h5 file import

I used a code to read a h5 file in matlab. However, there are 2000 different dataset names in the h5 file. To read one of the data, I have to use the code once:

double(h5read(tracename,'/Waveforms/Channel 1/Channel 1 Seg9995Data')

'Channel 1 Seg9995Data' is the name of one of the data. So I have to use this code 2000 times to read the whole data in the h5 file, which is very time consuming. Is there a simple way to do this? Thank you.

Answer 1

[Walter Roberson]

info = h5info(tracename);
trace_datasets = info.Datasets;
dataset_names = {trace_datasets.Name};

for K = 1 : length(dataset_names)
  double(h5read(tracename,dataset_names{K}))
end

[jUNSONG PENGT]

Dear Walter, Thank you for your answer. In the sentence''dataset_names = {trace_datasets.Name};'' What does ''Name'' correspond to? How can i find it? Thank you.

[per isakson]

In Walters answer, try to replace

info = h5info(tracename);

by

info = h5info(tracename,'/Waveforms/Channel 1');

[Walter Roberson]

If I read the documentation on h5info correctly, 'Name' is a field name in the Datasets field of the result of h5info() . {trace_datasets.Name} extracts all of those fields and puts them together in one cell array of strings.

I could be wrong about exactly what h5info returns; the documentation only implies the results rather than stating directly.

[per isakson]

That depends on the organization of the h5-file. It could be deeply nested, which gives a deeply nested structure, info. I guess, "/Waveforms/Channel 1" indicates a group named, "Channel 1", under a group named, "Waveforms". Furthermore, I guess, the datasets are in the group "Channel 1". (Too much guessing!)

Here is an example with a file I used for testing.

>> info0 = h5info('c:\m\PiaX\xtests\h5test.h5')
info0 = 
      Filename: 'C:\m\PiaX\xtests\h5test.h5'
          Name: '/'
        Groups: [2x1 struct]
      Datasets: []
     Datatypes: []
         Links: []
    Attributes: []
>>
>> {info0.Datasets.Name}
Attempt to reference field of non-structure array.
>>
>> {info0.Groups.Name}
ans = 
    '/group01_L1'    '/group02_L1'
>>
>> 
>> info1 = h5info('c:\m\PiaX\xtests\h5test.h5','/group01_L1')
info1 = 
      Filename: 'C:\m\PiaX\xtests\h5test.h5'
          Name: '/group01_L1'
        Groups: []
      Datasets: [2x1 struct]
     Datatypes: []
         Links: []
    Attributes: []
>> 
>> {info1.Datasets.Name}
ans = 
    'data11_L2'    'data12_L2'

[per isakson]

@jUNSONG, What does

info = h5info(tracename)

display in your command window?

Answer 2

[Lennart M]

Thanks to this thread, I managed to write a nice import function for my h5 files in MATLAB. I'll post it here for reference, maybe someone can use it. All files matching a file name pattern within a directory are imported and their data is merged into one large MATLAB struct. You may need to modify the code to fit your h5 file structure.

function h5data = loadh5bypattern(folder, filenamePattern)

% scan folder for matching files
filelist = dir(fullfile(folder,filenamePattern));

h5data = struct;

% parse all matching files...
for k = 1:numel(filelist)
    file = filelist(k);
    if ~file.isdir
        filename = filelist(k).name;

          fprintf(filename);

          info = h5info(filename);
          groups = info.Groups;
          nGroups = length(groups);
          for i=1:nGroups
              groupname = groups(i).Name;
              cleanname = regexprep(groupname,'[/]','');

              field = h5read(filename,strcat(groupname,'/data'));

              % the gps position group has some strange fields, remove them
              if strcmp(cleanname, 'gps_position')
                  field = rmfield(field,'altitude_units');
                  field = rmfield(field,'geo_sep_units');
                  field = rmfield(field,'lat_dir');
                  field = rmfield(field,'lon_dir');
              end

              % this field is not wanted either
              if isfield(field,'unit')
                  field = rmfield(field,'unit');
              end

              % first time: create new entry, after that append
              if ~isfield(h5data, cleanname)
                  h5data.(cleanname) = field;
              else
                  existingfields = fieldnames(h5data.(cleanname));
                  for j=1:length(existingfields)
                      entryname = existingfields{j};
                      h5data.(cleanname).(entryname) = ...
                          [h5data.(cleanname).(entryname); field.(entryname)];
                  end
              end

              % progress display
              if mod(i, floor(nGroups/15)) == 0
                  fprintf('.');
              end

          end
          fprintf('\n');
      end
  end

disp('Import of .h5 files succcessful! Storing as .mat file...');

% after successful import, store as .mat file
storename = regexprep(filenamePattern,'[*(.h5)]','');
save(storename,'h5data');

end

[Bobo]

@ LM. How can I use your code? Can you give a brief description? I mean what is folder (or do you mean folder name?). What is filenamePattern ?

[Walter Roberson]

"All files matching a file name pattern within a directory are imported"

"filelist = dir(fullfile(folder,filenamePattern));"

So in the first input you should be passing a directory name, as a character vector, and in the second input you should be passing a pattern accepted by your file system, such 'amoeba_*.h5', indicating that all files in the given directory that match the given pattern should be imported.

[Shakir Hussain]

HI Lennart

Though this post is three years old but it is useful for me now.

How I can you your code to import, read and merge in single file of matlab a bunch of MODIS snow cover MOD10C2 HDF-EOS https://modis.gsfc.nasa.gov/data/dataprod/mod10.php ?