annextimelog 0.9.0

Track time spent on projects, backed by Git Annex

⚠️ This tool is early development. The most basic time tracking feature (recording, deletion, time frame search) as well as syncing are implemented though.

annextimelog - ⏱️ Git Annex-backed Time Tracking

This is a brainstorm for a Git Annex-backed time tracker. The idea originated across some of my Mastodon threads:

• https://fosstodon.org/@nobodyinperson/109596495108921683
• https://fosstodon.org/@nobodyinperson/109159397807119512
• https://fosstodon.org/@nobodyinperson/111591979214726456

The gist is that I was (and still am) unhappy with the existing time tracking solutions. I worked with hledger's timeclock and timewarrior each for quite some time and built my own workflow and scripts around them.

✅ Requirements

Over the years, the below features turned out to be my personal requirements for a time-tracking system (TL;DR: easy and intuitive recording, hassle-free syncing, data export for further analysis). Here is a table comparing annextimelog with timewarrior and hledger timeclock:

✅ = feature available, 🟡 = partly available, ❌ = not available

feature	timewarrior	hledger timeclock	annextimelog
precise start and end times	✅	✅	✅ as git-annex metadata
tracking of overlapping/simultaneous periods	❌	🟡 (separate files)	✅ backend can do it
nice, colourful, graphical summary	✅	🟡	✅ with Python rich, more planned
plain text data storage	✅	✅	🟡 buried in git-annex branch
git-friendly, merge conflict free data format	🟡¹	🟡¹	✅ git-annex’ own merge strategy
arbitrary tags attachable to tracked periods	✅	🟡 hledger tags²	✅ just git-annex metadata
arbitrary notes attachable to tracked periods	🟡³	🟡 hledger tags²	✅ just git-annex metadata
tags can have values	❌	✅ hledger tags²	✅ just git-annex metadata
files attach-/linkable to tracked periods	❌	🟡 path as file: tag	🟡 annexed files, linking is planned
cli to start, stop, edit, etc. tracked periods	✅⁴	❌ own scripts needed	🟡 very basic, more planned
plugin system	🟡⁵	🟡⁶ (hledger’s own)	❌ git-style plugin system planned
data export to common format	✅ (JSON)	✅ (CSV, JSON)	✅ as timeclock, JSON, cli commands
syncing functionality built-in	❌	❌	✅ git-annex’s purpose is syncing
multi-user support	❌	❌	🟡 e.g. use tag user=NAME

¹last line is always modified, merge conflicts can arise when working from different machines

²hledger tags have limitations, e.g. no spaces, colons, commas, etc.

³timewarrior annotations can't contain newlines for example. I wrote an extension to edit your annotation in your $EDITOR and optionally GPG-encrypt it, which lets you add newlines. Quite an inconvenience.

⁴timewarrior’s cli has some nasty inconveniences (e.g. no shortcut for ‘yesterday’, must painfully type out the full date, no intelligence to operate only on yesterday, gets confused and errors out in certain combinations of start/end times, etc…)

⁵timewarrior extensions (here mine) are just fed the data via STDIN, not other command-line arguments. Not as useful as the git-style plugin system.

⁶for the analysis part, hledger plugins can be used. But as there is no actual cli to manage the data, there’s no plugin system for that.

🛠️ Implementation

To learn more about how annextimelog works under the hood with git-annex as backend, have a look at doc/internals.

📦 Installation

You can run this tool if you have nix installed:

# drop into a temporary shell with the command available
nix shell gitlab:nobodyinperson/annextimelog

# install it
nix profile install gitlab:nobodyinperson/annextimelog

On Arch Linux you can install from the AUR with your favorite helper, or directly with pacman from this user repository.

# use an AUR helper to install
paru -S annextimelog

Otherwise, you can install it like any other Python package, e.g. with pip or better pipx:

pipx install annextimelog

# latest development version
pipx install git+https://gitlab.com/nobodyinperson/annextimelog

Note that in this case you will need to install git-annex manually.

Any of the above makes the annextimelog (or atl) command available.

❓ Usage

# Show help page
atl --help
atl tr --help # subcommand help page

# Show exactly what's going on underneath (more -v → more detail)
atl -vvvvv ...

# Add a remote to sync to
atl git remote add myserver [e-mail]:...

# Sync status with git annex
atl sync

# Track a time period with metadata
atl track 10:00 15:00 work @home ="make a title with the equals sign" :"make a note with a colon"
atl tr 15:00 - 10min ago code @home # timewarrior-style time ranges

# List events, same language as 'atl tr' is used for matching
atl     # default: what happened today
atl ls  # default: what happened today
atl ls week 
atl ls project='my.*regex'
atl ls work @home last month

# Delete an event
atl rm 3QicA4G4

# Output formats
atl ls -a -O timeclock  # output all events in hledger timeclock format
atl ls -a -O timeclock | hledger -f timeclock:- bal --daily # analyse with hledger
atl ls -a -O json # output JSON

There's some example periods in the doc folder you can try out.

⚙️ Configuration

annextimelog reuses git's configuration system. You can configure annextimelog per-repo (atl config ...), per-user (atl config --global ...) or system-wide (typically in /etc/gitconfig).

# list all annextimelog configs known to git
atl config --get-regexp '^annextimelog'

# open the git config in your $EDITOR
atl config --edit

# no emojis in the output tables
atl config emojis false

# let the week start at Sunday
atl config weekstartssunday true

# Don't commit, except in 'annextimelog sync'.
# This speeds up things but reduces granularity to undo changes
atl config commit false

# Skip steps that are not strictly necessary.
# This speeds up things but might leave the repository in a less ordered state.
atl config fast true

🛠️ Development

This project uses poetry, so you can run the following in this repository to get into a development environment:

poetry install
poetry shell
# now you're in a shell with everything set up

Other:

# Auto-run mypy when file changes:
just watch-mypy

# Auto-run tests when file changes:
just watch-test

# Test how a sequence of command-line args is interpreted as event metadata
just test-tokens work @home note=bla myfield+=one,two,three 2h ago until now

# Run tests against a different Python version
just test-with-python-version 3.10