Problems with Concatenating Binary Data

Question

Hi
I am experiencing a problem with the unreliable concatenation of binary data.
I have a stored procedure that I wrote to perform cell encryption on long text fields (> 4000 characters in length). This takes the string, splits it into chunks of a set length, encrypts and concatenates the encrypted chunks into a varbinary(max) variable.
It then stores the length of the encrypted chunk at the beginning of the data (the length of output when encrypting strings of a set length is always be the same), e.g.
DECLARE @output VARBINARY(MAX), @length INT
SELECT @length = LEN(@encryptedchunk1)
SELECT @output = CONVERT(binary(8), @length) + @encryptedchunk1 + @encryptedchunk2 + @encryptedchunk3 + ...
So far so good, and when I decrypt the data, I can read the first 8 bytes of data at beginning of the binary data to get the length of the chunk:
-- get the length of the encrypted chunk
SELECT @length = CONVERT(INT, CONVERT(binary(8), LEFT(@encrypteddata, 8)))
-- then remove the first 8 bytes of data
SELECT @encrypteddata = CONVERT(VARBINARY(MAX), RIGHT(@encrypteddata, LEN(@encrypteddata) - 8))
<snip> code to split into chunks of length @length and decrypt
</snip>
This is where I am experiencing an issue. 99.4% of the time, the above code is reliable. However, 0.6% of the time this fails for one of two reasons:
the length is stored incorrectly. The length of the encrypted chunks is usually 4052 and substituting 4052 for the returned value (4051) allows the encrypted chunks to be decrypted.
the encrypted data sometimes starts at offset 8 instead of 9. The @length variable is correctly read at length 8 but only 7 bytes should be removed from the start, e.g.
SELECT @length = CONVERT(INT, CONVERT(binary(8), LEFT(@encrypteddata, 8)))
SELECT @encrypteddata = CONVERT(VARBINARY(MAX), RIGHT(@encrypteddata, LEN(@encrypteddata) - 7))
Has anyone any ideas on why this is happening and how the process can be made more reliable?
Julian

Answer 1

Use datalength, not len. Len() is designed for character strings and will ignore trailing bytes. And count characters. Datalength does exactly what you want: it counts bytes. Look at this:
DECLARE @b varbinary(200)
SELECT @b = 0x4142434420202020
SELECT @b, len(@b), datalength(@b)
Erland Sommarskog, SQL Server MVP, [email protected]