problem with std.variant rounding

[via Digitalmars-d-learn]

On Friday, 28 April 2017 at 16:42:28 UTC, Petar Kirov 
[ZombineDev] wrote:
> On Friday, 28 April 2017 at 16:24:55 UTC, Suliman wrote:
>> On Friday, 28 April 2017 at 15:45:25 UTC, Suliman wrote:
>>> I am using https://github.com/mysql-d/mysql-native
>>> It's return from DB variant data-type.
>>>
>>> My DB include value: 56.051151 (double type in DB)
>>>
>>> I need to extract it. I tried several variants:
>>>
>>> writeln(point[3].coerce!float);
>>> writeln(point[3].coerce!string);
>>> writeln(point[3].coerce!double);
>>>
>>> but all of them return me it as: 56.0512
>>>
>>> How to return exactly 56.051151 ?
>>
>> import std.stdio;
>> import std.variant;
>>
>> void main()
>> {
>> 	Variant b = 56.051151;
>> 	float x = b.coerce!float;
>> 	writeln(x);
>> }
>>
>>>56.0512
>
> The precision is still there, you're just not requesting it:
>
> import std.conv : text;
> import std.stdio;
> import std.variant;
>
> void main()
> {
>     Variant b = 56.051151;
>     float x = b.coerce!float;
>
>     foreach (i; 0 .. 10)
>         writefln(text("%.", i, "f"), x);
> }
>
> 56
> 56.1
> 56.05
> 56.051
> 56.0512
> 56.05115
> 56.051151
> 56.0511513
> 56.05115128
> 56.051151276
>
>> How to return exactly 56.051151 ?
>
> Specify the number of digits after the decimal point manually, 
> e.g. writefln("%.6f", x) will print 56.051151.

BTW, you should always try to use the same floating-point type, 
so if you use 64-bit doubles in the database, you should also use 
double in your D code, otherwise you may accumulate rounding 
errors after each conversion. When converting to smaller 
floating-point types (real -> double and double -> float) you are 
essentially throwing out precision. For example:

import std.conv : text;
import std.stdio;
import std.variant;

void main()
{
     Variant b = 56.051151;

     foreach (i; 0 .. 16)
     {
         writefln(text("%.", i, "f"), b.coerce!float);
         writefln(text("%.", i, "f"), b.coerce!double);
     }
}

Notice how the floats and doubles differ at i >= 6.

56
56
56.1
56.1
56.05
56.05
56.051
56.051
56.0512
56.0512
56.05115
56.05115
56.051151
56.051151
56.0511513
56.0511510
56.05115128
56.05115100
56.051151276
56.051151000
56.0511512756
56.0511510000
56.05115127563
56.05115100000
56.051151275635
56.051151000000
56.0511512756348
56.0511510000000
56.05115127563477
56.05115100000000
56.051151275634766
56.051150999999997