Package a Textual app with Hatch¶

Python apps may be distributed via PyPI so they can be installed via pip. This is known as packaging. The packaging process for Textual apps is much the same as any Python library, with the additional requirement that we can launch our app from the command line.

In this How To we will cover how to use Hatch to package an example application.

Hatch is a build tool (a command line app to assist with packaging). You could use any build tool to package a Textual app (such as Poetry for example), but Hatch is a good choice given its large feature set and ease of use.

Foreword¶

Packaging with Python can be a little intimidating if you haven't tackled it before. But it's not all that complicated. When you have been through it once or twice, you should find it fairly straightforward.

Example repository¶

See the textual-calculator-hatch repository for the project created in this How To.

The example app¶

To demonstrate packaging we are going to take the calculator example from the examples directory, and publish it to PyPI. The end goal is to allow a user to install it with pip:

pip install textual-calculator

Then launch the app from the command line:

calculator

Installing Hatch¶

There are a few ways to install Hatch. See the official docs on installation for the best method for your operating system.

Once installed, you should have the hatch command available on the command line. Run the following to check Hatch was installed correctly:

hatch

Hatch new¶

Hatch can create an initial directory structure and files with the new subcommand. Enter hatch new followed by the name of your project. For the calculator example, the name will be "textual calculator":

hatch new "textual calculator"

This will create the following directory structure:

textual-calculator
├── LICENSE.txt
├── README.md
├── pyproject.toml
├── src
│   └── textual_calculator
│       ├── __about__.py
│       └── __init__.py
└── tests
    └── __init__.py

This follows a well established convention when packaging Python code, and will create the following files:

• LICENSE.txt contains the license you want to distribute your code under.
• README.md is a markdown file containing information about your project, which will be displayed in PyPI and Github (if you use it). You can edit this with information about your app and how to use it.
• pyproject.toml is a TOML file which contains metadata (additional information) about your project and how to package it. This is a Python standard. This file may be edited manually or by a build tool (such as Hatch).
• src/textual_calculator/__about__.py contains the version number of your app. You should update this when you release new versions.
• src/textual_calculator/__init__.py and tests/__init__py indicate the directory they are within contains Python code (these files are often empty).

In the top level is a directory called src. This should contain a directory named after your project, and will be the name your code can be imported from. In our example, this directory is textual_calculator so we can do import textual_calculator in Python code.

Additionally, there is a tests directory where you can add any test code.

More on naming¶

Note how Hatch replaced the space in the project name with a hyphen (i.e. textual-calculator), but the directory in src with an underscore (i.e. textual_calculator). This is because the directory in src contains the Python module, and a hyphen is not legal in a Python import. The top-level directory doesn't have this restriction and uses a hyphen, which is more typical for a directory name.

Bear this in mind if your project name contains spaces.

Got existing code?¶

The hatch new command assumes you are starting from scratch. If you have existing code you would like to package, navigate to your directory and run the following command (replace <YOUR ROJECT NAME> with the name of your project):

hatch new --init <YOUR PROJECT NAME>

This will generate a pyproject.toml in the current directory.

Adding code¶

Your code should reside inside src/<PROJECT NAME>. For the calculator example we will copy calculator.py and calculator.tcss into the src/textual_calculator directory, so our directory will look like the following:

textual-calculator
├── LICENSE.txt
├── README.md
├── pyproject.toml
├── src
│   └── textual_calculator
│       ├── __about__.py
│       ├── __init__.py
│       ├── calculator.py
│       └── calculator.tcss
└── tests
    └── __init__.py

Adding dependencies¶

Your Textual app will likely depend on other Python libraries (at the very least Textual itself). We need to list these in pyproject.toml to ensure that these dependencies are installed alongside your app.

In pyproject.toml there should be a section beginning with [project], which will look something like the following:

[project]
name = "textual-calculator"
dynamic = ["version"]
description = 'A example app'
readme = "README.md"
requires-python = ">=3.8"
license = "MIT"
keywords = []
authors = [
  { name = "Will McGugan", email = "redacted@textualize.io" },
]
classifiers = [
  "Development Status :: 4 - Beta",
  "Programming Language :: Python",
  "Programming Language :: Python :: 3.8",
  "Programming Language :: Python :: 3.9",
  "Programming Language :: Python :: 3.10",
  "Programming Language :: Python :: 3.11",
  "Programming Language :: Python :: 3.12",
  "Programming Language :: Python :: Implementation :: CPython",
  "Programming Language :: Python :: Implementation :: PyPy",
]
dependencies = []

We are interested in the dependencies value, which should list the app's dependencies. If you want a particular version of a project you can add == followed by the version.

For the calculator, the only dependency is Textual. We can add Textual by modifying the following line:

dependencies = ["textual==0.47.1"]

At the time of writing, the latest Textual is 0.47.1. The entry in dependencies will ensure we get that particular version, even when newer versions are released.

See the Hatch docs for more information on specifying dependencies.

Environments¶

A common problem when working with Python code is managing multiple projects with different dependencies. For instance, if we had another app that used version 0.40.0 of Textual, it may break if we installed version 0.47.1.

The standard way of solving this is with virtual environments (or venvs), which allow each project to have its own set of dependencies. Hatch can create virtual environments for us, and makes working with them very easy.

To create a new virtual environment, navigate to the directory with the pyproject.toml file and run the following command (this is only require once, as the virtual environment will persist):

hatch env create

Then run the following command to activate the virtual environment:

hatch shell

If you run python now, it will have our app and its dependencies available for import:

$ python
Python 3.11.1 (main, Jan  1 2023, 10:28:48) [Clang 14.0.0 (clang-1400.0.29.202)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from textual_calculator import calculator

Running the app¶

You can launch the calculator from the command line with the following command:

python -m textual_calculator.calculator

The -m switch tells Python to import the module and run it.

Although you can run your app this way (and it is fine for development), it's not ideal for sharing. It would be preferable to have a dedicated command to launch the app, so the user can easily run it from the command line. To do that, we will need to add an entry point to pyproject.toml

Entry points¶

An entry point is a function in your project that can be run from the command line. For our calculator example, we first need to create a function that will run the app. Add the following file to the src/textual_calculator folder, and name it entry_points.py:

from textual_calculator.calculator import CalculatorApp


def calculator():
    app = CalculatorApp()
    app.run()

Then edit pyproject.toml to add the following section:

[project.scripts]
calculator = "textual_calculator.entry_points:calculator"

Each entry in the [project.scripts] section (there can be more than one) maps a command on to an import and function name. In the second line above, before the = character, calculator is the name of the command. The string after the = character contains the name of the import (textual_calculator.entry_points), followed by a colon (:), and then the name of the function (also called calculator).

Specifying an entry point like this is equivalent to doing the following from the Python REPL:

>>> import textual_calculator.entry_points
>>> textual_calculator.entry_points.calculator()

To add the calculator command once you have edited pyproject.toml, run the following from the command line:

pip install -e .

Now you can launch the calculator from the command line as follows:

calculator

Building¶

Building produces archive files that contain your code. When you install a package via pip or other tool, it will download one of these archives.

To build your project with Hatch, change to the directory containing your pyproject.toml and run the hatch build subcommand:

cd textual-calculator
hatch build

After a moment, you should find that Hatch has created a dist (distribution) folder, which contains the project archive files. You don't typically need to use these files directly, but feel free to have a look at the directory contents.

Publishing¶

After your project has been successfully built you are ready to publish it to PyPI.

If you don't have a PyPI account, you can create one now. Be sure to follow the instructions to validate your email and set up 2FA (Two Factor Authentication).

Once you have an account, login to PyPI and go to the Account Settings tab. Scroll down and click the "Add API token" button. In the "Create API Token" form, create a token with name "Uploads" and select the "Entire project" scope, then click the "Create token" button.

Copy this API token (long string of random looking characters) somewhere safe. This API token is how PyPI authenticates uploads are for your account, so you should never share your API token or upload it to the internet.

Run the following command (replacing <YOUR API TOKEN> with the text generated in the previous step):

hatch publish -u __token__ -a <YOUR API TOKEN>

Hatch will upload the distribution files, and you should see a PyPI URL in the terminal.

Managing API Tokens¶

Creating an API token with the "all projects" permission is required for the first upload. You may want to generate a new API token with permissions to upload a single project when you upload a new version of your app (and delete the old one). This way if your token is leaked, it will only impact the one project.

Publishing new versions¶

If you have made changes to your app, and you want to publish the updates, you will need to update the version value in the __about__.py file, then repeat the build and publish steps.

Installing the calculator¶

From the user's point of view, they only need run the following command to install the calculator:

pip install textual_calculator

They will then be able to launch the calculator with the following command:

calculator

Pipx¶

A downside of installing apps this way is that unless the user has created a virtual environment, they may find it breaks other packages with conflicting dependencies.

A good solution to this issue is pipx which automatically creates virtual environments that won't conflict with any other Python commands. Once PipX is installed, you can advise users to install your app with the following command:

pipx install textual_calculator

This will install the calculator and the textual dependency as before, but without the potential of dependency conflicts.

Summary¶

1. Use a build system, such as Hatch.
2. Initialize your project with hatch new (or equivalent).
3. Write a function to run your app, if there isn't one already.
4. Add your dependencies and entry points to pyproject.toml.
5. Build your app with hatch build.
6. Publish your app with hatch publish.

If you have any problems packaging Textual apps, we are here to help!