pip-mark-installed

A utility tool to mark packages as installed for pip or uv without actually installing them.

Description

pip-mark-installed creates the necessary metadata files to make pip believe that a package is installed, without actually installing any of the package code. This is particularly useful for resolving conflicts between different package variants that share the same namespace (e.g. GPU vs CPU, GUI vs headless variants) or when you want to prevent pip from installing certain packages that you have already managed through other means.

Installation

Grab the pip-mark-installed.py script or install it via pip:

pip install pip-mark-installed

Usage

pip-mark-installed PACKAGE_SPEC [PACKAGE_SPEC ...]

Where PACKAGE_SPEC can be either:

• PACKAGE_NAME (uses a default version of 9999.99.99)
• PACKAGE_NAME==VERSION

Examples

Mark a package as installed:

pip-mark-installed opencv-python

Mark multiple packages with specific versions:

pip-mark-installed opencv-python==4.5.1 opencv-contrib-python==4.5.1

Specify a custom site-packages directory:

pip-mark-installed --site-packages /path/to/site-packages some-package

Common Use Cases

This tool helps address situations where pip's dependency resolution doesn't behave as desired:

1. Conflicting Package Variants: When multiple packages offer the same functionality but under identical import names, conflicts can arise. A common example is the various OpenCV packages (e.g., opencv-python, opencv-contrib-python, opencv-python-headless). These all install to the cv2 namespace. You might want the opencv-contrib-python version, but pip might try to install opencv-python due to a dependency in another package. pip-mark-installed allows you to install your preferred version and then "trick" pip into thinking the others are already present, preventing unwanted installations.

   # Install the variant you actually want:
   pip install opencv-contrib-python
   
   # Mark conflicting variants as "installed" to prevent pip from installing them:
   pip-mark-installed opencv-python opencv-python-headless opencv-python-headless
   

2. Optional or broken dependencies: Sometimes, a package might list an optional dependency that you don't want installed, or a dependency known to cause issues in your specific environment. For instance, a package might list PyQt5 as a dependency, but you want to use PySide6. If these are not strictly necessary, or if you've already managed them outside of pip, you can use pip-mark-installed to satisfy the dependency check without actually installing the packages.

   # Mark PyQt5 as installed, to prevent a dependency from installing them.
   pip-mark-installed PyQt5
   

3. System-level installations: When a package exists at system level but you want to prevent pip from installing conflicting versions in your environment.

How It Works

The script creates the necessary .dist-info directory structure and metadata files that pip uses to determine if a package is installed.

Links

• GitHub Repository
• PyPI Package

License

MIT