parse-timelogs-for-upload/pomodoro_to_harvest.py

import pandas as pd
import numpy as np
import re
import settings

timelog = pd.read_csv(settings.pomodoro_logfile())

# Dump bad data.  The real solution here is to get rid of the damned 'Cancel'
# button on the Pomodoro Prompt dialog, but i don't know how to do that, so we
# need to drop the rows where the work task description is blank, which is
# coming in as not a number for reasons i'm not entirely clear on.  Maybe
# because it's the last row of the spreadsheet?  Anyway we cannot do anything
# with no data in the description, so drop them at the outset.
timelog = timelog.dropna()
timelog = timelog.reset_index(drop=True)

# For debugging, keep original description around.
timelog["orig_desc"] = timelog["description"]

# Clean up description before we go to work on it.
timelog['description'] = timelog['description'].str.strip()

# Allow multiple entries to be put into one prompt by splitting with semicolon.
# TODO make this a flag since it's possible to use semicolons without meaning
# to make multiple task entries at once.
timelog["description"] = list(timelog["description"].str.split(";"))
timelog = timelog.explode("description").reset_index()

timelog["started"] = pd.to_datetime(timelog["started"]).dt.tz_convert("US/Eastern")
timelog["recorded"] = pd.to_datetime(timelog["recorded"]).dt.tz_convert("US/Eastern")
timelog["time"] = 30
# A pomodoro started before 3am Eastern time is considered to be a continuation
# of the day before, so we are, effectively, on West Coast time for determining
# the day we want to associate a time entry with.  PomodoroPrompt saves as UTC.
timelog["date"] = timelog["started"].dt.tz_convert("US/Pacific").dt.date
timelog["day_of_week"] = pd.to_datetime(timelog["date"]).dt.day_name()

# If a project has been specified (task prefixed with a colon), then put the
# project in its own column.
timelog['project'] = (np.where(timelog['description'].str.contains(': '), timelog['description'].str.split(': ', 1).str[0], None))
timelog['description'] = (np.where(timelog['description'].str.contains(': '), timelog['description'].str.split(': ', 1).str[1], timelog['description']))

# Mid-work clean up of description and new project.
timelog['description'] = timelog['description'].str.strip()
timelog['project'] = timelog['project'].str.strip()

# If a multiplier has been provided (an asterisk and an integer at the end of a
# task), then multiply the time by it and remove it from the description.
# Ensure we're splitting on the same asterisk we found:  Use the end of string
# signifier in the regular expression ($), and split from the right.
p = re.compile(r'\*\s*\d$')
# On some systems, using np.where worked but others failed.  Why it worked is
# unknown but why it failed is because numpy where evaluates all parts, even
# the parts that will never get used because the where clause does not apply!
# This caused the chained strings to fail because— no string.
# timelog['tmp_multiplier'] = (np.where(timelog['description'].str.contains('\*\s*\d$'), timelog['description'].str.rsplit('*', 1).str[1].str.strip(), 1))
# timelog['description'] = (np.where(timelog['description'].str.contains('\*\s*\d$'), timelog['description'].str.rsplit('*', 1).str[0], timelog['description']))
timelog['tmp_multiplier'] = timelog['description'].apply(lambda x: x.rsplit('*', 1)[1].strip() if p.search(x) else 1)
timelog['description'] = timelog['description'].apply(lambda x: x.rsplit('*', 1)[0] if p.search(x) else x)
timelog["time"] = timelog["time"] * timelog['tmp_multiplier'].astype(int)
timelog.drop(columns=['tmp_multiplier'], inplace=True)

# Clean up description again, after it has been sliced and diced.
timelog['description'] = timelog['description'].str.strip()

# Replace irregular-but-known project names with ones timetracking tools use.
replacement_project_names = {
    "Find It Cambridge": ["Find It", "FIC", "Cambridge"],
    "The Propaganda Site": ["TPS", "Propaganda Site"],
    "MASS Design Group": ["MASS"],
    "Teachers with GUTS": ["TWIG", "GUTS"],
    "NICHQ Support": ["NICHQ", "NICHQ support", "nichq"],
    "Network engagement": ["Network Engagement", "network engagement", "Network engagment", "Social media", "Network building", "Agaric network engagement"],
    "Agaric internal": ["Agaric", "Internal"],
    "Agaric contrib": ["Contributing", "Contrib"],
    "Leads": ["Lead", "Agaric leads", "Lead followups"],
    "Learning": ["Personal learning"],
    "Drutopia": ["Drutopia improvements"],
    "Personal / external": ["Personal/external", "Personal", "External"],
    "Near North camp": ["Near North Camp", "Near North defense", "Encampment support", "Camp support"],
}
# TODO Probably put all alternatives in lower case and do str.lower() on
# project just before the "is in" check.
for preferred, alternatives in replacement_project_names.items():
    timelog.loc[timelog.project.isin(alternatives), "project"] = preferred

# Condense duplicate entries by date, summing the minutes spent, and listing
# the first started and last recorded times for each task.
# The fillna is essential or we drop entries with blank ('None') projects.
tl = timelog.groupby(["date", timelog.project.fillna(""), "description"]).agg({"time": 'sum', "started": 'min', "recorded": 'max'}).reset_index()

tl.to_csv('harvest-ready.csv', index = False)

settings.pomodoro_latest_recorded(tl.recorded.max())