Looking for numeric validation in c

[Legacy User]

While trying to enhance our not-so-good WinC API for numeric validation, I finally decided to give a try to an intrinsic solution, casting a string to a numeric field, forcing a NAN (Not a Number) return when the string field have other than numeric data. But while the check for NAN is a more or less solved issue since c++11, it is a fuzz issue for Visual Studio 2005. There is a function _isnan() available since VS2005, but only since VS2013 all types seems to be covered. For VS2005, there is a limited version, only based on double type, and not very accurate (at least in my test, never get a return > 0, which should be the response for a NAN). Based on my simple research (better discussion here: Checking if a double (or float) is NaN in C++ - Stack Overflow ), it could be a compiler parameters limit (based on /fp (Specify Floating-Point Behavior) ), and it is our case right now: I tried to add in Generate&Build window/custom parameters, for compiler parms the reccomended option /fp:precise in order to ask for DataTest != DataTest which should return true for a NAN, but I can't get any change in behavior.

Have someone some experience on this issue? some other way to solve it?

TIA

Jorge Ubeda

[Toshikazu Toyama]

Hi Jorge,

 

 

I feel sorry ab
I'm sorry, I do not have an answer about "function_isnan()".

However, (It is a reverse pattern,) The case which I know is a problem of "C format = Fixed_decimal". This is PTF of Plex which was revised not to return NaN.
Please see the following URLs.
https://support.ca.com/irj/portal/kbproblem?docid=630533&productcd=CPLEX&problemnbr=1808
https://support.ca.com/irj/portal/kbproblem?docid=632104&productcd=CPLEX&problemnbr=1933

Could you please contribute about optional addition to return NaN cord to an ideas?
Regards,

[Legacy User]

Hi Toshikazu

Thanks so much for your help. You have introduced the case for the c format fixed decimal...Second document says that its Oracle issue is solved since 7.0 build 001. I imagine it is referred to the involved cast from Oracle to Plex, but I suppose casting a character to a fixed decimal continues giving a NAN...

Certainly, with the advent of Plex 7.2.1 my issue becomes irrelevant, because I can use at this level c++ isnan routine.

Actually my question here was related to fine tune our compiling process, in order to apply the /fp:precise parameter to the compiler as reccomended in MSDN.

By the way, staying yet at VS 2005 level, we replaced our proposed test for NAN, with a detailed and ruled scan of a string, which works.

Jorge Ubeda

[Toshikazu Toyama]

Hi Jorge,
I looked for a solution, but it was not found.
You make an analysis program in VS2013, and can API call do it from Plex(VS2005)?
I hope that this helps you.
Regards,

[Legacy User]

Hi Toshikazu

Actually, as said, I solved my numeric validation at level 7.2 (VS2005) with a string based scan. Regarding the possibility of using isnan c++ template it requires at least VS2010. The case is that until Plex 7.2.1, it would be impossible to make an API covering numeric completeness well, but now it is solved. I ignore when we will move to 7.2.1, but there is a solution ahead now.

With regards

Jorge

[Legacy User]

second point: several conversations, like the one at StackOverflow, points to the query for NAN as a valid way to test numeric at VS2005 level. However, it is required to define the /fp parameter to the compiler, as said in MSDN:

The compiler correctly handles comparisons that involve NaN. For example, x != x evaluates to true if x is NaN and ordered comparisons involving NaN raise an exception.

Then, I have added to User Build Settings from Generate & Build options that /fp:precise compiler parameter, both for internal and external c++ functions, but source code continued not detecting NAN. Originally my question here was about the use and known results of that parameter. Indeed, if I get this option working, I would replace my current string focused validation by another one shorter.

Jorge