DumiFlex
9fe33ab4b2
A reference now carries at most one pick index and one axis, in either order - $o.0.SHOES and $o.SHOES.0 name the same value. Accepting both is deliberate: $o.0 is already meaningful on its own, so appending .SHOES is composition rather than a second syntax, and rejecting it would make the natural spelling the broken one. Two mirrored regex alternatives, so .0.1 and .SHOES.BELTS reject by not matching their second segment and the tail stays literal text - no validation pass. Both tokenizers normalise to one (name, index, axis) triple immediately, so nothing downstream learns which spelling was used. Replaces the var_nondigit_accessor_is_text corpus case, which pinned the behaviour this changes, with cases for both orders and both nonsense forms.
61 lines
2.3 KiB
Python
61 lines
2.3 KiB
Python
"""`$name` accessor grammar: at most one pick index and one axis, either order.
|
|
|
|
Both orders name the same value. Accepting both is deliberate — `$outfit.0` is
|
|
already a meaningful thing on its own, so a user who wants that pick's shoe will
|
|
reach for `$outfit.0.SHOES`, and rejecting it would make the natural spelling
|
|
the broken one. The tokenizer normalises to one shape immediately, so nothing
|
|
downstream learns which spelling was written.
|
|
"""
|
|
import pytest
|
|
|
|
from engine.syntax.tokenize import tokenize_text
|
|
from engine.syntax.types import TokenKind, parse_var_reference
|
|
|
|
|
|
def _var_meta(text):
|
|
toks = [t for t in tokenize_text(text) if t.kind is TokenKind.VAR]
|
|
return toks[0].meta if toks else None
|
|
|
|
|
|
@pytest.mark.parametrize("text,expected", [
|
|
("$outfit", {"name": "outfit"}),
|
|
("$outfit.0", {"name": "outfit", "index": 0}),
|
|
("$outfit.SHOES", {"name": "outfit", "axis": "SHOES"}),
|
|
("$outfit.0.SHOES", {"name": "outfit", "index": 0, "axis": "SHOES"}),
|
|
("$outfit.SHOES.0", {"name": "outfit", "axis": "SHOES", "index": 0}),
|
|
])
|
|
def test_accepted_forms(text, expected):
|
|
assert _var_meta(text) == expected
|
|
|
|
|
|
def test_both_orders_parse_identically():
|
|
assert _var_meta("$outfit.0.SHOES") == _var_meta("$outfit.SHOES.0")
|
|
|
|
|
|
@pytest.mark.parametrize("text", ["$outfit.0.1", "$outfit.SHOES.BELTS"])
|
|
def test_nonsense_forms_do_not_get_both_accessors(text):
|
|
"""Two indices or two axes is not a reference. The grammar rejects it by
|
|
not matching the second segment, so the tail stays literal text rather than
|
|
being silently absorbed."""
|
|
meta = _var_meta(text)
|
|
assert meta is not None
|
|
assert not ("axis" in meta and "index" in meta)
|
|
|
|
|
|
@pytest.mark.parametrize("raw,expected", [
|
|
("outfit", ("outfit", None, None)),
|
|
("outfit.0", ("outfit", 0, None)),
|
|
("outfit.SHOES", ("outfit", None, "SHOES")),
|
|
("outfit.0.SHOES", ("outfit", 0, "SHOES")),
|
|
("outfit.SHOES.0", ("outfit", 0, "SHOES")),
|
|
])
|
|
def test_string_path_parser(raw, expected):
|
|
assert parse_var_reference(raw) == expected
|
|
|
|
|
|
@pytest.mark.parametrize("raw", ["outfit.0.1", "outfit.SHOES.BELTS", "1bad"])
|
|
def test_string_path_parser_passes_junk_through_whole(raw):
|
|
"""An unparseable name is returned intact with no accessors, so an unusual
|
|
ctx key still resolves as itself rather than being truncated."""
|
|
assert parse_var_reference(raw) == (raw, None, None)
|