Floating point errors

[Anonymous]

Hi All,

There's an annoying bug I discovered and it's easy to replicate.

Create a measure

Test = (3/60)*0.80 > 0.04

It will return "True". Basically, Power BI is telling you that 0.04 is bigger than 0.04.

Regards,

Me

[v-yulgu-msft]

Hi @Anonymous ,

I have reported this problem internally, CRI:113755404. Would update here once I get any response.

Best regards,

Yuliana Gu

[v-yulgu-msft]

Hi @Anonymous ,

I got response from PG as follows:

This is a known limitation of decimal number data type. Power BI's decimal number data type is implemented as IEEE Standard 754 Floating Point Numbers. IEEE floating pointer numbers have rounding errors therefore comparison of two numbers with close values may lead to surprising results. See https://floating-point-gui.de/errors/comparison/ to examples. Users can switch to fixed decimal number data type or use epsilon comparison as described in https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/

Best regards,

Yuliana Gu

[Anonymous]

Excel doesn't have that problem though?

When you select the Format and decimal place under Modeling -> Formatin, it doesn't affect the floating point at all.

Like this abomination (below). Constant is 3.80 with formating selected to "Whole Number".

It's going to be a pain having to use Round() on every measure before comparing them...