It returns a string because 0.1, and powers thereof (which are used to display decimal fractions), are not representable (at least not with full accuracy) in binary floating-point systems.
For example, 0.1 is really 0.1000000000000000055511151231257827021181583404541015625, and 0.01 is really 0.01000000000000000020816681711721685132943093776702880859375. (Thanks to BigDecimal
for proving my point. :-P)
Therefore (absent a decimal floating point or rational number type), outputting it as a string is the only way to get it trimmed to exactly the precision required for display.