yf.download fails with pandas 2.0.0

[mndavidoff]

Pandas 2.0.0 was released today. It appears something changed in Pandas that breaks yf.download. I'm using:

• Python 3.11.2
• yfinance 0.2.14
• macOS 12.6.4

yf.download works with Pandas 1.5.3:

>>> import pandas as pd
>>> import yfinance as yf
>>> pd.__version__
'1.5.3'
>>> yf.__version__
'0.2.14'
>>> yf.download('MSFT', start='2023-04-03', end='2023-04-04')
[*********************100%***********************]  1 of 1 completed
                  Open        High         Low       Close   Adj Close    Volume
Date
2023-04-03  286.519989  288.269989  283.950012  287.230011  287.230011  24100907

yf.download fails with Pandas 2.0.0:

>>> import pandas as pd
>>> import yfinance as yf
>>> pd.__version__
'2.0.0'
>>> yf.__version__
'0.2.14'
>>> yf.download('MSFT', start='2023-04-03', end='2023-04-04')
[*********************100%***********************]  1 of 1 completed

1 Failed download:
- MSFT: No data found for this date range, symbol may be delisted
Empty DataFrame
Columns: [Open, High, Low, Close, Adj Close, Volume]
Index: []

[Kantubhukta-Hemanth]

I too got the Same problem. do anyone know how to fix it?

[alexdgithub]

It's not pandas, it's the python version. Leaving pandas alone, I switched to 3.10.0 with pyenv. Voila, it works.

[Kantubhukta-Hemanth]

May i know how exactly i can do that - "switched to 3.10.0 with pyenv. Voila"

[alexdgithub]

You can google pyenv. It has good documentation.

My findings so far:

1. With python set to 3.11.2 outside of a an env, opening a python console and running yf.download("SPY") returns "No data found for this date range, symbol may be delisted"
2. Running in a poetry environment with python set to 3.11.2 and running yf.download("SPY") returns "No data found for this date range, symbol may be delisted"
3. With python set to 3.10.0, opening a python console and running yf.download("SPY") returns the dataframe.
4. Running in a poetry environment with python set to 3.10.0 and running yf.download("SPY") returns "No data found for this date range, symbol may be delisted"

OS: Ventura 13.3 on M2
yfinance version: 0.2.14

[mndavidoff]

@alexdgithub,

It's not pandas, it's the python version. Leaving pandas alone, I switched to 3.10.0 with pyenv.

Your test results and my test results appear to be inconsistent.

1. Both tests in the original report were performed with Python 3.11.2. The only difference in the virtual environments was the version of Pandas.
2. I created two new virtual environments using Python 3.10.10. Download succeeds in the venv using Pandas 1.5.3. Download fails in the venv using Pandas 2.0.0.

I see no difference in the operation of yf.download when using Python 3.10 vs. Python 3.11. Perhaps you could double-check the version of Pandas in each of your Python 3.10 venvs?

[alexdgithub]

@mndavidoff Indeed it seems i had a global pandas 1.5.3 on the machine. I started with a clean poetry env with 3.10.0 and pandas 2. Doesn't work. Modified .toml to install pandas 1.5.3 and it works.

[mndavidoff]

I found the change in Pandas 2.0.0 that broke yf.download. Line 654 in yfinance/base.py says:

dt_now = end_dt.tzinfo.localize(_datetime.datetime.utcnow())

This works with Pandas 1.5.3. However, with Pandas 2.0.0, it raises AttributeError("'datetime.timezone' object has no attribute 'localize'"). Unfortunately, the exception handler at line 670 of base.py silently ignores this error.

In Pandas 1.5.3, end_dt.tzinfo is a pytz.UTC object. In Pandas 2.0.0, end_dt.tzinfo is a datetime.timezone object. This incompatible change to the Pandas API is documented at UTC and fixed-offset timezones default to standard-library tzinfo objects. There might be other issues with Pandas 2.0.0 as well.

[mndavidoff]

@Kantubhukta-Hemanth, to work around the problem, I suggest you create a Python virtual environment containing Pandas 1.5.3 instead of Pandas 2.0.0.

[ValueRaider]

This is the fix if someone wants to submit a pull request: #1268 (comment)

[Soel-Philippe]

@mndavidoff Hope they gonna fix it within a ouple of weeks.

[ValueRaider]

Be the change you want to see

[ef1313]

From my experience the issue appears to be any version of yfinance > 0.1.96. I've tested all yfinance versions above this using python 3.10.6 and all fail until downgrading yfinance to 0.1.96

pandas = 2.0.0