Performance improvements (#43)

* 100% threshold optimisation

* test slice off

* .

* Revert "100% threshold optimisation"

This reverts commit 6297a8e.

bigsi/graph/index.py

@@ -69,7 +69,7 @@ def __kmers_to_hashes(self, kmers):
         return d
 
     def __batch_get_rows(self, row_indexes):
-        return dict(zip(row_indexes, self.bitmatrix.get_rows(row_indexes)))
+        return dict(zip(row_indexes, self.bitmatrix.get_rows(row_indexes, slice=False)))
 
     def __bitwise_and_kmers(self, kmer_to_hashes, rows):
         d = {}

bigsi/matrix/bitmatrix.py

@@ -27,9 +27,14 @@ def create(cls, storage, rows, num_rows, num_cols):
     def get_row(self, row_index):
         return self.storage.get_bitarray(row_index)[: self.num_cols]
 
-    def get_rows(self, row_indexes):
+    def get_rows(self, row_indexes, slice=True):
+        ## Only need to slice for merging (it's a lot slower)
         # Takes advantage of batching in storage engine if available
-        return (ba[: self.num_cols] for ba in self.storage.get_bitarrays(row_indexes))
+        bitarrays=self.storage.get_bitarrays(row_indexes)
+        if slice:
+            return (ba[: self.num_cols] for ba in bitarrays)
+        else:
+            return bitarrays
 
     def set_row(self, row_index, bitarray):
         return self.storage.set_bitarray(row_index, bitarray)