h5save,h5size: handle scalar, vector more accurately

Author: scivision

+stdlib/+hdf5nc/h5save.m

@@ -20,26 +20,13 @@ function h5save(filename, varname, A, opts)
 
 filename = expanduser(filename);
 
-if isempty(opts.size)
-  if isvector(A)
-    sizeA = length(A);
-  else
-    sizeA = size(A);
-  end
-else
-  sizeA = opts.size;
-end
 % coerce if needed
 A = coerce_ds(A, opts.type);
-if ischar(A)
-  A = string(A);
-  sizeA = size(A);
-end
 
 if isfile(filename) && h5exists(filename, varname)
-  exist_file(filename, varname, A, sizeA)
+  exist_file(filename, varname, A, opts.size)
 else
-  new_file(filename, varname, A, sizeA)
+  new_file(filename, varname, A, opts.size)
 end
 
 end % function
@@ -57,6 +44,10 @@ function exist_file(filename, varname, A, sizeA)
   start = 1;
 end
 
+if isempty(sizeA)
+  sizeA = defaultSize(A);
+end
+
 if isscalar(A)
   h5write(filename, varname, A)
 elseif all(diskshape == sizeA)
@@ -71,24 +62,32 @@ function exist_file(filename, varname, A, sizeA)
 
 
 function new_file(filename, varname, A, sizeA)
-import stdlib.hdf5nc.auto_chunk_size
+
 
 folder = fileparts(filename);
 if strlength(folder) > 0 && ~isfolder(folder)
   error('hdf5nc:h5save:fileNotFound', '%s is not a folder, cannot create %s', folder, filename)
 end
 
-if isscalar(A) && ~isstring(A)
-  h5_write_scalar(filename, varname, A)
-elseif isvector(A)
-  h5create(filename, varname, sizeA, 'DataType', class(A))
+if isempty(sizeA)
+  if isscalar(A) && ~isstring(A)
+    h5_write_scalar(filename, varname, A)
+  elseif isvector(A)
+    h5create(filename, varname, length(A), 'DataType', class(A))
+  else
+    create_compress(filename, varname, A, size(A))
+  end
 else
-  % enable Gzip compression--remember Matlab's dim order is flipped from
-  % C / Python
-  h5create(filename, varname, sizeA, 'DataType', class(A), ...
-    'Deflate', 1, 'Fletcher32', true, 'Shuffle', true, ...
-    'ChunkSize', auto_chunk_size(sizeA))
-end % if
+  if isscalar(sizeA)
+    if sizeA == 0 && ~isstring(A)
+      h5_write_scalar(filename, varname, A)
+    else
+      h5create(filename, varname, sizeA, 'DataType', class(A))
+    end
+  else
+    create_compress(filename, varname, A, sizeA)
+  end
+end
 
 h5write(filename, varname, A)
 

+stdlib/+hdf5nc/h5size.m

@@ -18,7 +18,7 @@
 
 dsi = h5info(file, variable).Dataspace;
 if dsi.Type == "scalar"
-  fsize = 1;
+  fsize = [];
 else
   fsize = dsi.Size;
 end

+stdlib/+hdf5nc/private/coerce_ds.m

@@ -2,14 +2,17 @@
 % used by h5save and ncsave
 arguments
   A
-  dtype string
+  dtype string {mustBeScalarOrEmpty}
 end
 
-if isempty(dtype)
+if ischar(A)
+  A = string(A);
   return
 end
 
-assert(isscalar(dtype), "dtype must be scalar")
+if isempty(dtype)
+  return
+end
 
 switch dtype
   case ""

+stdlib/+hdf5nc/private/create_compress.m

@@ -0,0 +1,11 @@
+function create_compress(filename, varname, A, sizeA)
+%% enable Gzip compression
+% remember Matlab's dim order is flipped from C / Python
+
+import stdlib.hdf5nc.auto_chunk_size
+
+h5create(filename, varname, sizeA, 'DataType', class(A), ...
+  'Deflate', 1, 'Fletcher32', true, 'Shuffle', true, ...
+  'ChunkSize', auto_chunk_size(sizeA))
+
+end

+stdlib/+hdf5nc/private/defaultSize.m

@@ -0,0 +1,9 @@
+function s = defaultSize(A)
+
+if isvector(A)
+  s = length(A);
+else
+  s = size(A);
+end
+
+end

+stdlib/TestHDF5.m

@@ -116,8 +116,7 @@ function test_size(tc)
 basic = tc.TestData.basic;
 
 s = h5size(basic, '/A0');
-tc.verifyTrue(isscalar(s))
-tc.verifyEqual(s, 1)
+tc.verifyEmpty(s)
 
 s = h5size(basic, '/A1');
 tc.verifyTrue(isscalar(s))
@@ -162,6 +161,22 @@ function test_read(tc)
 end
 
 
+function test_shape(tc)
+import stdlib.hdf5nc.h5save
+import stdlib.hdf5nc.h5size
+basic = tc.TestData.basic;
+
+h5save(basic, "/vector1", 34, "size", 1)
+s = h5size(basic, '/vector1');
+tc.verifyEqual(s, 1);
+
+h5save(basic, "/scalar", 34, "size", 0)
+s = h5size(basic, '/scalar');
+tc.verifyEmpty(s);
+
+end
+
+
 function test_coerce(tc, type)
 import stdlib.hdf5nc.h5save
 import matlab.unittest.constraints.IsFile

CITATION.cff

@@ -5,4 +5,3 @@ authors:
     orcid: https://orcid.org/0000-0002-1637-6526
 title: Matlab-stdlib
 doi: 10.5281/zenodo.3964540
-date-released: 2021-11-17

Readme.md

@@ -62,6 +62,10 @@ Get the dataset size (shape)
 h5size(filename, dataset_name)
 ```
 
+A scalar dataset will return empty `[]`.
+
+---
+
 Get the names of all datasets in a file
 
 ```matlab

codemeta.json

@@ -4,7 +4,7 @@
     "license": "https://spdx.org/licenses/MIT",
     "codeRepository": "https://github.com/geospace-code/matlab-stdlib",
     "contIntegration": "https://github.com/geospace-code/matlab-stdlib/actions",
-    "dateModified": "2021-11-17",
+    "dateModified": "2022-03-29",
     "downloadUrl": "https://github.com/geospace-code/matlab-stdlib/releases",
     "issueTracker": "https://github.com/geospace-code/matlab-stdlib/issues",
     "name": "matlab-stdlib",