Common

class sklego.common.TrainOnlyTransformerMixin

Bases: sklearn.base.TransformerMixin

Allows using a separate function for transforming train and test data

Usage:

>>> from sklearn.base import BaseEstimator
>>> class TrainOnlyTransformer(TrainOnlyTransformerMixin, BaseEstimator):
...     def fit(self, X, y):
...         super().fit(X, y)
...
...     def transform_train(self, X, y=None):
...          return X + np.random.normal(0, 1, size=X.shape)
...
>>> X_train, X_test = np.random.randn(100, 4), np.random.randn(100, 4)
>>> y_train, y_test = np.random.randn(100), np.random.randn(100)
>>>
>>> trf = TrainOnlyTransformer()
>>> trf.fit(X_train, y_train)
>>>
>>> assert np.all(trf.transform(X_train) != X_train)
>>> assert np.all(trf.transform(X_test) == X_test)

Warning

Transformers using this class as a mixin should at a minimum:

• call super().fit in their fit method

• implement transform_train()

They may also implement transform_test(). If it is not implemented, transform_test will simply return the untransformed dataframe

fit(X, y=None)

Calculates the hash of X_train

transform(X, y=None)

Dispatcher for transform method.

It will dispatch to self.transform_train if X is the same as X passed to fit, otherwise, it will dispatch to self.trainsform_test

transform_test(X, y=None)

transform_train(X, y=None)

sklego.common.as_list(val)

Helper function, always returns a list of the input value.

Parameters

val – the input value.

Returns

the input value as a list.

Example

>>> as_list('test')
['test']

>>> as_list(['test1', 'test2'])
['test1', 'test2']

sklego.common.expanding_list(list_to_extent, return_type=<class 'list'>)

Make a expanding list of lists by making tuples of the first element, the first 2 elements etc.

Parameters

• list_to_extent –

• return_type – type of the elements of the list (tuple or list)

Example

>>> expanding_list('test')
[['test']]

>>> expanding_list(['test1', 'test2', 'test3'])
[['test1'], ['test1', 'test2'], ['test1', 'test2', 'test3']]

>>> expanding_list(['test1', 'test2', 'test3'], tuple)
[('test1',), ('test1', 'test2'), ('test1', 'test2', 'test3')]

sklego.common.flatten(nested_iterable)

Helper function, returns an iterator of flattened values from an arbitrarily nested iterable

>>> list(flatten([['test1', 'test2'], ['a', 'b', ['c', 'd']]]))
['test1', 'test2', 'a', 'b', 'c', 'd']

>>> list(flatten(['test1', ['test2']]))
['test1', 'test2']

sklego.common.sliding_window(sequence, window_size, step_size)

Returns sliding window generator object from a sequence

Parameters

• sequence (Iterable) – e.g. a list

• window_size (int) – the size of each window

• step_size (int) – the amount of steps to the next window

Returns

a sliding window generator object

Return type

Generator

Example

>>> generator = sliding_window([1,2,4,5], 2, 1)
>>> [i for i in generator]
[[1,2], [2,4], [4,5], [5]]