This error can be reproduced quite simply:

import pandas as pd
df = pd.DataFrame({'a': [1,2,3], 'b': [4,5,6]})
df.to_feather('out.f')

same error

See also pandas-dev/pandas#34670

Closing here as this is not a problem of pyabc. One should use the latest arrow packages (or feather, at some point) on python and R side, see the above link.

Here is a work-around, as I encounter this issue even though I am using the latest arrow(2.0.0)/feather(0.3.5) packages in R.
Using pyarrow(2.0.0) one can read and rewrite the feather file in "version 1", which is then readable in R.
import pyarrow.feather as feather
feather_db = feather.read_feather('my_db.feather')
feather.write_feather(feather_db, 'my_db_v1.feather', version=1)

Thanks to @nbungi, I find this link https://arrow.apache.org/docs/python/feather.html.

Writing Version 1 (V1) Files
For compatibility with libraries without support for Version 2 files, you can write the version 1 format by passing version=1 to write_feather. We intend to maintain read support for V1 for the foreseeable future.

For me, when using version=1 and I try to load the .feather file in R-Studio, RStudio crashes.

The feather format appears to be somewhat cumbersome, with different tools implementing/supporting different versions etc.. Not sure what lead to this error, but just to recapitulate a few options:

• As pointed out above by @nbungi , you can convert to version 1 post-hoc. Alternatively, you can also programmatically extract the database in https://github.com/ICB-DCM/pyABC/blob/main/pyabc/storage/export.py#L83 (this is the function called by abc-export) and then apply pandas.to_feather(filename, version=1).
• When importing in R, at least either feather::read_feather or arrow::read_feather can be used, with the latter in particular for version 1 exports.