def join(holdings_list, features_map, keyfun):
"""Join a list of holdings with a vector of arbitrary features.
This function joins a list of holdings with a features_map, where the join
key between the two mappings is provided by a configuration callable
'keyfun'. keyfun() is run on every holding, and the resulting key is looked
up from features_map.
All the values from features_map are first normalized to 100%, and when
aggregating, the 'market_value' attribute of the holdings is multiplied by
the normalized feature value.
Args:
holdings_list: A list of holdings.Holding instances.
features_map: A dict of some key to a dict of features. The values are
dicts of strings to a number.
keyfun: A function that produces the join key from a Holding instance.
The key is used to look up a feature vector from the features_map dict.
Returns:
A dict of labels (from the values of features_map) to a pair of: Decimal
number of market_value amounts, and a list of corresponding scaled
holdings.
"""
# Get the full list of features.
all_labels = set(label for features in features_map.values()
for label in features)
features_total = {label: D('0') for label in all_labels}
features_holdings = {label: [] for label in all_labels}
# Normalize the feature vectors.
norm_features_map = {
key: normalize_features(features)
for key, features in features_map.items()
}
# Accumulate the market value of each holding in buckets for each label.
for holding in holdings_list:
key = keyfun(holding)
if key is None:
logging.debug("Key not found: %s, %s, %s", holding.account,
holding.currency, holding.cost_currency)
try:
features = norm_features_map[key]
except KeyError as exc:
raise KeyError("Key '{}' not found in map {}".format(
key, norm_features_map))
for label, fraction in features.items():
if not holding.market_value:
continue
scaled_holding = holdings.scale_holding(holding, D(fraction))
features_total[label] += scaled_holding.market_value
features_holdings[label].append(scaled_holding)
return {
label: (features_total[label], features_holdings[label])
for label in all_labels
}
def test_scale_holding(self):
test_holding = holdings.Holding(
'Assets:US:Checking', D('100'), 'MSFT', D('54.34'), 'USD',
D('5434.00'), D('6000.00'), D('60'), datetime.date(2012, 5, 2))
expected_holding = holdings.Holding(
'Assets:US:Checking', D('70.0'), 'MSFT', D('54.34'), 'USD',
D('3803.80'), D('4200.00'), D('60'), datetime.date(2012, 5, 2))
self.assertEqual(expected_holding, holdings.scale_holding(test_holding, D('0.7')))