[LambdaLambda]

Hey,

I bet I again made silly mistake, but weird things is happening. I got maths formula that I tested on:
PHP
Java
calculator

and works great, but on SP it outputs me 0.

PHP Code:

         int pointsNew = (65 / 65 / (1 + 25) * 10000); 


What am I doing wrong?

[Dr. Greg House]

int/+*-%int returns int.

[LambdaLambda]

Right, shame on me. Cheers!

[ddhoward]

Going into more detail for any newbies who happen across this thread.

You have to think about each individual operation.

65 / 65 / (1 + 25) * 10000
65 / 65 / 26 * 10000
1 / 26 * 10000

Here's where it gets tricky. Since 1 and 26 are both integers, the answer can only be an integer. The real value of 1/26 is around 0.0384, which cannot be represented by an integer. So we round down to the nearest integer, which is zero.

0 * 10000

= 0

[WildCard65]

Originally Posted by ddhoward Going into more detail for any newbies who happen across this thread. You have to think about each individual operation. 65 / 65 / (1 + 25) * 10000 65 / 65 / 26 * 10000 1 / 26 * 10000 Here's where it gets tricky. Since 1 and 26 are both integers, the answer can only be an integer. The real value of 1/26 is around 0.0384, which cannot be represented by an integer. So we round down to the nearest integer, which is zero. 0 * 10000 = 0

I'm pretty sure C++(which is the code running SM plugins) truncates floats when doing int dividing by int. eg:
5 /3 = 1 (1.6 repeat truncates to 1)

[Powerlord]

Originally Posted by WildCard65 I'm pretty sure C++(which is the code running SM plugins) truncates floats when doing int dividing by int. eg: 5 /3 = 1 (1.6 repeat truncates to 1)

SourceMod has a Pawn VM built into it, so I wouldn't be so quick to assume that's why this is happening.

No, it's in the Pawn language definition. Just like it is in language definitions for C, C++, C#, and Java just to name a few.

As a general rule¹, a strongly typed language is going to return the result of a simple math operation */+- as the type with the largest range used in the operation. For Pawn, this means int / int == int, float / int == float

¹That's not always the case, however... for instance, you'll be stung by this in C# if you try adding a float/double and a decimal as there is no implicit conversion from float/double to decimal... you have to explicitly convert the float/double first.

[LambdaLambda]

Yeah, that was the problem, as I said on my previous post. This used to be converting by itself on old syntax, so I was a bit confused. But I done some research, I'm on a right track now. Thank you! : )

[Chdata]

I'm not sure if it actually always converts in the old syntax. I've seen weird results where after an update or something my functions that used to mix floats and ints would get messed up and give results like 253,793 or something, in the exact same scenarios as when they used to work.