from __future__ import absolute_import, division

import numpy as np
import pandas as pd

from .. import logging
from ..dataset import Davis
from .abstract import AbstractStorage


class LocalStorage(AbstractStorage):
    """ Local storage of the results.

    This class encapsulates the storage of the results into a pandas DataFrame.
    """

    def __init__(self):
        self.report = pd.DataFrame(columns=self.COLUMNS)
        logging.verbose('Report DataFrame created')
        self.annotated_frames = pd.DataFrame(
            columns=['sequence', 'scribble_idx', 'frame', 'override'])
        logging.verbose('Annotated frames created')

    def store_interactions_results(self, user_id, session_id, sequence,
                                   scribble_idx, interaction, timing,
                                   objects_idx, frames, jaccard, contour):
        """ The information of a single interaction is given and stored.

        # Arguments
            user_id: String. User identifier. As it is in local, this value is
                ignored.
            session_id: String. Session identifier.
            sequence: String. Sequence name.
            scribble_idx: Integer. Scribble index of the sample.
            interaction: Integer. Interaction number.
            timing: Float. Timing in seconds that lasted the interaction.
            objects_idx: List of Integers. List of the objects identifiers that
                match with the jaccard metric.
            frames: List of Integers: List of frame index matching with the
                jaccard metric.
            jaccard: List of Floats: List of jaccard metric.
            contour: List of Floats: List of contour metric.
        """
        # Check the data.
        objects_idx = list(objects_idx)
        frames = list(frames)
        jaccard = list(jaccard)
        contour = list(contour)
        assert len(jaccard) == len(contour)
        j_and_f = [.5 * j + .5 * f for j, f in zip(jaccard, contour)]
        if (min(jaccard) < 0.) or (max(jaccard) > 1.):
            raise ValueError('Jaccard values must be between 0 and 1')
        if (min(contour) < 0.) or (max(contour) > 1.):
            raise ValueError('Jaccard values must be between 0 and 1')

        nb = len(jaccard)
        if len(objects_idx) != nb or len(frames) != nb:
            raise ValueError('`jaccard`, `frames` and `objects_idx` must '
                             'have the same length')

        # Check previous entries
        if self.report.loc[(self.report.sequence == sequence) &
                           (self.report.scribble_idx == scribble_idx) &
                           (self.report.interaction == interaction)].size > 0:
            raise RuntimeError(('For {} and scribble {} already exist a '
                                'result for interaction {}').format(
                                    sequence, scribble_idx, interaction))
        if interaction > 1 and self.report.loc[
            (self.report.sequence == sequence) &
            (self.report.scribble_idx == scribble_idx) &
            (self.report.interaction == interaction - 1)].size == 0:
            raise RuntimeError(('For {} and scribble {} does not exist a '
                                'result for previous interaction {}').format(
                                    sequence, scribble_idx, interaction - 1))

        sample = {}
        sample['session_id'] = [session_id] * nb
        sample['sequence'] = [sequence] * nb
        sample['scribble_idx'] = [scribble_idx] * nb
        sample['interaction'] = [interaction] * nb
        sample['timing'] = [timing] * nb
        sample['object_id'] = objects_idx
        sample['frame'] = frames
        sample['jaccard'] = jaccard
        sample['contour'] = contour
        sample['j_and_f'] = j_and_f
        sample = pd.DataFrame(data=sample, columns=self.COLUMNS)
        self.report = pd.concat([self.report, sample], ignore_index=True)
        logging.info('Successfully stored sample interaction entry')

        return True

    def get_report(self, session_id=None, **kwargs):
        """ Return current report.

        # Arguments
            session_id: String. Session identifier

        # Returns
            Pandas DataFrame. Report in the form of the DataFrame.
        """
        df = self.report
        df = df.loc[df['session_id'] == session_id].reset_index(drop=True)
        return df

    def get_annotated_frames(self, session_id, sequence, scribble_idx):
        """Get the previous annotated frames for the given iteration.

        # Arguments
            session_id: String. Ignored.
            sequence: String. Sequence name.
            scribble_idx: Integer. Scribble index of the sample.

        # Returns
            List of Integers. List of the frames that have been previously
                annotated in the current iteration.
        """
        del session_id

        df = self.annotated_frames.copy()
        prev_frames = df.loc[
            (df['sequence'] == sequence) &
            (df['scribble_idx'] == scribble_idx)]['frame'].values
        prev_frames = np.unique(prev_frames)

        if len(prev_frames) == Davis.dataset[sequence]['num_frames']:
            return tuple()

        return tuple(prev_frames)

    def store_annotated_frame(self, session_id, sequence, scribble_idx,
                              annotated_frame, override):
        """ Get and store the frame to generate the scribble.

        This function will check all the previous generated scribbles frames
        and return the frame with lower metric that the robot hasn't generated
        a scribble.

        # Arguments
            session_id: String. Ignored.
            sequence: String. Sequence name.
            scribble_idx: Integer. Scribble index of the sample.
            annotated_frame: Integer. Index of the frame of the next scribble
                iteration.
            override: Boolean. Whether or not the annotated frame was override
                by the user or not.
        """
        del session_id

        new_row = pd.DataFrame(
            [[sequence, scribble_idx, annotated_frame, override]],
            columns=['sequence', 'scribble_idx', 'frame', 'override'])
        self.annotated_frames = pd.concat([self.annotated_frames, new_row],
                                          ignore_index=True)