MRG: fix multigather output by adding md5sum along with -U/--output-add-query-md5sum

doc/command-line.md

@@ -347,6 +347,10 @@ metagenome and genome bin analysis.  (See
 [Classifying Signatures](classifying-signatures.md) for more
 information on the different approaches that can be used here.)
 
+`sourmash gather` takes exactly one query and one or more
+[collections of signatures](#storing-and-searching-signatures). Please see
+[`sourmash multigather`](#sourmash-multigather-do-gather-with-many-queries) if you have multiple queries!
+
 If the input signature was created with `-p abund`, output
 will be abundance weighted (unless `--ignore-abundances` is
 specified).  `-o/--output` will create a CSV file containing the
@@ -534,6 +538,47 @@ This combination of commands ensures that the more time- and
 memory-intensive `gather` step is run only on a small set of relevant
 signatures, rather than all the signatures in the database.
 
+### `sourmash multigather` - do gather with many queries
+
+The `multigather` subcommand runs `sourmash gather` on multiple
+queries.  (See
+[`sourmash gather` docs](#sourmash-gather-find-metagenome-members) for
+specifics on what gather does, and how!)
+
+Usage:
+```
+sourmash multigather --query <queries ...> --db <collections>
+```
+
+Note that multigather is single threaded, so it offers no substantial
+efficiency gains over just running gather multiple times!  Nontheless, it
+is useful for situations where you have many sketches organized in a
+combined file, e.g. sketches built with `sourmash sketch
+... --singleton`).
+
+#### `multigather` output files
+
+multigather produces three output files for each
+query:
+
+* `<output_base>.csv` - gather CSV output
+* `<output_base>.matches.sig` - all matching outputs
+* `<output_base>.unassigned.sig` - all remaining unassigned hashes
+
+As of sourmash v4.8.7, `<output_base>` is set as follows:
+* the filename attribute of the query sketch, if it is not empty or `-`;
+* the query sketch md5sum, if the query filename is empty or `-`;
+* the query filename + the query sketch md5sum
+  (`<query_file>.<md5sum>`), if `-U/--output-add-query-md5sum` is
+  specified;
+
+By default, `multigather` will complain and exit with an error if
+the same `<output_base>` is used repeatedly and an output file is
+going to be overwritten.  With `-U/--output-add-query-md5sum` this
+should only happen when identical sketches are present in a query
+database. Use `--force-allow-overwrite-output`
+to allow overwriting of output files without an error.
+
 ## `sourmash tax` subcommands for integrating taxonomic information into gather results
 
 The `sourmash tax` subcommands support taxonomic analysis of genomes

src/sourmash/cli/multigather.py

@@ -79,6 +79,11 @@ def subparser(subparsers):
         action="store_true",
         help="stop at databases that contain no compatible signatures",
     )
+    subparser.add_argument(
+        "--force-allow-overwrite-output",
+        action="store_true",
+        help="allow output files to be overwritten",
+    )
     subparser.add_argument(
         "--no-fail-on-empty-database",
         action="store_false",
@@ -92,6 +97,19 @@ def subparser(subparsers):
         "--outdir",
         help="output CSV results to this directory",
     )
+    subparser.add_argument(
+        "-U",
+        "--output-add-query-md5sum",
+        action="store_true",
+        help="add md5sum of each query to ensure unique output file names",
+    )
+    subparser.add_argument(
+        "-E",
+        "--extension",
+        type=str,
+        default=".sig",
+        help="write signature files with this extension ('.sig' by default)",
+    )
 
     add_ksize_arg(subparser)
     add_moltype_args(subparser)

src/sourmash/commands.py

@@ -1162,6 +1162,7 @@
     # run gather on all the queries.
     n = 0
     size_may_be_inaccurate = False
+    output_base_tracking = set()  # make sure we are not reusing 'output_base'
     for queryfile in inp_files:
         # load the query signature(s) & figure out all the things
         for query in sourmash_args.load_file_as_signatures(
@@ -1228,21 +1229,42 @@
             result = None
 
             query_filename = query.filename
-            if not query_filename:
+            if not query_filename or query_filename == "-":
                 # use md5sum if query.filename not properly set
-                query_filename = query.md5sum()
+                output_base = query.md5sum()
+            elif args.output_add_query_md5sum:
+                # Uniquify the output file if all signatures were made from the same file (e.g. with --singleton)
+                assert query_filename and query_filename != "-"  # first branch
+                output_base = os.path.basename(query_filename) + "." + query.md5sum()
+            else:
+                output_base = os.path.basename(query_filename)
 
-            output_base = os.path.basename(query_filename)
             if args.output_dir:
                 output_base = os.path.join(args.output_dir, output_base)
-            output_csv = output_base + ".csv"
+
+            # track overwrites of output files!
+            if output_base in output_base_tracking:
+                error(
+                    f"ERROR: detected overwritten outputs! '{output_base}' has already been used. Failing."
+                )
+                if args.force_allow_overwrite_output:
+                    error("continuing because --force-allow-overwrite was specified")
+                else:
+                    error(
+                        "Consider using '-U/--output-add-query-md5sum' to build unique outputs"
+                    )
+                    error("and/or '--force-allow-overwrite-output'")
+                    sys.exit(-1)
+
+            output_base_tracking.add(output_base)
 
             output_matches = output_base + ".matches.sig"
             save_sig_obj = SaveSignaturesToLocation(output_matches)
             save_sig = save_sig_obj.__enter__()
             notify(f"saving all matching signatures to '{output_matches}'")
 
-            # track matches
+            # write out basic CSV file
+            output_csv = output_base + ".csv"
             notify(f'saving all CSV matches to "{output_csv}"')
             csv_out_obj = FileOutputCSV(output_csv)
             csv_outfp = csv_out_obj.__enter__()
@@ -1330,31 +1352,32 @@
                 notify("nothing found... skipping.")
                 continue
 
-            output_unassigned = output_base + ".unassigned.sig"
-            with open(output_unassigned, "w"):
-                remaining_query = gather_iter.query
-                if noident_mh:
-                    remaining_mh = remaining_query.minhash.to_mutable()
-                    remaining_mh += noident_mh.downsample(scaled=remaining_mh.scaled)
-                    remaining_query.minhash = remaining_mh
+            output_unassigned = output_base + f".unassigned{args.extension}"
+            remaining_query = gather_iter.query
+            if noident_mh:
+                remaining_mh = remaining_query.minhash.to_mutable()
+                remaining_mh += noident_mh.downsample(scaled=remaining_mh.scaled)
+                remaining_query.minhash = remaining_mh
 
-                if is_abundance:
-                    abund_query_mh = remaining_query.minhash.inflate(orig_query_mh)
-                    remaining_query.minhash = abund_query_mh
+            if is_abundance:
+                abund_query_mh = remaining_query.minhash.inflate(orig_query_mh)
+                remaining_query.minhash = abund_query_mh
 
-                if found == 0:
-                    notify("nothing found - entire query signature unassigned.")
-                elif not remaining_query:
-                    notify("no unassigned hashes! not saving.")
-                else:
-                    notify(f'saving unassigned hashes to "{output_unassigned}"')
+            if found == 0:
+                notify("nothing found - entire query signature unassigned.")
+            elif not remaining_query:
+                notify("no unassigned hashes! not saving.")
+            else:
+                notify(f'saving unassigned hashes to "{output_unassigned}"')
+
+            with SaveSignaturesToLocation(output_unassigned) as save_sig:
+                save_sig.add(remaining_query)
 
-                with SaveSignaturesToLocation(output_unassigned) as save_sig:
-                    # CTB: note, multigather does not save abundances
-                    save_sig.add(remaining_query)
             n += 1
 
         # fini, next query!
+
+    # done! report at end.
     notify(f"\nconducted gather searches on {n} signatures")
     if size_may_be_inaccurate:
         notify(

tests/conftest.py

@@ -84,6 +84,11 @@ def sig_save_extension(request):
     return request.param
 
 
+@pytest.fixture(params=["sig", "sig.gz", "zip", ".d/"])
+def sig_save_extension_abund(request):
+    return request.param
+
+
 # --- BEGIN - Only run tests using a particular fixture --- #
 # Cribbed from: http://pythontesting.net/framework/pytest/pytest-run-tests-using-particular-fixture/
 def pytest_collection_modifyitems(items, config):