Recently I discover a new package manager for Python, uv, which could be a pip alternative written in rust. Just like all the “Rust version” remakes, it has an incredible speed boost compared to pip or poetry. UV can even generate a virtual environment (like the python venv).

alt text
This is the official benchmark from the uv project description.
Just by the glance of this graph I’m alredy amazed by the claim.

Official Website

Intro

Installation

1
2
3

pip install uv
# or 
$ curl -LsSf https://astral.sh/uv/install.sh | sh

Virtual Environment

Usage

1
2
3
4

# original
python -m venv $virtual_env_name
# uv
uv venv $virtual_env_name

Local Performance

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

$ time python3 -m venv env
python3 -m venv env  2.11s user 0.58s system 72% cpu 3.695 total

# vs 

$ time uv venv env
Using Python 3.10.13 interpreter at: /Users/username/.pyenv/versions/3.10.13/bin/python3
Creating virtualenv at: env
Activate with: source env/bin/activate
uv venv env  0.01s user 0.02s system 5% cpu 0.537 total

Package Installation

Usage

1
2
3
4

# original
pip install $package_name
# uv
uv pip install $package_name

Local Performance

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21

$ time pip3 install --no-cache-dir requests
Collecting requests
  Downloading requests-2.31.0-py3-none-any.whl (62 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 62.6/62.6 kB 1.1 MB/s eta 0:00:00
...

Successfully installed certifi-2024.2.2 charset-normalizer-3.3.2 idna-3.6 requests-2.31.0 urllib3-2.2.1
pip3 install --no-cache-dir requests  0.74s user 0.18s system 31% cpu 2.907 total

# vs

$ time uv pip install --no-cache-dir requests
Resolved 5 packages in 325ms
Downloaded 5 packages in 256ms
Installed 5 packages in 7ms
 + certifi==2024.2.2
 + charset-normalizer==3.3.2
 + idna==3.6
 + requests==2.31.0
 + urllib3==2.2.1
uv pip install --no-cache-dir requests  0.08s user 0.13s system 16% cpu 1.228 total

Package Installation Using Same Requirements.txt

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

$ time pip3 install -r requirements.txt
Collecting anyio==3.7.1
  Using cached anyio-3.7.1-py3-none-any.whl (80 kB)

...

pip3 install -r requirements.txt  13.57s user 3.98s system 69% cpu 28.342 total

# vs 

$ time uv pip install -r requirements.txt --no-cache-dir
Resolved 77 packages in 22ms
Installed 77 packages in 171ms

...

uv pip install -r requirements.txt  0.05s user 0.29s system 40% cpu 0.849 total

Note that uv pip & pip are seperate package managers

In the official uv benchmarks, uv is 8-10x faster than pip and pip-tools without caching, and 80-115x faster when running with a warm cache (e.g., recreating a virtual environment or updating a dependency)

Poetry Compatibility

Poetry is eliminating all pip usage in installations, it seems they don’t plan to support uv in near future: issue link