Safe and correct way to convert REAL to Int

mktdata
mktdata Contributor
edited May 20 in TREP APIs

Hello, We use RFA 8.1 C++ api. We have a use case where we want to convert REAL type to int using DataBuffer. We accept the fact that doing this will result in data truncation. I can do it in two ways but what is the most accurate way to do it?

Assuming DataBuffer contains REAL type -

A. dataBuffer.getInt()

B. static_cast<int>(dataBuffer.getDouble());

Which one of the above is more accurate? Is there any other way to do this?

Tagged:

Answers

  • mktdata
    mktdata Contributor

    I just tried something and realized that following does not work correctly

    dataBuffer.getInt()

  • Jirapongse
    Jirapongse ✭✭✭✭✭

    @mktdata

    Thank you for reaching out to us.

    It could look like this:

    			write("\n%s\n", dataBuffer.getAsString().c_str());						
			write("%lld\n", dataBuffer.getInt());			
			write("%d\n", (int)round(dataBuffer.getDouble()));

    If you would like to get a round number, you can use the round method in <cmath>.

  • mktdata
    mktdata Contributor
    edited May 20

    HI @Jirapongse When value is "45.110" then dataBuffer.getInt() returns 45110 which is not expected in our case I guess.

    Also REAL64 can have exponent part, so how would it work in that case?

  • Jirapongse
    Jirapongse ✭✭✭✭✭

    @mktdata

    I used this code.

    		case DataBuffer::RealEnum:
			printf("\n%s\n", dataBuffer.getAsString().c_str());						
			printf("%lld\n", dataBuffer.getInt());
			printf("%d\n", (int)round(dataBuffer.getDouble()));
			break;

    It returned the correct values. 

    2025 May 20 17:26:32.441 ST GMT+07:00 998C 8F08   594 TRACE <- Received MMT_MARKET_PRICE Update JPY=.
    
    FieldList: standard data count = 2
    
        FieldEntry [    22] BID
    
45.110
    
45
    
45

    I am using rfa8.2.5.L3.

  • mktdata
    mktdata Contributor
    edited May 26

    @Jirapongse can you cofirm if round() is needed to handle IEEE754 binary floating point representation? So is it possible that when we have to convert REAL64 value 12.0 , it can be represented as 11.9999999998 after converting to double from REAL64? If so then I have to always use round after REAL to int conversion to get corrrect 12 value?

  • Jirapongse
    Jirapongse ✭✭✭✭✭

    @mktdata

    The round() is not part of our library. You may need to find and check it online regarding IEEE754.

    Moreover, you may need to run a lot of test cases to make sure that the real values can be converted to integer values properly.

Categories