I would go even further and say that any equality comparison of float numbers has to be a one-off special case. You need to know how much error can arise in you

I would go even further and say that any equality comparison of float numbers has to be a one-off special case. You need to know how much error can arise in your calculations, and you need to know how far apart legitimately different numbers will for your particular data. And of course the former has to be smaller than the latter.