1adrianb / 1adrianb/face-alignment

Retroactive change in version 1.3.5 package

Open
#341 0 comments 3 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
7.5k
Forks
1.4k
PR merge metrics
No merged PRs in 30d

Description

This has been a particularly interesting one to debug! Possibly related to https://github.com/1adrianb/face-alignment/issues/340?

I've got some machine instances that were set up last week and are running fine on version 1.3.5 - I spun up a couple more this afternoon and they're all failing at `face_alignment.LandmarksType._2D`. Immediate assumption was just a version mismatch in my setup scripts somewhere, but `pip show` reports 1.3.5 on all instances, and looking at `__init__.py` confirms that.

Long story short, `api.py` differs on machines where the library was installed a few days ago from ones where it was installed today, despite `__init__.py` matching for both. PyPI is still [showing](https://pypi.org/project/face-alignment/1.3.5/#history) the 1.3.5 release date as 2021-09-14, but installing version 1.3.5 today gives a version that uses the `TWO_D` naming convention (rather than `_2D`), which was introduced in commit https://github.com/1adrianb/face-alignment/commit/34e436a103879c35ac58b7d2648af73b68e1112c on 2022-06-20.

It looks as though it's now installing as a wheel package rather than an egg package, which could explain how there are two 1.3.5s, although I'm not familiar enough with the inner workings of `pip`/PyPI to know why they would differ.

Contributor guide

No contributing guide indexed for this repository

Research direction

Check the PyPI release history and the package's wheel vs egg distribution. Look at api.py and __init__.py in the 1.3.5 tag to see the naming discrepancy (TWO_D vs _2D). Understand how pip installs different artifacts for the same version. The fix likely involves updating the package metadata or re-releasing a corrected version.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
build-system, release
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.