def setUp(self):
self.ml_model = factories.MLModelFactory(name="test_estimator",
is_principal=True,
used_in_competitions=True)
candy = CandyStore(seasons=TIP_SEASON_RANGE)
fixtures = data_factories.fake_fixture_data(fixtures=candy.fixtures(
to_dict=None))
predictions = data_factories.fake_prediction_data(
match_data=candy.fixtures(to_dict=None))
match_results = data_factories.fake_match_results_data(
match_results=candy.match_results(to_dict=None))
self.fixture_data = [
fixtures.query("year == @season")
for season in range(*TIP_SEASON_RANGE)
]
self.prediction_data = [
predictions.query("year == @season")
for season in range(*TIP_SEASON_RANGE)
]
self.match_results_data = [
match_results.query("year == @season")
for season in range(*TIP_SEASON_RANGE)
]
self.api = api
def test_seasons_out_of_range():
current_year = date.today().year
# When more than two seasons are given, it raises an exception
seasons = (FIRST_AFL_SEASON, current_year + 1)
with pytest.raises(AssertionError, match=r"seasons must be in the range"):
CandyStore(seasons=(seasons[0] - 1, seasons[1]))
with pytest.raises(AssertionError, match=r"seasons must be in the range"):
CandyStore(seasons=(seasons[0], seasons[1] + 1))
def fake_fixture_data(
fixtures: Optional[pd.DataFrame] = None,
seasons: Union[Tuple[int, int], int] = 1,
) -> pd.DataFrame:
"""
Return minimally-valid data for fixture data.
These matches are usually unplayed, future matches, but it is also possible to get
data for past fixtures.
"""
fixtures = (
CandyStore(seasons=seasons).fixtures(to_dict=None)
if fixtures is None
else fixtures
)
return (
fixtures.rename(columns={"season": "year", "round": "round_number"}).drop(
"season_game", axis=1, errors="ignore"
)
# Recreates data cleaning performed in data_import
.assign(
date=lambda df: pd.to_datetime(df["date"], utc=True),
# Team name translations happen in augury's data pipeline
home_team=_translate_team_names("home"),
away_team=_translate_team_names("away"),
)
)
def fake_match_data(
match_results: Optional[pd.DataFrame] = None,
seasons: Union[Tuple[int, int], int] = 1,
) -> pd.DataFrame:
"""Return minimally-valid dummy match results data."""
match_results = (
CandyStore(seasons=seasons).match_results(to_dict=None)
if match_results is None
else match_results
)
return (
match_results.rename(
columns={
"season": "year",
"home_points": "home_score",
"away_points": "away_score",
}
)
# Recreates data cleaning performed in data_import
.assign(
date=lambda df: pd.to_datetime(df["date"], utc=True),
# Team name translations happen in augury's data pipeline
home_team=_translate_team_names("home"),
away_team=_translate_team_names("away"),
)
)
def test_tuple_season_count(tuple_seasons):
data = CandyStore(seasons=tuple_seasons).fixtures()
data_frame = pd.DataFrame(data)
first_season, last_season = tuple_seasons
# It generates seasons across the given season range
assert len(data_frame["season"].drop_duplicates()) == last_season - first_season
def test_too_long_tuple_seasons(tuple_seasons):
current_year = date.today().year
# When more than two seasons are given, it raises an exception
seasons = tuple(
sorted(tuple_seasons + (np.random.randint(FIRST_AFL_SEASON, current_year + 1),))
)
with pytest.raises(AssertionError, match=r"provide two seasons"):
CandyStore(seasons=seasons)
def setUp(self):
candy = CandyStore(seasons=MATCH_SEASON_RANGE)
fixtures = data_factories.fake_fixture_data(
fixtures=candy.fixtures(to_dict=None)
)
predictions = data_factories.fake_prediction_data(
fixtures=candy.fixtures(to_dict=None)
)
self.fixture_return_values = [
fixtures.query("year == @season") for season in MOCK_IMPORT_SEASONS
]
self.prediction_return_values = [
predictions.query("year == @season") for season in MOCK_IMPORT_SEASONS
]
self.mock_submitter = FootyTipsSubmitter()
self.mock_submitter.submit_tips = MagicMock()
self.api = api
def fake_match_results_data(
match_results: Optional[pd.DataFrame] = None,
seasons: Union[Tuple[int, int], int] = 1,
) -> pd.DataFrame:
"""Return minimally-valid dummy match results data."""
match_results = (CandyStore(seasons=seasons).match_results(
to_dict=None) if match_results is None else match_results)
return match_results.assign(date=lambda df: pd.to_datetime(df["date"]).map(
timezone.make_aware)).rename(
columns={
"home_points": "home_score",
"away_points": "away_score",
"season": "year",
})
def fake_fixture_data(
fixtures: Optional[pd.DataFrame] = None,
seasons: Union[Tuple[int, int], int] = 1) -> pd.DataFrame:
"""
Return minimally-valid data for fixture data.
These matches are usually unplayed, future matches, but it is also possible to get
data for past fixtures.
"""
fixtures = (CandyStore(seasons=seasons).fixtures(
to_dict=None) if fixtures is None else fixtures)
return (fixtures.rename(columns={
"season": "year",
"round": "round_number"
}).drop("season_game", axis=1, errors="ignore")
# Recreates data cleaning performed in views.fixtures
.assign(date=lambda df: pd.to_datetime(df["date"], utc=True)))
replace(".", "_").lower())
with localconverter(ro.default_converter + pandas2ri.converter):
fitzroy = packages.importr("fitzRoy")
fixtures = clean_column_names(fitzroy.get_fixture())
match_results = clean_column_names(fitzroy.get_match_results())
players = clean_column_names(
fitzroy.get_afltables_stats(start_date="2019-01-01",
end_date="2019-12-31"))
betting_odds = clean_column_names(
fitzroy.get_footywire_betting_odds(start_season="2019",
end_season="2019"))
data_factory = CandyStore(seasons=np.random.randint(1, 10))
@pytest.mark.parametrize(
["data_type", "fitzroy_data"],
[
("fixtures", fixtures),
("betting_odds", betting_odds),
("match_results", match_results),
("players", players),
],
)
def test_matching_column_names(data_type, fitzroy_data):
data = getattr(data_factory, data_type)(to_dict=None)
data_columns = set(data.columns)
def test_int_season_count(int_seasons):
data = CandyStore(seasons=int_seasons).fixtures()
data_frame = pd.DataFrame(data)
# It generates one season per requested season count
assert len(data_frame["season"].drop_duplicates()) == int_seasons
def test_unknown_seasons_format(tuple_seasons):
seasons = list(tuple_seasons)
with pytest.raises(TypeError, match=r"seasons argument must be"):
CandyStore(seasons=seasons)
def test_seasons_out_of_order(tuple_seasons):
seasons = tuple(reversed(tuple_seasons))
with pytest.raises(AssertionError, match=r"First season must be less"):
CandyStore(seasons=seasons)
def test_non_postive_seasons():
# When seasons is <= 0, it raises an exception
seasons = np.random.randint(-10, 1)
with pytest.raises(AssertionError, match=r"at least one season"):
CandyStore(seasons=seasons)
def data_factory():
seasons = np.random.randint(1, 10)
return CandyStore(seasons=seasons)