[jira] [Commented] (MATH-1493) Eigendecomposition Float comparison with ==; this is bad

["Gilles (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MATH-1493?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16877306#comment-16877306 ] 

Gilles commented on MATH-1493:
------------------------------

You might be pointing to bugs, but the construct could also be intentional.
This is very old code, and it would be dangerous to change it without a demonstrated failure.
Unit tests are most welcome.

> Eigendecomposition Float comparison with ==; this is bad
> --------------------------------------------------------
>
>                 Key: MATH-1493
>                 URL: https://issues.apache.org/jira/browse/MATH-1493
>             Project: Commons Math
>          Issue Type: Bug
>    Affects Versions: 3.6.1
>         Environment: I use very small and very big float numbers and number which are very near to 0.0 like -10^30;
> so this effect me.
>            Reporter: Nietzsche
>            Priority: Major
>
> If you compare two float numbers you should not use ==. I think this lead to some arithmetik mistakes. Maybe you should use epsilon comparison so not x == 0 but
> x < 0.00000000000001
>  
>  
> Effected methods:
> findEigenVectors;
> row 671
> if (e[i + 1] == 0.0) {
> row 687
> if (t == 0.0 && i >= j) 
>  
> isNonSingular(); row 522
> largestEigenvalueNorm == 0.0
>  
>  



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)