I am using apache cassandra 3.x version. I am bit confused regarding when should I use decimal vs float types?
Is there any specific use-cases/differences when should go for float or avoid decimal and vice-versa?
I have gone through some quick tutorial none covered this difference. can anyone help me understand this ?
From the book Learning Apache Cassandra By Mat Brown:
Cassandra has three types that store non-integer numbers:
- The
floattype stores 32-bit IEEE-754 floating point numbers.- The
doubletype stores 64-bit IEEE-754 floating point numbers.- The
decimaltype stores variable-precision decimal numbers, with no upper bound on size. Unlike a floating point number, a variable-precision decimal will never suffer from base 10 rounding errors in the fractional part of the number.
But decimal is likely to take up more space compared to other two. So, if it is a matter of precision, you can go for decimal. Otherwise, float/double are good enough in most of the cases.
My observation was that float at max holds value 5 digits after decimal beyond that it approximates it. Ex if you have 1.00001 it would get inserted as it is but if you have 1.000001 it's inserted as 1 . It rounds up based on precision. If you have clear idea about about mantisa and precision (how many digits before and how many digits after decimal ) you can go with decimal . (Float and double behave the same way) .
