IMP: Adds warnings when importing potentially problematic TSVTaxonomyFormat files. (#379)

Author: Macabe222

q2_types/feature_data/_deferred_setup/_validators.py

@@ -6,14 +6,17 @@
 # The full license is in the file LICENSE, distributed with this software.
 # ----------------------------------------------------------------------------
 
+import warnings
 import pandas as pd
 
 from qiime2.plugin import Properties, ValidationError
 
-from .. import FeatureData, SequenceCharacteristics
+from .. import FeatureData, SequenceCharacteristics, Taxonomy
 
 from ...plugin_setup import plugin
 
+from qiime2.core.exceptions import RachisWarning
+
 
 @plugin.register_validator(FeatureData[SequenceCharacteristics %
                                        Properties("length")])
@@ -37,3 +40,34 @@ def validate_seq_char_len(data: pd.DataFrame, level):
     if not (data['length'] > 0).all():
         raise ValidationError('Column "length" cannot contain negative '
                               'values.')
+
+
+@plugin.register_validator(FeatureData[Taxonomy])
+def _check_single_taxon(data: pd.DataFrame, level):
+    '''
+    Raises a warning if the taxonomy's depth is one.
+    '''
+    max_depth = 0
+    for taxon in data['Taxon']:
+        if taxon.count(';') > max_depth:
+            max_depth = taxon.count(';')
+
+    if max_depth == 0:
+        warnings.warn(
+            'Importing taxonomy with taxonomic depth of one.',
+            RachisWarning
+        )
+
+
+@plugin.register_validator(FeatureData[Taxonomy])
+def _check_trailing_semicolon(data: pd.DataFrame, level):
+    '''
+    Raises a warning if a taxon string in the taxonomy has a trailing
+    semicolon.
+    '''
+    for taxon in data['Taxon']:
+        if taxon.rstrip().endswith(';'):
+            warnings.warn(
+                'Importing taxonomy with a trailing semicolon.',
+                RachisWarning
+            )

q2_types/feature_data/tests/data/taxonomy/one-depth.tsv

@@ -0,0 +1,3 @@
+Feature ID	Taxon
+seq1;	k__Bacteria
+seq2;	k__Archaea

q2_types/feature_data/tests/data/taxonomy/trailing-semicolon.tsv

@@ -0,0 +1,4 @@
+Feature ID	Taxon	Confidence
+seq1	k__Bacteria	0.3
+seq2;	k__Bacteria; p__Proteobacteria	0.5
+seq3	k__Bacteria; p__Proteobacteria;	0.5

q2_types/feature_data/tests/test_formats.py

@@ -35,6 +35,8 @@
 )
 from qiime2.plugin.testing import TestPluginBase
 from qiime2.plugin import ValidationError
+from qiime2.core.exceptions import RachisWarning
+from qiime2.sdk.result import Artifact
 
 
 class TestTaxonomyFormats(TestPluginBase):
@@ -48,7 +50,6 @@ def test_taxonomy_format_validate_positive(self):
 
         for filepath in filepaths:
             format = TaxonomyFormat(filepath, mode='r')
-
             format.validate()
 
     def test_taxonomy_format_validate_negative(self):
@@ -116,6 +117,31 @@ def test_headerless_tsv_taxonomy_directory_format(self):
 
         format.validate()
 
+    def test_tsv_taxonomy_format_warnings(self):
+        '''
+        Tests that the taxonomic depth of one and trailing semicolon warnings
+        are raised under the proper conditions.
+        '''
+        filepath = self.get_data_path(
+            os.path.join('taxonomy', 'one-depth.tsv')
+        )
+        format = TSVTaxonomyFormat(filepath, mode='r')
+
+        with self.assertWarnsRegex(
+            RachisWarning, 'Importing taxonomy with taxonomic depth of one.'
+        ):
+            Artifact.import_data('FeatureData[Taxonomy]', format)
+
+        filepath = self.get_data_path(
+            os.path.join('taxonomy', 'trailing-semicolon.tsv')
+        )
+        format = TSVTaxonomyFormat(filepath, mode='r')
+
+        with self.assertWarnsRegex(
+            RachisWarning, 'Importing taxonomy with a trailing semicolon.'
+        ):
+            Artifact.import_data('FeatureData[Taxonomy]', format)
+
     def test_tsv_taxonomy_format_validate_positive(self):
         filenames = ['2-column.tsv', '3-column.tsv', 'valid-but-messy.tsv',
                      'many-rows.tsv']