Using np.where with a pandas column. How do you fill the column with the previous value until condition is met again

I have this code:

df['cross'] = np.where((df['mean_close_spread'].shift(1) < 0) & (df['mean_close_spread'] > 0), 'cross', 'none')

df['cross_price'] = np.where((df['cross'] == 'cross'), df['close'], 'none')

The above code gives me the below dataframe

Instead of using the "none" string in the cross_price column when there is no cross in the cross column I would like to use the last price from the last cross until the next cross happens.

Here is an example of what i want:

>Solution :

Use Series.where for NaNs if not match cross, so possible forward missing values:

df['cross_price'] = df['close'].where(df['cross'] == 'cross').ffill() 
print (df)
    close  cross  cross_price
0  0.3434   none          NaN
1  0.3435   none          NaN
2  0.3433   none          NaN
3  0.3434  cross       0.3434
4  0.3433   none       0.3434
5  0.3432   none       0.3434
6  0.3431  cross       0.3431
7  0.4330   none       0.3431

Then replacement to none not recommended, because get mixed numeric and strings values.

Alternative without cross column:

m = (df['mean_close_spread'].shift(1) < 0) & (df['mean_close_spread'] > 0)
df['cross_price'] = df['close'].where(m).ffill()