correctly load 2D char arrays (#62)

* small change to correctly load in 2D char arrays

* python 3.7 has been removed from runners

see actions/runner-images#10893

* Update mat73/core.py

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

* Update test_mat73.py

* Update create_mat.m

* Update core.py

* fix string array reading

* remove dependency on StringDType numpy>=2

---------

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Simon Kern <14980558+skjerns@users.noreply.github.com>

Author: 3 people

mat73/core.py

@@ -266,9 +266,31 @@ def convert_mat(self, dataset, depth, MATLAB_class=None):
                 return None
 
         elif mtype=='char':
-            string_array = np.ravel(dataset)
-            string_array = ''.join([chr(x) for x in string_array])
-            return string_array
+            codes = np.asarray(dataset, dtype=np.uint16)
+
+            # object dtype → keeps '\x00'
+            # see https://github.com/numpy/numpy/issues/28964
+            to_char = np.vectorize(chr, otypes=[object])
+            arr = to_char(codes)
+
+            char_axis = 0 if arr.ndim < 3 else -2
+            char_arr = np.apply_along_axis(lambda x: ''.join(x), axis=char_axis, arr=arr)
+
+            string_list = char_arr.tolist()
+
+            if arr.ndim==2 and arr.shape[1]==1:
+                string_list = string_list[0]
+
+            if arr.ndim>2:
+                # print warning to be sure. I haven't encountered any char
+                # arrays with ndim>2 in the wild yet so can't be sure that
+                # they are actually the way I synthesized them
+                logging.warning(f"Loading char array '{dataset.name}' with {arr.ndim} dimensions "
+                                f"might be wrong stacked (i.e. dimensions scrambled). "
+                                f"please check variable is correct and report errors "
+                                f"on github.com/skjerns/mat7.3")
+
+            return string_list
 
         elif mtype=='bool':
             return bool(dataset)
@@ -361,8 +383,5 @@ def savemat(filename, verbose=True):
 
 if __name__=='__main__':
     # for testing / debugging
-    d = loadmat('../tests/testfile11.mat', only_include='foo')
-
-
-    # file = '../tests/testfile8.mat'
-    # data = loadmat(file)
+    d = loadmat('../tests/testfile16.mat')
+    print(d)

tests/create_mat.m

@@ -65,3 +65,20 @@
 clear all
 
 
+%% file to test 2D char arrays
+char_arr_1d = ['abcd']
+char_arr_2d = [ ...
+    'PSTH tensor for image sequences (averaged across frames):'; ...
+    'dimension 1: 2 scales (zoom1x, zoom2x)                   '; ...
+    'dimension 2: 3 category (natural, synthetic, contrast)   '; ...
+    'dimension 3: 10 movies                                   '; ...
+    'dimension 4: sorted units                                '; ...
+    'dimension 5: PSTH time bins                              '];
+
+char_arr_3d = cat(3, ...
+    ['abcd'; 'defg'], ...  % First "page"
+    ['ghij'; 'jklm'], ...  % Second "page"
+    ['mnöp'; 'pqrs'])     % Third "page"
+
+save('testfile16.mat','char_arr_1d', 'char_arr_2d', 'char_arr_3d', '-v7.3')
+clear all

tests/test_mat73.py

@@ -15,7 +15,7 @@
     version = pkg_resources.get_distribution('mat73').version
 except:
     version = '0.00'
-    
+
 try:
     repo = Repository('.')
     head = repo.head
@@ -35,7 +35,7 @@ class Testing(unittest.TestCase):
 
     def setUp(self):
         """make links to test files and make sure they are present"""
-        for i in range(1, 16):
+        for i in range(1, 17):
             file = 'testfile{}.mat'.format(i)
             if not os.path.exists(file):
                 file = os.path.join('./tests', file)
@@ -390,7 +390,7 @@ def test_load_specific_vars(self):
         elapsed2 = time.time()-start
         assert elapsed2<elapsed1, 'loading specific var was not faster'
 
-    def test_file10_nullchars(self):
+    def test_file8_nullchars(self):
         """test if null chars are retained in char arrays"""
         data = mat73.loadmat(self.testfile8)
         self.assertEqual(len(data['char_array']), 7, 'not all elements loaded')
@@ -503,12 +503,27 @@ def test_file15_strip(self):
                     'x_1_1_10_1_1': (1, 1, 10),
                     'x_10_1_1_10': (10, 1, 1, 10),
                     }
-        
-        
+
+
         for var, shape in expected.items():
             self.assertEqual(data[var].shape, shape)
             self.assertEqual(data[var].ndim, len(shape))
 
+    def test_file16_2d_char_array(self):
+        """Test loading of 2D char array"""
+        # the matlab array has shape (6 X 57)
+
+        data = mat73.loadmat(self.testfile16)
+
+        expected = ['PSTH tensor for image sequences (averaged across frames):',
+                    'dimension 1: 2 scales (zoom1x, zoom2x)                   ',
+                    'dimension 2: 3 category (natural, synthetic, contrast)   ',
+                    'dimension 3: 10 movies                                   ',
+                    'dimension 4: sorted units                                ',
+                    'dimension 5: PSTH time bins                              ']
+        self.assertEqual(data['char_arr_1d'], 'abcd')
+        self.assertEqual(data['char_arr_2d'], expected)
+        self.assertEqual(data['char_arr_3d'], [['abcd', 'defg'], ['ghij', 'jklm'], ['mnöp', 'pqrs']])
 
 if __name__ == '__main__':