Pip does unnecessary git clone for github dependencies

[gukoff]

• Pip version: 9.0.1
• Python version: 2.7.14
• Operating system: OSX

Description:

Pip installs github dependencies through git clone. Cloning takes a long time for large projects. For example, the clone part of this command is ~2 minutes:

pip install --no-cache-dir git+https://github.com/django/[email protected]

At the same time, one can download source files via a magic link in seconds flat: https://github.com/django/django/archive/2.0.1.tar.gz. This works for tags, commits and branches. And in case link stops working one can easily fallback to git clone.

Can we make use of this?

[pradyunsg]

It's mostly that currently, there are tools that rely on the fact that pip clones the entire repository. We don't want to break those tools since many projects depend on them and they depend on having the entire history of the project. For more on this discussion, you'd want to look at #2432.

[ghost]

Also, this specific optimization has no benefit vs. git clone --depth=1.

[cjerdonek]

Can we make use of this?

You can. You just have to specify this behavior, e.g.:

$ pip install https://github.com/django/django/archive/2.0.1.tar.gz

[pradyunsg]

Closing as a duplicate of #2432.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.