How to read 8-byte floating point numbers from binary？use fread

fseek(fid,56,'bof');

x = fread(fid,1,'float64','b')

[Cris LaPierre]

What is or isn't happening? Why do you think this code is incorrect?

[aijiangzhao aixuwu]

This is correct, but a new question has arisen. For example, how can "DC 88 12 C0 C1 4F 7C A9" be replaced with "C0 12 88 DC A9 7C 4F C1"? Then read as double-precision floating point

[Walter Roberson]

When you fopen you should specify the byte order as the third parameter to fopen, unless you are reading from something with mixed orders.

If you are reading from something with mixed orders but know the order for one particular binary fread then specify the byte order as a parameter to fread.

If you are reading a sequence of bytes at one time and regrouping them (such as if you were reading a structure with mixed datatype) then if necessary use swapbytes after any necessary typecast()

Answer 1

[Jan]

With 'ieee-le' format the bytes are read in the order [8,7,6,5,4,3,2,1], with 'ieee-be' as [1,2,3,4,5,6,7,8]. There is no format to store doubles with the byte order [4,3,2,1,8,7,6,5]. Yo you have to convert the order manually by reading the bytes, mix them as wanted and use typecast for the conversion.

[EDITED] As code:

bytes = fread(fid, 8, '*uint8');
bytes = bytes([4,3,2,1,8,7,6,5]);
value = typecast(bytes, 'double')

[aijiangzhao aixuwu]

Yes, but how to implement the 8-byte big-endian conversion in matlab?

[Walter Roberson]

If the data is in big-endian then either specify ieee-be or swapbytes()

format long g
T = -pi

T = 
         -3.14159265358979

fprintf('source data\n');

source data

t64 = typecast(T, 'uint64');
fprintf('%016x\n', t64(1));

c00921fb54442d18

t8 = typecast(T, 'uint8');
fprintf('%02x ', t8(1:8)); fprintf('\n');

18 2d 44 54 fb 21 09 c0 

fn = tempname;
fid = fopen(fn, 'w', 'ieee-be');
fwrite(fid, T, 'double');
fclose(fid);

fprintf('reading from file\n');

reading from file

dinfo = dir(fn);
fprintf('file is %d bytes long\n', dinfo.bytes);

file is 8 bytes long

fid = fopen(fn, 'r');
in_bytes = fread(fid, [1 inf], '*uint8');
fprintf('%02x ', in_bytes); fprintf('\n');

c0 09 21 fb 54 44 2d 18 

frewind(fid)
in_64 = fread(fid, [1 inf], '*double');
in_64

in_64 = 
     3.20737563067642e-192

swapbytes(in_64)

ans = 
         -3.14159265358979

fprintf('%016x\n', typecast(in_64, 'uint64'));

182d4454fb2109c0

frewind(fid)
in_64 = fread(fid, [1 inf], '*double', 'ieee-be');
in_64

in_64 = 
         -3.14159265358979

swapbytes(in_64)

ans = 
     3.20737563067642e-192

fprintf('%016x\n', typecast(in_64, 'uint64'));

c00921fb54442d18

fclose(fid);

[aijiangzhao aixuwu]

Thank you for your answer. Sorry to reply you so late, my question now is how to turn 1 2 3 4 5 6 7 8 into 5 6 7 8 1 2 3 4?

[Walter Roberson]

typecast to uint32, exchange pairs of entries, typecast back to double .

format long g
in_64 = 0x0102030405060708

in_64 = uint64
   72623859790382856

d64 = typecast(in_64, 'double')

d64 = 
     8.20788039913184e-304

num2hex(d64)

ans = '0102030405060708'

temp_in = typecast(d64, 'uint32');
temp_out = temp_in([2:2:end; 1:2:end]);
d64swap = typecast(temp_out, 'double')

d64swap = 
     1.85164890819352e-284

num2hex(d64swap)

ans = '0506070801020304'

[Jan]

bytes = fread(fid, 8, '*uint8');
bytes = bytes([5,6,7,8,1,2,3,4]);
value = typecast(bytes, 'double')

[aijiangzhao aixuwu]

The picture is my hexadecimal file, the shaded part is 8 bytes as a parameter, my original code is as follows:

fseek(fid,56,'bof');
x = fread(fid,1,'float64','b');

The output result is -8.041e27, which is obviously wrong and should be of the order of e6；

Now we need to convert between big and little endian, C5 B9 FB 60 C1 4B FD CA convert to

C1 4B FD CA C5 B9 FB 60(it should be this)，and then read as a double-precision floating-point number.

Thank you very much for your answer and look forward to your reply.

[Jan]

As I have suggested already, except for the changed byte order:

% fseek(fid, 56, 'bof');
% bytes = fread(fid, 8, '*uint8');
% Manually:
bytes = uint8(sscanf('C5 B9 FB 60 C1 4B FD CA', '%x')).';
value = typecast(fliplr(bytes), 'double')  % As your fread('float64')

value = -8.0410e+27

% [197, 185, 251, 96, 193, 75, 253, 202]
bytes = bytes([4,3,2,1,8,7,6,5]);
value = typecast(bytes, 'double')

value = -3.6689e+06

Now the byte order and the order of 4-Byte-blocks is swapped.

[aijiangzhao aixuwu]

Thank you very much for solving the problem that caused me a lot of headaches.