jmansilla

import numpy as np

class DirectTransformer:
    """Utility for building class-like features from a single-point function, but that may need
    some general configuration first (you usually override __init__ for that)
    """

    def fit(self, X, y=None):
        return self

jmansilla

import numpy as np

class OneHotTransformer:
    def __init__(self, func):
        self.f = func

    def fit(self, X, y=None):
        unseen = object()
        seen = set()
        for x in X:

jmansilla

from sklearn.feature_extraction.text import CountVectorizer

class ProjectionCountVectorizer(CountVectorizer):

    def __init__(self, projection_path, *args, **kwargs):
        self.projection_path = projection_path.split('/')
        super().__init__(*args, **kwargs)

    def build_preprocessor(self):
        built = super().build_preprocessor()