Lag features on test data

[CallmeAmmo]

Hi , went through your code. But i find one problem over hear , you have created lag features first and then you have split the data into training and testing . Doesn't it count as data leak as for the unseen data you would not have sales for all the unseen day then there this creation of lag would become a problem. Can you help with this and how could you create lag for the upcoming days for prediction as there are important features?