bpo-34010: Fix tarfile read performance regression (GH-8020)
During buffered read, use a list followed by join instead of extending a bytes object.
This is how it was done before but changed in commit b506dc32c1a.
(cherry picked from commit 12a08c47601cadea8e7d3808502cdbcca87b2ce2)
Co-authored-by: hajoscher <hajoscher@gmail.com>
diff --git a/Lib/tarfile.py b/Lib/tarfile.py
index 85119a4..edd31e9 100755
--- a/Lib/tarfile.py
+++ b/Lib/tarfile.py
@@ -532,7 +532,7 @@
if not buf:
break
t.append(buf)
- buf = "".join(t)
+ buf = b"".join(t)
else:
buf = self._read(size)
self.pos += len(buf)
@@ -545,6 +545,7 @@
return self.__read(size)
c = len(self.dbuf)
+ t = [self.dbuf]
while c < size:
buf = self.__read(self.bufsize)
if not buf:
@@ -553,26 +554,27 @@
buf = self.cmp.decompress(buf)
except self.exception:
raise ReadError("invalid compressed data")
- self.dbuf += buf
+ t.append(buf)
c += len(buf)
- buf = self.dbuf[:size]
- self.dbuf = self.dbuf[size:]
- return buf
+ t = b"".join(t)
+ self.dbuf = t[size:]
+ return t[:size]
def __read(self, size):
"""Return size bytes from stream. If internal buffer is empty,
read another block from the stream.
"""
c = len(self.buf)
+ t = [self.buf]
while c < size:
buf = self.fileobj.read(self.bufsize)
if not buf:
break
- self.buf += buf
+ t.append(buf)
c += len(buf)
- buf = self.buf[:size]
- self.buf = self.buf[size:]
- return buf
+ t = b"".join(t)
+ self.buf = t[size:]
+ return t[:size]
# class _Stream
class _StreamProxy(object):