def test_nan_insertion(self):
'''when the external data source returns a subset of the requested dates
NaNs are inserted into the database for the dates not returned.
'''
connection = SQLiteTimeseries.connect(':memory:')
driver = SQLiteTimeseries(connection=connection,
table='price',
metric='Adj Close')
symbol = 'MSFT'
missing_date = datetime.datetime(2012, 12, 4, tzinfo=pytz.UTC)
missing_dates = {missing_date}
returned_dates = {
datetime.datetime(2012, 12, 1, tzinfo=pytz.UTC),
datetime.datetime(2012, 12, 2, tzinfo=pytz.UTC),
datetime.datetime(2012, 12, 3, tzinfo=pytz.UTC),
}
requested_dates = missing_dates | returned_dates
mock_yahoo = mock.Mock()
mock_yahoo.return_value = ((date, 10.) for date in returned_dates)
cache = FinancialDataTimeSeriesCache(gets_data=mock_yahoo,
database=driver)
cached_values = list(
cache.get(symbol=symbol, dates=list(requested_dates)))
db_val = connection.execute(
"SELECT value FROM price WHERE date = '{}'".format(
missing_date)).next()
self.assertEqual(db_val['value'], 'NaN')
cache_value_dict = {date: value for date, value in cached_values}
assert np.isnan(cache_value_dict[missing_date])
def test_sqlite(self):
cache = FinancialDataTimeSeriesCache.build_sqlite_price_cache(sqlite_file_path=':memory:',
table='price',
metric='Adj Close')
self.run_load_from_cache_yahoo(cache=cache)
cache = FinancialDataTimeSeriesCache.build_sqlite_price_cache(sqlite_file_path=':memory:',
table='price',
metric='Adj Close')
self.run_load_from_cache_multiple_tickers(cache=cache)
def test_load_from_cache(self):
cache = FinancialDataTimeSeriesCache(gets_data=None, database=None)
test_date, test_price = datetime.datetime(2012, 12, 3, tzinfo=pytz.UTC), 100
cache.get = lambda *args, **kwargs : [(test_date, test_price),
(datetime.datetime(2012, 12, 4, tzinfo=pytz.UTC), 101),
(datetime.datetime(2012, 12, 5, tzinfo=pytz.UTC), 102),
]
symbol = 'ABC'
df = cache.load_from_cache(start=datetime.datetime(2012, 11, 30, tzinfo=pytz.UTC),
end=datetime.datetime(2013, 1, 1, tzinfo=pytz.UTC),
stocks=[symbol])
self.assertIsInstance(df, pd.DataFrame)
self.assertIn(symbol, df.keys())
self.assertEqual(df[symbol][test_date], test_price)
def sqlite_price_cache(db_file_path=DEFAULT_PRICE_PATH):
'''Return a cache that persists prices downloaded from yahoo.
'''
return FinancialDataTimeSeriesCache.build_sqlite_price_cache(sqlite_file_path=db_file_path,
table='prices',
metric='Adj Close')
def mongo_price_cache(mongo_host='localhost', mongo_port=27017):
client = pymongo.MongoClient(mongo_host, mongo_port)
collection = client.prices.prices
db = MongoTimeseries(mongo_collection=collection, metric='price')
cache = FinancialDataTimeSeriesCache(gets_data=get_prices_from_yahoo,
database=db)
return cache
def test_indexes(self):
cache = FinancialDataTimeSeriesCache.build_sqlite_price_cache(sqlite_file_path=':memory:',
table='price',
metric='Adj Close')
df = cache.load_from_cache(indexes={'SPX' : '^GSPC'},
start=datetime.datetime(2012, 12, 1, tzinfo=pytz.UTC),
end=datetime.datetime(2012, 12, 31, tzinfo=pytz.UTC))
self.assertEqual(df['SPX'][datetime.datetime(2012, 12, 3, tzinfo=pytz.UTC)], 1409.46)
def run_load_from_cache_multiple_tickers(self, cache):
cache = FinancialDataTimeSeriesCache.build_sqlite_price_cache(sqlite_file_path=':memory:',
table='price',
metric='Adj Close')
symbols = ['GOOG', 'AAPL']
df = cache.load_from_cache(stocks=symbols,
start=datetime.datetime(2012, 12, 1, tzinfo=pytz.UTC),
end=datetime.datetime(2012, 12, 31, tzinfo=pytz.UTC))
self.assertEqual(df['GOOG'][datetime.datetime(2012, 12, 3, tzinfo=pytz.UTC)], 695.25)
self.assertEqual(df['AAPL'][datetime.datetime(2012, 12, 31, tzinfo=pytz.UTC)], 522.16)
def test_load_from_cache(self):
cache = FinancialDataTimeSeriesCache(gets_data=None, database=None)
test_date, test_price = datetime.datetime(2012, 12, 3,
tzinfo=pytz.UTC), 100
cache.get = lambda *args, **kwargs: [
(test_date, test_price),
(datetime.datetime(2012, 12, 4, tzinfo=pytz.UTC), 101),
(datetime.datetime(2012, 12, 5, tzinfo=pytz.UTC), 102),
]
symbol = 'ABC'
df = cache.load_from_cache(start=datetime.datetime(2012,
11,
30,
tzinfo=pytz.UTC),
end=datetime.datetime(2013,
1,
1,
tzinfo=pytz.UTC),
stocks=[symbol])
self.assertIsInstance(df, pd.DataFrame)
self.assertIn(symbol, df.keys())
self.assertEqual(df[symbol][test_date], test_price)
def test_nan_insertion(self):
'''when the external data source returns a subset of the requested dates
NaNs are inserted into the database for the dates not returned.
'''
connection = SQLiteTimeseries.connect(':memory:')
driver = SQLiteTimeseries(connection=connection,
table='price',
metric='Adj Close')
symbol = 'MSFT'
missing_date = datetime.datetime(2012, 12, 4, tzinfo=pytz.UTC)
missing_dates = {missing_date}
returned_dates = {datetime.datetime(2012, 12, 1, tzinfo=pytz.UTC),
datetime.datetime(2012, 12, 2, tzinfo=pytz.UTC),
datetime.datetime(2012, 12, 3, tzinfo=pytz.UTC),
}
requested_dates = missing_dates | returned_dates
mock_yahoo = mock.Mock()
mock_yahoo.return_value = ((date, 10.) for date in returned_dates)
cache = FinancialDataTimeSeriesCache(gets_data=mock_yahoo, database=driver)
cached_values = list(cache.get(symbol=symbol, dates=list(requested_dates)))
db_val = connection.execute("SELECT value FROM price WHERE date = '{}'".format(missing_date)).next()
self.assertEqual(db_val['value'], 'NaN')
cache_value_dict = {date : value for date, value in cached_values}
assert np.isnan(cache_value_dict[missing_date])
def test_indexes(self):
cache = FinancialDataTimeSeriesCache.build_sqlite_price_cache(
sqlite_file_path=':memory:', table='price', metric='Adj Close')
df = cache.load_from_cache(indexes={'SPX': '^GSPC'},
start=datetime.datetime(2012,
12,
1,
tzinfo=pytz.UTC),
end=datetime.datetime(2012,
12,
31,
tzinfo=pytz.UTC))
self.assertEqual(
df['SPX'][datetime.datetime(2012, 12, 3, tzinfo=pytz.UTC)],
1409.46)
def run_load_from_cache_multiple_tickers(self, cache):
cache = FinancialDataTimeSeriesCache.build_sqlite_price_cache(
sqlite_file_path=':memory:', table='price', metric='Adj Close')
symbols = ['GOOG', 'AAPL']
df = cache.load_from_cache(stocks=symbols,
start=datetime.datetime(2012,
12,
1,
tzinfo=pytz.UTC),
end=datetime.datetime(2012,
12,
31,
tzinfo=pytz.UTC))
self.assertEqual(
df['GOOG'][datetime.datetime(2012, 12, 3, tzinfo=pytz.UTC)],
695.25)
self.assertEqual(
df['AAPL'][datetime.datetime(2012, 12, 31, tzinfo=pytz.UTC)],
522.16)
def _build_mongo_cache(self):
db_driver = MongoTimeseries(mongo_collection=self.collection,
metric='Adj Close')
cache = FinancialDataTimeSeriesCache(
gets_data=prices.get_prices_from_yahoo, database=db_driver)
return cache