"""Defines preprocessing filters which can be applied by preprocessing.get_data.

Kyle Roth. 2019-02-12.
"""


import librosa
import numpy as np
from scipy import fft
from scipy import signal


# used by mel function
_mel = librosa.filters.mel(16000, 2048, n_mels=80)  # linear transformation, (80,1025)


def mel(X, y):
    """Mel spectrogram preprocessing."""
    out = []
    for x in X:
        spectrum = np.log(np.abs(fft(x))[:len(x)//2])
        spectrum = signal.resample(spectrum, 1025)
        x_mel = np.dot(_mel, spectrum)
        out.append(x_mel)
    X = np.array(out)
    y = np.array(y)

    # remove NaNs
    is_nan = np.isnan(X).any(axis=1)
    X, y = X[~is_nan], y[~is_nan]

    return X, y


def resample(data, y):
    """Resample audio to 800 points."""
    out = []
    for thing in data:
        out.append(signal.resample(thing, 800))
    return np.array(out), y


def get_filter(name):  # pylint: disable=too-many-return-statements
    """Get the preprocessor specified by the name given.

    Args:
        name (str): name of model constructor to be called.
    Returns:
        (callable): model constructor to be called.
    """
    name = name.lower()
    if name == 'mel':
        return mel
    if name == 'resample':
        return resample
    raise ValueError("filter must be one of {mel, resample}")