parse-timelogs-for-upload/pomodoro_to_harvest.py

import pandas as pd
import numpy as np
import glob
import re
import sys
# Import our local settings management.
import settings

if hasattr(sys, 'ps1'):
    import copy
    debug = True
else:
    debug = False

if settings.pomodoro_logfile():
    # This works for one file:
    timelog = pd.read_csv(settings.pomodoro_logfile())
else:
    # For multiple files:
    path = settings.pomodoro_logpath()
    all_files = glob.glob(path + "*.csv")

    li = []

    for filename in all_files:
        df = pd.read_csv(filename, index_col=None, header=0)
        li.append(df)

    timelog = pd.concat(li, axis=0, ignore_index=True)

if debug:
   imported = copy.deepcopy(timelog)

timelog.drop_duplicates(inplace=True)

if debug:
   nodupes = copy.deepcopy(timelog)

# 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.
# We can allow no data in the 'intention' so define the three columns to check:
timelog = timelog.dropna(subset=['started', 'recorded', 'description'])
timelog = timelog.reset_index(drop=True)

if debug:
   dropna = copy.deepcopy(timelog)

   # For debugging, keep originals around.
   timelog["orig_desc"] = timelog["description"]
   timelog["orig_started"] = timelog["started"]
   timelog["orig_recorded"] = timelog["recorded"]

# 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()

if debug:
    mess = copy.deepcopy(timelog)

timelog["started"] = pd.to_datetime(timelog["started"], errors='coerce').dt.tz_convert("US/Eastern")
timelog["recorded"] = pd.to_datetime(timelog["recorded"]).dt.tz_convert("US/Eastern")

latest_recorded = settings.pomodoro_latest_recorded()
if latest_recorded:
    timelog = timelog[timelog.recorded > pd.to_datetime(latest_recorded)]

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.
harvest_project_names = {
    "Boston Modern Orchestra Project": ["BMOP", "BMOP.org"],
    "CRLA.org upgrade": ["CRLA", "CRLA upgrade"],
    "Contrib": ["Contributing", "Agaric contrib", "Agaric contributions"],
    "Drutopia": ["Drutopia improvements", "Drutopia overhead"],
    "EC Connect": ["eccconectcolorado.org", "Denver Early Childhood", "ECconnect", "ECconnectColorado"],
    "encuentro 5 sites": ["e5", "Encuentro"],
    "Family & Home": ["Family and Home", "Family home"],
    "Find It Cambridge": ["Find It", "FIC", "Cambridge"],
    "GEO Support": ["GEO", "GEO.coop", "Grassroots Economic Organizing"],
    "Internal": ["Agaric", "Agaric internal"],
    "Leads": ["Lead", "Agaric leads", "Lead followups"],
    "Internal: Personal Learning": ["Learning", "Personal learning"],
    "MASS Continuous Improvement": ["MASS Design Group", "MASS", "MASS Design"],
    "NICHQ Data Upgrade": ["NICHQ Data"],
    "NICHQ Support": ["NICHQ", "NICHQ support"],
    "NICHQ FL CMS LAN": ["FL CMS LAN", "flcmslan", "NICHQ FLCMSLAN"],
    "Internal: Network Engagement": ["Network Engagement", "network engagement", "Network engagment", "Social media", "Network building", "Agaric network engagement"],
    "SCDTDP Collaboratory Data Site System Security": ["SCDTDP", "NICHQ SCDTDP", "NICHQ security"],
    "Teachers with GUTS": ["TWIG", "GUTS"],
    "The Propaganda Site": ["TPS", "Propaganda Site", "The Propganda Site", "Murat & Clay"],
}
other_project_names = {
    "Near North camp": ["Near North Camp", "Near North defense", "Encampment support", "Camp support", "NN camp defense", "NN camp", "NN defense", "Near North camp defense"],
    "Personal": ["Personal/external", "Personal / external", "External"],
}

replacement_project_names = harvest_project_names.copy | other_project_names

for preferred, alternatives in replacement_project_names.items():
    # We compare all alternatives to lower case versions, and add the
    # preferred output to this list for that purpose, but note that what we use
    # as preferred retains its capitalization.
    alternatives.append(preferred)
    alternatives = [item.lower() for item in alternatives]
    timelog.loc[timelog.project.str.lower().isin(alternatives), "project"] = preferred

# If a compound project was specified, break that out into a sub-project (in
# Harvest, we use Task, which is really task type, for this.
timelog['subproject'] = (np.where(timelog['project'].str.contains(': '), timelog['project'].str.split(': ', 1).str[1], None))
timelog['project'] = (np.where(timelog['project'].str.contains(': '), timelog['description'].str.split(': ', 1).str[0], timelog['project']))

# 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()

# We're doing the final conversion to Harvest as a separate step because we
# want to factor out all of the above non-Harvest-specific logic.

latest = tl.recorded.max()

# Filter out any blank projects and any projects we know are not in Harvest.
# We also do the opposite to get a CSV of the excluded items.
non_harvest_list = ["", "Personal", "Near North camp"] 
harvest = tl[tl.project.isin(harvest_project_names.keys())]
other = tl[tl.project.isin(other_project_names.keys())]
unknown = tl[~tl.project.isin(replacement_project_names.keys())]


if not debug:
    harvest.to_csv('harvest-timesheets.csv', index=False)
    other.to_csv('not-harvest.csv', index=False)
    settings.pomodoro_latest_recorded(latest)
else:
    harvest_grouped = harvest.groupby("project").agg({"time": "sum"})["time"]/60
    other_grouped = other.groupby("project").agg({"time": "sum"})["time"]/60
    unknown_grouped = unknown.groupby("project").agg({"time": "sum"})["time"]/60
    print("We do not write to the harvest-ready.csv nor update the latest recorded setting when run interactively in the python shell.")