[MRG] fix two bugs in gather --output-unassigned (#1156) (#1160)

* add test for no unassigned hashes to output with gather

* add test for gather --output-unassigned with protein query

* fix two bugs in gather --output-unassigned

* Update tests/test_sourmash.py

Co-authored-by: Olga Botvinnik <[email protected]>

Co-authored-by: Olga Botvinnik <[email protected]>

Co-authored-by: Olga Botvinnik <[email protected]>

sourmash/commands.py

@@ -694,22 +694,13 @@ def gather(args):
             sig.save_signatures([ r.match for r in found ], fp)
 
     if args.output_unassigned:
-        if not len(query.minhash):
-            notify('no unassigned hashes! not saving.')
+        if not len(next_query.minhash):
+            notify('no unassigned hashes to save with --output-unassigned!')
         else:
             notify('saving unassigned hashes to "{}"', args.output_unassigned)
 
-            with_abundance = next_query.minhash.track_abundance
-            e = MinHash(ksize=query.minhash.ksize, n=0, max_hash=new_max_hash,
-                        track_abundance=with_abundance)
-            if with_abundance:
-                abunds = next_query.minhash.hashes
-                e.set_abundances(abunds)
-            else:
-                e.add_many(next_query.minhash.hashes)
-
             with FileOutput(args.output_unassigned, 'wt') as fp:
-                sig.save_signatures([ sig.SourmashSignature(e) ], fp)
+                sig.save_signatures([ next_query ], fp)
 
 
 def multigather(args):

tests/test_sourmash.py

@@ -3216,6 +3216,35 @@ def test_gather_metagenome_output_unassigned():
                     'NC_011294.1' in out))
 
 
+def test_gather_metagenome_output_unassigned_none():
+    # test what happens when there's nothing unassigned to output
+    with utils.TempDirectory() as location:
+        testdata_glob = utils.get_test_data('gather/GCF_*.sig')
+        testdata_sigs = glob.glob(testdata_glob)
+
+        query_sig = utils.get_test_data('gather/combined.sig')
+
+        cmd = 'gather {} {} -k 21'.format(query_sig, " ".join(testdata_sigs))
+        cmd += ' --output-unassigned=unassigned.sig'
+        cmd += ' --threshold=0'
+        status, out, err = utils.runscript('sourmash', cmd.split(' '),
+                                           in_directory=location)
+
+        print(out)
+        print(err)
+
+        assert 'found 12 matches total' in out
+        assert 'the recovered matches hit 100.0% of the query' in out
+        assert all(('4.9 Mbp       33.2%  100.0%' in out,
+                    'NC_003198.1 Salmonella enterica subsp...' in out))
+        assert all(('4.5 Mbp        0.1%    0.4%' in out,
+                    'NC_004631.1 Salmonella enterica subsp...' in out))
+
+        # now examine unassigned
+        assert not os.path.exists(os.path.join(location, 'unassigned.sig'))
+        assert 'no unassigned hashes to save with --output-unassigned!' in err
+
+
 @utils.in_tempdir
 def test_gather_metagenome_output_unassigned_nomatches(c):
     # test --output-unassigned when there are no matches
@@ -3234,6 +3263,28 @@ def test_gather_metagenome_output_unassigned_nomatches(c):
     assert x.minhash == y.minhash
 
 
+@utils.in_tempdir
+def test_gather_metagenome_output_unassigned_nomatches_protein(c):
+    # test --output-unassigned with protein signatures
+    query_sig = utils.get_test_data('prot/protein/GCA_001593925.1_ASM159392v1_protein.faa.gz.sig')
+    against_sig = utils.get_test_data('prot/protein/GCA_001593935.1_ASM159393v1_protein.faa.gz.sig')
+
+    c.run_sourmash('gather', query_sig, against_sig,
+                   '--output-unassigned', 'foo.sig')
+
+    print(c.last_result.out)
+    assert 'found 0 matches total;' in c.last_result.out
+
+    c.run_sourmash('sig', 'describe', c.output('foo.sig'))
+    print(c.last_result.out)
+
+    x = sourmash.load_one_signature(query_sig, ksize=57)
+    y = sourmash.load_one_signature(c.output('foo.sig'))
+
+    assert x.minhash == y.minhash
+    assert y.minhash.moltype == "protein"
+
+
 def test_gather_metagenome_downsample():
     # downsample w/scaled of 100,000
     with utils.TempDirectory() as location: