I am using a DECIMAL (19,8) column, in a division having approximate results, example:
select col1, col1/1000 as col1div from table1 where col3 = 173286399
col1 col1div 1.000097 0.10001
In addition, the original value of col1 in the csv file, was :
but importend into the DECIMAL(19,8) datatype, it becomes: 1.000097 (using COPY FROM) .
Does the DECIMAL datatype provide the same - not exact - numerical representation as the FLOAT datatype in MapD despite having 8 bytes for memory size?
In that case, the only exact decimal representation would be the DOUBLE datatype with the performance issue raised in this post.
Strange enough, I still didn’t get to find a way to use a both fast AND exact decimal representation in this analytical, performance oriented database.