function m = loadbinary(file,precision,msize,lastdimrange,dim)
<file> is a pattern matching one or more files (see matchfiles.m)
<precision> is something like 'int16'
<msize> (optional) is the expected dimensions of the matrix for one file.
one of the dimensions can be 0, in which case we figure out what
that number should be. default: [1 0].
<lastdimrange> (optional) is
[A B] where A<=B and A and B are indices referring to the last dimension
of <m> (i.e. the last entry in <msize>). this indicates that we
want the portion of the matrix that lies between A to B (inclusive.)
-V where V is a vector of indices referring to the last dimension of <m>.
this indicates that we want exactly the indices specified by V. V can
have indices in any order and may include repeats.
we read and return only the portion of the matrix specified by
<lastdimrange>. default is [], which means return the whole matrix.
<dim> (optional) is the dimension along which to concatenate
matrices from different files. default: 1.
read <file> and return a matrix. we assume that each file contains
some data (i.e. it isn't empty). for the machine format, we assume
IEEE floating point with little-endian byte ordering ('l'; see fopen).
see also savebinary.m.
example:
savebinary('test','uint8',repmat(0:255,[2 1]));
isequal(loadbinary('test','uint8',[0 256],[255 256]),repmat([254 255],[2 1]))0001 function m = loadbinary(file,precision,msize,lastdimrange,dim) 0002 0003 % function m = loadbinary(file,precision,msize,lastdimrange,dim) 0004 % 0005 % <file> is a pattern matching one or more files (see matchfiles.m) 0006 % <precision> is something like 'int16' 0007 % <msize> (optional) is the expected dimensions of the matrix for one file. 0008 % one of the dimensions can be 0, in which case we figure out what 0009 % that number should be. default: [1 0]. 0010 % <lastdimrange> (optional) is 0011 % [A B] where A<=B and A and B are indices referring to the last dimension 0012 % of <m> (i.e. the last entry in <msize>). this indicates that we 0013 % want the portion of the matrix that lies between A to B (inclusive.) 0014 % -V where V is a vector of indices referring to the last dimension of <m>. 0015 % this indicates that we want exactly the indices specified by V. V can 0016 % have indices in any order and may include repeats. 0017 % we read and return only the portion of the matrix specified by 0018 % <lastdimrange>. default is [], which means return the whole matrix. 0019 % <dim> (optional) is the dimension along which to concatenate 0020 % matrices from different files. default: 1. 0021 % 0022 % read <file> and return a matrix. we assume that each file contains 0023 % some data (i.e. it isn't empty). for the machine format, we assume 0024 % IEEE floating point with little-endian byte ordering ('l'; see fopen). 0025 % 0026 % see also savebinary.m. 0027 % 0028 % example: 0029 % savebinary('test','uint8',repmat(0:255,[2 1])); 0030 % isequal(loadbinary('test','uint8',[0 256],[255 256]),repmat([254 255],[2 1])) 0031 0032 % constants 0033 machineformat = 'l'; 0034 0035 % input 0036 if ~exist('msize','var') || isempty(msize) 0037 msize = [1 0]; 0038 end 0039 if ~exist('lastdimrange','var') || isempty(lastdimrange) 0040 lastdimrange = []; 0041 end 0042 if ~exist('dim','var') || isempty(dim) 0043 dim = 1; 0044 end 0045 0046 % massage input 0047 if ~isempty(lastdimrange) 0048 if any(lastdimrange < 0) 0049 lastdimrange = -lastdimrange; 0050 else 0051 lastdimrange = lastdimrange(1):lastdimrange(2); 0052 end 0053 end 0054 0055 % get file name 0056 file = matchfiles(file); 0057 assert(length(file) >= 1,'<file> does not match at least one file'); 0058 0059 % loop through files 0060 for p=1:length(file) 0061 0062 % open file 0063 fid = fopen(file{p},'r',machineformat); 0064 assert(fid ~= -1,'<file> could not be opened for reading'); 0065 0066 % handle case of no specific range 0067 if isempty(lastdimrange) 0068 0069 % read it all in 0070 m0 = fread(fid,Inf,['*' precision],0,machineformat); 0071 0072 % figure out msize 0073 if any(msize==0) 0074 assert(sum(msize==0)==1); % make sure only one is 0 0075 msize(msize==0) = prod(size(m0))/prod(msize(msize~=0)); % calc appropriate value 0076 assert(all(isint(msize)),'<msize> is not correct'); 0077 end 0078 0079 % reshape 0080 m0 = reshape(m0,msize); 0081 0082 % handle case of specific range 0083 else 0084 0085 % peek to find out the byte size of a word 0086 temp = fread(fid,1,['*' precision],0,machineformat); 0087 wordsize = getfield(whos('temp'),'bytes'); 0088 0089 % check how big the data segment is 0090 assert(fseek(fid,0,'bof')==0); 0091 pos1 = ftell(fid); assert(pos1~=-1); 0092 assert(fseek(fid,0,'eof')==0); 0093 pos2 = ftell(fid); assert(pos2~=-1); 0094 0095 % calculate number of words 0096 numwords = (pos2-pos1)/wordsize; assert(isint(numwords)); 0097 0098 % figure out msize 0099 if any(msize==0) 0100 assert(sum(msize==0)==1); % make sure only one is 0 0101 msize(msize==0) = numwords/prod(msize(msize~=0)); % calc appropriate value 0102 assert(all(isint(msize)),'<msize> is not correct'); 0103 end 0104 0105 % calc slice size in terms of number of words 0106 slicesize = prod(msize(1:end-1)); 0107 0108 % ok, now we have to do fancy handling to deal with arbitrary vectors of indices 0109 0110 % process chunks of consecutive indices 0111 lastdimrange_sorted = sort(lastdimrange); 0112 cur = 1; % pos in sorted list that we are on currently 0113 m0 = cast([],precision); % initialize 0114 while cur <= length(lastdimrange_sorted) 0115 ff = find(diff(lastdimrange_sorted(cur:end))~=1); % ff(1) tells us how many consecutive integers we have 0116 if isempty(ff) % in this case, the entire list is consecutive integers 0117 rng = [lastdimrange_sorted(cur) lastdimrange_sorted(end)]; 0118 cur = cur + diff(rng)+1; 0119 else 0120 rng = [lastdimrange_sorted(cur) lastdimrange_sorted(cur)+ff(1)-1]; 0121 cur = cur + ff(1); 0122 end 0123 0124 % calc number of slices wanted 0125 numslices = rng(2)-rng(1)+1; 0126 0127 % read data and reshape 0128 assert(fseek(fid,slicesize*(rng(1)-1)*wordsize,'bof')==0); 0129 m0 = cat(length(msize),m0,reshape(fread(fid,numslices*slicesize,['*' precision],0,machineformat),[msize(1:end-1) numslices])); 0130 end 0131 0132 % now, return exactly what the user wanted 0133 m0 = subscript(m0,[repmat({':'},[1 length(msize)-1]) {calcposition(lastdimrange_sorted,lastdimrange)}]); 0134 0135 end 0136 0137 % close file 0138 assert(fclose(fid)==0); 0139 0140 % save 0141 if p==1 0142 m = m0; % get the datatype right instead of initializing to double via [] 0143 else 0144 m = cat(dim,m,m0); 0145 end 0146 0147 end