question

Henrik Staun Poulsen avatar image
Henrik Staun Poulsen asked

how do I convert a hex to a floating point number

How do I convert a hex to a floating point number? as in 0x43F8E354 should be 497.776 This is SQL Server 2008 R2 I've tried [ http://www.novicksoftware.com/UDFofWeek/Vol1/T-SQL-UDF-Volume-1-Number-23-fn_replinttobitstring.htm][1], but it does not look as if fn_replinttobitstring exists in Sql Server 2008 R2. In SSIS I can do like this: Public Function ReadDouble(ByVal blobdata As Byte()) As Double ' Convert gain Dim MemoryStream As New System.IO.MemoryStream(blobdata) Dim BinaryReader As New System.IO.BinaryReader(MemoryStream) ReadDouble = BinaryReader.ReadDouble() End Function But how to do this in T-SQL? [1]: http://www.novicksoftware.com/UDFofWeek/Vol1/T-SQL-UDF-Volume-1-Number-23-fn_replinttobitstring.htm
sql-server-2008t-sqlsql-server-2008-r2math
3 comments
10 |1200 characters needed characters left characters exceeded

Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

@Henrik Staun Poulsen The function you have in your question does not do what you need even in SSIS because it converts the input to long integer not the floating point number. As @bopeavy already stated in the answer below, conversion to/from int (or bigint) from/to hex is already available in T-SQL. The only problem is that the statement select convert(bigint, 0x43F8E354) returns **1140384596** not the desired **497.77600098** Since your input is 4 bytes, 497.77600098 is not an accurate representation of the input. When you declare a variable as **float** then **float(53)** is assumed and **8** bytes of storage are needed. If your input is 4 bytes then your output cannot be a default float but is rather a single floating point number, which translates in T-SQL to **float(24**) and is sometimes called **real**. This one needs only 4 bytes of storage but the number of significant digits is only 7, so 497.77600098 does not comply with this limitation. Here is the sample showing that default definition of float needs 8 bytes of storage:
declare @f float = 0.0;
select datalength(@f);

-- the above returns 8
Please confirm that you need a conversion to float(24)
0 Likes 0 ·
The number 497.76600098 is taken from another tool for which I do not have the source. But I can see that it converts the number to a FLOAT, not a REAL. I had shown the wrong function. That should be fixed now.
0 Likes 0 ·
@Henrik Staun Poulsen I added some comments to my answer related to your SSIS function. I hope they make sense.
0 Likes 0 ·
bopeavy avatar image
bopeavy answered
This would be a good place for you to start: [Convert Hex to int][1]. [1]: http://stackoverflow.com/questions/703019/sql-query-convert-integer-to-hex-and-hex-to-integer
1 comment
10 |1200 characters needed characters left characters exceeded

Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

yes, I was able to find that link, but not get any use of it.
0 Likes 0 ·
Oleg avatar image
Oleg answered
There is a great discussion of this topic on the main SSC site which includes Jeff Moden, Peter Larsson (whom we know as @Peso on this site) and Martin Wills. Here is the link: [ http://www.sqlservercentral.com/Forums/Topic356756-8-1.aspx][1]. I just tried the version which Martin Wills posted on page 3 which was in turn a recoding of Peter's code. Here is the paste from the post referenced above: /* Hats off to Peter Larsson and Martin Wills for this one */ CREATE FUNCTION dbo.fnBinaryReal2Real ( @BinaryFloat BINARY(4) ) RETURNS REAL AS BEGIN DECLARE @Mantissa REAL, @Exponent SMALLINT, @IntValue INT, @Real2 REAL SELECT @Real2 = CAST(2.0 AS REAL), @IntValue = CAST(@BinaryFloat AS INT), @Exponent = (@IntValue & 0x7f800000) / 0x00800000, @Mantissa = 1.0 + (@IntValue & 0x007FFFFF) * POWER(@Real2, -23) RETURN SIGN(@IntValue) * @Mantissa * POWER(@Real2, @Exponent - 127) END GO Here is the test: declare @input binary(4) = 0x43F8E354; select dbo.fnBinaryReal2Real(@input) result; result ------------- 497.776 Please note that the desired number 497.77600098 is not real (in both senses) because it has way too many digits for a real (a.k.a. float(24)) number to afford. However, if the accurate binary representation **0x43F8E354** of the single floating point number **497.776** is sufficient then the function above should help. <\!-- **Begin Edit** Here is the link to Peter's post where he restated both binary to real and binary to float as inline scalar functions: [ http://www.sqlteam.com/forums/topic.asp?TOPIC_ID=81849][2]. Scroll to the bottom of the post to see them. **End Edit** --> <\!-- **Begin Edit 2** I am glad that the links helped. However, I would like to comment on the function you use in SSIS. Please don't use it as is if you can because it leaks memory. Not much memory to speak of, but it still does which is never a good thing. Both MemoryStream and BinaryReader are disposable types but the code does not close and dispose any of them. Thus, the code will leak some memory and might eventually trash the server if the function is called frequently enough for a long enough period of time. Usually the way any conversion works is consistent: it always converts as many bytes as needed. For example if you need to convert a byte array to REAL (System.Single in VB or float in C#) then only 4 bytes of the array will be read and converted. Conversion to FLOAT (System.Double in VB or double in C#) will take to read and then convert 8 bytes of the byte array input. Here is the function written in C# which you can use in place of the one you have (you can use online [telerik code converter][3] to convert it to VB from C#): /// /// Converts the byte array to a double precision floating point number. /// These have a maximum number of significant digits limited to 15. /// In T-SQL this type translates to float(53) or simply float by default. /// /// System.Double (called double in C#) needs 8 bytes of /// storage which means that the input should be limited to 8 bytes. /// Double precision floating point number static double ReadDouble(byte[] blob) { if (BitConverter.IsLittleEndian) Array.Reverse(blob); return BitConverter.ToDouble(blob, 0); } Similarly, if you need to convert a byte array to a single precision floating point number, you can feed the 4 bytes long byte array to a similar function like this: /// /// Converts the byte array to a single precision floating point number. /// These have a maximum number of significant digits limited to 7. /// In T-SQL this type translates to float(24) sometimes called real. /// /// System.Single (called float in C#) needs 4 bytes of /// storage which means that the input should be limited to 4 bytes. /// Single precision floating point number static float ReadSingle(byte[] blob) { if (BitConverter.IsLittleEndian) Array.Reverse(blob); return BitConverter.ToSingle(blob, 0); } As you can see, the code is pretty simple, and does not use any disposable types. Testing the latter using the bytes in your original question: byte[] blob = new byte[] { 0x43, 0xF8, 0xE3, 0x54 }; float f = ReadSingle(blob); Console.WriteLine(f); Console.ReadLine(); will print 497.776 in the console window as expected. **End Edit 2** --> Oleg [1]: http://www.sqlservercentral.com/Forums/Topic356756-8-1.aspx [2]: http://www.sqlteam.com/forums/topic.asp?TOPIC_ID=81849 [3]: http://converter.telerik.com/
2 comments
10 |1200 characters needed characters left characters exceeded

Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.

Hi Oleg, That is two very useful links. The first link to a function that solves my problem 300%. 100% to be able to decode the value, 100% for something that is fast, and 100% for beeing convertable to a view. The second link to 26 pages of excellent code. Whoa. You've made my day. Thank you very much. Best regards, Henrik
0 Likes 0 ·
Hi Oleg, Bummer, so that is what has been taking our server down... Well, we've noticed a memory leak, but it was so small it did not matter (much (on a 512 GB machine)). I'll fix that in 3 weeks time, when I'm back from summerholidays. Again, Thank you very much for your help. Best regards, Henrik
0 Likes 0 ·

Write an Answer

Hint: Notify or tag a user in this post by typing @username.

Up to 2 attachments (including images) can be used with a maximum of 512.0 KiB each and 1.0 MiB total.