modify handling of taxonomy

backend/scripts/populate_db/import_igc_data.py

@@ -65,9 +65,7 @@ class ImportIGCGenes(object):
             if len(resp_dict['results']) > 1:
                 logger.warning(f"More than 1 result found for phylum {phylum}. First result is kept.")
         if resp_dict.get('count', 0) > 0:
-            gene_dict.update(
-                {'taxonomy': resp_dict['results'][0]['tax_id']}
-            )
+            gene_dict.update({'taxonomy': resp_dict['results'][0]['tax_id']})
         return gene_dict
 
     def _parse_gene(self, raw_line, selected_keys=SELECTED_KEYS):
@@ -83,8 +81,10 @@ class ImportIGCGenes(object):
         gene_dict['gene_name'] = gene_dict['gene_id']
         gene_dict['gene_id'] = slugify(gene_dict['gene_id'])
         gene_dict['functions'] = gene_dict.pop('kegg_ko')
-        if self.skip_tax:
+        if gene_dict.get('taxonomy', None) == 'unknown' or self.skip_tax:
             gene_dict.pop('taxonomy')
+        else:
+            gene_dict = self._select_taxonomy(gene_dict)
         if self.skip_functions or 'unknown' in gene_dict['functions']:
             gene_dict.pop('functions')
         return gene_dict
@@ -95,7 +95,7 @@ class ImportIGCGenes(object):
                 chunk_genes = list(islice(file, chunk_size))
                 if not chunk_genes:
                     break
-                genes = [self._clean_gene(self._select_taxonomy(self._parse_gene(i))) for i in chunk_genes]
+                genes = [self._clean_gene(self._parse_gene(i)) for i in chunk_genes]
                 try:
                     response = self.metagenedb_gene_api.put(genes)
                     self.created_genes += response.get('created').get('count')

backend/scripts/populate_db/test_import_igc_data.py

@@ -72,6 +72,7 @@ class TestCleanGene(TestCase):
 
     def setUp(self):
         self.import_igc_genes = ImportIGCGenes('test', 'test')
+        self.import_igc_genes._select_taxonomy = lambda x: x  # Mock to return same dict
         self.gene_dict = {
             'gene_id': 'gene.01',
             'length': 135,