Merge pull request #87 from dib-lab/partition/determ

Make "kevlar partition" order deterministic

kevlar/assemble.py

@@ -152,8 +152,12 @@ def main(args):
         debugout = args.logfile
 
     reads, kmers = load_reads_and_kmers(kevlar.open(args.augfastq, 'r'),
-                                        debugout)
+                                        args.logfile)
     inputreads = list(reads)
+    message = 'loaded {:d} reads'.format(len(inputreads))
+    message += ' and {:d} interesting k-mers'.format(len(kmers))
+    print('[kevlar::assemble]', message, file=args.logfile)
+
     graph = kevlar.overlap.graph_init_strict(reads, kmers, args.min_abund,
                                              args.max_abund, debugout)
     if args.gml:

kevlar/overlap.py

@@ -257,7 +257,10 @@ def write_partitions(read_graph, reads, ccprefix, logstream):
     n = 0
     reads_in_ccs = 0
     cclog = open(ccprefix + '.cc.log', 'w')
-    for n, cc in enumerate(networkx.connected_components(read_graph)):
+    ccs = sorted(networkx.connected_components(read_graph), reverse=True,
+                 # Sort first by number of reads, then by read names
+                 key=lambda c: (len(c), sorted(c)))
+    for n, cc in enumerate(ccs):
         print('CC', n, len(cc), cc, sep='\t', file=cclog)
         reads_in_ccs += len(cc)
         outfilename = '{:s}.cc{:d}.augfastq.gz'.format(ccprefix, n)

kevlar/seqio.py

@@ -97,7 +97,7 @@ def load_reads_and_kmers(instream, logstream=None):
     kmers = defaultdict(set)
     for n, record in enumerate(kevlar.parse_augmented_fastx(instream), 1):
         if logstream and n % 10000 == 0:  # pragma: no cover
-            print('[kevlar::assemble]    loaded {:d} reads'.format(n),
+            print('[kevlar::seqio]    loaded {:d} reads'.format(n),
                   file=logstream)
         reads[record.name] = record
         for kmer in record.ikmers: