#!/usr/bin/env python
# encoding: utf-8

# The MIT License (MIT)

# Copyright (c) 2018 CNRS

# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:

# The above copyright notice and this permission notice shall be included in
# all copies or substantial portions of the Software.

# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.

# AUTHORS
# Hervé BREDIN - http://herve.niderb.fr

import numpy as np
from typing import Union
from typing import List
from typing import Text
from typing import Optional
from ..segment import Segment
from ..segment import SlidingWindow
from ..timeline import Timeline
from ..annotation import Annotation
from ..feature import SlidingWindowFeature
from .generators import string_generator


def one_hot_encoding(
    annotation: Annotation,
    support: Union[Segment, Timeline],
    window: Union[SlidingWindow, SlidingWindowFeature],
    labels: Optional[List[Text]] = None,
    mode: Text = "center",
) -> SlidingWindowFeature:
    """Convert annotation to one-hot-encoded numpy array

    Parameters
    ----------
    annotation : `pyannote.core.Annotation`
    support : `pyannote.core.Timeline` or `pyannote.core.Segment`
    window : `SlidingWindow`
        Use this `window`.
    labels : list, optional
        Predefined list of labels. Defaults to using labels in `annotation`.

    Returns
    -------
    y : SlidingWindowFeature
        (N, K) array where y[t, k] > 0 when labels[k] is active at timestep t.
        y[t, k] = -1 means we have no idea.

    See also
    --------
    See `one_hot_decoding` to convert `y` back to a `pyannote.core.Annotation`
    instance
    """

    if not isinstance(window, SlidingWindow):
        if hasattr(window, "sliding_window"):
            window = window.sliding_window
        else:
            msg = (
                f"`window` must be an instance of `SlidingWindow` "
                f"or have an attribute called 'sliding_window'."
            )
            raise TypeError(msg)

    if isinstance(support, Segment):
        extent = support
        support = Timeline([support])
    else:
        extent = support.extent()

    window = SlidingWindow(
        start=extent.start, step=window.step, duration=window.duration
    )

    n_samples = window.samples(extent.duration, mode=mode)

    # defaults to `labels` contained by `annotation`
    labels = annotation.labels() if labels is None else labels
    indices = {label: i for i, label in enumerate(labels)}

    # one-hot encoding
    # -1 = unknown / +1 = active / 0 = inactive
    y = -np.ones((n_samples, len(labels)), dtype=np.int8)
    for i, j in window.crop(support, mode=mode, return_ranges=True):
        i = max(0, min(n_samples, i))
        j = max(0, min(n_samples, j))
        y[i:j, :] = 0

    for label in annotation.labels():
        try:
            k = indices[label]
        except KeyError as e:
            msg = f'List of `labels` does not contain label "{label}".'
            print(indices.keys())
            raise ValueError(msg)

        for i, j in window.crop(
            annotation.label_timeline(label), mode=mode, return_ranges=True
        ):
            i = max(0, min(n_samples, i))
            j = max(0, min(n_samples, j))
            y[i:j, k] += 1

    y = np.minimum(y, 1, out=y)

    return SlidingWindowFeature(y, window, labels=labels)


def one_hot_decoding(y, window, labels=None):
    """Convert one-hot-encoded numpy array to annotation

    Parameters
    ----------
    y : (N, K) or (N, ) `numpy.ndarray`
        When y has shape (N, K), y[t, k] > 0 means kth label is active at
        timestep t. When y has shape (N, ), y[t] = 0 means no label is active
        at timestep t, y[t] = k means (k-1)th label is active.
    window : `SlidingWindow`
        Use this `window`.
    labels : list, optional
        Predefined list of labels.  Defaults to labels generated by
        `pyannote.core.utils.generators.string_generator`.

    Returns
    -------
    annotation : `pyannote.core.Annotation`

    See also
    --------
    `one_hot_encoding`
    """

    if not isinstance(window, SlidingWindow):
        if hasattr(window, "sliding_window"):
            window = window.sliding_window
        else:
            msg = (
                f"`window` must be an instance of `SlidingWindow` "
                f"or have an attribute called 'sliding_window'."
            )
            raise TypeError(msg)

    # if y has shape (N, ), convert it to (N, K) shape
    if len(y.shape) < 2:
        (N,) = y.shape

        # estimate the number of classes
        if labels is not None:
            K = len(labels)
        else:
            K = np.max(y)

        # convert to (N, K) shape
        # y[t, k] = 1 means kth class is active
        y_ = np.zeros((N, K), dtype=np.int64)
        for t, k in enumerate(y):
            if k > 0:
                y_[t, k - 1] = 1
        y = y_

    N, K = y.shape

    # generate label names when not provided
    if labels is None:
        labels = string_generator()
        labels = [next(labels) for _ in range(K)]

    annotation = Annotation()

    y_off = np.zeros((1, K), dtype=np.int64)
    y = np.vstack((y_off, y, y_off))
    diff = np.diff(y, axis=0)
    for k, label in enumerate(labels):
        for t in np.where(diff[:, k] != 0)[0]:
            if diff[t, k] > 0:
                onset_t = window[t].middle
            else:
                segment = Segment(onset_t, window[t].middle)
                annotation[segment, k] = label

    return annotation