I have two questions concerning primitives conversions.

1. If you have
float f = 3.4;
the compiler will complain since 3.4 will default to a double and you cannot assign a double to a float without an explicit cast.However, if you have
byte b = 3;
I cant see why the compiler does not issue an error - I was given to understand that 3 (or any integer literal) would default to an int - in that case, how can it be assigned to a byte without an explicit cast ?

2.Consider the following code :

int i = 0; int j = 20; long l = 22;
i = j << l;

I cant understand how the compiler forgives this. I thought j would be promoted to a long and the expression would be evaluated to a long. In that case, how can it be assigned to an int ? And why does the compiler complain when the + is used instead of << - after all, they are both binary operands !!!

It is not integer. It is compile time integer constant
Because

byte b = 1; //is ok
int i=1;
byte b = i; // is not ok