Ahh I never wrote the follow-up post 😓, but this Great Tables post does one better, and shows how we used singledispatch with another tool (databackend), to support multiple dataframes without a hard dependency on Pandas or Polars:

posit-dev.github.io/great-tables...

Thanks for this deep dive! partridge is a bit hard for me to grok, since I'm noticing a lot of for loops, but more used to data wrangling looking a bit more like SQL transformations / tidyverse pipes.

(I wonder if Polars would help with this, since it makes easier to keep things table-y, SQL-y)

I think it's good when a system has a real story, but that most good systems are totally inscrutable 😭.

verraes.net/2013/08/john...

It seems similar to how memory champions use spatial locations to memorize large lists of numbers.

A system explained in the abstract is a list of arbitrary facts and rules to learn. A story might be the same system with all kinds of juicy extra context / structure as a mnemonic aid?!

omg this is incredible

pins-python uses fsspec under the hood. afaict since fsspec has both iceberg and duckdb support, this wouldn't be too hard to test out!