bpo-34043: Optimize tarfile uncompress performance (GH-8089)

tarfile._Stream has two buffer for compressed and uncompressed data. Those buffers are not aligned so unnecessary bytes slicing happens for every reading chunks. This commit bypass compressed buffering. In this benchmark [1], user time become 250ms from 300ms. [1]: https://bugs.python.org/msg320763

Author: methane

Lib/tarfile.py

@@ -513,21 +513,10 @@ def seek(self, pos=0):
  raise StreamError("seeking backwards is not allowed")
  return self.pos
 
- def read(self, size=None):
- """Return the next size number of bytes from the stream.
- If size is not defined, return all bytes of the stream
- up to EOF.
- """
- if size is None:
- t = []
- while True:
- buf = self._read(self.bufsize)
- if not buf:
- break
- t.append(buf)
- buf = b"".join(t)
- else:
- buf = self._read(size)
+ def read(self, size):
+ """Return the next size number of bytes from the stream."""
+ assert size is not None
+ buf = self._read(size)
  self.pos += len(buf)
  return buf
 
@@ -540,9 +529,14 @@ def _read(self, size):
  c = len(self.dbuf)
  t = [self.dbuf]
  while c < size:
- buf = self.__read(self.bufsize)
- if not buf:
- break
+ # Skip underlying buffer to avoid unaligned double buffering.
+ if self.buf:
+ buf = self.buf
+ self.buf = b""
+ else:
+ buf = self.fileobj.read(self.bufsize)
+ if not buf:
+ break
  try:
  buf = self.cmp.decompress(buf)
  except self.exception:

Misc/NEWS.d/next/Library/2018-07-04-21-14-35.bpo-34043.0YJNq9.rst

@@ -0,0 +1 @@
+Optimize tarfile uncompress performance about 15% when gzip is used.