Support parsing `snan`

[tannergooding]

The IEEE 754:2008 specification requires that we be able to recover a string matching "snan" (case-insensitive) with an optional preceding sign. We are not required to recover this to an actual snan and are permitted to silently convert it to a qnan (such as double.NaN or float.NaN) instead.

@jkotas raised some concern around this on https://github.com/dotnet/coreclr/issues/19802 given that it may require a new public surface and due to it not being unusual to have NaN checked for by string comparison

At a minimum, we should update the invariant culture to check for snan and return double.NaN or float.NaN. Optionally, and if it is determined to not be breaking, we can add a NumberFormatInfo.SignallingNaN and provide culture specific snan information.

I don't believe either of these should be considered breaking since we are only parsing new values. What could be breaking is if we start formatting snan to NumberFormatInfo.SignallingNaN, as that may break existing code that checks for NaN by string comparison (we currently don't, as IEEE allows us to silently convert a signalling NaN into the quiet nan string: "NaN").

[danmoseley]

@tannergooding I marked this Future

[tannergooding]

I think future is fine. The actual fix here is pretty trivial, we just need to decide on what to do.

I think, the easiest fix, would be to support this only for the invariant culture and only support it for parsing (that is, you can would never get SNaN as a result of calling ToString()).

• The spec indicates that we need to support parsing snan, but we can return qnan as the result. It also specifies that we can return just "NaN" as the result of formatting
• Only supporting it for the invariant culture means we don't need to worry about exposing a new SignallingNaN property on NumberFormatInfo, while still giving users a way to support the value

@jkotas, do you think the above would alleviate the concerns you raised?

[jkotas]

I would wait for somebody to ask for this. It does not look important to me to fix.

What does Java or Python do for parsing / formatting snans?

[tannergooding]

Both format quiet and signalling NaN to the same result NaN on Java and nan on Python. Neither look to support parsing snan in any casing and Java looks to be case sensitive when parsing NaN. Both support parsing the sign prefix of + or -.

I agree that it isn't particularly important to fix, but it would be nice to check the box off indicating that we support this particular IEEE requirement, especially since we are allowed to quietly convert it to a quiet NaN (and therefore don't need any special support).

Tagging subscribers to this area: @tannergooding
Notify danmosemsft if you want to be subscribed.

[dotnet-policy-service]

Due to lack of recent activity, this issue has been marked as a candidate for backlog cleanup. It will be closed if no further activity occurs within 14 more days. Any new comment (by anyone, not necessarily the author) will undo this process.

This process is part of our issue cleanup automation.