Pyarrow cannot write numpy complex64, so we'd probably have to serialize ourselves

NTNU-IndEcol / df_file_interchange

File interchange code to consistently save and reload Pandas dataframes with metadata.

BSD 3-Clause "New" or "Revised" License

0 stars 0 forks source link

Pyarrow cannot write numpy complex64, so we'd probably have to serialize ourselves #2

Open EsmeMaxwell opened 8 months ago

EsmeMaxwell commented 8 months ago

See, for example: https://github.com/apache/arrow/issues/12674

Attempting to do so in our code gets

ArrowNotImplementedError: ('Unsupported numpy type 14', 'Conversion failed for column F_np_complex64 with type complex64')

To fix will require serializing that column in the data itself, which is a different sort of feature. Just disabling in the tests for now, can do this when have more time.