Broken link: scripts/create_venv.sh is not found

[aalexandersson]

How to update an existing installation of splink?

I am a beginner Python user. I have used splink 3.4.3 and a related requirements.txt file (e.g., with duckdb==0.6.0) in a virtual environment for Python 3.11 on Windows for 6+ months. How do I update to the new v3.9.0? I want to update because the new version has a new critical feature to me, Damerau-Levenshtein comparator in duckdb==0.8.0.

The current instructions at https://moj-analytical-services.github.io/splink/dev_guides/changing_splink/building_env_locally.html state

To automatically create a simple virtual environment using venv, simply run this script:
source scripts/create_venv.sh

I tried running the command in Windows PowerShell but it gives me this error message:

The term 'scripts/create_venv.sh' is not recognized as the name of a cmdlet, function, script file, or operable
program.

Also, the web page has a broken link. It only has the content Not Found. Maybe this is the correct URL with the script?:
https://github.com/moj-analytical-services/splink/blob/master/scripts/create_venv.sh

If that is the script, how do I run scripts/create_venv.sh from Windows PowerShell to update splink?

I am trying to learn about the script. In the meanwhile, my "old" method of editing requirements.txt for updating to splink==3.9.0 and duckdb==0.8.0 still worked:

pip3 install -r requirements.txt

Update: The link is broken. Otherwise my question seems to be a duplicate of issue 1031. I updated the discussion to reflect this.

[RossKen]

Hey @aalexandersson, your "old method" is the correct way of doing it. The source scripts/create_venv.sh command is intended for people developing splink, not for users - so for use when you clone the splink repo to add a feature/bug fix etc. Since the core team work on macs, it's intended to be run in unix-based environment, hence the errors you're seeing.

You are correct that Damerau-Levenshtein requires duckdb 0.8.0, however we haven't made that a strict requirement for installing splink. This is because we know of users who have struggled to install duckdb in their specific environments and given Damerau-Levenshtein is an optional feature we decided not to force everyone to upgrade to 0.8.0. Anyone requiring the most up to date version can then upgrade their duckdb install as they wish to access the latest features. However, I admit that we could have documented this better.

[aflaxman]

In case it is useful to others, I just ran into this on Google Colab yesterday, and was able to use D-L by starting my notebook with the following cell

!pip install --upgrade pseudopeople splink duckdb