Data type demotions

Data type demotions in C/C++

You previously saw that demotions are not detected in C and only in a few cases in C++.

To learn more about demotions, consider the small C++ program below.

Use a text editor to copy the code to a text file named demotiontest.cpp:

    

        
        
            #include <cmath>
#include <cstdio>
#include <cstdint>

int main() {
    double w = 1.0e50;
    printf("w = %f\n", w);
    float y = w;
    float z{w};
    printf("y = %f\n", y);
    printf("z = %f\n", z);

    int16_t a = -16380;
    int32_t b = a*1000;
    int64_t c = b*1000;

    int16_t d1 = c*2;
    int16_t d2{c*2};
    printf("a = %d, b = %d, c = %ld, d1 = %d, d2 = %d\n", a, b, c, d1, d2);

    int32_t e1 =w;
    int32_t e2{w};
    printf("e1 = %d, e2 = %d\n", e1, e2);
}
        
    

Compile the example:

    

        
        
            g++ demotiontest.cpp -o demotiontest
        
    

You will most likely get the following warnings but the program will compile successfully:

    

        
        
            demotiontest.cpp: In function ‘int main()’:
demotiontest.cpp:9:13: warning: narrowing conversion of ‘w’ from ‘double’ to ‘float’ [-Wnarrowing]
    9 |     float z{w};
      |             ^
demotiontest.cpp:18:17: warning: narrowing conversion of ‘(c * 2)’ from ‘int64_t’ {aka ‘long int’} to ‘int16_t’ {aka ‘short int’} [-Wnarrowing]
   18 |     int16_t d2{c*2};
      |                ~^~
demotiontest.cpp:22:16: warning: narrowing conversion of ‘w’ from ‘double’ to ‘int32_t’ {aka ‘int’} [-Wnarrowing]
   22 |     int32_t e2{w};
      |
        
    

Now run the program:

    

        
        
            ./demotiontest 
        
    

The output is:

    

        
        w = 100000000000000007629769841091887003294964970946560.000000
y = inf
z = inf
a = -16380, b = -16380000, c = 799869184, d1 = 4608, d2 = 4608
e1 = 2147483647, e2 = 2147483647

        
    

Obviously w is a huge value and it does not fit in a float (remember that the largest positive float is 3.4e+38), but the compiler only complains about z which uses bracket initialization and not y which uses assignment.

Similarly d1, with type int16_t, uses assignment and does not generate a warning but d2, also an int16_t, uses bracket initialization and triggers a compiler warning.

The same happens with the demotion/conversion of the double w to the int32_t variables e1, e2.

This means that some demotions/conversions that happen through assignment can pass the compiler without a warning, leading to wrong values and difficult to track bugs.

In general, demotions are risky and you should always take a second look when a demotion takes place. When using C++, try to use bracket initialization to trigger compiler warnings.