search

kedro-org / kedro

Kedro is a toolbox for production-ready data science. It uses software engineering best practices to help you create data engineering and data science pipelines that are reproducible, maintainable, and modular.
https://kedro.org
Apache License 2.0
9.48k stars 874 forks source link

Get rid of GitHub magic in README #3416

Open astrojuanlu opened 6 months ago

astrojuanlu commented 6 months ago

@marrrcin reported that he saw the README like this

image

and it's because of a bad combination of the magic we're using to detect light/dark theme

https://github.com/kedro-org/kedro/blob/8257d55ec96abbc1ce18ae229da8c773c60bfc1b/README.md?plain=1#L1-L2

and how GitHub assumes night = dark

image

This has created other problems in the past https://github.com/kedro-org/kedro/pull/2961#issuecomment-1688542270

Given that this is broken on PyPI and half-broken on GitHub, I'd say let's get rid of this.

A strawman proposal would be something like

image

But better to have this properly designed.

stichbury commented 5 months ago

Moving this to Viz as it'll be part of design initially and quite likely require web savvy too.

rashidakanchwala commented 4 months ago

@SajidAlamQB also reported this issue on pypi - https://pypi.org/project/kedro/

astrojuanlu commented 4 months ago

We don't need tech design for this. We need to just do it.

ravi-kumar-pilla commented 3 months ago

I encountered this when publishing github action. The way we do conditional images is deprecated. This link should help.

astrojuanlu commented 3 months ago

@ravi-kumar-pilla could you check if the <picture> thing is rendered correctly on PyPI?

ravi-kumar-pilla commented 3 months ago

@ravi-kumar-pilla could you check if the <picture> thing is rendered correctly on PyPI?

@astrojuanlu I see an open ticket - https://github.com/pypi/warehouse/issues/11251 . Needs some more searching/testing when we work on the ticket.

I tried testing the html and picture tag seems to work fine -

image

image

astrojuanlu commented 3 months ago

The way to test this would be to upload a test version to https://test.pypi.org/project/kedro/ or a newly registered package only for testing purposes.