import numpy as np

import argparse

from train_test_split import load_data, save_data

def remove_sample(X, y, removing_indices: list):

mask = np.ones(X.shape[0], dtype=bool)

mask[removing_indices] = False

X_out = X[mask]

y_out = y[mask]

X_remain = X[~mask]

y_remain = y[~mask]

return X_out, y_out, X_remain, y_remain

if __name__ == '__main__':

parser = argparse.ArgumentParser()

parser.add_argument('data_path', type=str)

parser.add_argument('-if', '--input-fmt', type=str, default='libsvm')

parser.add_argument('-of', '--output-fmt', type=str, default='csv')

parser.add_argument('--scale-y', action='store_true')

parser.add_argument('-s', '--seed', type=int, default=0)

parser.add_argument('-r', '--remove_ratio', type=float, default=0)

args = parser.parse_args()

X, y = load_data(data_path=args.data_path, data_fmt=args.input_fmt, scale_y=args.scale_y, output_dense=True)

n_remove = int(args.remove_ratio * X.shape[0])

X_out, y_out, X_remain, y_remain = remove_sample(X, y, removing_indices=list(range(n_remove)))

print(f"Removed {n_remove} instances.")

save_data(X_out, y_out, save_path=f"{args.data_path}.remain_{args.remove_ratio:.0e}", save_fmt=args.output_fmt)

save_data(X_remain, y_remain, save_path=f"{args.data_path}.delete_{args.remove_ratio:.0e}", save_fmt=args.output_fmt)