Release date: 2022-11-29
This release addresses various floating point accuracy issues by using arbitrary precision arithmetic for all calculations where such issues may occur.
All methods now expect columns which were previously double precision to be of type decimal and uses this type when outputting columns which would have previously been of double precision. This type provides arbitrary precision arithmetic with support for increasing the precision where calculations require.
No type casting is performed on either the input or output data frames. Thus providing columns using double precision where decimal types are now expected will cause an error as with any other incorrect data types.
To avoid loss of precision, the exact precision of the output data frames is unspecified. This can be enforced by explicitly casting the output from the methods.