harsh is habit tracking for geeks. A minimalist, command line tool for tracking and understanding your habits.
Succinctly: it's quick and gets out of your way. And gives you excellent visibility on your habits.
There're 3 commands: ask, log, and todo.
(and one subcommand log stats)
As of v0.10.6``)askorlogabout a particular habit if you want to fill in just a single habit outcome or drill down on a single habit. Useharsh ask andharsh log .harsh
logwithout a string shows you your full consistency graph.harsh ask`
withouy a string asks you about all your listed habits.
Designed for simplicity, visibility, and longevity, harsh uses simple text files for tracking that are human-grokable and editable in your favourite text editor. It's simpler, less messy, and more portable than commercial or mobile applications and less fussy to manage than emacs habit tracking (imho). While quantified individual tracking is exhaustive, important habits get lost in the data deluge, so this provides deliberate, explicit, habits to track and progress.
It's written in Go and adds features and fixes on top of habitctl (its
inspiration) such as skips, streak break warnings, stats, quantities, and as of
0.10.0 targets over intervals (like 3 times in a week). If you're a geek,
I think you'll like it. Despite trying an exhaustive number of habit trackers,
this was what worked for me. YMMV. If you're interested in why I wrote it,
there's a launch post about my motivations on my
blog.
My biggest hope that it helps you get done what you're trying to get done in your life and live a better one.
Harsh lets you see your habits in a consistency graph (aka Seinfeld chain) from left to right over the time you've been doing the habit. The interface shows the last 100 days (but dynamically adapts to a smaller day length if your terminal window is too small.).
For each day, you enter a yes/no/skip on whether you did the habit in question, and the horizontal line for ach habit is like an x-axis time graph showing you how your habits have fared. There's a sparkline graph at the top of the interface to show how you are doing percentage wise each day over the 100 days, and some interface hintint with a M(onday), W(ednesday), F(riday) label below that to jelp you figure out which day goes with what (very helpful if you find particular days are where your habits falldown consistently since you can then fix whatever that is in that day that messes things up.).
Consistency graphs are also known as the "Seinfield Method" due to an apocryphal story about getting better at writing jokes, which I think is attributed to Brad Isaac in a (possibly never happened) piece of advice he got from the comedian Jerry Seinfeld. Basically, the idea was he had a big year long calendar with a day for every year. Every day he wrote jokes, he put an X down. He could see how consistent he was being. The idea is never to bfreak the chain. Harsh works the same way. The idea is turning something into a consistent habit makes you good -- eventually. The original place I saw it (Lifehacker) has killed the original post (largely I think because Seinfeld repudiated it) - but you can see here
harsh is available on OSX (and via homebrew), Linux (also as a Snap and homebrew), FreeBSD, OpenBSD, and Windows. A specific goal was increasing uptake and adoption of a portable, command line, text-based approach to habit tracking. harsh also supports ARM architectures for OSX (M1, M2, and M3 chipped Macs and Linux and BSDs) as of 0.8.8. Binaries for FreeBSD and OpenBSD are also available as of 0.8.23.
Homebrew tap (Mac and Linux):
brew install wakatara/tap/harsh(this will also nicely alert you of updates when you brew update)
Snap (Linux only):
sudo snap install harsh(you'll also get alerted when there are updates)
If you're not using a package manager, by far the easiest way to install harsh and get started is to use this one-line bash command if you're on OSX or Linux:
curl -sSLf https://raw.githubusercontent.com/wakatara/harsh/master/install.sh | shThe script downloads the latest release directly from github, verifies it
cryptographically, and then moves the executable to usr/local/bin ready for
immediate use on the command line.
You may prefer a direct Go install if you have a working Go environment and Go 1.14+.
go install github.com/wakatara/harsh
from the command line. Unlike a package manager like brew or snap, you won't get informed of new version releases.
Alternatively, you can download the pre-compiled binaries from
the releases page
and copy to the desired location (/usr/local/bin recommended),
making sure it's in your $PATH.
If you want to build from source cause you like that sort of thing, follow these steps:
Clone:
git clone https://github.com/wakatara/harsh
cd harshGet the dependencies:
go get ./...Build:
go build -o harsh .Verify it works:
./harsh --versionYou can then move the file to the desired location in your $PATH.
When you run harsh ask for the first time, it will set up the required files:
$ harsh ask
Welcome to harsh!
Created /Users/daryl/.config/harsh/habits This file lists your habits.
Created /Users/daryl/.config/harsh/log This file is your habit log.
What? No habits yet?
Open the habits file and edit the habit list using a text editor.
Then run `harsh ask` to start tracking
'harsh todo` will show you undone habits for today.
`harsh log` will show you a consistency graph of your efforts
(trust me, it looks way cooler looking over time)
Happy tracking! I genuinely hope this helps you get better.On OSX and Linux based systems, the habits and log files will be under
~/.config/harsh/. On Windows, you can find the files under %APPDATA%\harsh.
Alternatively, you can set a different directory using the HARSHPATH environment variable.
Ubuntu's snap overrides (sensible) defaults and forcibly places harsh's config and log files in ~/snap/harsh/current/. If you snap remove the harsh app, snap's uninstaller will nuke your config and log files with the sandbox and you may lose your config and log data if it's not backed up (please always exercise a good backup regime). For this reason, we highly recommend snap users set the HARSHPATH env variable to ~/.config/harsh/ and move config and log files there right after installing to protect them. Or never uninstall harsh. π |
Open the habits file in your text editor of choice (nano, vim, VS Code, Sublime, or emacs).
You'll see an example file like this:
harsh
# This is your habits file.
# It tells harsh what to track and how frequently in days.
# 1 means daily, 7 means weekly, 14 every two weeks.
# You can also track targets within a set number of days
# For example, Gym 3 times a week would translate to 3/7
# 0 is for tracking a habit. 0 frequency habits will not warn or score.
# Examples:
! Dailies
Gymmed: 1
Bed by midnight: 1
! Weeklies
Gymmed: 3/7
Called Mom: 7
! Monthly+
Did Finances: 30
New Skill: 90
! Tracking
Too much coffee: 0
Used harsh: 0A big change starting in version 0.10.0, harsh now allows you to pick a target
number of times you want to perform a habit over a set interval of time (in
days). Examples would be 3/7 for 3 times a week, 2/30 for twice a month etc
etc.
So, for example, wanting to make it to the gym 3 times a week (ie. 7 days) would translate into a line in the habit file like:
Gymmed: 3/7You can also simply pick a number of days you want to repeat a habit (also, for backwards compatibility with older habit files).
Note that this uses a rolling window of days than actual week delimiters (mostly because having hard start of week, month, and quarter breaks broke the point of consistency graphs when I implemented it. This is much nicer. Trust me.).
If it's not obvious from the example file, habits can have any character that is
not a : as that delimits the period. We also use : as the separator in log
files as well for easy parsing.
Headings are denoted by a "!" at the start of a line and allow you to categorize habits visually (useful if you have a lot of them).
Comments can go in the habits file by starting a line with "#" and are not parsed by the program.
The real trick of tracking is figuring out what habits you want to track building or breaking. Too many, you'll fail. Too few, and the app loses its edge. Too short-term, you feel good but fail on longer-term objectives.
If you're getting started, try 5-8 and mix short term and long term and see how you go. Tracking your habits is strangely also a habit you need to build. There're no right answers, but if this is new, focus on foundational keystone habits that will feed future ones. If you're coming into this cold, I'd also recommend a good read of James Clear's Atomic Habits.
Here are some ideas of what to track:
- Went to bed on time
- Got X hours of sleep
- Inbox zero
- Practiced Italian
- Morning Pages
- Blogged
- Studied astrophysics
- Socialized
- Stuck to budget
- Coded
- 2 coffees only
- Thanked someone
- Went for a walk
- Told SO they're amazing
Simply run harsh ask regularly, specify whether you did the habit from the
prompt on a particular day (or needed to skip the habit for some reason - eg.
could not clean apartment because you were away for week), and get pretty
graphs!
As of v0.10.6 you can also add a fragment of a habit after harsh ask which
will then only ask you about that the outcomes for that single habit.
Taking the example of the Pullups habit below, if you had Pullups: 5/7
in your habit file, typing harsh ask pul (or any variations of the habit
name's string) will ask you only about the outcome for the unanswered todos
for that habit (an oft-requested feature).
harsh ask allows you to pick between [y/n/s/β] which is yes/no/skip/don't
answer right now. CTRL-c breaks you out of the ask cycle at any point and
returns you to your prompt.
As of version 0.9.0, to support longer term pattern tracking (and fallible
memories), you can follow any of the y | n | s options with an optional typed
@ symbol to then denote a quantity you want to track for the daily habit
(example: number of words written, km's run, pullups performed etc), and/or an
optional typed # symbol for a comment. Primarily for analysis at a later date.
The log stats subcommand will also now total up any amounts you've entered for
a habit and show you the total along with your streaks, skips, breaks, and days
tracked.
An example of how to use this for a habit like PullUps might be:
$ harsh ask
2024-01-05:
Dailies
PullUps β β β ββ β β β β βββββββββββ β β β[y/n/s/β] y @ 15 # Crushed workout today!This will, besides adding y to the log file, also record "15" and the comment
"Crushed workout today" to your log file. The feature is backwards compatible
even with older log files so even if you don't use it, you're alright. 0.9.0
will also parse older logs. You must use a valid number in the @ position and
the only other caveat is if you use both an @ and # in your prompt response,
the @ must come before the #. The only disallowed character in a prompt is
: as it is used as a field separator in the log files (to make for easy
importing and parsing by spreadsheet and other tools like pandas.)
The annotating features are primarily for analysis at a later date to help uncover patterns between events, reasons you may have written a comment, and good or bad knock-on effects.
Personally, I use comments sparingly and to denote why you had to skip, broke a
consistency chain with an n, or for when you're trying to figure out
something notable on a day so when you look back you can see why habit X may
or may not have succeeded.
As of v0.10.6, harsh log supports the same fragment querying harsh ask
does. You can add a fragment of a habit after harsh log which will then only
show you the consistency graph outcomes for that single habit. This supercedes
the harsh log check <habit substring> subcommand (which is not deprecated).
Taking the example of the Pullups habit mentioned above, if you typed harsh log pul (or any variations of the habit name's string) you'll see just the
consistency graph for that single habit along along with the sparkline for all
habits and usual log command metrics.
harsh log by itself shows your consistency graph for the last 100 days.
$ harsh ask
2020-01-05:
Dailies
Meditated β β β ββ β β β β βββββββββββ β β β[y/n/s/β] y
Weeklies
Cleaned the apartment βββββββ βββββββ βββββββ β’Β·Β·Β·Β·Β·Β·[y/n/s/β] n
Tracking
Had a headache β β ββ ββ β ββ [y/n/s/β] n
Used harsh β βββ β βββ β β β β β β β β ββ β β ββββ β [y/n/s/β] y
... some habits omitted ...(Some weeks later)
$ harsh log
βββββββββββ
ββββββββββββββ
ββββ
ββ
βββββββ
ββ
β
β
βββββββ
β
βββ
βββββββ
β
Meditated β β β ββ β β β β βββββββββββ β β ββ
Cleaned the apartment βββββββ βββββββ βββββββ β’Β·Β·Β·Β·Β·Β·
Had a headache β β ββ ββ β ββ
Used harsh β βββ β βββ β β β β β β β β ββ β β ββββ β β
... some habits omitted ...
Yesterday's score: 88.3%The sparkline at the top give a graphical representation of each day's score. The M W F stands for M(onday), W(ednesday), and F(riday) to provide visual hinting as to which days are which and possibly help diagnoze whether particular days are villains or heros in building your habits (How is this ueful? As an example, because of then way I had meetings structured on Tuesday, I often fell off the truck on certain habits on Tuesdays. Seeing this clearly in the graphs let me restructure my Tuesday work days to get me back on track.)
The score at the bottom specifies how many of your habits you met that previous day of total possible and removes any you may have skipped from the calculation.
Run harsh log stats gives an analysis of your entire log file and a quantified
idea of days you've been on streak, how many days you broke your consistency
chain, and how many days you may have skipped -- as well as the total number of
days you've been tracking a particular habit for (note: I swap out my file every
year, but at least one person wanted this feature to track over 800+ days of log
files they'd converted from another app.). It's also nicely coloured to help
visually separate the information.
In particular, when you've been tracking for longer than 100 days (the visual length of the consistency graph), you can get summary stats for your entire log (tbh, I did not think this was useful until I implemented it. I was wrong.). This can be surprisingly useful to see longer trends quantified.
Slept 7h+ Streaks 173 days Breaks 147 days Skips 1 days Tracked 320 days
Morning Pages Streaks 310 days Breaks 9 days Skips 2 days Tracked 320 days
Share daily Streaks 320 days Breaks 0 days Skips 1 days Tracked 320 days
Workouts Streaks 246 days Breaks 27 days Skips 48 days Tracked 320 days
Read Daily Streaks 302 days Breaks 18 days Skips 0 days Tracked 320 days
Write daily Streaks 320 days Breaks 0 days Skips 1 days Tracked 320 days
TIL Streaks 285 days Breaks 26 days Skips 9 days Tracked 320 days
Running Streaks 117 days Breaks 56 days Skips 148 days Tracked 320 days
Blog Fortnightly Streaks 314 days Breaks 6 days Skips 0 days Tracked 320 days
...As you can see here, I need to work on sleep more than anything, but digging down on these stats I shocked myself at skipped and breaks in workouts (covid vaccine related in some cases), and how I need to rethink how some of these were set up or I'm doing them (running, blogging etc.). The point is not how terrible I am, but that looking into the numbers revealed patterns (sleep, affects workouts, running, and TIL - today I learned - rather terribly). YMMV.
Run harsh log <habit search term> gives a slightly more in depth analysis of
individual habits in conjunction with your topline aparkline. The idea here is
that you can examine individual habits graphically against your topline to see
if there are patterns of correlation between the variation in an individual
habit and your overall daily score over time.
Say you expect there's a close correlation between you not getting good sleep
and your habits generally falling from a state of grace. It's easy to check
with a log <habit substring> and any string you type from your habits. In my
case, I have a habit around getting to bed on time and getting a solid 7 I call
Bed by 12+Slept 7h+. I can just use the handy bed as a short match term for
the habit and get the result.
$ harsh log bed
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
Bed by 12+Slept 7h+ β β β ββββ βββ ββ β βββ ββ’ ββββββββ ββββ ββββ
As we can see, there is a pretty close correlation here to not getting enough sleep (or going to bed too late) and me hitting all my daily habits.
A done habit gives you a nice bright β on the consistency graph line. It's done.
Additionally, the app checks in future days if you are still within the "every
x days" period of performing the habit by drawing a dimmer β after the done
marker to let you know you've satisfied the requirement for that habit.
Sometimes, it's impossible to exercise a habit cause life happens. If cleaning
the house is a habit you want to exercise, but you happen to be away on
a business trip, that is an impossibility. And sometimes, you decide to skip and
push the habit to the next period (or a simple day or so). Skips being selected
(s in the prompt) allows this to happen. A skip is denoted by a bright β’.
Much like satisfied habits where you've performed them once in the period,
"skipified" habits let you know you're still withing the calculated grace period
of the skip with a lighter dot Β·.
harsh also has a warnings feature to help flag to you when you're in danger of breaking your consistency graph. Harsh will give you a warning by showing a "!" symbol in your upcoming habits.
For habits of every less than 7 days period, you get a warning sigil on the day
the chain will break if you do not perform the habit. For a week or longer,
you'll start to see a warning sigil of 1 + days/7 rounded down (eg. so,
2 weeks' warning would get you the sigil 3 days ahead of breaking the chain
etc.).
Enter harsh help if you're lost:
Ξ» ~/harsh help
NAME:
Harsh - habit tracking for geeks
USAGE:
harsh [global options] command [command options] [arguments...]
VERSION:
0.10.0
DESCRIPTION:
A simple, minimalist CLI for tracking and understanding habits.
COMMANDS:
ask, a Asks and records your undone habits
log, l Shows graph of habits
todo, t Shows undone habits for today.
help, h Shows a list of commands or help for one command
GLOBAL OPTIONS:
--help, -h show help (default: false)
--no-color, -n no colors in output (default: false)
--version, -v print the version (default: false)
As you increasingly use harsh for tracking, you'll inevitably end up making
small errors in your log file or will want to edit your config file. While
people have suggested adding a harsh config command to the app, this feels
like overkill for the intended audience (geeks) but you can get the same effect
through using aliases.
As a simple quality of life improvement, add the following to your bash,
zsh, fish (what I use), or shell of choice:
alias h="harsh"
alias hc="nvim ~/.config/harsh/habits"
alias hl="nvim ~/.config/harsh/log"For me, this equates to me ending up typing h log for the graph, h ask etc
etc. When I need to edit the log file because I was a bit too itchy with my
trigger finger (or decide I should add a note), hl is my friend.
New from 0.8.22: if you are logging the output of harsh log stat and other
features which contains colour codes, you can instead use --no-color flag via
harsh --no-color [command] (or harsh -n [command]) to suppress colourized
output. This is also very helpful in n/vim with the :.! harsh -n log stats
command stanza to record harsh's output to n/vim's buffer to augment your
weekly, monthly, or daily tracking (learned that vim snippet from the feature
requester!).
Much like the above feature of accessing config and log files, you can alias
harsh -n in your shell if you prefer to suppress all colour output coming from
all harsh commands.
harsh is free software. You can redistribute it and/or modify it under the terms of the MIT License.
Primo, check out the Contributing guidelines.
- Fork it (https://github.com/wakatara/harsh/fork)
- Create your feature branch (
git checkout -b my-new-feature) - Commit your changes (
git commit -am 'Add some feature') - Push to the branch (
git push origin my-new-feature) - Create a new Pull Request
- Daryl Manning - creator, maintainer, and evil mastermind
- Aisling - improved configuration discovery.
- vchslv13 - improved shell installation script.
- manu-cyber - documentation fixes.