how to support precision of 38 decimal

[danhuiwang]

999995881114.744959800000000134 ,use decimal the result is 999995881114.7449598000000001

[schungx]

You only get 28 digits of precision for this type because it is based on 96 bits. You cannot get more precision than 28.

[paupino]

Hi @danhuiwang,

As @schungx points out - this library only supports 28 digits of precision at present. When parsing a string it tries to round the value (safely) to fit into the provided format - which in this case drops the last two digits of precision. If you want to avoid this implicit rounding you can instead use from_str_exact which will instead error if it can't parse the entire decimal.

A maximum scale of 28 is a hard limit of this library at present - it is something I'd like to investigate with version 2 of this library using const generics however this requires a few more nightly features to be stabilized first.

If you need a library with a larger precision then I believe you can utilize bigdecimal. This uses a BigInteger behind the scenes which can grow in size accordingly.

Let me know if you have any further questions!

[Tony-Samuels]

A maximum scale of 28 is a hard limit of this library at present - it is something I'd like to investigate with version 2 of this library using const generics however this requires a few more nightly features to be stabilized first.

If this comes out, it is likely to be as a new crate, at least initially. Closing this out for now.