import pandas as pd
from statsmodels.tsa.tsatools import freq_to_period
from statsmodels.tsa.stattools import acf, pacf
from ..idsc import IDSC

import logging

from .visualiser import Visualiser


class Analyser():
    def __init__(self) -> None:
        logging.debug('Analyser init')
        self.idsc = IDSC()
        self.viz = Visualiser()

    def fit(self, data):
        self.data = data

    def profiling(self):
        # --------------------------------- #
        # Profiling data with IDSC profiler #
        # --------------------------------- #
        self.quantity_change_point = []
        self.quantity_predictability = []
        self.characteristic = None

        self.profile = self.idsc.profiling(
            self.data[['y']].rename(columns={'y': 'target'}).dropna().to_json())
        # Parse profile result
        self.intermittent_change_points_idx = self.profile['change_point_res']['inter_order_cpi']
        self.quantity_change_points_idx = self.profile['change_point_res']['order_quantity_cpi']

        # Set the change points to datetime index value
        self.intermittent_change_points = self.data.iloc[self.intermittent_change_points_idx].index
        self.quantity_change_points = self.data.iloc[self.quantity_change_points_idx].index

        self.characteristic = self.profile['classification_res']['time_series_class']['overall_characteristic']
        self.intermittent_predictability = [
            round(r['predictability'], 3)
            for r in
            self.profile['predictability_res']['predictability_result']['inter_order']
        ]
        self.quantity_predictability = [
            round(r['predictability'], 3)
            for r in
            self.profile['predictability_res']['predictability_result']['order_quantity']
        ]
        # ----- End of [Profiling data with IDSC profiler] ----- #

    def analyse(self, data: pd.DataFrame):
        '''
        data: pd.DataFrame, required
            Data must contains datetime index, and the index must be proper datetime index with freq value
            data should complete, no missing value and contain at least "y" columns as the time-series data itself
        '''
        logging.debug('Analysing ...')
        self.data = data
        self.freq: str = pd.infer_freq(data.index)
        self.period = freq_to_period(self.freq)

        self.acf = pd.DataFrame(
            [acf(data[col]) for col in data.columns],
            index=data.columns).T

        self.pacf = pd.DataFrame(
            [pacf(data[col]) for col in data.columns],
            index=data.columns).T

        self.viz.fit(data)

        return {
            'freq': self.freq,
            'period': self.period,
            'change_points': self.quantity_change_point,
            'predictability': self.quantity_predictability,
            'characteristic': self.characteristic
        }