Create a dataframe that doesn't execute commands but rather logs.
The core python operators would be tracked via the overridden magic methods.
df + 1 would use __add__.
as far as I can tell, the biggest parts after that are things like
pd.rolling_moments
and
np.isnan
It'd be possible to create a np equivalent that requests operators of the dataframe.
I suppose the same could be said of pd.rolling_sum.
I thought of maybe doing this via __array_prepare__ but that was a dead end. I can't stop the calculation and can't expect that every ufunc has a context object.
Create a dataframe that doesn't execute commands but rather logs.
The core python operators would be tracked via the overridden magic methods.
df + 1would use__add__.as far as I can tell, the biggest parts after that are things like
pd.rolling_momentsandnp.isnanIt'd be possible to create a np equivalent that requests operators of the dataframe. I suppose the same could be said of
pd.rolling_sum.I thought of maybe doing this via
__array_prepare__but that was a dead end. I can't stop the calculation and can't expect that every ufunc has acontextobject.