Reading a Windows 'binary' float into a ASP jscript variable

Question

I need to read files produced by a legacy Windows app that stores real numbers (the 8-byte "double" type) in binary - i.e. as a packed array of 8 bytes. I can read the 8 byte group OK but how can I present it to my ASP JScript code such I can get the real number back again.

Or to put it another way:

Say a file was produced by a Windows (Delphi) program:

Assign (f, 'test.bin') ;
rewrite (f, 1) ;
r := 1234.56E78 ;
BlockWrite (f, r, SizeOf (Double)) ;
Close (f) ;

Inspection of the file will show it contains 8 bytes, being:

94 0E 4C CA C2 97 AD 53

which is the real number in IEEE format. Assuming I can read these 8 bytes back in ASP, is there a simple way of getting the real number back again?

Answer 1

You could use an ADO stream. It is used to read, write, and manage a stream of binary data or text.

Dim objStream As ADODB.Stream
objStream.Type = 1
objStream.LoadFromFile path

Dim Number : Number = BytesToNumEx(objStream.Read 8, 1, 8, False)

You'll need the following function though:

Function BytesToNumEx(ByteArray(), StartRec, EndRec, UnSigned) 
    Dim i
    Dim lng256 : lng256 = 1
    Dim lngReturn : lngReturn = 0

    If EndRec < 1 Then
        EndRec = UBound(ByteArray)
    End If

    If StartRec > EndRec Or StartRec < 0 Then
        BytesToNumEx = -1
        Exit Function
    End If

    lngReturn = lngReturn + (ByteArray(StartRec))
    For i = (StartRec + 1) To EndRec
        lng256 = lng256 * 256
        If i < EndRec Then
            lngReturn = lngReturn + (ByteArray(i) * lng256)
        Else
            If ByteArray(i) > 127 And UnSigned = False Then
             lngReturn = (lngReturn + ((ByteArray(i) - 256) * lng256))
            Else
                lngReturn = lngReturn + (ByteArray(i) * lng256)
            End If
        End If
    Next 

    BytesToNumEx = lngReturn
End Function

Answer 2

Look at BitConverter.ToDouble().

In an ASP.NET page, this could look like:

<%@ Page Language="JScript"
    AutoEventWireup="true"
    CodeBehind="Default.aspx.cs"
    Inherits="WebApplication1._Default" %>
<%
    var bytes: byte[] = [0x94, 0x0e, 0x4c, 0xca, 0xc2, 0x97, 0xad, 0x53];
    var d = BitConverter.ToDouble(bytes, 0);
    Response.Write(d);
%>

Which gets you the output:

1.2345678E+95

(You can check for correctness at http://babbage.cs.qc.edu/IEEE-754/)

Answer 3

Thanks Oren for your help. The babbage stuff was particularly useful. For what it's worth here the code I had to write (not thoroughly tested for anything other than norms):

function ReadFileDouble (Stream)
{ 
var Bytes0To3 = ReadFileLongword (Stream) ;
var Bytes4To5 = ReadFileWord (Stream) ;
var Bytes6To7 = ReadFileWord (Stream) ;

var Mantissa = Bytes0To3 + (Bytes4To5 * 65536 * 65536) + ((Bytes6To7 & 0x0F) * 65536 * 65536 * 65536) ;
var Fraction = Mantissa / 0x0010000000000000 ;
var Exponent = ((Bytes6To7 >>> 4) & 0x07FF) ;
var SignBit  = ((Bytes6To7 & 0x8000) != 0) ;
if (SignBit)
    {
    Sign = -1 ;
    }
else
    {
    Sign = +1 ;
    }    

var Bias     = 1023 ;

if (Exponent == 0)
    {
    if (Mantissa == 0)
        {
        return (0.00 * Sign) ;
        }
    else
        {
        return ((Fraction * (Math.pow (2.0, - Bias + 1))) * Sign) ;
        }        
    }
else if (Exponent == 1023)
    {
    if (Mantissa == 0)
        {
        return (Number.POSITIVE_INFINITY * Sign) ;
        }
    else if (Mantissa < 0x0008000000000000)
        {
        return (Number.NaN * Sign) ;
        }
    else
        {
        return (Number.NaN * Sign) ;
        }        
    }    
else 
    {
    return ((1.0 + Fraction) * (Math.pow (2.0, Exponent - Bias)) * Sign) ;
    }    
}