Developer guide

Contributors of all levels are most welcome in the hledger project. This guide is action-oriented: below you'll find useful links, then procedures, then general info.

IRC Join #hledger (chat log; see also #ledger)
Mail list (Gmane)
Twitter #hledger, see also #ledgercli, #plaintextaccounting, @LedgerTips
hledger-web demo
hledger-api demo, in swagger editor
Trello old backlog/wishlist planning board
Github simonmichael/hledger (alias:, forks
commits, COMMITS!
open bugs, open wishes, open pull requests, all issues
issues with bounty tag, bountysource bounties, codemill bounties, codefund bounties
stars: (#99 of ~30k starred haskell projects in 2016/04, #71 in 2016/12, #65 in 2017/3)
Throughput Graph
Travis CI travis ubuntu build status
Appveyor CI disabled
Hackage packages: hledger-lib, hledger, hledger-ui, hledger-web, hledger-api, hledger-diff, hledger-interest, hledger-irr, *hledger*
GHC compatibility: hledger-lib, hledger, hledger-ui, hledger-web, hledger-api
reverse deps: hledger-lib, hledger, hledger-ui, hledger-web, hledger-api
Stackage build-constraints.yaml, open hledger-related issues
Debian source packages: haskell-hledger-lib, bugs, haskell-hledger, bugs, haskell-hledger-ui, bugs, haskell-hledger-web, bugs
binary packages:
 stable hledger, bugs, hledger-web, bugs
 testing hledger, bugs, hledger-web, bugs
 unstable hledger, bugs, hledger-ui, bugs, hledger-web, bugs
 experimental hledger, bugs, hledger-ui, bugs, hledger-web, bugs
 all *hledger*
popularity stats: hledger, hledger-ui, hledger-web
PTS help
Ubuntu source packages: haskell-hledger-lib, bugs, haskell-hledger, bugs, haskell-hledger-ui, bugs, haskell-hledger-web, bugs
binary packages: *hledger*
Gentoo hledger, hledger-web, *hledger*
Fedora hledger, *hledger*, hledger (package db), Haskell SIG
Nix hydra build status
Homebrew hledger
Sandstorm hledger web app & reviews, issues
Reference GHC Can I Use

How to..

Get help

For quick help or support, ask on the #hledger ( IRC channel. If you don't get an answer quickly (depends on time of day), you can leave the window open and check back later, type "sm:" to alert me, and/or leave your email address.

For slightly less quick help, but more eyeballs, ask on the mail list (

Also, always search the issue tracker. You may find an explanation or workaround, or when the fix will be released.

Search the issue tracker

The hledger project's issue tracker is on github. It contains:

Use these shortcut urls for quick access:

Labels are used to categorise:

Clicking blue topic labels is a good way to review issues in a topic you're interested in.

Milestones are used to organise releases. Click a milestone to see issues fixed in/planned for that release.

Time annotations: some issue names might have a suffix noting estimated and spent time. Basic format: [ESTIMATEDTOTALTASKTIME|TIMESPENTSOFAR]. Examples:

[2]       two hours estimated, no time spent
[..]      half an hour estimated (a dot is ~a quarter hour, as in timedot format)
[1d]      one day estimated (a day is ~4 hours)
[1w]      one week estimated (a week is ~5 days or ~20 hours)
[3|2]     three hours estimated, about two hours spent so far  
[1|1w|2d] first estimate one hour, second estimate one week, about two days spent so far 

Estimates are always for the total time cost (not time remaining). Estimates are not usually changed, a new estimate is added instead. Numbers are very approximate, but better than nothing.

Report bugs

If you found a similar bug report already in the tracker, add any new information there. Otherwise, open a new bug by clicking "New issue", or

Suggest enhancements

Some enhancement requests land in the bug tracker; these will get the WISH tag to avoid obscuring bugs. But for general brainstorming and idea capture, consider using

Do user testing

Install stack and git

stack is the recommended tool for building hledger from source. It builds haskell projects, installing required haskell libraries as needed. It can also install GHC (the compiler) and (on windows) git, if needed.

You don't need to use stack, if you are already expert with the older cabal tool, or even just GHC, but I won't attempt to document those procedures; these docs assume you have downloaded and installed stack.

On Windows, you should choose the 64-bit stack download if you will be processing >50,000 transactions at a time with hledger (#275).

git is the revision control tool you'll need to fetch the latest hledger source and submit changes. On windows, stack can install it for you. These docs assume you have installed git and know a little about how to use it.

Install other optional tools

Up-to-date alex, happy, and haddock tools are required, but stack should install those for you.

Here are some optional extra tools:

You can install them all with:

$ stack install shelltestrunner hasktags profiteur hpack hoogle

Get the latest hledger source

$ git clone hledger    # aka
$ cd hledger

Use the Makefile

A Makefile is provided to make common developer tasks easy to remember, and to insulate us a little from the ever-evolving Haskell tools ecosystem. Using it is entirely optional, but recommended. You'll need GNU Make installed.

The Makefile is self-documenting. Run make to see a list of the main make rules:

$ make
Makefile:37: -------------------- hledger make rules --------------------
Makefile:39: make [help] -- list documented rules in this makefile. make -n RULE shows more detail.
Makefile:204: (INSTALLING:)
Makefile:206: make install -- download dependencies and install hledger executables to ~/.local/bin or equivalent (with stack)
Makefile:231: (BUILDING:)
Makefile:235: make build -- download dependencies and build hledger executables (with stack)
Makefile:304: make hledgerdev -- quickly build the hledger executable (with ghc and -DDEVELOPMENT)

To see what a make rule will do without actually doing it, use the -n flag:

$ make build -n
stack build
$ make test -n
(stack test \
		&& echo pkgtest PASSED) || echo pkgtest FAILED
(stack exec hledger test \
		&& echo builtintest PASSED) || echo builtintest FAILED
(COLUMNS=80 PATH=`pwd`/bin:/home/simon/src/hledger/bin:/home/simon/.local/bin:/home/simon/.cabal/bin:/opt/ghc/7.10.1/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/var/lib/gems/1.9.1/bin stack exec -- shelltest --execdir -- -j16 --hide-successes tests \
		&& echo functest PASSED) || echo functest FAILED

Build or install hledger

Run make to see a list of build rules. You probably want build or install.

make build runs stack build, which downloads required haskell dependencies and builds all hledger packages. The resulting executables will be somewhere under .stack-work, eg in .stack-work/install/i386-linux/lts-3.0/7.10.2/bin/.

$ make build
stack build
hledger-lib-0.27: configure
hledger-lib-0.27: build
hledger-lib-0.27: install
hledger-0.27: configure
hledger-0.27: build
Progress: 1/4

Note stack will install required haskell libraries, but not C libraries such as curses or terminfo. If you get a build error, it is likely because one of these is missing, in which case you must identify and install it yourself using your system's package manager. This is usually a bit harder on Windows.

make install runs stack install, which does everything stack build does and also copies the executables to ~/.local/bin or the Windows equivalent. You should make sure this directory is in your $PATH, so that you can just type hledger to run the latest.

$ make install
stack install
NOTE: the install command is functionally equivalent to 'build --copy-bins'
hledger-0.27: build
Copied executables to /Users/simon/.local/bin/:
- hledger-web
- hledger-ui
- hledger

You can save time and effort by building just the package(s) you're interested in. To install just the hledger command-line tool, but not hledger-ui or (especially costly) hledger-web, do:

$ stack install hledger

(This looks like the download page command for installing the latest hledger release from Stackage. The difference is, here we are running it inside the hledger source tree, so the source version will be installed.)

Run benchmarks

Benchmarks are standard performance measurements, which we define using bench declarations in cabal files. There is one in hledger.cabal, with related code and data files in hledger/bench/.

To run the standard hledger benchmark, use stack bench hledger. This installs haskell dependencies (but not system dependencies) and rebuilds as needed, then runs hledger/bench/bench.hs, which by default shows quick elapsed-time measurements for several operations on a standard data file:

$ stack bench hledger
NOTE: the bench command is functionally equivalent to 'build --bench'
hledger-0.27: benchmarks
Running 1 benchmarks...
Benchmark bench: RUNNING...
Benchmarking hledger in /Users/simon/src/hledger/hledger with timeit
read bench/10000x1000x10.journal        [1.57s]
print                                   [1.29s]
register                                [1.92s]
balance                                 [0.21s]
stats                                   [0.23s]
Total: 5.22s
Benchmark bench: FINISH

bench.hs has some other modes, which you can use by compiling and running it directly. --criterion reports more detailed and dependable measurements, but takes longer:

$ cd hledger; stack exec -- ghc -ibench bench/bench && bench/bench --criterion
Linking bench/bench ...
Benchmarking hledger in /Users/simon/src/hledger/hledger with criterion
benchmarking read bench/10000x1000x10.journal
time                 1.414 s    (1.234 s .. 1.674 s)
                     0.996 R²   (0.989 R² .. 1.000 R²)
mean                 1.461 s    (1.422 s .. 1.497 s)
std dev              59.69 ms   (0.0 s .. 62.16 ms)
variance introduced by outliers: 19% (moderately inflated)

benchmarking print
time                 1.323 s    (1.279 s .. 1.385 s)
                     1.000 R²   (0.999 R² .. 1.000 R²)
mean                 1.305 s    (1.285 s .. 1.316 s)
std dev              17.20 ms   (0.0 s .. 19.14 ms)
variance introduced by outliers: 19% (moderately inflated)

benchmarking register
time                 1.995 s    (1.883 s .. 2.146 s)
                     0.999 R²   (0.998 R² .. NaN R²)
mean                 1.978 s    (1.951 s .. 1.995 s)
std dev              25.09 ms   (0.0 s .. 28.26 ms)
variance introduced by outliers: 19% (moderately inflated)

benchmarking balance
time                 251.3 ms   (237.6 ms .. 272.4 ms)
                     0.998 R²   (0.997 R² .. 1.000 R²)
mean                 260.4 ms   (254.3 ms .. 266.5 ms)
std dev              7.609 ms   (3.192 ms .. 9.638 ms)
variance introduced by outliers: 16% (moderately inflated)

benchmarking stats
time                 325.5 ms   (299.1 ms .. 347.2 ms)
                     0.997 R²   (0.985 R² .. 1.000 R²)
mean                 329.2 ms   (321.5 ms .. 339.6 ms)
std dev              11.08 ms   (2.646 ms .. 14.82 ms)
variance introduced by outliers: 16% (moderately inflated)

--simplebench shows a table of elapsed-time measurements for the commands defined in bench/default.bench. It can also show the results for multiple h/ledger executables side by side, if you tweak the bench.hs code. Unlike the other modes, it does not link with the hledger code directly, but runs the "hledger" executable found in $PATH (so ensure that's the one you intend to test).

$ cd hledger; stack exec -- ghc -ibench bench/bench && bench/bench --simplebench
Benchmarking /Users/simon/.local/bin/hledger in /Users/simon/src/hledger/hledger with simplebench and shell
Using bench/default.bench
Running 4 tests 1 times with 1 executables at 2015-08-23 16:58:59.128112 UTC:
1: hledger -f bench/10000x1000x10.journal print	[3.27s]
1: hledger -f bench/10000x1000x10.journal register	[3.65s]
1: hledger -f bench/10000x1000x10.journal balance	[2.06s]
1: hledger -f bench/10000x1000x10.journal stats	[2.13s]

Summary (best iteration):

|                                         || hledger |
| -f bench/10000x1000x10.journal print    ||    3.27 |
| -f bench/10000x1000x10.journal register ||    3.65 |
| -f bench/10000x1000x10.journal balance  ||    2.06 |
| -f bench/10000x1000x10.journal stats    ||    2.13 |

bench's --simplebench mode is based on a standalone tool, tools/simplebench.hs. simplebench.hs is a generic benchmarker of one or more executables (specified on the command line) against one or more sets of command-line arguments (specified in a file). It has a better command-line interface than bench.hs, so you may find it more convenient for comparing multiple hledger versions, or hledger and ledger. Eg:

$ stack exec -- ghc tools/simplebench
[1 of 1] Compiling Main             ( tools/simplebench.hs, tools/simplebench.o )
Linking tools/simplebench ...
$ tools/simplebench -h
tools/simplebench -h
simplebench: at least one executable needed
bench [-f testsfile] [-n iterations] [-p precision] executable1 [executable2 ...]

Run some functional tests with each of the specified executables,
where a test is "zero or more arguments supported by all executables",
and report the best execution times.

  -f testsfile   --testsfile=testsfile    file containing tests, one per line, default: bench.tests
  -n iterations  --iterations=iterations  number of test iterations to run, default: 2
  -p precision   --precision=precision    show times with this precision, default: 2
  -v             --verbose                show intermediate results
  -h             --help                   show this help

- executables may have arguments if enclosed in quotes
- tests can be commented out with #
- results are saved in benchresults.{html,txt}
cd hledger; $ ../tools/simplebench -f bench/default.bench hledger ledger
Using bench/default.bench
Running 4 tests 2 times with 2 executables at 2015-08-24 04:24:37.257068 UTC:

Summary (best iteration):

|                                         || hledger | ledger |
| -f bench/10000x1000x10.journal print    ||    3.24 |   0.43 |
| -f bench/10000x1000x10.journal register ||    3.80 |   3.48 |
| -f bench/10000x1000x10.journal balance  ||    2.05 |   0.18 |
| -f bench/10000x1000x10.journal stats    ||    2.10 |   0.19 |

Finally, for quick, fine-grained performance measurements when troubleshooting or optimising, I use dev.hs.

Generate sample journal files

Synthetic data files like examples/100x100x10.journal are useful for benchmarks and testing. The numbers describe the number of transactions, number of accounts, and maximum account depth respectively. They are generated by tools/generatejournal.hs. They should be built as needed, if not you can use make samplejournals rule:

$ make samplejournals
ghc tools/generatejournal.hs
[1 of 1] Compiling Main             ( tools/generatejournal.hs, tools/generatejournal.o )
Linking tools/generatejournal ...
tools/generatejournal 100 100 10 >examples/100x100x10.journal
tools/generatejournal 1000 1000 10 >examples/1000x1000x10.journal
tools/generatejournal 1000 10000 10 >examples/1000x10000x10.journal
tools/generatejournal 10000 1000 10 >examples/10000x1000x10.journal
tools/generatejournal 10000 10000 10 >examples/10000x10000x10.journal
tools/generatejournal 100000 1000 10 >examples/100000x1000x10.journal
tools/generatejournal 3 5 5 >examples/ascii.journal
tools/generatejournal 3 5 5 --chinese >examples/chinese.journal
tools/generatejournal 3 5 5 --mixed >examples/mixed.journal

Run tests

This command will install haskell dependencies (you might need to install additional system dependencies yourself) and run the package test suites. Currently these consist of hledger-lib's unit tests, hledger's unit tests, and hledger-web's functional tests:

$ stack test [PKG]

Run the hledger-lib and hledger unit tests as a built-in hledger command:

$ [stack exec] hledger test

Run the hledger functional tests:

$ stack install shelltestrunner  # if not already done
$ make functest

Run both unit and functional tests:

$ make test

Test haddock doc generation:

$ make haddocktest

Add a test

Use the REPL (GHCI)

These all work from the main hledger source directory (at least).

First, ensure all required dependencies are installed with these commands. (You might also need to install some system libs like terminfo or curses.)

$ stack test
$ stack bench

Get a GHCI prompt for hledger-lib:

$ cd hledger-lib; stack ghci hledger-lib

Changing into the package directory isn't actually needed, but it enables a custom .ghci script which sets a more useful short prompt.

Get a GHCI prompt for hledger:

$ cd hledger; stack ghci hledger

Get a GHCI prompt for hledger-ui:

$ cd hledger-ui; stack ghci hledger-ui

Get a GHCI prompt for hledger-web:

$ cd hledger-web; stack ghci hledger-web

hledger-web also needs to find some things in its current directory (like the static/ directory). This normally just works, if not please send details.

Improve the documentation

Fix a bug or add a feature

Get your changes accepted

Follow the usual github workflow:

If you're new to this process, may be useful.

Become a contributor

Do code review

Help with packaging

Help with project management

Do a major release

Do a minor release

Differences from a major release: work in a release branch, set PACKAGES only to the affected package(s), don't run make setversion.

  1. cleanup
    • review working copies (laptop, server, website) & branches, commit pending changes
  2. document
    • */*.cabal for affected package(s) (descriptions, tested-with, files..)
    • */CHANGES for affected package(s)
    • site/
    • site/ (commands, options, --help, ledger compatibility..)
    • site/
    • site/how-to-*
  3. test
    • make unittest
    • make functest
    • make haddocktest
  4. branch
    • switch to release branch (git checkout X.Y)
  5. version
    • edit .version (don't make setversion)
    • manually bump version for affected package(s): cabal files, manual..
  6. package
    • set Makefile's PACKAGES to affected package(s)
    • make cabalsdist
  7. test
    • install from tarball(s) into a clean directory
  8. tag
    • make tagrelease
  9. push
    • git push --tags
  10. upload
    • make cabalupload
  11. announce
    • [email hledger]
    • [tweet]

Project overview

Mission, principles, goals

The hledger project aims to produce:

Roles and activities


Project documentation lives in a number of places:


The hledger repo is hosted on Github, at You can also jump there via[/commits].

Quality control

Relevant tools include:

Code reviews

We have held one code review party, in July 2014, on the mail list and IRC channel. Here's the original proposal giving some motivation, and the discussion logs, note these are a good source of hledger development tips:

Dev sprint/party #2 was on 2015/10/10.

Implementation notes


There are two core cabal packages:

hledger-lib - data model, parsing, manipulation, standard reports (github)
hledger - command line interface, reusable cli options & helpers (github)

Most data types are defined in hledger-lib:Hledger.Data.Types, while functions that operate on them are defined in hledger-lib:Hledger.Data.TYPENAME. Here's a diagram of the main data model:

hledger parses the journal file into a Journal, which contains a list of Transactions, each containing multiple Postings of some MixedAmount (multiple single-CommoditySymbol Amounts) to some AccountName. Commands get and render Reports from the Journal, or sometimes from a Ledger, which contains Accounts representing the summed balances and other details of each account.

After surveying the packages, modules, and data types, try tracing the execution of a hledger command:

  1. CLI stuff is in hledger:Hledger.Cli.
  2. hledger:Hledger.Cli.Main:main parses the command line to select a command, then
  3. gives it to hledger:Hledger.Cli.Utils:withJournalDo, which runs it after doing all the initial parsing.
  4. Parsing code is under hledger-lib:Hledger.Read, eg the hledger-lib:Hledger.Read.JournalReader.
  5. Commands extract useful information from the parsed data model using hledger-lib:Hledger.Reports, and
  6. render it to the console.
  7. Everything uses the types and data utilities under hledger-lib:Hledger.Data, and the general helpers from hledger-lib:Hledger.Utils and below.


hledger-web is in a third cabal package:

hledger-web - web interface (github)

It is a single-executable web application using the yesod framework. It runs a built-in web server serving some views of the journal file, reading it at startup and again whenever it changes. It can also append new transactions to the journal file. There are two main views, which can be filtered with query arguments:

There is also:

Most of the action is in

Handler module and function names end with R, like the Yesod-generated route type they deal with.

Dynamically generated page content is mostly inline hamlet. Lucius/Julius files and widgets generally are not used, except for the default layout.

The quickest way to test changes is cd hledger-web; stack ghci hledger-web, :main --serve, control-C, :r, repeat. No linking is required, and changes to static files like hledger.js are visible after reloading a page.

Another way is yesod devel, which rebuilds automatically when files change, including config files, templates and static files (but only in the hledger-web package).

A third way is make autoweb, if you can get it working (see the makefile for instructions). This rebuilds automatically when haskell files change in any of the hledger{-lib,,-web} packages.