question: Float formatting

[xhpohanka]

Hello,
can I somehow control formatting of floats in resulting json string? Basically I would like to limit the maximal count of digits...

Similar code

float val = 40.9;
string key = "Humidity";
tao::json::value v = { { key, val } };
auto str = tao::json::to_string(v);
cout << str << endl;

results in
{"Humidity":40.900001525878906}

best regards
Jan

[d-frey]

Do you actually need to control the number of digits or are you looking for the shortest (but correct) representation?

@ColinH Would an update of the conversion libraries help to improve the current overlong representations?

[xhpohanka]

I would prefer something like %0.2f in printf :D
Seriously - we use JSON also for human readable output from our sensor that definitely has lesser accuracy than double/float .

[ColinH]

@xhpohanka This is not currently possible, but should be easy to add given the modularity of the library.
@d-frey I'm wondering why 40.9 produces such a long output, we need to check the bit-pattern whether it's one of the innocent-looking but complicated-to-encode-in-base-2 numbers, in other word we (also) need to check whether our current output library is doing the right thing (in general).

[ColinH]

The number 40.9 does not have a direct representation as double as it seems to translate to an infinitely repeating sequence in binary (just like 1/3 doesn't have a finite decimal representation). I believe that the library we use to format floating point numbers does not choose to format the double that 40.9 converts to as 40.9 because while it would be correct in the sense that it is a shorter decimal number that converts to the same bit pattern, it does not faithfully represent the binary number as precisely as possible...

[ColinH]

I've added an example that shows how to use printf() for doubles, does that work for now?