Question

I wrote a function in plpgsql unpack_numeric_bytes, the most important part of the functions is:

FOR i IN 1..v_length LOOP
    v_val := v_val + (get_byte(v_bytes, v_byte_index) << v_bit_shift);
    v_bit_shift := v_bit_shift + 8;
    v_byte_index := v_byte_index + 1;
END LOOP;

It has worked fine until I came across value that it didnt decode correctly.. and it might be too obvious but I am not seeing it.

The call is: select unpack_numeric_bytes(E'g\\363I\\274', array[4], 'f');

From the code v_length is 4 bytes and it tries to decode E'g\\363I\\274'
Which comes up to : {-1136004249}.. I noticed that
2^32 – 1136004249 = 3158963047, and that is the right answer I am looking for ! (tested it with python unpack function).

What is throwing it off? and what am I doing wrong? This is the first time the function failed me.

Thanks!

Answer 1

Found what the problem was:

select ('10111100000000000000000000000000'::bit(32))::integer;
int4 | -1140850688
-- versus
select (('10111100000000000000000000000000'::bit(32))::bigint);
int8 | 3154116608

Fixed this line:

v_val := v_val + (get_byte(v_bytes, v_byte_index)::BIGINT << v_bit_shift)::BIGINT;