Floating point precision - does/ will this library support this? about stdlib HOT 3 CLOSED

@bguiz Agreed. These libraries would be nice to have, but we first want to cover all our bases and get the basics right before moving on to higher order concerns. Much to do, so little time!

Also, if you have any other general questions, feel free to also seek us out and leave a message on Gitter.

from stdlib.

@bguiz Yeah, thanks for the question. This behavior is expected.

> 0.1 + 0.2
0.30000000000000004

This is not because JavaScript implementors have failed to properly implement floating-point math. The reason the answer is not exactly 0.3 is because not all numbers are exactly representable as double-precision floating-point numbers (see here). In short, addition is working as expected.

In general, this type of imprecision is not an issue, as floating-point rounding errors tend to cancel out. What we are concerned about is the types of algorithms used to implement basic math functions, such as trig and other special functions. The errors arising from these algorithms are of a different kind than the errors arising from floating-point math.

To "address" floating-point errors, you can use symbolic or arbitrary precision arithmetic libraries, which is not currently within the scope of this library, but may at some point in the future.

from stdlib.

Thanks for the answer!

To "address" floating-point errors, you can use symbolic or arbitrary precision arithmetic libraries, which is not currently within the scope of this library, but may at some point in the future.

With the name stdlib I think including this at some point in the future might make sense. For example, you already have some util functions which overlap with these libraries, for example this one which converts a higher precision float to a lower precision one: https://github.com/stdlib-js/stdlib/tree/develop/lib/node_modules/%40stdlib/math/base/utils/float64-to-float32

from stdlib.