@thentangler KNIME would support a lot of variable types moving between the KNIME nodes and Python code.
I have set up an example where I try a few things. Some types are not natively supported in the new framework; although you could send them ‘thru’.
What does work is to do some date and time manipulation and then send the result to a parquet file while allowing for some truncation of Datetime to milliseconds (which should be OK in most cases)
df_time.to_parquet(v_parquet_file, compression='gzip'
, allow_truncated_timestamps=True
, coerce_timestamps='ms')
What seems to get lost is some information about an Asia time zone. So you absolutely can mix KNIME and Python code and also date and time variables. Sometimes my impression is when dealing with different systems and data and time variables is to transfer them as a string or (long) integer with explicit values (not UNIX-timestamps or some other system).