Switch to typed parquet and drop nulls

#3

Hi. Thanks for the dataset! I've translated your dataset a few days ago and have some ready-to-merge changes to make your dataset a lot user-friendly:

  • Removed 2 nulls from dataset
  • Switched from .csv to .parquet with typed ingredients column, so users can skip parsing items
  • Updated .gitattributes

Code: https://static.marimo.app/static/cocktail-recipe-modifications-us0q

HF have a really bad UX for dataset PRs, so I had to manually upload dataset and them Copy-Paste some info. But I've already:

  • Tested that you can pull a new version from HF without any issues
  • Checked that everything is typed and super duper smooth

Code: https://static.marimo.app/static/toiletsandpaper-fix-cocktails-gs8h

p.s. pulling this PR dataset is in the last cells

toiletsandpaper changed pull request status to open

Code: https://static.marimo.app/static/toiletsandpaper-fix-cocktails-gs8h

Btw, the same code can also fix https://huggingface.co/datasets/erwanlc/cocktails_recipe_no_brand with a little changes in code for you. (I don't really want to deal with PRs here)

Ready to merge
This branch is ready to get merged automatically.

Sign up or log in to comment