Open allefeld opened 3 years ago
We do have a tag and schema in the core standard that supports tabular data:
Currently the only implementation is for astropy.table.Table.
What's missing is an extension that knows how to translate between that schema and pandas.Dataframe
. I suppose we'll also need a feature in this library that allows users to select which table implementation to create when reading a file -- I don't think we've had this kind of collision before, and I'm not sure what will happen when both astropy and pandas are installed.
In another issue I made a side comment regarding
pandas.DataFrame
, but now I think it deserves a separate issue.Originally posted by @allefeld in https://github.com/asdf-format/asdf/issues/918#issuecomment-769276112
The reply was:
Originally posted by @perrygreenfield in https://github.com/asdf-format/asdf/issues/918#issuecomment-769306761
I would say that a language-neutral scientific data format does need to support this. Not specifically for
pandas.DataFrame
objects, but for tabular data, which is a structure that occurs in many areas of science.This would correspond to Python's
pandas.DataFrame
, R'sdata.frame
, Matlab'stable
, etc. With this feature, an ASDF file could hold the data that is otherwise kept in CSV/TSV files, Excel sheets, and I'm guessing Apache Parquet and many others.