The current description states that the argument of sq() can have “any data type”, whereas only numeric types can be used.
The current description states that the returned value is a “double”, but this is only true if the argument is itself a double.
This pull request attempts to fix the types. It is, however, impossible to give a completely accurate description of the returned type without fully stating all the C++ rules of integral promotion, and this would be too much for the intended audience.
As a compromise, the proposed description is:
Data type: int if the argument type is smaller than an int (e.g. char), otherwise the returned value has the same type as the argument.
This captures the gist of integral promotion, and should be accurate on ILP32 (ARM). On IP16 (AVR), however, this simplified description implies that sq(short_var) is a short, whereas it is actually an int. I deem the inaccuracy acceptable given that, on IP16, short and int are basically the same thing (they have the same underlying representation).
The current description states that the argument of
sq()can have “any data type”, whereas only numeric types can be used.The current description states that the returned value is a “
double”, but this is only true if the argument is itself adouble.This pull request attempts to fix the types. It is, however, impossible to give a completely accurate description of the returned type without fully stating all the C++ rules of integral promotion, and this would be too much for the intended audience.
As a compromise, the proposed description is:
This captures the gist of integral promotion, and should be accurate on ILP32 (ARM). On IP16 (AVR), however, this simplified description implies that
sq(short_var)is ashort, whereas it is actually anint. I deem the inaccuracy acceptable given that, on IP16,shortandintare basically the same thing (they have the same underlying representation).