strings_as_dictionary fails if reading more than 1023 leading nulls

[batterseapower]

This fails:

    with contextlib.closing(connect()) as conn, contextlib.closing(conn.cursor()) as cursor:
        sql = '\nunion all\n'.join((['select null ccy'] * 1024) + ["select 'CNY'"])
        cursor.execute(sql)
        print(cursor.fetchallarrow(strings_as_dictionary=True).to_pandas())

With an exception as follows:

  File "C:\Users\mboling\AppData\Local\Continuum\anaconda3\envs\py3\lib\site-packages\turbodbc\cursor.py", line 374, in fetchallarrow
    adaptive_integers).fetch_all()
turbodbc_intern.InterfaceError: Fetching Arrow result set failed.

If I change 1024 to 1023, set strings_as_dictionary=False, or move the select 'CNY' so it is the first item in the list, then turboodbc works as expected.

numpy==1.18.1
pandas==0.25.3
pyarrow==0.16.0
turbodbc==3.3.0