Have a Python dataframe that I’m looking to forward fill across rows.

Data looks like this:

index   trade_date  clean_pub_date  day_lag ticker  transaction_type    asset_type  clean_amt   d0  d0_1m   d0_3m   d0_6m   d0_12m  d0_t    d0_1m_t d0_3m_t d0_6m_t d0_12m_t
136 2023-01-13  2023-02-09  27  UL  P   ST  32500   47.2510986328125    46.9969444274902    51.7693290710449    50.7017021179199    49.7376365661621    48.7582054138184    49.1847496032715    51.7501983642578    49.8611183166504    47.3593482971191
142 2023-01-13  2023-02-06  24  CRM P   ST  7500    168.829467773438    182.711318969727    197.641815185547    215.778137207031    285.457092285156    149.31494140625 170.856811523438    193.766891479492    226.983489990234    268.838836669922
169 2023-06-09  2023-06-09  0   PTON    P   ST  7500    8.3100004196167 8.39000034332275    5.82000017166138    6.07999992370606    NaN 8.3100004196167 8.39000034332275    5.82000017166138    6.07999992370606    NaN
170 2023-06-09  2023-07-06  27  TMUS    P   ST  7500    138.005340576172    135.516159057617    136.914474487305    161.267501831055    NaN 130.270050048828    137.192138671875    136.14094543457 154.882934570313    NaN
171 2023-06-09  2023-06-12  3   EMR P   ST  7500    82.3291549682617    90.2124633789063    98.5805587768555    88.8684310913086    NaN 82.4564666748047    87.5781631469727    97.8716278076172    86.8675918579102    NaN

And looking to replace the NaNs with preceding within-row entry via this:

t = t.ffill(axis=1)

This works as desired except data types change with the application of ffill:

Before

t.dtypes
Out[278]: 
name                        object
trade_date          datetime64[ns]
clean_pub_date      datetime64[ns]
day_lag                      int64
ticker                      object
transaction_type            object
asset_type                  object
clean_amt                    int64
d0                         float64
d0_1m                      float64
d0_3m                      float64
d0_6m                      float64
d0_12m                     float64
d0_t                       float64
d0_1m_t                    float64
d0_3m_t                    float64
d0_6m_t                    float64

And after:

t = t.ffill(axis=1) 

t.dtypes
Out[280]: 
name                        object
trade_date          datetime64[ns]
clean_pub_date      datetime64[ns]
day_lag                     object
ticker                      object
transaction_type            object
asset_type                  object
clean_amt                   object
d0                          object
d0_1m                       object
d0_3m                       object
d0_6m                       object
d0_12m                      object
d0_t                        object
d0_1m_t                     object
d0_3m_t                     object
d0_6m_t                     object

I don’t see why this would happen as all values replacing NaNs are floats. Also don’t see any option in the documentation to address this.

Any ideas how/why this is happening?