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

# In[1]:


import pandas as pd  # pandas import for data manipulation
import numpy as np
import ipywidgets as widgets
# Jupyter import for iteractive widgets ()
from ipywidgets import interact, interact_manual, interactive, HBox, VBox, Layout
import matplotlib.pyplot as plt
from IPython.display import display

# import bokeh palette to color graphs
from bokeh.palettes import Category10_10 as palette
import itertools  # import itertool to loop over palette

# %matplotlib inline
raw_data = pd.read_csv('data.csv',
                       sep=',', skiprows=0, index_col=False)  # read data


# In[2]:


def wm(df):
    wm = sum(df['price']*df['results'])/sum(df['results'])
    return wm


def std(df):
    p_mean = wm(df)
    std = (sum(df['results']*(df['price']-(p_mean))**2) /
           (sum(df['results'])-1))**0.5
    return std


# In[3]:


def color_picker(models, model, colors):
    if len(models) <= len(palette):
        # give each model in the list sequential palette color
        current_color = palette[models.index(model)]
    else:
        # cycle through palette if more models than colors
        current_color = next(colors)
    return current_color


# In[4]:


def default_graph():
    fig, ax = plt.subplots(1, 1, figsize=(8, 6))
    ax.grid(True)
    ax.set_title('Trend', fontsize=16)
    ax.set_xlabel('Year', fontsize=16)
    ax.set_ylabel('Price', fontsize=16)
    ax.autoscale(tight=True)

    return fig, ax


def default_dist():
    fig, ax = plt.subplots(1, 1, figsize=(8, 6))
    ax.grid(True)
    ax.set_title('Distribution', fontsize=16)
    ax.set_xlabel('Price', fontsize=16)
    ax.set_ylabel('Offers', fontsize=16)
    ax.autoscale(tight=True)
    return fig, ax


# In[5]:


# make a cyclable palette for large models lists
colors = itertools.cycle(palette)


def update(models=['renault clio (FR)'], year=2012, checkbox_status=False):

    # create 2 output widgets to intercept graphs
    out1 = widgets.Output()
    out2 = widgets.Output()

    # create a Tab widget to arrange Output widgets
    tab = widgets.Tab(children=[out2, out1])
    tab.set_title(0, 'Distribution')
    tab.set_title(1, 'Trend')

    # redirect mpl output to one Output widget (tabs separation)
    with out1:
        fig, ax = default_graph()
        for model in models:     # loop through the selected models

            # set color scheme for selected models
            current_color = color_picker(models, model, colors)

            price_by_year = pd.DataFrame()  # create empty data frame

            price_by_year['year'] = raw_data[raw_data['model']
                                             == model]['year'].unique()  # get years for a specific model

            price_by_year['price'] = raw_data[raw_data['model'] == model][[
                'year', 'results', 'price']].groupby('year').apply(wm).values  # get weighted mean price per year

            price_by_year['std'] = raw_data[raw_data['model'] == model][[
                'year', 'results', 'price']].groupby('year').apply(std).values  # get standard deviation

            price_by_year['lower'] = price_by_year['price'] - \
                price_by_year['std']  # get lower band limit

            price_by_year['upper'] = price_by_year['price'] + \
                price_by_year['std']  # get upper band limit

            if checkbox_status:

                ax.fill_between(price_by_year['year'], price_by_year['lower'],
                                price_by_year['upper'], facecolor=current_color, alpha=0.3)
            ax.plot(price_by_year['year'], price_by_year['price'], lw=3,
                    color=current_color, marker=".", markersize=10, label=model)

        # display(fig)
        plt.show(fig)

    # redirect mpl output to the second Output widget (tabs separation)
    with out2:
        fig2, ax2 = default_dist()

        for model in models:     # loop through the selected models

            # set color scheme for selected models
            current_color = color_picker(models, model, colors)

            # prepare price distribution for a selected year

            df = raw_data[(raw_data['model'] == model) & (
                raw_data['year'] == year)]  # create filtered df
            # introduce shift of bars in bar plot for better visibility
            shift = models.index(model)*200/len(models)
            # smoothen results by rolling mean with window 5
            x = df['price'].values+shift
            # create array of bar width parameters for mpl bar plot
            width_arr = np.ones(len(x))*150
            # calculate sliding window mean with winndow size 5
            y_rolling_mean = df['results'].rolling(5).mean().values
            # create bar plot
            ax2.bar(x, y_rolling_mean, color=current_color,
                    width=width_arr, linewidth=1, label=model, alpha=1.0)

        plt.show(fig2)

    # add legends to the plots
    ax2.legend(loc='best')
    ax.legend(loc='best')
    # display graphs wrapped in Tab widgets
    display(tab)

    return


# create instance of Interactive widget acting as a container for other widgets
widget = interactive(update, models=widgets.SelectMultiple(
    options=list(raw_data['model'].unique()),
    value=[],
    # rows=8,
    description='Model',
    disabled=False),
    year=widgets.IntSlider(
    value=2012,
    min=2008,
    max=2016,
    step=1,
    description='year',
    disabled=False,
    continuous_update=True,
    orientation='horizontal',
    readout=True,
    readout_format='d'),
    checkbox_status=widgets.Checkbox(
    value=False,
    description='Show trend error bar',
    disabled=False)
)

# create vertical column with control elements
controls = VBox(widget.children[:-1],
                layout=Layout(width='35%', height='400px'))
# declare output window as the Output widget (last in child list) of interactive widget
output = widget.children[-1]

# display created design
display(HBox([controls, output]))

#show(update(), notebook_handle=True)


# In[ ]:





# In[ ]:





# In[ ]: