One of the many reasons you should stay away from real world languages when talking about algorithms unless you are an expert in the language.
The code snippet that the article claims is C:
int32_t compute_magic(void) {
double sigma = 0.0450465;
double expression = 1.5 * pow(2.0, 23.0) * (127.0 - sigma);
int32_t i = expression;
return i;
}
If you use a union, your code won't read or write out of bounds, since it will save space for the larger type. Ideally you would also specify _Float32_t and int32_t for some pseudoportability (at least before endianness gets involved), although I'm aware this was not available in 1999.
Otherwise, carry on.
Whether it is bad practice or not, people used to do this all the time, and so the compilers all tend to support type conversion via pointer and type punning via union. A lot of critical code in the real world would break if it suddenly didn’t work.
C++ bit-cast is brand new (C++20) and until now the suggestion in C++ was to use memcpy, and hope and pray it gets elided. That might be correct but boy is it ugly and gross.