Join (23) Joins on Double or Float are not recommended due to rounding error.

Becky

I often meet following alert and wondering what solution is best. 

Join (23) Joins on Double or Float are not recommended due to rounding error.

However, I'm not sure there is any solution about this now.

Actually, I feel final result is little bit strange after the decimal point but difference is not big so ignore them....

Usually everyone ignore this alert or deal with some solution???

JoeM

@Becky,Great question! My best explanation: What is 2/3? It's 0.666666666666666666666... repeating. The result of 2/3 shown as a decimal has an infinite amount of 6's, but realistically, a computer can store only so much. So the computer naturally rounds to 0.666666666666667 (through complex logic found here). If I were to demonstrate what happens if I join  0.666666666666667 to 2/3 (0.666666666666667), you would see that the values, although appearing identical, are not. The reason that these are different is because how they are computed and stored on your computer with the hyperlinked logic above. See the attached workflow to view this behavior.

What can I do to resolve this error? Well, if we join the results of 2/3 and 2/3 together, this would probably work. However, we recommend that if you are trying to join with the utmost accuracy, readjust the double to your own specified rounded precision, or if applicable, integer by using the select tool to change the datatypes upstream of the join tool.

join_demo.yxmd