chart_rainbow

#!/usr/bin/python3

import sys, datetime, os, math
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker

n_days = 65
# Y units are about <y_units_per_x_unit> denser than X units.
# This is used to calculate the angle of annotations.
y_units_per_x_unit = 4
data = {}

def parse_date(s):
    return datetime.datetime.strptime(s[:8], '%Y%m%d').date()

def record(as_of, date_of_death, deaths):
    if date_of_death not in data:
        data[date_of_death] = []
    data[date_of_death].append((as_of, deaths))

def dump():
    for date_of_death in sorted(data.keys()):
        print(f'{date_of_death}', end='')
        for (as_of, deaths) in data[date_of_death]:
            print(f' {as_of}:{deaths}', end='')
        print('')

def col(i, max_days):
    return plt.cm.jet(round(255 * i / max_days))

def chart(fname):
    (fig, ax) = plt.subplots(dpi=300)
    max_days = 1 + n_days
    i = 0
    for date_of_death in sorted(data.keys())[-max_days:]:
        # data may not be sorted
        data[date_of_death].sort()
        days_past_date = [(x[0] - date_of_death).days for x in data[date_of_death]]
        deaths = [x[1] for x in data[date_of_death]]
        if len(days_past_date) > 1:
            angle = math.atan((deaths[-1] - deaths[-2]) / y_units_per_x_unit)
            xytext = (int(math.cos(angle) * 14), int(math.sin(angle) * 14))
            angle = angle / math.pi * 180
            color = col(i, max_days)
            ax.annotate(str(date_of_death), (days_past_date[-1], deaths[-1]),
                    ha='center', va='center', fontsize=4, rotation=angle,
                    xytext=xytext, textcoords='offset points', alpha=color[3])
            ax.plot(days_past_date, deaths, label=str(date_of_death),
                    color=color, linewidth=1.0)
        i += 1
    ax.grid(True, which='minor', linewidth=0.1)
    ax.grid(True, which='major', linewidth=0.3)
    ax.spines['top'].set_visible(False)
    ax.spines['bottom'].set_visible(True)
    ax.spines['left'].set_visible(True)
    ax.spines['right'].set_visible(False)
    ax.set_ylabel('Number of deaths reported')
    ax.set_xlabel('Days after date of death')
    ax.xaxis.set_minor_locator(ticker.MultipleLocator(base=1))
    ax.yaxis.set_minor_locator(ticker.MultipleLocator(base=5))
    ax.set_xlim(left=0)
    ax.set_ylim(bottom=0)
    fig.suptitle('Florida COVID-19 deaths by day')
    ax.text(
            -0.05, -0.14,
    'Each curve represents "Deaths by Day", the number of deaths on a given day. The Florida Department of Health\n'
    'labels this data table Florida_COVID_19_Deaths_by_Day. As time goes by, more deaths are processed and assigned\n'
    'to the correct calendar day, so the curve goes up. FDOH warns that recent data is incomplete. However the initial\n'
    'steepness of the curve gives an intuitive sense of how many deaths will be approximately reported.\n'
    'Source: https://github.com/mbevand/florida-covid19-deaths-by-day    Created by: Marc Bevand — @zorinaq',
            transform=ax.transAxes, fontsize='x-small', verticalalignment='top',
    )
    ax.legend(bbox_to_anchor=(-.03, -.30), loc='upper left', fontsize='x-small', ncol=5)
    fig.savefig(fname, bbox_inches='tight')
    plt.close()

def main():
    for f in sorted(os.listdir('.')):
        if not f.endswith('.csv'):
            continue
        df = pd.read_csv(f)
        # Parse date out of "deathsbydateexport_20200623.csv"
        as_of = parse_date(f.split('_')[1])
        #if as_of.weekday() != 2:
        #    continue
        found_date_of_death_for_same_day = False
        for _, row in df.iterrows():
            date_of_death = datetime.datetime.utcfromtimestamp(row['Date'] / 1000).date()
            record(as_of, date_of_death, int(row['Deaths']))
            found_date_of_death_for_same_day |= (as_of == date_of_death)
        if not found_date_of_death_for_same_day:
            # if the CSV file has no deaths recorded for the date of death the CSV file was published,
            # it means it is zero
            record(as_of, as_of, 0)
    #dump()
    chart('chart_rainbow.png')

if __name__ == '__main__':
    main()