num_digits: double → int

[DimitriPapadopoulos]

It looks like a bad idea to apply operators ++ and -- to a double, or to compare a double to integer values.

[LB--]

Interesting, I wonder why it was ever a double to begin with. Do you think we need to cast it to a double again during the pow call near line 384?

json-parser/json.c

Line 834 in 60fd8a7

top->u.dbl += num_fraction / pow (10.0, num_digits);

I get the feeling there may be circumstances where the call could be ambiguous without the cast, though perhaps I am wrong. Maybe keeping it as an int would benefit the code instead. Thoughts?

[DimitriPapadopoulos]

The call to pow() will implicitly cast to a double anyway:

double pow(double x, double y);

We could make the cast explicit, but I don't see any reason to do that.

[LB--]

When compiling as C++, pow might have numerous overloads even without the std:: prefix, which is why I was concerned. Though, this might technically only be an issue in non-conforming implementations, I can't remember if unqualified pow is required to be unambiguous in C++.

[DimitriPapadopoulos]

The C++ pow() with overloads would be in <cmath>, wouldn't it?

We include <math.h> which defines pow() as:

double pow (double base, double exponent);

https://www.cplusplus.com/reference/cmath/pow/

[LB--]

Ah, right you are! Somehow of all things I considered, I forgot <math.h> is being used and not <cmath>.