Can unions be used to convert char arrays to float?

Question

I was wondering whether or not is possible to use unions to get a float from a received char array. Let's say that I have defined the following struct

typedef union {
  float f;
  char c[4];
} my_unionFloat_t;

If I receive a char array encoding a float like this ( the numbers are made up)

data[4] = {32,45,56,88};

Can I do the following?

my_unionFloat_t c2f;

c2f.c[0] = data[0];
c2f.c[1] = data[1];
c2f.c[2] = data[2];
c2f.c[3] = data[3];

float result = c2f.f;

yes, if sender respects IEEE and endianness. but `memcpy(&the_float,the_string_data,sizeof(float))` achieves the same thing — Jean-François Fabre, May 04 '17 at 17:10
The proposed dup is not using a `char` array. You should make sure you use `unsigned char` (and not try to initialise 5 elements of a 4 element array). — Weather Vane, May 04 '17 at 17:13
@WeatherVane Whether it's an array or another type is irrelevant. — Justin, May 04 '17 at 17:15
My understanding is that you can do this in c++ if you use an unsigned char array. — Bathsheba, May 04 '17 at 17:16
It looks like in C++, if you want to do this, [the best way is to use `memcpy`](http://stackoverflow.com/a/17790026/1896169) as @Jean-FrançoisFabre suggested. You have to be careful, as in C++ it's really easy to break the strict-aliasing rules — Justin, May 04 '17 at 17:26
Note: no need to hard code 4. Use `char c[sizeof (float)];` BTW, it is better to use `unsigned char` - less complications. — chux - Reinstate Monica, May 04 '17 at 17:29
@Jean-FrançoisFabre in your solution shouldn't I indicate the size of the_string_data? — ndarkness, May 04 '17 at 17:34
@chux do you mean `char c[sizeof(float)]` in the definition of my union, don't you? Regarding to use `unsigned char` also in the same definition right? — ndarkness, May 04 '17 at 17:36
@ndarkness Yes, `typedef union { float f; unsigned char uc[sizeof (float)]; } my_unionFloat_t;` This is fine in C. Notice commented links about additional C++ issues. — chux - Reinstate Monica, May 04 '17 at 18:34

chtz · Answer 1 · 2019-06-25T19:43:42.563

1

The easiest way to achieve that in C++ is to use reinterpret_cast:

unsigned char data[4] = {32,45,56,88};
float f = reinterpret_cast<const float&>(data);
const unsigned char* ch =  reinterpret_cast<const unsigned char*>(&f);
for(int i=0; i<4; ++i)
    std::cout << +ch[i] << ' ';
std::cout << f << '\n';

edited Jun 25 '19 at 19:43

answered May 04 '17 at 18:36

chtz

17,329
4
26
56

1

Note that you want to be very careful; if you aren't casting to/from a `char *`, this is undefined behaviour due to strict-aliasing. – Justin May 04 '17 at 18:39
You forgot your `std::` on the second `cout`. – S.S. Anne Jun 25 '19 at 19:00
Thanks @JL2210. Don't you have enough reputation (yet) to directly edit? – chtz Jun 25 '19 at 19:44
@chtz Almost. Need 271 more. – S.S. Anne Jun 26 '19 at 13:12

Can unions be used to convert char arrays to float?

1 Answers1