Trying to get my head around the correct way to do calculations when the number I am wanting is a Fl...

# getting-started

[Ben Edwards]

Trying to get my head around the correct way to do calculations when the number I am wanting is a Float but the numbers I am using to do the calculation are integers> i.e.
val purchasedTickets: Int = 10
val totalTickets: Int = 100
percentagePurchased = (purchasedTickets/ totalTickets) * 100
This is obviously going to truncate the numbers inside the brackets so the answer will be wrong but what is the correct way to handle this?

[Ciaran Sloan]

val purchasedTickets: Int = 10
    val totalTickets: Int = 100
    val percentagePurchased = (purchasedTickets.toFloat()/ totalTickets.toFloat()) * 100F

is this what you're looking for?

[Ben Edwards]

Thanks, or maybe

percentagePurchased = (purchasedTickets.toFloat() / totaltickets) * 100

However, still testing.

[Ciaran Sloan]

yeah that should work

[ephemient]

I would default to Double unless you have a reason to use Float

[ephemient]

Float can only represent integers in the range [-2^24, 2^24] exactly; integers outside of that range get rounded to a multiple of a power of 2:

val x = 50000000
x + 1 == x // false, of course
x.toFloat() + 1f == x.toFloat() // true due to running out of precision

[ephemient]

this can happen with Double too, but you have up to 2^53

[Ben Edwards]

Yes, double safer. Whats the perfomance hit?

[ephemient]

depends on hardware

[ephemient]

on x86-32 it doesn't matter: everything goes through the 80-bit x87 coprocessor anyway

[ephemient]

on x86-64 double tends to be faster

[ephemient]

but double takes twice as much memory, so if you have large arrays of them, float may be better

[ephemient]

(apparently there's ARM processors without native Double support though, so Float is way faster for them)

[Ben Edwards]

OK, so doble generally seems the way to go.

[Michael de Kaste]

double for generic use, float for speed, BigDecimal for precision. Thats my general usage