "." * count)
+spaces = st.sampled_from((" ", "  ", "\t", "\n"))
+
+
+@seed(20260712)
+@settings(max_examples=100, deadline=None)
+@given(
+    mode=st.sampled_from(("ppmck", "pyxel")),
+    notes=st.lists(
+        st.tuples(note_names, accidentals, lengths, dots), min_size=1, max_size=20
+    ),
+    separator=spaces,
+)
+def test_generated_valid_notes_round_trip_to_ast_and_ir(mode, notes, separator):
+    rendered = [
+        name + accidental + length + dot for name, accidental, length, dot in notes
+    ]
+    if mode == "pyxel":
+        rendered = [note.upper() for note in rendered]
+        source = "0: " + separator.join(rendered)
+    else:
+        source = "A " + separator.join(rendered)
+    _, program, sequence, errors = parse_pipeline(source, mode)
+    assert not any(error.severity == "error" for error in errors)
+    ast_notes = [
+        stmt for stmt in program.tracks[0].statements if isinstance(stmt, NoteStmt)
+    ]
+    ir_notes = [
+        event
+        for event in sequence.channels["Pulse1"].events
+        if isinstance(event, NoteEvent)
+    ]
+    assert len(ast_notes) == len(notes) == len(ir_notes)
+
+
+@seed(20260712)
+@settings(max_examples=50, deadline=None)
+@given(
+    mode=st.sampled_from(("ppmck", "pyxel")),
+    bad=st.sampled_from(("?", "{", "}", ",", "♬")),
+    index=st.integers(min_value=0, max_value=4),
+)
+def test_generated_single_token_corruption_is_not_silent(mode, bad, index):
+    note = "cdefg"[index]
+    source = f"A {note}{bad}" if mode == "ppmck" else f"0: {note.upper()}{bad}"
+    _, _, _, errors = parse_pipeline(source, mode)
+    assert any(error.severity == "error" for error in errors)
+
+
+@pytest.mark.parametrize(
+    ("mode", "source"),
+    [("ppmck", "A l" + "9" * 1000 + " c"), ("pyxel", "0: L" + "9" * 1000 + " C")],
+)
+def test_very_large_numbers_are_reported_without_crashing(mode, source):
+    _, _, _, errors = parse_pipeline(source, mode)
+    assert ErrorCode.SEMANTIC_VALUE_OUT_OF_RANGE in error_codes(errors)