[R] data.table exact matching numeric keys

[James Hirschorn]

I have looked over the documentation and did not see an answer to this 
seemingly basic question:

I noticed that if I have a data.table with a key containing numeric 
values, then a close number is considered a match when searching. For 
example,

 > dt[J(a)]
#     certainty probability
# 1: 0.8596491          -1
# 2: 0.8596491          -1

But in fact the two certainty values differ by 5.797585e-13.

Is there a way to only return exact matches, or set the epsilon value 
(to something like machine epsilon)?