def test_parse_multiple(eat_fst_txt: str):
"""
Test that the parser will disallow more than one Foma FST in one file.
Foma will allow multiple, but we'll explicitly reject that.
"""
invalid_fst = eat_fst_txt * 2
with pytest.raises(FSTParseError):
parse_text(invalid_fst)
def test_parse_states() -> None:
"""
Ensure we parse all kinds of FST states/arcs.
"""
fst_text = """##foma-net 1.0##
##props##
2 390211 90019 390213 5 -1 1 2 2 1 0 2
##sigma##
0 @_EPSILON_SYMBOL_@
##states##
0 0 0 0
0 1
0 0 2
1 -1 -1 1
2 0 0 2 1
-1 -1 -1 -1 -1
##end##
"""
parse = parse_text(fst_text)
s0, s1, s2 = (StateID(i) for i in range(3))
assert parse.arcs == {
Arc(s0, Epsilon, Epsilon, s0),
Arc(s0, Epsilon, Epsilon, s1),
Arc(s0, Epsilon, Epsilon, s2),
Arc(s2, Epsilon, Epsilon, s2),
}
assert parse.accepting_states == {1, 2}
def test_parse_bad_symbols() -> None:
"""
Makes sure we crash if the FST tries to define a symbol entry twice.
"""
bad_fst = """##foma-net 1.0##
##props##
2 390211 90019 390213 5 -1 1 2 2 1 0 2
##sigma##
0 @_EPSILON_SYMBOL_@
3 @P.UN.ON@
0 +Err/Orth
##states##
-1 -1 -1 -1 -1
##end##
"""
with pytest.raises(FSTParseError):
parse_text(bad_fst)
def test_parse_simple(eat_fst_txt: str):
"""
Parse a simple FST WITHOUT flag diacritics.
"""
result = parse_text(eat_fst_txt)
assert len(result.sigma) == 15
multichar_symbols = set("+3P +Mass +N +Past +PastPart "
"+PresPart +Sg +V".split())
graphemes = set("eats eaten eating ate".replace(" ", ""))
assert len(multichar_symbols) + len(graphemes) == len(result.sigma)
assert stringified_set(result.multichar_symbols) == set(multichar_symbols)
assert stringified_set(result.graphemes) == graphemes
assert len(result.states) == 15
assert len(result.arcs) == 19
assert result.accepting_states == {14}
def test_parse_fst_with_flag_diacritics(english_flags_fst_txt: str) -> None:
"""
Parse a Foma FST with flag diacritics.
"""
result = parse_text(english_flags_fst_txt)
assert len(result.sigma) == 22
flag_diacritics = set("@C.UN@ @D.UN@ @P.UN.ON@".split())
multichar_symbols = set("+Adj +Inf +Pl +V UN+ ".split())
graphemes = set("a b d e i k l n o p r s u y".split())
assert len(multichar_symbols) + len(graphemes) + len(
flag_diacritics) == len(result.sigma)
assert stringified_set(result.multichar_symbols) == set(multichar_symbols)
assert stringified_set(result.flag_diacritics) == flag_diacritics
assert stringified_set(result.graphemes) == graphemes
assert len(result.states) == 21
assert len(result.arcs) == 27
assert result.accepting_states == {20}
def test_parse_symbols() -> None:
"""
Ensures we parse symbols properly
"""
parse = parse_text("""##foma-net 1.0##
##props##
2 390211 90019 390213 5 -1 1 2 2 1 0 2
##sigma##
0 @_EPSILON_SYMBOL_@
3 @P.UN.ON@
4 +Err/Orth
5 î
##states##
-1 -1 -1 -1 -1
##end##
""")
assert parse.has_epsilon
assert all(isinstance(sym, Symbol) for sym in parse.sigma.values())
assert stringified_set(parse.sigma) == {"@P.UN.ON@", "+Err/Orth", "î"}
def test_parse_whitespace_in_sigma() -> None:
"""
Ensures that whitespace within sigma is parsed correctly.
"""
result = parse_text("""##foma-net 1.0##
##props##
2 390211 90019 390213 5 -1 1 2 2 1 0 2
##sigma##
0 @_EPSILON_SYMBOL_@
1 @_UNKNOWN_SYMBOL_@
2 @_IDENTITY_SYMBOL_@
3 \u0020
4 \u00A0
5 \u00AD
##states##
-1 -1 -1 -1 -1
##end##
""")
assert len(result.sigma) == 3
assert stringified_set(result.graphemes) == {
' ', '\N{NO-BREAK SPACE}', '\N{SOFT HYPHEN}'
}
def test_parse_at_symbol():
# It used to throw "NotImplementedError"
parse_text(FST_WITH_AT_TEXT)
