Installing Beancount

Martin Blais - Updated: June 2024

http://furius.ca/beancount/doc/install

Instructions for downloading and installing Beancount on your computer.

Releases

Beancount is a mature project: the first version was written in 2008. The current version of Beancount — branch "v3" — is stable and under continued maintenance and development. There is a mailing-list and a PyPI page. The "master" branch is the development branch.

(Note: If you used the "v2" branch in the past, many of the tools have been removed from branch "v3" and moved to their own dedicated github projects under http://github.com/beancount. Some of the tools, e.g. bean-report, bean-web, have been deprecated.)

Where to Get It

This is the official location for the source code:

https://github.com/beancount/beancount

Download it like this, by using Git to make a clone on your machine:

git clone https://github.com/beancount/beancount

How to Install

Installing Python

Beancount uses Python 3.8 or above, which is a pretty recent version of Python (as of this writing), and a few common library dependencies. I try to minimize dependencies, but you do have to install a few. This is very easy.

First, you should have a working Python install. Install the latest stable version >=3.8 using the download from python.org. Make sure you have the development headers and libraries installed as well (e.g., the “Python.h” header file). For example, on a Debian/Ubuntu system you would install the python3-dev package.

Beancount supports setuptools since Feb 2016, and you will need to install dependencies. You will want to have the “pip3” tool installed. It’s installed by default along with Python3 by now—test this out by invoking “python3 -m pip --help” command.

Installing Beancount

Installing Beancount using pipx

If you haven’t already, first install pipx.

Next, use pipx to install Beancount. This should automatically download and install all the dependencies in a virtual environment.

pipx install beancount

Confirm Beancount has been installed.

pipx list

Installing Beancount using pip

Installing Beancount using pip is no longer recommended, however you may install Beancount using

sudo -H python3 -m pip install beancount

This should automatically download and install all the dependencies.

Note, however, that this will install the latest version that was pushed to the PyPI repository and not the very latest version available from source. Releases to PyPI are made sporadically but frequently enough not to be too far behind. Consult the CHANGES file if you’d like to find out what is not included since the release date.

Warning: Proceed with caution this may break system libraries.

Installing Beancount using pip from the Repository

You can also use pip to install Beancount from its source code repository directly:

sudo -H python3 -m pip install git+https://github.com/beancount/beancount#egg=beancount

Installing Beancount from Source

Installing from source offers the advantage of providing you with the very latest version of the stable branch (“master”). The master branch should be as stable as the released version most of the time.

Get the source code from the official repository:

git clone https://github.com/beancount/beancount

You might need to install some non-Python library dependencies, such as bison and flex and perhaps a few more (you'll find out when you try to build). It should be obvious what’s missing. If on Ubuntu, use apt get to install those.

If installing on Windows, see the Windows section below.

Build and Install Beancount from source using pip3

You can then install all the dependencies and Beancount itself using pip:

cd beancount
sudo -H python3 -m pip install .
Installing for Development

If you want to execute the source in-place for making changes to it, you can use the setuptools “develop” command to point to it:

cd beancount
sudo python3 setup.py develop

Warning: This modifies a .pth file in your Python installation to point to the path to your clone. You may or may not want this. I don't do this myself; the way I work with it is the "old school" way; I just build it locally and modify my shell's environment to find its libraries. You build it like this:

cd beancount
python3 setup.py build_ext -i  # or "make build"

and then both the PATH and PYTHONPATH environment variables need to be updated for it like this:

export PATH=$PATH:/path/to/beancount/bin
export PYTHONPATH=$PYTHONPATH:/path/to/beancount
Dependencies for Development

Beancount needs a few more tools for development. If you’re reading this, you’re a developer, so I’ll assume you can figure out how to install packages, skip the detail, and just list what you might need:

  • pytest: for unit tests

  • ruff: for linting

  • GNU flex: This lexer generator is needed if you intend to modify the lexer. It generated C code that chops up the input files into tokens for the parser to consume.

  • GNU bison: This old-school parser generator is needed if you intend to modify the grammar. It generates C code that parses the tokens generated by flex. (I like using old tools like this because they are pretty low on dependencies, just C code. It should work everywhere.)`

  • python-dateutil : to run the beancount.scripts.example example generator script.

Installing from Distribution Packages

Various distributions may package Beancount. Here are links to those known to exist:

Windows Installation

Native

Installing this package by pip requires compiling some C++ code during the installation procedure which is only possible if an appropriate compiler is available on the computer, otherwise you will receive an error message about missing vsvarsall.bat or cl.exe.

To be able to compile C++ code for Python you will need to install the same major version of the C++ compiler as your Python installation was compiled with. By running python in a console and looking for a text similar to [MSC v.1900 64 bit (AMD64)] you can determine which compiler was used for your particular Python distribution. In this example it is v.1900.

Using this number find the required Visual C++ version here. Since different versions seem to be compatible as long as the first two digits are the same you can in theory use any Visual C++ compiler between 1900 and 1999.

According to my experience both Python 3.8 and 3.6 was compiled with MSC v.1900 so you can do either of the following to satisfy this requirement:

  • Install the standalone Build Tools for Visual Studio 2017 or

  • Install the standalone Visual C++ Build Tools 2015 or

  • Modify an existing Visual Studio 2017 installation

    • Start the Visual Studio 2017 installer from Add or remove programs

    • Select Individual components

    • Check VC++ 2017 version 15.9 v14.16 latest v141 tools or newer under Compilers, build tools, and runtimes

    • Install

  • Visual Studio 2019

    • add C++ build tools: C++ core features, MSVC v142 build tools

If cl.exe is not in your path after installation, run Developer Command Prompt for Visual Studio and run the commands there.

With Cygwin

Windows installation is, of course, a bit different. It’s a breeze if you use Cygwin. You just have to prep your machine first. Here’s how.

  • Install the latest Cygwin. This may take a while (it downloads a lot of stuff), but it is well worth it in any case. But before you kick off the install, make sure the following packages are all manually enabled in the interface provided by setup.exe (they’re not selected by default):

    • python3

    • python3-devel

    • python3-setuptools

    • git

    • make

    • gcc-core

    • flex

    • bison

  • Start a new Cygwin bash shell (there should be a new icon on your desktop) and install the pip3 installer tool by running this command:
    easy_install-3.4 pip
    Make sure you invoke the version of easy_install which matches your Python version, e.g. easy_install-3.8 if you have Python 3.8 installed, or more.

At this point, you should be able to follow the instructions from the previous sections as is, starting from “Install Beancount using pip”.

With WSL

The newly released Windows 10 Anniversary Update brings WSL 'Windows Subsystem for Linux' with bash on Ubuntu on Windows (installation instructions and more at https://msdn.microsoft.com/commandline/wsl/about).

This makes beancount installation easy, from bash:

sudo apt-get install python3-pip
sudo pip3 install m3-cdecimal
sudo pip3 install beancount --pre

This is not totally “Windows compatible”, as it is running in a pico-process, but provides a convenient way to get the Linux command-line experience on Windows. (Contrib: willwray)

Checking your Install

You should be able to run the binaries from this document. For example, running bean-check should produce something like this:

$ bean-check

usage: bean-check [-h] [-v] filename
bean-check: error: the following arguments are required: filename

If this works, you can now go to the tutorial and begin learning how Beancount works.

Reporting Problems

If you need to report a problem, either send email on the mailing-list or file a ticket on Github. Running the following command lists the presence and versions of dependencies installed on your computer and it might be useful to include the output of this command in your bug report:

python3 -m beancount.scripts.deps

Editor Support

There is support for some editors available:

If You Have Problems

If you run into any installation problems, file a ticket or email the mailing-list.

Post-Installation Usage

Normally the scripts located under beancount/bin should be automatically installed somewhere in your executable path. For example, you should be able to just run “bean-check” on the command-line.