split.py

from numpy import array
from pickle import dump

import pandas as pd

def load_csv(filename):
    datafile = pd.read_csv(filename)
    
    return datafile.close.tolist()


# save a list of sequences to file as a pkl
def save_sequence(sequence, filename):
    dump(sequence, open(filename, 'wb'))
    print('Saved: %s' % filename)

# Splits a sequence into samples. Each sample is `step_count` elements in length.
# The X is the sample, the y is the value that comes after the i-th sample
def split_into_samples(sequence, step_count):
    X_y = []

    for i in range(len(sequence)):
        # Find the end of this particular pattern (e.g. 5 elements per step)
        end_index = i + step_count

        if end_index > len(sequence) - 1:
            break

        seq_x, seq_y = sequence[i:end_index], sequence[end_index]

        # Append the X, y tuple to the array
        X_y.append((array(seq_x), seq_y))

    return array(X_y, dtype=object)

def main(in_file: "Sequence input in non-binary format (e.g. CSV)",
         out_file: "Out files title (X_y file will be saved as ____.pkl)",
         step_count: ("The number of steps per sub-sequence", 'option', 's') = 5):

    sequence = load_csv(in_file)

    X_y = split_into_samples(sequence=sequence, step_count=step_count)

    # print(X_y)

    save_sequence(X_y, out_file + '.pkl')

if __name__ == "__main__":
    import plac

    plac.call(main)