Is NaN ==/!= NaN guaranteed to be consistent in all Groovy versions?

["Winnebeck, Jason" <Ja...@windstream.com>]

I see in Groovy that == and != work to compare variables against NaN. The rationale seems to be that compareTo is used instead of primitive comparisons or equals. Is this behavior guaranteed to be defined this way in all future versions of Groovy?

I've come across some Groovy code that is doing "x != Float.NaN" and in all of the test cases it seems to work. I ask because I'm afraid of primopts or static compiler in the future perhaps replacing the float to float comparison with a primitive !=, which returns false always in Java, but if this is a locked in behavior for Groovy, then I won't worry about refactoring existing code.

I see that https://issues.apache.org/jira/browse/GROOVY-3309 addresses this and Jochen responded to it but it was left open instead of closed as "won't fix", which concerns me that Groovy team may introduce a breaking change in the future.

Jason

----------------------------------------------------------------------
This email message and any attachments are for the sole use of the intended recipient(s). Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message and any attachments.

RE: Is NaN ==/!= NaN guaranteed to be consistent in all Groovy versions?

["Winnebeck, Jason" <Ja...@windstream.com>]

I just discovered that while this works in dynamic Groovy, in at least in Groovy 2.4.3 when using @CompileStatic, it works as Java does and not as dynamic Groovy does. Therefore, I will assume it is not safe to rely on ==/!= Float.NaN working properly. Even if this is to be considered a bug in static compiler, the fact that a released version of Groovy is inconsistent means that we should never use this code pattern.

Jason

From: Winnebeck, Jason [mailto:Jason.Winnebeck@windstream.com]
Sent: Monday, September 21, 2015 10:15 AM
To: users@groovy.incubator.apache.org
Subject: Is NaN ==/!= NaN guaranteed to be consistent in all Groovy versions?

I see in Groovy that == and != work to compare variables against NaN. The rationale seems to be that compareTo is used instead of primitive comparisons or equals. Is this behavior guaranteed to be defined this way in all future versions of Groovy?

I've come across some Groovy code that is doing "x != Float.NaN" and in all of the test cases it seems to work. I ask because I'm afraid of primopts or static compiler in the future perhaps replacing the float to float comparison with a primitive !=, which returns false always in Java, but if this is a locked in behavior for Groovy, then I won't worry about refactoring existing code.

I see that https://issues.apache.org/jira/browse/GROOVY-3309 addresses this and Jochen responded to it but it was left open instead of closed as "won't fix", which concerns me that Groovy team may introduce a breaking change in the future.

Jason
________________________________
This email message and any attachments are for the sole use of the intended recipient(s). Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message and any attachments.