binary file write/read operation in Matlab - matlab

Question

I am trying to write some 2D coordinates into a binary file. However, what I read from the file that had been writen is quite different from the original data. Details are given here.
For example, I have 45 (X,Y) points. Both X and Y are integral number less than 600. The simulation requires store each of them with two bytes (8 bits) and 2 upper bits of each byte is reserved (for X, the reserved bits are filled by .mrk which is 1 or 2; for Y, simply use 0 instead). In this case, 14 bits binary number is able to represent the maximum, 16383. I write the data in several ways:
in_tmp is structure consisted of points number (.nm), points reserved mark (.mrk) and points coordinates (.coor)
for i=1:in_tmp.nm
x1 = dec2bin(in_tmp.coor(i,1));
y1 = dec2bin(in_tmp.coor(i,2));
t1 = in_tm.mrk(i);
if(t1==1)
t2 = '01';
t2b = 1;
elseif(t1==2)
t2 = '10';
t2b = 2;
end
lenx = 16-length(x1);
leny = 16-length(y1);
x1hl = strcat(t2, '00000000000000'); % High and low
y1hl = '0000000000000000';
x1a = strcat(x1hl(1:lenx), num2str(x1));
y1a = strcat(y1hl(1:leny), num2str(y1));
y1a(1:2) = '00';
% x1b = in_tmp.coor(i,1);
% y1b = in_tmp.coor(i,2);
% fwrite(fp1, t2b, 'ubit2');
% fwrite(fp1, x1b, 'ubit14');
%
% fwrite(fp1, 0, 'ubit2');
% fwrite(fp1, y1b, 'ubit14');
fwrite(fp1, bin2dec(x1a), 'uint16');
fwrite(fp1, bin2dec(y1a), 'uint16');
% fwrite(fp1, bin2dec(x1a(1:8)), 'uint8');
% fwrite(fp1, bin2dec(x1a(9:end)), 'uint8');
% fwrite(fp1, bin2dec(y1a(1:8)), 'uint8');
% fwrite(fp1, bin2dec(y1a(9:end)), 'uint8');
% x1c = in_tmp.coor(i,1);
% y1c = in_tmp.coor(i,2);
%
% x1hex = dec2hex(x1c);
% y1hex = dec2hex(y1c);
% if(length(x1hex)>2)
% x1h = x1hex(1:end-2);
% x1l = x1hex(end-1:end);
% else
% x1h = dec2hex(0);
% x1l = x1hex;
% end
%
% tx1h = dec2bin(hex2dec(x1h));
% l1 = length(tx1h);
% bin0 = dec2bin(128); % '10000000'
% if(t1==1)
% bin0(end-l1+1:end) = tx1h;
% bin0(1)=0;
% bin0(2)=1;
%
% elseif(t1==2)
% bin0(end-l1+1:end) = tx1h;
% end
% x1h = bin2dec(tx1h);
%
% if(length(y1hex)>2)
% y1h = y1hex(1:end-2);
% y1l = y1hex(end-1:end);
% else
% y1h = dec2hex(0);
% y1l = y1hex;
% end
% fwrite(fp1, x1h, 'uint8');
% fwrite(fp1, hex2dec(x1l), 'uint8');
% fwrite(fp1, hex2dec(y1h), 'uint8');
% fwrite(fp1, hex2dec(y1l), 'uint8');
end
The way I read it
for i=1:mt.nm % nm points.
mred(i,6) = fread(fp1, 1, 'uint8'); % Raw X coordinates.
mred(i,7) = fread(fp1, 1, 'uint8'); % upper 2 bits are reserved info.
tmpx = [dec2bin(mred(i,6)), dec2bin(mred(i,7))];
if(length(tmpx)==16)
mred(i,4) = bin2dec(tmpx(1:2)); % Real Mark.
mred(i,1) = bin2dec(tmpx(3:end)); % Real X.
elseif(length(tmpx)==15)
mred(i,4) = bin2dec(tmpx(1)); % Real Type.
mred(i,1) = bin2dec(tmpx(2:end)); % Real X.
else
mred(i,4) = bin2dec(tmpx(1:2)); % Type unknown.
mred(i,1) = bin2dec(tmpx(3:end)); % Real X.
end
mred(i,8) = fread(fp1, 1, 'uint8'); % Y coordinates.
mred(i,9) = fread(fp1, 1, 'uint8'); % upper 2 bits are reserved.
tmpy = [dec2bin(mred(i,8)), dec2bin(mred(i,9))];
if(length(tmpy)==16)
mred(i,10) = bin2dec(tmpy(1:2)); % Real reserved.
mred(i,2) = bin2dec(tmpy(3:end)); % Real Y.
elseif(length(tmpy)==15)
mred(i,10) = bin2dec(tmpy(1)); % Real reserved.
mred(i,2) = bin2dec(tmpy(2:end)); % Real Y.
else
mred(i,10) = -1; % Reserved unknown.
mred(i,2) = bin2dec(tmpy); % Real Y.
end
end
The read() function works well for a given software which is implemented via C++. The software generates coordinates series in such a format. Then, I prepare a read() to get the information in the binary file generated by C++ software. Then, I want to implement the write() with Matlab in that format, but the read() fails to obtain what I had written to the binary file. Anybody help? Thanks.

Answer 1

The problem is probably (at least in part) an endian issue.
On intel architecture (little endian), reading two bytes as uint8 followed by appending the binary representations is not going to give you the same result in general as reading two bytes as uint16.
The following script illustrates how swapping the order of the two uint8's will do the trick.
I changed some variable names and made things more concise for readability.
% declare an input int and the header (in binary string rep)
v = 68;
t2 = '01';
% write to file
x1 = dec2bin(v);
lenx = 16-length(x1);
x1hl = strcat(t2, '00000000000000'); % High and low
x1a = strcat(x1hl(1:lenx), num2str(x1));
% x1a = 0100000001000100
fp1=fopen('temp','w+');
fwrite(fp1, bin2dec(x1a), 'uint16');
% read the file
frewind(fp1);
vtest = fread(fp1, 1, 'uint16'); % upper 2 bits are reserved info.
dec2bin(vtest)
% ans = 100000001000100
% now read *two* bytes as two uint8
frewind(fp1);
byte1 = fread(fp1, 1, 'uint8'); % Raw X coordinates.
byte2 = fread(fp1, 1, 'uint8'); % upper 2 bits are reserved info.
tmpx = [dec2bin(byte2) dec2bin(byte1)]; % <-- swap the order
% tmpx = 10000001000100
Alternately just read the whole 2 bytes as uint16 and work from there.