I m finding a strange behaviour with floats in Kotlin 1 5 mu kotlinlang #getting-started

Channels

fluid-libraries

jhipster-kotlin

effectivekotlin

effective-kotlin

android-architecture

algolialibraries

stayhungrystayfoolish

pattern-matching

kotlin-data-storage

skia-wasm-interop-temp

squarelibraries

beepiz-libraries

cscenter-course-2016

kotlin-foundation

compose-android

lambdaworld_cadiz

kotlin-pakistan

java-to-kotlin-refactoring

event21-community-content

fb-internal-demo

kotlinprogrammers

kotlinx-datetime

framework-elide

github-workflows-kt

library-development

arrow-contributors

spring-security

pocketgitclient

domain-driven-design

swiss-user-group

refactoring-to-kotlin

ohio-kotlin-users

introduce-yourself

kotlin-by-example

kotlin-multiplatform-contest

graviton-browser

tricity-kotlin-user-group

kotlin-in-action

android-databinding

kotlintest-devs

cincinnati-user-group

codemash-precompiler

kotlingforbeginners

prozis-android-backup

realworldkotlin

connect-audit-events

russian-kotlinasfirst

androidgithubprojects

intellij-tricks

machinelearningbawas

swagger-gradle-codegen

100daysofkotlin

competitive-programming

compose-ui-showcase

androidx-xprocessing

navigation-architecture-component

strife-discord-lib

embedded-kotlin

image-processing

myndocs-oauth2-server

leedskotlinusergroup

kotest-contributors

practical-functional-programming

kotlinultimatechallenge

k2-early-adopters

kotlinforbeginners

compose-desktop

competitivecoding

100daysofkotlin-2021

refreshversions

language-evolution

jvm-ir-backend-feedback

python-contributors

codingconventions

sofl-user-group

getting-started

language-proposals

intellij-plugins

#getting-started

Title

# getting-started

m

Michael Evans

01/26/2022, 11:59 PM

I'm finding a strange behaviour with floats in Kotlin 1.5 multiplatform

1F - 0.9F == 0.100000024F

anyone know why the precision is getting messed up. I would expect

1F - 0.9F == 0.1F

p

Pratik Tandel

01/27/2022, 12:21 AM

https://blog.demofox.org/2017/11/21/floating-point-precision/

you’re better off doing FP comparisons using abs(target, current) < EPS EPS can be a small value like 1e-9

m

Michael Evans

01/27/2022, 12:23 AM

awesome thanks for the resource

i had assumed it was a kotlin bug, looks like it's inherent to floats though huh?

p

Pratik Tandel

01/27/2022, 12:33 AM

correct

r

Rob Elliot

01/27/2022, 12:46 AM

If you're on the JVM use BigDecimal if you need precise arithmetic.

g

gildor

01/27/2022, 3:55 AM

https://floating-point-gui.de/formats/fp/

it’s inherent to floats though

Or more precisly it’s a behaviour of floats according to IEEE 754 standard

3 Views