ValueError: datetime64/timedelta64 values and dtype must have a unit specified

drdan

I'm getting the below behavior when calling get_news_headlines in the python api. I'm running python 3.10.7, and eikon version 1.1.16, windows 11 Pro

Input:

> import eikon as ek
> ek.get_news_headlines('ORDER IMBALANCE', date_from='2022-10-19T00:00:00', date_to='2022-10-19T23:59:59', count=100)

Error:

  File "C:\Python310\lib\site-packages\eikon\news_request.py", line 126, in get_news_headlines
    return get_data_frame(result)
  File "C:\Python310\lib\site-packages\eikon\news_request.py", line 144, in get_data_frame
    headlines_dataframe = pd.DataFrame([], numpy.array(first_created, dtype='datetime64'), Headline_Selected_Fields)
  File "C:\Python310\lib\site-packages\pandas\core\frame.py", line 770, in __init__
    mgr = dict_to_mgr(
  File "C:\Python310\lib\site-packages\pandas\core\internals\construction.py", line 451, in dict_to_mgr
    index = ensure_index(index)
  File "C:\Python310\lib\site-packages\pandas\core\indexes\base.py", line 7371, in ensure_index
    return Index._with_infer(index_like, copy=copy)
  File "C:\Python310\lib\site-packages\pandas\core\indexes\base.py", line 718, in _with_infer
    result = cls(*args, **kwargs)
  File "C:\Python310\lib\site-packages\pandas\core\indexes\base.py", line 488, in __new__
    result = klass(data, copy=copy, name=name, **kwargs)
  File "C:\Python310\lib\site-packages\pandas\core\indexes\datetimes.py", line 360, in __new__
    dtarr = DatetimeArray._from_sequence_not_strict(
  File "C:\Python310\lib\site-packages\pandas\core\arrays\datetimes.py", line 308, in _from_sequence_not_strict
    subarr, tz, inferred_freq = _sequence_to_dt64ns(
  File "C:\Python310\lib\site-packages\pandas\core\arrays\datetimes.py", line 2092, in _sequence_to_dt64ns
    data = astype_overflowsafe(data, dtype=DT64NS_DTYPE)
  File "pandas\_libs\tslibs\np_datetime.pyx", line 288, in pandas._libs.tslibs.np_datetime.astype_overflowsafe
  File "pandas\_libs\tslibs\np_datetime.pyx", line 321, in pandas._libs.tslibs.np_datetime.astype_overflowsafe
ValueError: datetime64/timedelta64 values and dtype must have a unit specified

The error is dependent on the data being returned. I do not get this error for headline searches with a different query. Any ideas?

zoya faberov

Hello @drdan ,

I was not able to reproduce the issue in my local environment or CodeBook, with the same request, at this time, as expect not much time has passed so the result should be close.

It seems from the error that it does not originate with eikon library, rather with one of the eikon dependencies, suspecting numpy or pandas.

Would suggest to verify the versions of pandas and numpy against CodeBook versions:

and if found different, try upgrading to these.

Please let us know how this works on your side?

noel.randewich01

Hi, I am getting the same error code.

Using the below as an example, I believe the error only happens when no headlines were published during the time period specified in the search.

ek.get_news_headlines(query = 'Topic:RTRS Language:LEN ', date_from = '10-Sep-2022 01:00:00',  date_to = '10-Sep-2022 01:50:00', count = 100)

This is new. I just went back to some old code that used to work fine.

I am currently using:

Eikon version 1.1.16

Pandas version 1.5.1

Python 3.10.7

Unfortunately I can't say what versions I was using when the code last ran okay.

raksina.samasiri

Hi @noel.randewich01 ,

I'm sorry to hear that you're facing an issue. However, I was not able to reproduce the issue in my local environment or CodeBook, with the same request.

It seems from the error that it does not originate with the eikon library, but rather with one of the eikon dependencies, suspecting NumPy or pandas.

Would suggest verifying the versions of pandas and NumPy against CodeBook versions as mentioned by Zoya in the previous comment

To be of help, below are the current version of the libraries you mentioned in the Codebook app

• Eikon version 1.1.14
• Pandas version 1.3.5
• NumPy version 1.22.4
• Python 3.8.13