Fold _encode_tail_segment in to _encode_segment

src/allmydata/immutable/encode.py

@@ -255,11 +255,11 @@ class Encoder(object):
             # captures the slot, not the value
             #d.addCallback(lambda res: self.do_segment(i))
             # use this form instead:
-            d.addCallback(lambda res, i=i: self._encode_segment(i))
+            d.addCallback(lambda res, i=i: self._encode_segment(i, is_tail=False))
             d.addCallback(self._send_segment, i)
             d.addCallback(self._turn_barrier)
         last_segnum = self.num_segments - 1
-        d.addCallback(lambda res: self._encode_tail_segment(last_segnum))
+        d.addCallback(lambda res: self._encode_segment(last_segnum, is_tail=True))
         d.addCallback(self._send_segment, last_segnum)
         d.addCallback(self._turn_barrier)
 
@@ -317,8 +317,24 @@ class Encoder(object):
             dl.append(d)
         return self._gather_responses(dl)
 
-    def _encode_segment(self, segnum):
-        codec = self._codec
+    def _encode_segment(self, segnum, is_tail):
+        """
+        Encode one segment of input into the configured number of shares.
+
+        :param segnum: Ostensibly, the number of the segment to encode.  In
+            reality, this parameter is ignored and the *next* segment is
+            encoded and returned.
+
+        :param bool is_tail: ``True`` if this is the last segment, ``False``
+            otherwise.
+
+        :return: A ``Deferred`` which fires with a two-tuple.  The first
+            element is a list of string-y objects representing the encoded
+            segment data for one of the shares.  The second element is a list
+            of integers giving the share numbers of the shares in the first
+            element.
+        """
+        codec = self._tail_codec if is_tail else self._codec
         start = time.time()
 
         # the ICodecEncoder API wants to receive a total of self.segment_size
@@ -350,9 +366,11 @@ class Encoder(object):
         # footprint to 430KiB at the expense of more hash-tree overhead.
 
         d = self._gather_data(self.required_shares, input_piece_size,
-                              crypttext_segment_hasher)
+                              crypttext_segment_hasher, allow_short=is_tail)
         def _done_gathering(chunks):
             for c in chunks:
+                # If is_tail then a short trailing chunk will have been padded
+                # by _gather_data
                 assert len(c) == input_piece_size
             self._crypttext_hashes.append(crypttext_segment_hasher.digest())
             # during this call, we hit 5*segsize memory
@@ -365,31 +383,6 @@ class Encoder(object):
         d.addCallback(_done)
         return d
 
-    def _encode_tail_segment(self, segnum):
-
-        start = time.time()
-        codec = self._tail_codec
-        input_piece_size = codec.get_block_size()
-
-        crypttext_segment_hasher = hashutil.crypttext_segment_hasher()
-
-        d = self._gather_data(self.required_shares, input_piece_size,
-                              crypttext_segment_hasher, allow_short=True)
-        def _done_gathering(chunks):
-            for c in chunks:
-                # a short trailing chunk will have been padded by
-                # _gather_data
-                assert len(c) == input_piece_size
-            self._crypttext_hashes.append(crypttext_segment_hasher.digest())
-            return codec.encode(chunks)
-        d.addCallback(_done_gathering)
-        def _done(res):
-            elapsed = time.time() - start
-            self._times["cumulative_encoding"] += elapsed
-            return res
-        d.addCallback(_done)
-        return d
-
     def _gather_data(self, num_chunks, input_chunk_size,
                      crypttext_segment_hasher,
                      allow_short=False):