IMP: add skbio.DistanceMatrix validation for LSMatFormat

q2_types/distance_matrix/_deferred_setup/_transformers.py

@@ -56,3 +56,24 @@ def _4(data: pd.Series) -> LSMatFormat:
     with ff.open() as fh:
         dm.write(fh, format='lsmat')
     return ff
+
+
+@plugin.register_transformer
+def _5(ff: LSMatFormat) -> pd.DataFrame:
+    dm = skbio.DistanceMatrix.read(str(ff), format='lsmat', verify=False)
+    return dm.to_data_frame()
+
+
+@plugin.register_transformer
+def _6(data: pd.DataFrame) -> LSMatFormat:
+    # Raise if index != columns because columns will be ignored later.
+    if not data.index.equals(data.columns):
+        raise ValueError(
+            "Distance matrix DataFrame index and columns must contain "
+            "the same IDs in the same order."
+        )
+    dm = skbio.DistanceMatrix(data, ids=data.index)
+    ff = LSMatFormat()
+    with ff.open() as fh:
+        dm.write(fh, format='lsmat')
+    return ff

q2_types/distance_matrix/_formats.py

@@ -6,14 +6,23 @@
 # The full license is in the file LICENSE, distributed with this software.
 # ----------------------------------------------------------------------------
 
-import skbio.io
+import skbio
+from skbio.io import FileFormatError
+from skbio.stats.distance import PairwiseMatrixError
+
 import qiime2.plugin.model as model
+from qiime2.plugin import ValidationError
 
 
 class LSMatFormat(model.TextFileFormat):
-    def sniff(self):
-        sniffer = skbio.io.io_registry.get_sniffer('lsmat')
-        return sniffer(str(self))[0]
+    """An lsmat distance matrix with non-empty square float data, unique IDs,
+    symmetric values, and a zero-valued diagonal.
+    """
+    def _validate_(self, level):
+        try:
+            skbio.DistanceMatrix.read(str(self), format='lsmat')
+        except (FileFormatError, PairwiseMatrixError, ValueError) as error:
+            raise ValidationError(error) from error
 
 
 DistanceMatrixDirectoryFormat = model.SingleFileDirectoryFormat(

q2_types/distance_matrix/tests/data/duplicate-ids-distance-matrix.tsv

@@ -0,0 +1,3 @@
+	s1	s1
+s1	0.0	42.0
+s1	42.0	0.0

q2_types/distance_matrix/tests/data/non-numeric-distance-matrix.tsv

@@ -0,0 +1,3 @@
+	s1	s2
+s1	0.0	invalid
+s2	42.0	0.0

q2_types/distance_matrix/tests/data/non-symmetrical-distance-matrix.tsv

@@ -0,0 +1,3 @@
+	s1	s2
+s1	0.0	42.0
+s2	41.0	0.0

q2_types/distance_matrix/tests/data/nonzero-diagonal-distance-matrix.tsv

@@ -0,0 +1,3 @@
+	s1	s2
+s1	1.0	42.0
+s2	42.0	0.0

q2_types/distance_matrix/tests/test_formats.py

@@ -13,6 +13,8 @@
 from q2_types.distance_matrix import LSMatFormat, DistanceMatrixDirectoryFormat
 from qiime2.plugin.testing import TestPluginBase
 from qiime2.plugin import ValidationError
+from skbio.io import FileFormatError
+from skbio.stats.distance import DistanceMatrixError, PairwiseMatrixError
 
 
 class TestFormats(TestPluginBase):
@@ -32,9 +34,53 @@ def test_lsmat_format_validate_negative(self):
         filepath = self.get_data_path('not-lsmat')
         format = LSMatFormat(filepath, mode='r')
 
-        with self.assertRaisesRegex(ValidationError, 'LSMat'):
+        with self.assertRaises(ValidationError) as error:
             format.validate()
 
+        self.assertIsInstance(error.exception.__cause__.__cause__,
+                              FileFormatError)
+
+    def test_lsmat_format_validate_rejects_non_symmetrical(self):
+        filepath = self.get_data_path('non-symmetrical-distance-matrix.tsv')
+        format = LSMatFormat(filepath, mode='r')
+
+        with self.assertRaisesRegex(ValidationError, 'symmetric') as error:
+            format.validate()
+
+        self.assertIsInstance(error.exception.__cause__.__cause__,
+                              DistanceMatrixError)
+
+    def test_lsmat_format_validate_rejects_nonzero_diagonal(self):
+        filepath = self.get_data_path('nonzero-diagonal-distance-matrix.tsv')
+        format = LSMatFormat(filepath, mode='r')
+
+        with self.assertRaisesRegex(ValidationError, 'hollow') as error:
+            format.validate()
+
+        self.assertIsInstance(error.exception.__cause__.__cause__,
+                              DistanceMatrixError)
+
+    def test_lsmat_format_validate_rejects_non_numeric_value(self):
+        filepath = self.get_data_path('non-numeric-distance-matrix.tsv')
+        format = LSMatFormat(filepath, mode='r')
+
+        with self.assertRaisesRegex(ValidationError,
+                                    'could not be read') as error:
+            format.validate()
+
+        self.assertIsInstance(error.exception.__cause__.__cause__, ValueError)
+
+    def test_lsmat_format_validate_rejects_duplicate_ids(self):
+        filepath = self.get_data_path('duplicate-ids-distance-matrix.tsv')
+        format = LSMatFormat(filepath, mode='r')
+
+        with self.assertRaisesRegex(ValidationError,
+                                    'IDs must be unique') as error:
+            format.validate()
+
+        self.assertIsInstance(error.exception.__cause__.__cause__,
+                              PairwiseMatrixError)
+
     def test_distance_matrix_directory_format(self):
         # This test exists mainly to assert that the single-file directory
         # format is defined and functional. More extensive testing is performed

q2_types/distance_matrix/tests/test_transformers.py

@@ -71,6 +71,37 @@ def test_pd_series_to_skbio_distance_matrix(self):
 
             self.assertEqual(obs, exp)
 
+    def test_lsmat_format_to_pd_data_frame(self):
+        filenames = ('distance-matrix-1x1.tsv', 'distance-matrix-2x2.tsv',
+                     'distance-matrix-NxN.tsv')
+        for filename in filenames:
+            input, obs = self.transform_format(LSMatFormat,
+                                               pd.DataFrame,
+                                               filename=filename)
+            exp = skbio.DistanceMatrix.read(str(input)).to_data_frame()
+            pd.testing.assert_frame_equal(obs, exp)
+
+    def test_pd_data_frame_to_skbio_distance_matrix(self):
+        transformer = self.get_transformer(pd.DataFrame, LSMatFormat)
+
+        filenames = ('distance-matrix-1x1.tsv', 'distance-matrix-2x2.tsv',
+                     'distance-matrix-NxN.tsv')
+        for filename in filenames:
+            exp = skbio.DistanceMatrix.read(self.get_data_path(filename))
+            obs = transformer(exp.to_data_frame())
+            obs = skbio.DistanceMatrix.read(str(obs))
+            self.assertEqual(obs, exp)
+
+    def test_pd_data_frame_to_lsmat_format_rejects_misordered_columns(self):
+        transformer = self.get_transformer(pd.DataFrame, LSMatFormat)
+        dm = skbio.DistanceMatrix.read(
+            self.get_data_path('distance-matrix-NxN.tsv'))
+        data = dm.to_data_frame().reindex(columns=dm.ids[::-1])
+
+        with self.assertRaisesRegex(
+                ValueError, "same IDs in the same order"):
+            transformer(data)
+
 
 if __name__ == "__main__":
     unittest.main()