AppStore Sessions Ad-Hoc Python Plots

plot.py

import pygal
import datetime
import statistics

apple_filename = "rosta_–_duty_calendar-sessions-20161001-20200629.csv"


def every_n_dates_criteria(n):
    return lambda date, data: len(data) == n


class PlotParams:
    def __init__(self, data_point_criteria=None, aggregator=None):
        self.data_point_criteria = data_point_criteria
        self.aggregator = aggregator


class Data:
    def __init__(self, dates, data, label):
        self.dates = dates
        self.data = data
        self.label = label


class DateRange:
    def __init__(self, from_date=None, to_date=None):
        self.from_date = from_date
        self.to_date = to_date


class Criteria:
    def __init__(self, name,
                 strategy,
                 x_label_formatter=lambda d: f'{d.strftime("%b/%Y")}'):
        self.name = name
        self.strategy = strategy
        self.x_label_formatter = x_label_formatter


def apple_data_points(data_point_criteria=lambda date, data: date.day == 1,
                      aggregator=lambda data: max(data),
                      date_range=None):
    dates = []
    sessions = []
    users = []
    with open(apple_filename) as f:
        for line in f:
            data = line.replace("'", "").split(",")
            try:
                current_date = datetime.datetime.strptime(data[0], "%m/%d/%y")
            except ValueError:
                continue
            if date_range:
                if date_range.from_date and current_date < date_range.from_date:
                    continue
                if date_range.to_date and current_date > date_range.to_date:
                    break
            date = current_date
            users.append(int(eval(data[1])))

            if data_point_criteria(date, users):
                sessions.append(aggregator(users))
                dates.append(date)
                users = []
    if not sessions and users:
        sessions.append(aggregator(users))
        dates.append(date)
    return (dates, sessions)


def create_plot(title, dates, data, plot_name, x_label_formatter):
    """Plot implementation with pygal."""
    line_chart = pygal.Line(x_label_rotation=90, fill=True)
    line_chart.title = title
    line_chart.x_labels = map(x_label_formatter, dates)
    for d in data:
        line_chart.add(d.label, d.data)
    line_chart.render_to_file(plot_name)


if __name__ == "__main__":
    # Just daily session as indicator by now
    indicators = {"daily sessions": None}
    # How to aggregate each period data
    aggregators = {
        "min": lambda data: min(data),
        "max": lambda data: max(data),
        "avg": lambda data: int(statistics.mean(data)),
        # # "mode": lambda data: statistics.mode(data),
        # "median": lambda data: statistics.median(data),
        # # "sum": lambda data: sum(data)
    }
    # Time criteria use to aggregate the data
    criterias = [
        Criteria("monthly", lambda date, data: date.day == 1),
        Criteria("bimonthly", lambda date, data: len(data) == 30 * 2),
        Criteria("quarterly", lambda date, data: len(data) == 30 * 3),
        Criteria("semester", lambda date, data: len(data) == 30 * 6),
        Criteria("weekly", lambda date, data: len(data) == 7,
                 lambda d: f'{d.strftime("%b/%d/%y")}'),
        Criteria("fortnight", lambda date, data: len(data) == 14,
                 lambda d: f'{d.strftime("%b/%d/%y")}'),
    ]
    # Just official AppStore Data.
    # There's a draft impl of fabric->firebase but is less consistent
    # and data is only available since 5/2020
    sources = {
        # For now just ignore fabric/firebase sources.
        # "Fabric_Firebase": data_points,
        "AppStore": apple_data_points
    }
    # Date ranges of interest
    date_ranges = [
        # From start to PT translation
        DateRange(to_date=datetime.datetime(2019, 10, 9, 0, 0)),
        # # From start to COVID
        DateRange(to_date=datetime.datetime(2020, 3, 20, 0, 0)),
        # From COVID to now 1/7/2020
        DateRange(from_date=datetime.datetime(2020, 2, 17, 0, 0)),
        # # All available data
        DateRange()
    ]

    # Multiply parameters to produce the plots:
    # Source x indicator x aggregator x time criteria x time range
    for source_k, source_v in sources.items():
        for indicator_k, indicator_v in indicators.items():
            for aggregator_k, aggregator_v in aggregators.items():
                for criteria in criterias:
                    for date_range in date_ranges:
                        # 1. Produce the plot source data
                        dates, daus = source_v(
                            criteria.strategy, aggregator_v, date_range)
                        # 2. Additional plot parameters
                        label = indicator_k.title()
                        dates_range_text = f"{criteria.x_label_formatter(dates[0])} – {criteria.x_label_formatter(dates[-1])}"
                        title = f"{indicator_k}({criteria.name} {aggregator_k}) {dates_range_text}".title(
                        )
                        data = Data(dates, daus, label)
                        # 3. Create the plot
                        create_plot(title,
                                    dates,
                                    [data],
                                    f"{source_k}_{title.replace(' ', '_').replace('/', '_')}.svg",
                                    criteria.x_label_formatter)

requirements.txt

pygal

rosta_–_duty_calendar-sessions-20161001-20200629.csv