How to add an "int" to a "float" in the B programming language?

Question

C has data types, so if you add an int to a float, for example:

int i = 1;
float f = 2.5;
float r = i + f;

The C compiler will know that it should generate machine instructions that adds an int to a float because the C compiler can see that i is an int and f is a float.

---

But there is no data types in the B programming language, so if we did the following (not sure if this is a correct syntax):

auto i = 1;
auto f = 2.5;
auto r = i + f;

How would the B compiler know that it should generate machine instructions that adds an int to a float?

Answer 1

B was a typeless language so i + f was always an integer addition operation. As it was typeless this also means the value in f wouldn't be converted to integer, it was just assumed to already be an integer. The bits stored in f would give it its value.

However the original B implementation didn't support floating point, so there was no way to even have floating point literals. There was at least one implementation of B, for GCOS8, where you could in fact do f = 2.5;, but i + f still worked as described above. If you wanted to do a floating point addition instead you'd do i #+ f, but that had the same problem. It assumed i was already a floating point value and wouldn't convert it.

So to get the same behaviour as in C, in addition to using the #+ floating point addition operator, you'd have to explicitly convert i to a floating point value:

auto r;
r = #i #+ f;

Where # is the unary operator that converts integer values to floating point. B also doesn't seem to have supported initialization of auto variables, so two statements are necessary here.