NSNumber from CGFloat

Question

Is there a safe way to "convert" a CGFloat to a NSNumber ?

NSNumber has the numberWithFloat: and numberWithDouble: methods but CGFloat being defined as float or double depending on the platform, it seems risky to use either one of them.

Or is numberWithDouble: safe to use with a CGFloat, as it is the one with the more precision ?

Fafnir

If you have a enough high compiler version, just write @(myNumber) and it will be resolved at compile time. However you can safely call [NSNumber numberWithDouble: myNumber], if CGFloat is defined to be double you do it correctly, it CGFloat is defined to be float you still haven't lost any precision. – Microelectronics

safe? can you explain? – Barmecide

safe as "without precision loss". Is this why I was downvoted ? – Fafnir

I down voted it because IMHO you are just over thinking. C is full of functions that handle doubles, not overridden to work also with floats just because if you pass a float instead of a double nothing harmful can happen. It will just be implicitly casted to a double, which uses extra precision. – Microelectronics

I'm over thinking it because my code will handle real money. So better be safe (pun not intended) than sorry. (and I also don't think this was worth a downvote, but eh...) – Fafnir

This casting wastes cycles, and can be significant in code that does lots of floating point arithmetic. In terms of math functions, I recommend using tgmath.h and letting the compiler choose the right function for you at compile time. – Bursary

If you're handling money, you may wish to consider using a fixed point or decimal representation. See NSDecimal and NSDecimalNumber. – Bursary

For which device support you are bulding this app? – Barmecide

Don't use floating point for real money. – Neuropath

@LithuT.V iPad but I didn't want it to be relevant, I wanted portable-ish code. @Bursary already using these, but wanted a convenient way to init them from CGFloat. – Fafnir

well iphone is a 32 bit device .CGFloat is a regular float on 32-bit systems and a double on 64-bit systems.thats why i asked – Barmecide

Answer 1

I believe @ NSNumber literal is good enough

@(myCGFloat)

Read more here: http://clang.llvm.org/docs/ObjectiveCLiterals.html

Balustrade

Thanks for the solution and the link. This was enlightening :) – Fafnir

Answer 2

This is how I handled it:

@interface NSNumber (CGFloatAdditions)

+ (NSNumber*)numberWithCGFloat: (CGFloat)value;
- (CGFloat)CGFloatValue;

@end

@implementation NSNumber (CGFloatAdditions)

+ (NSNumber*)numberWithCGFloat: (CGFloat)value
{
#if CGFLOAT_IS_DOUBLE
    return [NSNumber numberWithDouble: (double)value];
#else
    return [NSNumber numberWithFloat: value];
#endif
}

- (CGFloat)CGFloatValue
{
#if CGFLOAT_IS_DOUBLE
    return [self doubleValue];
#else
    return [self floatValue];
#endif
}


@end

CGFLOAT_IS_DOUBLE is defined in CGBase.h conditionally by platform.

Bursary

This is not even necessary, it's over-coding. If you call [NSNumber numberWithDouble: value] (for compiler versions that don't support literal numbers) you don't risk to lose any precision whether if the number is float or double. – Microelectronics

Yes, but you waste cycles to convert between float and double on platforms where CGFloat is a float. But whatever. Use this, don't use this. I wrote it a long time ago and I continue to use it. It works for me. – Bursary

Thnaks @ipmcc, I'll use Lukas' solution but yours is what I was searching for originally. – Fafnir

I find it odd this is not NSNumber already, since they did add methods for NSInteger. – Atreus