Script that computes the trend for new authors in my chat.

author_trend.py

import json
import datetime
import numpy as np
import matplotlib
from itertools import groupby
from operator import itemgetter
from matplotlib import pyplot as plt
from time import perf_counter

matplotlib.rcParams['font.size'] = 18

class Trace(object):
    def __init__(self):
        self.trace = []

    def step(self, label):
        self.trace.append((label,perf_counter()))

    def print_stats(self):
        if len(self.trace) < 2:
            print("Not enough tracing steps")
        else:
            print("***********************")
            print("Tracing statistics")
            for prev,post in zip(self.trace[:-1],self.trace[1:]):
                print("{} -> {}: {:8.2f} s".format(
                    prev[0], post[0], post[1]-prev[1])
                )
            print("***********************")

def chat_stats(messages):
    stats = []
    for k,g in groupby(sorted(messages,key=itemgetter('author')),
                       key=itemgetter('author')):
        timestamps = sorted([
            datetime.datetime.strptime(m['date'],"%b %d %Y %H:%M:%S")
            for m in g])
        stats.append((k, timestamps[0], len(timestamps)))

    return stats

def monthly_active_users(messages):
    active_users = []
    engagement = []
    times = []
    last_visit = {}
    any_visit = {}

    threshold = datetime.timedelta(days=30)
    
    for m in reversed(messages):
        todelete = []
        author = m['author'],
        timestamp = datetime.datetime.strptime(m['date'],"%b %d %Y %H:%M:%S")
        last_visit[author] = timestamp
        any_visit[author] = timestamp

        # Clean user list
        for user in last_visit:
            if (timestamp - last_visit[user]) > threshold:
                todelete.append(user)

        for user in todelete:
            del last_visit[user]

        active_users.append(len(last_visit))
        engagement.append(len(last_visit)/len(any_visit))
        times.append(timestamp)

    return active_users, engagement, times
    

if __name__ == '__main__':
    trace = Trace()
    trace.step('start')

    with open('messages.json') as f:
        messages = json.load(f)

    trace.step('IO')

    stats = chat_stats(messages)
    dates = sorted([fa[1] for fa in stats])
    timedeltas_seconds = [(date-dates[0]).total_seconds() for date in dates]

    # Largest contributors
    print('Users that have contributed more than 20 messages')
    for s in sorted(stats,key=itemgetter(2), reverse=True):
        if s[2] > 20:
            print('{}: {}'.format(s[0],s[2]))
    
    x = np.linspace(0,len(dates),len(dates))
    p = np.polyfit(x,timedeltas_seconds,2)

    seconds_elapsed = np.polyval(p,x)
    newdates = [dates[0] + datetime.timedelta(seconds = s) for s in seconds_elapsed]
    trace.step('newusers')

    active_users, engagement, timestamps = monthly_active_users(messages)

    trace.step('monthly users')
    trace.print_stats()

    plt.figure(1)
    plt.xticks(rotation=70)
    plt.plot_date(dates, x,'-',linewidth=4)
    plt.plot_date(newdates, x, 'k--', linewidth=2)
    plt.subplots_adjust(bottom=0.21, left=0.15)
    plt.ylabel('Number of total users.')

    plt.figure(2)
    plt.xticks(rotation=70)
    plt.plot_date(timestamps, active_users,'-',linewidth=4)
    plt.subplots_adjust(bottom=0.21, left=0.15)
    plt.ylabel('Number of monthly active users.')

    plt.figure(3)
    plt.xticks(rotation=70)
    plt.plot_date(timestamps, engagement,'-',linewidth=4)
    plt.subplots_adjust(bottom=0.21, left=0.15)
    plt.ylim([0,1.1])
    plt.ylabel('Engagement.')
    
    plt.show()