Hi all,

We support binning in transform encode and NaN if we try to encode
values not in any Bin.

e.g. if the bins are :

0-1,1-2,2-3,3-4,4-5,5-6,6-7,7-8,8-9,9-10

If we try to encode 0.1 we return 1 for bin 1, or 8.8 we return 9.

The issue I would like to address comes with values close to the
boundaries where potential rounding errors in double values make us
return Nan on inputs like 10.0000000001.

I suggest that we open the end boundaries to encode any value that is
above or below the end buckets like:

-inf-1,1-2,2-3,3-4,4-5,5-6,6-7,7-8,8-9,9-inf

This meanse that -13 is in bucket 1 and 1324 is in bucket 10, while we
keep NaN mapping to NaN.

Best regards
Sebastian



Reply via email to