Split string of hex shorts into cell values

I have a field from an ODBC connection where the data of a curve is stored in the following way:

• range for single value is signed 16bit (-32768 to 32767)
• representation in HEX 4 "digits" with padding left zeros (8000 to 7FFF)
• all values are concatenated to a string in the database
• no delimiters
• number of elements is string length/4

Example:

• Values (123, 456, -17000)
• String contains: "007B01C8BD98"

Now I would like to us this data inside Matlab.

I could prepare the data and create a translation database. But I would prefere a direct approach.

All string splitting operators make use of delimiters. But this does not work here.

What is the Matlab/Database Explorer way to get access to that data ? Are there any ready to use functions ?

Answer 1

[Guillaume]

Simply reshape the string into 4 columns, pass it through dec2hex and account for the fact that dec2hex assumes unsigned integer by casting from unsigned to signed:

s = '007B01C8BD98';  %demo data
typecast(uint16(hex2dec(reshape(s, 4, [])')), 'int16')

Answer 2

[Jan]

Based on Guillaume's solution:

typecast(uint16(sscanf(s, '%4x'), 'int16')

sscanf is much faster and nicer to convert hex to dec values.

[Guillaume]

Yep, much simpler. I never think about sscanf.

Answer 3

[ES]

One suggestion would be, since the width of each value is fixed, you can do something like this..

string = '007B01C8BD98ABCDEFA';%Example data
idx = 1:4:length(string);
for iloop=1:length(idx)-1
    value(iloop) = hex2dec(string(idx(iloop):idx(iloop+1)-1))
end

[Guillaume]

You could avoid the loop entirely (see my answer), but if you're going to use a loop, you should at least predeclare value

value = zeros(1, numel(string)/4);

Clearly your code does not decode properly negative values.

Note that using string as a variable nam is not recommended in version >= R2016b, since it's now the name of a very useful class.

[Bartosch Slomak]

This answer also works but the "one line approach" by Guillaume seems more elegant. Thank you!