Sharing Computational Environments with Pixi

Agenda

Duration	Content
15min	Introduction and Installation
5min	Break
15min	Learner example
5min	Break
15min	Instructor example

Computational environment

Shareable

Why?

Operating system (OS) and programming language agnostic.
Avoid user privilege escalation, i.e. sudo.
Avoid users to have to walk pass the dependency hell.
- Long chains of dependencies
- Conflicting dependencies
Avoid users reporting bugs that only work on their computer.
Share alias for the command line interface (cli).

File system used by Pixi

File system used by Conda

Use containers instead

If your project has one of the following requirements

run two database serves simultaneous
use a different version of C standard library

you should use containers (for example, Podman, Docker, Apptainer, …) instead of Pixi.

Also, you must use containers if you want complete isolation from the host machine.

How?

Pixi uses its own package repository.
Pixi uses its own copy of the dependencies.
Pixi locks the full dependency tree.
Pixi will check if all the dependencies are satisfied before executing tasks.

Requirements

Package repository

Because Pixi primarily install packages from conda-forge.
Computation (as in number of CPU)

Because Pixi will check the dependencies before each task execution.
Storage (as in size of HDD)

Because Pixi will duplicate the dependencies for each project.
Transfer (as in speed of Internet connection)

Because some dependencies can be in the 3 digits MB.

Where are we

	1995 (Windows 95)	2009 (Windows 7)	2021 (Windows 11)
CPU	Intel® 200 MHz Pentium®	Intel® Core™ 2.6GHz	Intel® Core™ 5.2GHz
RAM	32 MB	3GB	16GB
HDD	3.8 GB	500GB	1TB
Internet	30 Kb/s	10 Mb/s	100 Mb/s

`pixi.toml`

Pixi uses the file pixi.toml as the human readable configuration file.

Tip

pixi.toml can be created using

pixi init

Example of `pixi.toml`

[workspace]
authors = ["Jane Doe <Jane.Doe@digital-research.academy>"]
channels = ["conda-forge"]
name = "example"
platforms = ["linux-64"]
version = "0.1.0"

[tasks]

[dependencies]
python = ">=3.14.0,<3.15"

`pixi.lock`

Pixi uses the file pixi.lock as the machine readable configuration file and includes a full description of the environment.

Important

pixi.lock must be included in your version controlled repository.

Example of `pixi.lock`

version: 6
environments:
  default:
    channels:
    - url: https://conda.anaconda.org/conda-forge/
    packages:
      linux-64:
      ...
      - conda: https://conda.anaconda.org/conda-forge/linux-64/openssl-3.6.0-h26f9b46_0.conda
      - conda: https://conda.anaconda.org/conda-forge/linux-64/python-3.14.0-h32b2ec7_102_cp314.conda
      ...
packages:
...
- conda: https://conda.anaconda.org/conda-forge/linux-64/openssl-3.6.0-h26f9b46_0.conda
  sha256: a47271202f4518a484956968335b2521409c8173e123ab381e775c358c67fe6d
  md5: 9ee58d5c534af06558933af3c845a780
  depends:
  - __glibc >=2.17,<3.0.a0
  - ca-certificates
  - libgcc >=14
  license: Apache-2.0
  license_family: Apache
  size: 3165399
  timestamp: 1762839186699
- conda: https://conda.anaconda.org/conda-forge/linux-64/python-3.14.0-h32b2ec7_102_cp314.conda
  build_number: 102
  sha256: 76d750045b94fded676323bfd01975a26a474023635735773d0e4d80aaa72518
  md5: 0a19d2cc6eb15881889b0c6fa7d6a78d
  depends:
  ...
  license: Python-2.0
  size: 36681389
  timestamp: 1761176838143
  python_site_packages_path: lib/python3.14/site-packages

`.pixi`

Pixi uses the directory .pixi to store a working copy of all the dependencies of the project.

Tip

.pixi must be excluded from your version controled repository.

Pixi will configure your .gitignore when you run

pixi init

Example of `.pixi`

.pixi/
└── envs
    └── default
        ├── bin
        ⋮   ⋮
        │   ├── openssl
        │   ├── python -> python3.14
        │   ├── python3 -> python3.14
        │   ├── python3.1 -> python3.14
        │   ├── python3.14
        ⋮   ⋮
        ├── include
        ⋮   ⋮
        │   ├── openssl
        │   │   ├── aes.h
        ⋮   ⋮    ⋮
        │   ├── python3.14
        │   │   ├── abstract.h
        ⋮   ⋮    ⋮
        ├── lib
        │   ├── libpython3.14.so -> libpython3.14.so.1.0
        │   ├── libpython3.14.so.1.0
        │   ├── libpython3.so
        ⋮   ⋮
        │   ├── python3.14
        │   │   ├── abc.py
        ⋮   ⋮    ⋮

Installation

Windows

On PowerShell, run

powershell -ExecutionPolicy ByPass `
  -c "irm -useb https://pixi.sh/install.ps1 | iex"

Pay attention to the backtick (`) at the end of the first line.

Unix-like

On your terminal emulator, run

curl -fsSL https://pixi.sh/install.sh | sh

Other Methods

Visit the Installation section on Pixi’s documentation.

Sarah’s use case

Load course material
Learn programming using Jupyter Notebook
Data analysis with R
Data visualisation with ggplot2

Sarah’s use case - Cheat sheet

Copy course material:

git clone https://github.com/rgaiacs/dra-ttt-course-pixi-mwe.git

Change directory
```
cd dra-ttt-course-pixi-mwe
```
Check tasks created by instructor:
```
pixi task list
```
Start Jupyter Lab:
```
pixi run lab
```

Convert Notebook to R script

pixi run jupyter nbconvert notes.ipynb --to script

Renata’s use case

Course development
Data analysis with R
Data visualisation with ggplot2

Renata’s use case - Cheat sheet

Start environment:
```
pixi init
```

Add dependencies:

pixi add jupyterlab r-base r-ggplot2 xeus-r

Create alias for tasks:

pixi task add lab "jupyter lab --no-browser"

Share files:

git add . && git commit -m "Add pixi config" && git push origin main

Environment activation - Cheat sheet

Activate environment:
```
pixi shell
```
Run R command:
```
Rscript -e "1 + 1"
```

Important

Pixi does not isolate the environment. Tools installed globally will be available.

More features

Integration with
- VSCode
- RStudio
- JupyterLab
- GitHub Actions
- …
Support to multi platform.

For example, it is possible to specify Python 3.14 for GNU/Linux and Python 3.13 for Windows.

That’s All Folks

Read Pixi’s documentation.
Ask question at Pixi’s Discussions hosted on GitHub.

Multi Environment (1/3)

Create hello.py with

try:
    print("Hello!")
except RuntimeError, ValueError:
    print('Except.')

Start environment:
```
pixi init
```
Create task:
```
pixi task add hello "python hello.py"
```

Multi Environment (2/3)

Edit pixi.toml to include

[feature.py313.dependencies]
python = "~=3.13.0"

[feature.py314.dependencies]
python = "~=3.14.0"

[environments]
py313 = ["py313"]
py314 = ["py314"]

Multi Environment (3/3)

Run task:
```
pixi run -e py314 hello
```
Run task:
```
pixi run -e py313 hello
```

Camila’s use case

New application written in C

Camila’s C code

#include <stdio.h>

int main() {
    printf("Hello World");

    return 0;
}

Camila’s use case - Cheat sheet

Start environment:
```
pixi init
```
Add dependencies:
```
pixi add gcc
```

Create task:

pixi task add make "gcc -o hello hello.c"

Run task:
```
pixi run make
```

Caution when creating tasks

Pixi uses Deno’s implementation of shell.

Start environment:
```
pixi init
```
Create task:
```
pixi task add myhead "head --help"
```
Run task:
```
pixi run myhead
```

Paige’s use case

Image processing with Python
Jupyter Notebook for exploratory analysis

Paige’s use case - Cheat sheet

Start environment:
```
pixi init --format pyproject
```

Add dependencies:

pixi add Pillow requests jupyter notebook

Create task:
```
pixi run jupyter notebook
```

Paige’s use case - Jupyter Notebook

import requests
from PIL import Image
from io import BytesIO

with requests.get(
    'https://github.com/python-pillow/Pillow/' +
    'blob/main/Tests/images/hopper.jpg?raw=true'
    ) as r:
    if r.status_code == 200:
        hopper = Image.open(BytesIO(r.content))

hopper

Sharing Computational Environments with Pixi

Agenda

Computational environment

Shareable

Why?

File system used by Pixi

File system used by Conda

Use containers instead

How?

Requirements

Where are we

pixi.toml

Example of pixi.toml

pixi.lock

Example of pixi.lock

.pixi

Example of .pixi

Installation

Windows

Unix-like

Other Methods

Sarah’s use case

Sarah’s use case - Cheat sheet

Renata’s use case

Renata’s use case - Cheat sheet

Environment activation - Cheat sheet

More features

That’s All Folks

Multi Environment (1/3)

Multi Environment (2/3)

Multi Environment (3/3)

Camila’s use case

Camila’s C code

Camila’s use case - Cheat sheet

Caution when creating tasks

Paige’s use case

Paige’s use case - Cheat sheet

Paige’s use case - Jupyter Notebook

`pixi.toml`

Example of `pixi.toml`

`pixi.lock`

Example of `pixi.lock`

`.pixi`

Example of `.pixi`