That is not very well designed code: the regexp is overkill, and the str2num with cellfun is likely not very efficient (as well as hiding evil eval inside...).
The code below imports the header into a structure hdr and the numeric data into matrix out (with class single to save memory):
str = fileread('zero_mac_cluster1.asc');
str = strrep(str,',','.');
% Parse Header:
opt = {'MultipleDelimsAsOne',true};
hdr = textscan(str,'%s%f',6,opt{:});
hdr = cell2struct(num2cell(hdr{2}),hdr{1});
% Parse Matrix:
opt = {'HeaderLines',numel(fieldnames(hdr)), 'CollectOutput',true};
fmt = repmat('%f32',1,hdr.ncols);
out = textscan(str,fmt,opt{:});
out = out{1};
out(out==hdr.NODATA_value) = NaN;
And checking:
>> hdr
hdr =
ncols: 10139
nrows: 7013
xllcorner: 1557267.2310345
yllcorner: 4850030.77
cellsize: 10
NODATA_value: -9999
>> size(out) % matches the file rows, which is NOT what the header states!
ans =
6065 10139
>> out(~isnan(out))
ans =
10
1.25
21.75
24.25
22.25
1.25
6.25
24.25
0
38.25
7.5
1.75
1
1.25
0.25
2.25
7
0.25
36.75
0.5
0
20.5
2.75
22.25
4.5
1.5
3.25
1.75
0.25
0
1.25
2.5
30.25
27.25
2.75
0
1
0.3846154
0.75
1
3.75
12.77778
28.75
12
0
3.181818
1.5
0
5.5
12
66.92308
0
0
1.75
0.25
0
5
45.75
9.75
0
20.5
24.25
0
19.75
2.25
0
1
23.5
0
0
0.25
1
0
12
21.38889
1.75
0.2777778
34.5
8
0.5
0.25
16.75
4
0
-99
2.25
2.5
45.25
0.25
0
5.5
2.25
3.75
3.75
6.75
0
1.75
Note that the memory requirements are not insignificant:
>> whos out
Name Size Bytes Class Attributes
out 6065x10139 245972140 single
