# -*- coding: utf-8 -*-
"""
Created on Wed Oct 16 2019

@author: rorymorrison

You may need to pip install some of the modules
"""
import bs4, os, requests
import numpy as np
import pandas as pd
pd.set_option('display.max_rows', 500)
pd.set_option('display.max_columns', 500)
import datetime
from datetime import timedelta

os.chdir('C:\\Users\\Rory\\.spyder-py3\\running_analysis')

# Set up a function to get the data from PowerOf10

def get_table(link, event):
    res = requests.get(link)
    soup = bs4.BeautifulSoup(res.content, "html.parser")
    headings = soup.find("tr", class_="rankinglistheadings")
    headings = [i.getText() for i in headings]
    rows = soup.find_all("tr", class_="rlr") + soup.find_all("tr", class_="rlra")
    print("number of rows found: " + str(len(rows)))
    
    rows_text = []
    for i in rows:
        rows_text.append([j.getText() for j in i.find_all("td")])
    
    df = pd.DataFrame(rows_text, columns=headings)
    df['Name'].replace('', np.nan, inplace=True)
    df.dropna(subset=['Name'], inplace=True)
    timestamps = []
    for k in range(len(df.Perf)):
        try:
            date_str = df.Perf[k]
            entry = datetime.datetime.strptime(date_str, "%M:%S.%f")
            entry = timedelta(minutes=entry.minute, seconds=entry.second)
            timestamps.append(entry)
        except:
            timestamps.append("")
    time = "Time_"+event
    df[time] = pd.Series(timestamps, index = df.index)
    df[time] = pd.to_datetime(df[time],errors='coerce')
    df.dropna(inplace=True)
    df.sort_values(time)
    rank = "Rank_" + event
    df[rank] = pd.to_numeric(df['Rank'], errors='coerce')
    
    return df

# Get the 5km data
link_5k = "https://www.thepowerof10.info/rankings/rankinglist.aspx?event=5000&agegroup=ALL&sex=M&year=2019"    
table_5k = get_table(link_5k, "5k")

# Get the 10km data
link_10k = "https://www.thepowerof10.info/rankings/rankinglist.aspx?event=10000&agegroup=ALL&sex=M&year=2019"    
table_10k = get_table(link_10k, "10k")

# Merge the two tables together using the 'Name' column as the reference

mutual = pd.merge(table_5k, table_10k, on='Name')

# =============================================================================
# Linear regression
# =============================================================================

from scipy import stats

x = []
for i in mutual.PB_x.tolist():
    try:
        a = int(i[:2]) + int(i[3:5])/60
        x.append(a)
    except:
        continue

y = []
for i in mutual.PB_y.tolist():
    try:
        a = int(i[:2]) + int(i[3:5])/60
        y.append(a)
    except:
        continue

# =============================================================================
# Filter the model to remove 5km times which are less than twice the 10km time
# i.e. this time is not normal
# =============================================================================
x_filt = []
y_filt = []
for i in range(0, len(x)):
    if x[i]*2 < y[i]:
        x_filt.append(x[i])
        y_filt.append(y[i])
x = np.array(x_filt)
y = np.array(y_filt)
slope, intercept, r_value, p_value, std_err = stats.linregress(x,y)

x_lr = []
for i in np.linspace(13, 19, 100):
    x_lr.append(i)
y_lr = [(intercept + slope*i) for i in x_lr]
df_lr = pd.DataFrame({'x': x_lr, \
                      'y': y_lr})
df_lr['x_mins'] = [datetime.timedelta(minutes=i) for i in df_lr['x']]
df_lr['y_mins'] = [datetime.timedelta(minutes=i) for i in df_lr['y']]

# =============================================================================
# Make up a table to store my times
# =============================================================================
my_5k = "18:40.00"
my_10k = "39:35.00"
my_time = pd.DataFrame({'Name' : ['Rory Morrison'], \
                        'Perf_x' : [my_5k], \
                        'Perf_y' : [my_10k]})
my_5k = datetime.datetime.strptime(my_5k, "%M:%S.%f")
my_5k = timedelta(minutes=my_5k.minute, seconds=my_5k.second)
my_time['Time_5k'] = pd.Series([my_5k]) 
my_time['Time_5k'] = pd.to_datetime(my_time['Time_5k'],errors='coerce')
my_10k = datetime.datetime.strptime(my_10k, "%M:%S.%f")
my_10k = timedelta(minutes=my_10k.minute, seconds=my_10k.second)
my_time['Time_10k'] = pd.Series([my_10k]) 
my_time['Time_10k'] = pd.to_datetime(my_time['Time_10k'],errors='coerce')
    
# plot the data
from bokeh.io import output_file, show
from bokeh.models import ColumnDataSource, HoverTool, LinearColorMapper
from bokeh.palettes import plasma
from bokeh.plotting import figure, output_file, show
from bokeh.transform import transform

list_x = mutual['Time_5k'].tolist()
list_y = mutual['Time_10k'].tolist()
Time_5k = mutual['Perf_x'].tolist()
Time_10k = mutual['Perf_y'].tolist()
Rank_5k = mutual['Rank_5k'].tolist()
Rank_10k = mutual['Rank_10k'].tolist()
desc = mutual['Name'].tolist()

source = ColumnDataSource(data=dict(x=list_x, y=list_y, desc=desc, \
                                    Rank_5k=Rank_5k, Rank_10k=Rank_10k, \
                                    Time_5k=Time_5k, Time_10k=Time_10k))
my_source = ColumnDataSource(data=dict(desc=my_time['Name'].tolist(), \
                                    x=my_time['Time_5k'].tolist(), \
                                    y=my_time['Time_10k'].tolist(), \
                                    Time_5k=my_time['Perf_x'].tolist(), \
                                    Time_10k=my_time['Perf_y'].tolist()))
hover = HoverTool(tooltips=[
    ('Name', '@desc'),
    ('Rank 5k', '@Rank_5k'),
    ('5k Time', '@Time_5k'),
    ('Rank 10k', '@Rank_10k'),
    ('10k Time', '@Time_10k')
])
mapper = LinearColorMapper(palette=plasma(256), low=min(Rank_5k), high=max(Rank_5k))

p = figure(plot_width=600, plot_height=600, tools=[hover, 'pan', 'wheel_zoom'], title="10k vs 5k PBs", \
           x_axis_label='5km (minutes)', y_axis_label='10km (minutes)', \
           x_axis_type='datetime', y_axis_type='datetime')
p.circle('x', 'y', size=10, source=source,
         fill_color=transform('Rank_5k', mapper), legend='PowerOf10')
p.circle('x', 'y', size=10, fill_color='red', source=my_source, legend='My time')
p.line(df_lr['x_mins'].tolist(), df_lr['y_mins'].tolist(), line_dash=[4, 4], line_width=4, \
       line_color='orange', legend='best fit of PowerOf10')
p.ygrid.grid_line_dash = [6, 4]
p.xgrid.grid_line_dash = [6, 4]

p.legend.location = "top_left"
p.legend.click_policy="hide"
output_file('test.html')
show(p)