Dealing with Rounding Errors

Hi,

I've just started experiencing rounding errors in my work and I wanted to get your guys opinion on the general rule of thumb to dealing with them.


They've made an appearance as I've begun testing:

if a point is on a line
if a point is in a triangle

So, I've been fooling around with different settings of rounding and trying to get reliable results without any false positives but still not missing any positives. Do you just play with the number of decimal places to round to until you're comfortable or what's the standard approach.

So far, I seem to find rounding my main data calculations to 12 decimal places and setting the point-line tests rounded to 4 decimal places seems to be giving accurate results.

Am I going about this wrong or is this just a fact of math?

thanks for any advice

This is more of what the person wants.. I always use rounding to two decimal places.. I dont think there is any thumb rule of that sort.

Deep

Any time you round, you are going to find errors at some point or other, so unless there is a great need to round, why do it?

I hardly ever round, but just set the number of decimal points to display