def test_equality(self):
pc1 = PatternCreator()
pc2 = PatternCreator()
pc2.add("ETYPE", "hello")
assert pc1 == pc1
assert pc1 != "wrong type"
assert pc1 != pc2 # different columns
def test_errors(self):
pc = PatternCreator()
with pytest.raises(TypeError):
pc.add("ETYPE", 234324) # invalid type
pc.add("ETYPE", "hello")
with pytest.raises(ValueError):
pc.add("ETYPE", "hello", check_exists=True) # duplicate
with pytest.raises(ValueError):
pc.add("etype", [{"a": 1, "b": 2}]) # wrong contents
def test_row2raw(self):
# unsupported value_type - eval fails
with pytest.raises(NameError):
PatternCreator.row2raw(
pd.Series(
{
"label": "et1",
"attribute_0": "TEXT",
"value_0": "aaa",
"value_type_0": "wrong_type",
"op_0": "",
}
)
)
# already the first token is invalid
with pytest.raises(ValueError):
PatternCreator.row2raw(
pd.Series(
{
"label": "et1",
"attribute_0": np.nan,
"value_0": "aaa",
"value_type_0": "wrong_type",
"op_0": "",
}
)
)
res = PatternCreator.row2raw(
pd.Series(
{
"label": "et1",
"attribute_0": "TEXT",
"value_0": "aaa",
"value_type_0": "str",
"op_0": "",
"attribute_1": np.nan,
"value_1": "bbb",
"value_type_1": "int",
"op_1": "!",
}
)
)
assert res == {"label": "et1", "pattern": [{"TEXT": "aaa"}]}
def test_raw2row(self):
# pattern not a list
with pytest.raises(TypeError):
PatternCreator.raw2row({"label": "ET1", "pattern": {"LOWER": "TEXT"}})
# label not a str
with pytest.raises(TypeError):
PatternCreator.raw2row({"label": 232, "pattern": [{"LOWER": "TEXT"}]})
# element not dictionary
with pytest.raises(TypeError):
PatternCreator.raw2row({"label": "etype", "pattern": [11]})
def test_to_df(self):
pc = PatternCreator()
pc.add("ET1", "hello")
pc.add("ET1", "there")
df_1 = pc.to_df()
df_2 = pc.to_df()
df_2.loc[0, "label"] = "REPLACED_LABEL"
df_3 = pc.to_df()
assert not df_1.equals(df_2)
assert df_1.equals(df_3)
def test_call(self):
pc = PatternCreator()
pc.add("new_entity_type", "tall")
text = "I saw a tall building."
doc = pc(text)
assert len(doc.ents) == 1
assert list(doc.ents)[0].label_ == "new_entity_type"
pc.drop(0)
doc2 = pc(text)
assert len(doc2.ents) == 0
def test_overall(self, tmpdir):
tmpdir_p = pathlib.Path(str(tmpdir)) / "patterns.json"
pc = PatternCreator()
assert len(pc.to_df()) == 0
pc.add("NEW_ENTITY_TYPE", "cake")
assert len(pc.to_df()) == 1
assert set(pc.to_df().columns) == {
"label",
"attribute_0",
"value_0",
"value_type_0",
"op_0",
}
pc.add("COOL_ENTITY_TYPE", {"LEMMA": "pancake", "OP": "*"})
assert len(pc.to_df()) == 2
pc.add("SOME_ENTITY_TYPE", [{"TEXT": "good"}, {"TEXT": "pizza"}])
assert len(pc.to_df()) == 3
assert set(pc.to_df().columns) == {
"label",
"attribute_0",
"value_0",
"value_type_0",
"op_0",
"attribute_1",
"value_1",
"value_type_1",
"op_1",
}
pc.to_jsonl(tmpdir_p)
pc_loaded = PatternCreator.from_jsonl(tmpdir_p)
pc_manual = PatternCreator(storage=pc.to_df())
assert pc == pc_loaded == pc_manual
def test_to_list(self):
pc = PatternCreator()
pc.add("ET1", "hello")
pc.add("ET2", {"TEXT": "there"})
pc.add("ET3", [{"TEXT": {"IN": ["world", "cake"]}}])
pc.add("ET4", [{"TEXT": {"IN": ["aa", "bbb"]}}, {"TEXT": {"REGEX": "^s"}}])
res = pc.to_list()
assert len(res) == 4
def test_drop(self):
pc = PatternCreator()
pc.add("ET1", "hello")
pc.add("ET1", "there")
pc.add("ET2", "world")
pc.add("ET4", "dog")
assert pc.to_df().index.to_list() == [0, 1, 2, 3]
pc.drop([1, 2])
assert pc.to_df().index.to_list() == [0, 1]
def test_raw2row2raw(self, raw):
assert raw == PatternCreator.row2raw(PatternCreator.raw2row(raw))