Is there public evidence of anyone using extension slots? is there public evidence of someone asking for this slot and showing a meaningful use case?

[matentzn]

Asked by @cthoyt here: #478 (comment)

[matentzn]

is there public evidence of anyone using extension slots? is there public evidence of someone asking for this slot and showing a meaningful use case?

#328

The general issue and it need has been discussed a huge deal and implemented a long time ago!

It emerged (for Monarch) because of the OHDSI OMOP community contribution model. They tried to adopt a variety of SSSOM features but had very specific requirements for metadata for their mapping sets. It was our hope back them to convert that community to SSSOM entirely, and we are slowly making progress, but extension slots were a hard requirement for this to work.

It now took me 20 minutes to find any evidence of anything, but here is an excerpt of their community contribution template. They wanted to be able to reflect slots like relationship_valid_start_date etc which is integral to their way of thinking.

https://docs.google.com/spreadsheets/d/1ucBUDaTofXizwWYUdr40BXt9eRswVRGQ/edit?gid=373557228#gid=373557228

Anyways this is nothing near SSSOM tbh.

I cant find examples now but I remember seeing many cases where people just "added a column to a SSSOM table" - I think its a common enough use case. I tbh did not use it (yet)

[gouttegd]

Here is also an explicit request for extension support in Python. The author references an issue where the fact that sssom-py drops any non-standard slot is described as “trouble with the SSSOM package”.

[gouttegd]

I would also point to the original other slot, which AFAIK has been present since the beginning of SSSOM – indicating that the ability to represent non-SSSOM metadata within a SSSOM set has always been on SSSOM’s requirements.

“Extension slots” are merely a better, semantically more meaningful way of doing just that.