BUG: read_gbq return timestamp as long int

[itay-jether]

Hey, I've tried to read parquet files using gbq, and it returns an integer instead of timestamps. example attached below
it works for me on python 3.10, same pandas and pandas_gbq version

Environment details

• OS type and version: ubuntu 4.22
• Python version: python 3.11.4
• pip version: 22.3.1
• pandas-gbq version: 0.19.2

Steps to reproduce

1. install python 3.11

Code example

import pandas as pd


s = pd.Series(0, index=pd.date_range('2021-01-01 01', periods=24, freq='H', tz='UTC', name='ts')).to_frame().reset_index()
save_path = 'gs://my_bucket/test_dir'
s.to_parquet(f'{save_path}/test_gbq.parquet', index=False)
s_res = pd.read_gbq('select * from test_data', 
            configuration={'query': {
                'tableDefinitions': 
                    {'test_data': {'sourceFormat': 'PARQUET',
                                   'sourceUris': [f'{save_path}/*.parquet'],
                                   'autodetect': True}}}},
            location='eu'
)
print(s)
#                          ts  0
# 0 2021-01-01 01:00:00+00:00  0
# 1 2021-01-01 02:00:00+00:00  0
# 2 2021-01-01 03:00:00+00:00  0
# 3 2021-01-01 04:00:00+00:00  0
# 4 2021-01-01 05:00:00+00:00  0
print(s_res)
#                     ts  0
# 0  1609462800000000000  0
# 1  1609466400000000000  0
# 2  1609470000000000000  0
# 3  1609473600000000000  0
# 4  1609477200000000000  0

Thanks!

[itay-jether]

I see that it it's probably related to the new pyarrow version.
I didn't notice that my 3.11 environment is with pyarrow 13.0.0 and my 3.10 environment is with 12.0.1.
I'll leave open in case this is something that needs to be fixed here as well

[chalmerlowe]

thanks for this heads up.
were you able to discern the nature of the problem ... does the new version of pyarrow do some sort of behind the scenes conversion?

[itay-jether]

Unfortunately I didn't dive into the changes in Pyarrow, I just validated that this is the only change in my environment.
To be clear I installed pyarrow==12.0.1 on my py311 environment and it worked as expected

[chalmerlowe]

Gonna be honest, I am struggling to get pyarrow=13.0.0 with Python 3.11 at all.
I am using the mamba resolver to evaluate package dependencies and it is not successfully resolving this.

I was able to easily install pyarrow=12.x and that ran the code successfully, as expected.
This issue feels out of scope for my project and I think I need to close this issue for now as non-reproducible.