Hi @Walter_C 

Thanks for the reply. I can specify the schema. But that would not be an ideal scenario in our case. The reason being:

1. We have around 20 different sources to read data from. Each source has different column having such values. And the precision can be different as well. For some columns it can be upto 3 decimal points and while for others it can have upto 5 decimal points. 

2. The schema can change as per time. 

So, I need a generic solution where I can read the data, retaining the precision, in the bronze layer and then can apply the schema in the silver layer. That's why I thought of reading it as a string.

Is there any other way where we can retain precision ?