[Python-checkins] cpython (2.7): Issue #15068: Got rid of excessive buffering in the fileinput module.
serhiy.storchaka
python-checkins at python.org
Tue Mar 8 11:37:18 EST 2016
https://hg.python.org/cpython/rev/5fbd16326353
changeset: 100459:5fbd16326353
branch: 2.7
parent: 100453:ff2368d5c1d3
user: Serhiy Storchaka <storchaka at gmail.com>
date: Tue Mar 08 18:35:45 2016 +0200
summary:
Issue #15068: Got rid of excessive buffering in the fileinput module.
The bufsize parameter is no longer used.
files:
Doc/library/fileinput.rst | 6 +
Lib/fileinput.py | 150 +++++++++++-------------
Lib/test/test_fileinput.py | 72 +++++++++++-
Misc/NEWS | 3 +
4 files changed, 147 insertions(+), 84 deletions(-)
diff --git a/Doc/library/fileinput.rst b/Doc/library/fileinput.rst
--- a/Doc/library/fileinput.rst
+++ b/Doc/library/fileinput.rst
@@ -60,6 +60,9 @@
.. versionchanged:: 2.5
Added the *mode* and *openhook* parameters.
+ .. versionchanged:: 2.7.12
+ The *bufsize* parameter is no longer used.
+
The following functions use the global state created by :func:`fileinput.input`;
if there is no active state, :exc:`RuntimeError` is raised.
@@ -143,6 +146,9 @@
.. versionchanged:: 2.5
Added the *mode* and *openhook* parameters.
+ .. versionchanged:: 2.7.12
+ The *bufsize* parameter is no longer used.
+
**Optional in-place filtering:** if the keyword argument ``inplace=1`` is passed
to :func:`fileinput.input` or to the :class:`FileInput` constructor, the file is
moved to a backup file and standard output is directed to the input file (if a
diff --git a/Lib/fileinput.py b/Lib/fileinput.py
--- a/Lib/fileinput.py
+++ b/Lib/fileinput.py
@@ -64,13 +64,6 @@
disabled when standard input is read. XXX The current implementation
does not work for MS-DOS 8+3 filesystems.
-Performance: this module is unfortunately one of the slower ways of
-processing large numbers of input lines. Nevertheless, a significant
-speed-up has been obtained by using readlines(bufsize) instead of
-readline(). A new keyword argument, bufsize=N, is present on the
-input() function and the FileInput() class to override the default
-buffer size.
-
XXX Possible additions:
- optional getopt argument processing
@@ -86,6 +79,7 @@
_state = None
+# No longer used
DEFAULT_BUFSIZE = 8*1024
def input(files=None, inplace=0, backup="", bufsize=0,
@@ -207,17 +201,15 @@
self._files = files
self._inplace = inplace
self._backup = backup
- self._bufsize = bufsize or DEFAULT_BUFSIZE
self._savestdout = None
self._output = None
self._filename = None
- self._lineno = 0
+ self._startlineno = 0
self._filelineno = 0
self._file = None
+ self._readline = self._start_readline
self._isstdin = False
self._backupfilename = None
- self._buffer = []
- self._bufindex = 0
# restrict mode argument to reading modes
if mode not in ('r', 'rU', 'U', 'rb'):
raise ValueError("FileInput opening mode must be one of "
@@ -242,22 +234,18 @@
return self
def next(self):
- try:
- line = self._buffer[self._bufindex]
- except IndexError:
- pass
- else:
- self._bufindex += 1
- self._lineno += 1
+ line = self._readline()
+ if line:
self._filelineno += 1
return line
- line = self.readline()
- if not line:
+ if not self._file:
raise StopIteration
- return line
+ self.nextfile()
+ # Recursive call
+ return self.next()
def __getitem__(self, i):
- if i != self._lineno:
+ if i != self.lineno():
raise RuntimeError, "accessing lines out of order"
try:
return self.next()
@@ -277,7 +265,8 @@
output.close()
finally:
file = self._file
- self._file = 0
+ self._file = None
+ self._readline = self._start_readline
try:
if file and not self._isstdin:
file.close()
@@ -289,75 +278,72 @@
except OSError: pass
self._isstdin = False
- self._buffer = []
- self._bufindex = 0
def readline(self):
- try:
- line = self._buffer[self._bufindex]
- except IndexError:
- pass
+ while 1:
+ line = self._readline()
+ if line:
+ self._filelineno += 1
+ return line
+ if not self._file:
+ return line
+ self.nextfile()
+ # repeat with next file
+
+ def _start_readline(self):
+ if not self._files:
+ return ""
+ self._filename = self._files[0]
+ self._files = self._files[1:]
+ self._startlineno = self.lineno()
+ self._filelineno = 0
+ self._file = None
+ self._isstdin = False
+ self._backupfilename = 0
+ if self._filename == '-':
+ self._filename = '<stdin>'
+ self._file = sys.stdin
+ self._isstdin = True
else:
- self._bufindex += 1
- self._lineno += 1
- self._filelineno += 1
- return line
- if not self._file:
- if not self._files:
- return ""
- self._filename = self._files[0]
- self._files = self._files[1:]
- self._filelineno = 0
- self._file = None
- self._isstdin = False
- self._backupfilename = 0
- if self._filename == '-':
- self._filename = '<stdin>'
- self._file = sys.stdin
- self._isstdin = True
+ if self._inplace:
+ self._backupfilename = (
+ self._filename + (self._backup or os.extsep+"bak"))
+ try: os.unlink(self._backupfilename)
+ except os.error: pass
+ # The next few lines may raise IOError
+ os.rename(self._filename, self._backupfilename)
+ self._file = open(self._backupfilename, self._mode)
+ try:
+ perm = os.fstat(self._file.fileno()).st_mode
+ except OSError:
+ self._output = open(self._filename, "w")
+ else:
+ fd = os.open(self._filename,
+ os.O_CREAT | os.O_WRONLY | os.O_TRUNC,
+ perm)
+ self._output = os.fdopen(fd, "w")
+ try:
+ if hasattr(os, 'chmod'):
+ os.chmod(self._filename, perm)
+ except OSError:
+ pass
+ self._savestdout = sys.stdout
+ sys.stdout = self._output
else:
- if self._inplace:
- self._backupfilename = (
- self._filename + (self._backup or os.extsep+"bak"))
- try: os.unlink(self._backupfilename)
- except os.error: pass
- # The next few lines may raise IOError
- os.rename(self._filename, self._backupfilename)
- self._file = open(self._backupfilename, self._mode)
- try:
- perm = os.fstat(self._file.fileno()).st_mode
- except OSError:
- self._output = open(self._filename, "w")
- else:
- fd = os.open(self._filename,
- os.O_CREAT | os.O_WRONLY | os.O_TRUNC,
- perm)
- self._output = os.fdopen(fd, "w")
- try:
- if hasattr(os, 'chmod'):
- os.chmod(self._filename, perm)
- except OSError:
- pass
- self._savestdout = sys.stdout
- sys.stdout = self._output
+ # This may raise IOError
+ if self._openhook:
+ self._file = self._openhook(self._filename, self._mode)
else:
- # This may raise IOError
- if self._openhook:
- self._file = self._openhook(self._filename, self._mode)
- else:
- self._file = open(self._filename, self._mode)
- self._buffer = self._file.readlines(self._bufsize)
- self._bufindex = 0
- if not self._buffer:
- self.nextfile()
- # Recursive call
- return self.readline()
+ self._file = open(self._filename, self._mode)
+
+ self._readline = self._file.readline
+ return self._readline()
def filename(self):
return self._filename
def lineno(self):
- return self._lineno
+ return self._startlineno + self._filelineno
def filelineno(self):
return self._filelineno
diff --git a/Lib/test/test_fileinput.py b/Lib/test/test_fileinput.py
--- a/Lib/test/test_fileinput.py
+++ b/Lib/test/test_fileinput.py
@@ -5,7 +5,7 @@
import unittest
from test.test_support import verbose, TESTFN, run_unittest
-from test.test_support import unlink as safe_unlink
+from test.test_support import unlink as safe_unlink, check_warnings
import sys, re
from StringIO import StringIO
from fileinput import FileInput, hook_encoded
@@ -28,6 +28,42 @@
for name in names:
safe_unlink(name)
+class LineReader:
+
+ def __init__(self):
+ self._linesread = []
+
+ @property
+ def linesread(self):
+ try:
+ return self._linesread[:]
+ finally:
+ self._linesread = []
+
+ def openhook(self, filename, mode):
+ self.it = iter(filename.splitlines(True))
+ return self
+
+ def readline(self, size=None):
+ line = next(self.it, '')
+ self._linesread.append(line)
+ return line
+
+ def readlines(self, hint=-1):
+ lines = []
+ size = 0
+ while True:
+ line = self.readline()
+ if not line:
+ return lines
+ lines.append(line)
+ size += len(line)
+ if size >= hint:
+ return lines
+
+ def close(self):
+ pass
+
class BufferSizesTests(unittest.TestCase):
def test_buffer_sizes(self):
# First, run the tests with default and teeny buffer size.
@@ -228,7 +264,7 @@
f.write('\x80')
self.addCleanup(safe_unlink, TESTFN)
- fi = FileInput(files=TESTFN, openhook=hook_encoded('ascii'), bufsize=8)
+ fi = FileInput(files=TESTFN, openhook=hook_encoded('ascii'))
# The most likely failure is a UnicodeDecodeError due to the entire
# file being read when it shouldn't have been.
self.assertEqual(fi.readline(), u'A\n')
@@ -239,6 +275,38 @@
list(fi)
fi.close()
+ def test_readline_buffering(self):
+ src = LineReader()
+ fi = FileInput(files=['line1\nline2', 'line3\n'], openhook=src.openhook)
+ self.assertEqual(src.linesread, [])
+ self.assertEqual(fi.readline(), 'line1\n')
+ self.assertEqual(src.linesread, ['line1\n'])
+ self.assertEqual(fi.readline(), 'line2')
+ self.assertEqual(src.linesread, ['line2'])
+ self.assertEqual(fi.readline(), 'line3\n')
+ self.assertEqual(src.linesread, ['', 'line3\n'])
+ self.assertEqual(fi.readline(), '')
+ self.assertEqual(src.linesread, [''])
+ self.assertEqual(fi.readline(), '')
+ self.assertEqual(src.linesread, [])
+ fi.close()
+
+ def test_iteration_buffering(self):
+ src = LineReader()
+ fi = FileInput(files=['line1\nline2', 'line3\n'], openhook=src.openhook)
+ self.assertEqual(src.linesread, [])
+ self.assertEqual(next(fi), 'line1\n')
+ self.assertEqual(src.linesread, ['line1\n'])
+ self.assertEqual(next(fi), 'line2')
+ self.assertEqual(src.linesread, ['line2'])
+ self.assertEqual(next(fi), 'line3\n')
+ self.assertEqual(src.linesread, ['', 'line3\n'])
+ self.assertRaises(StopIteration, next, fi)
+ self.assertEqual(src.linesread, [''])
+ self.assertRaises(StopIteration, next, fi)
+ self.assertEqual(src.linesread, [])
+ fi.close()
+
class Test_hook_encoded(unittest.TestCase):
"""Unit tests for fileinput.hook_encoded()"""
diff --git a/Misc/NEWS b/Misc/NEWS
--- a/Misc/NEWS
+++ b/Misc/NEWS
@@ -58,6 +58,9 @@
Library
-------
+- Issue #15068: Got rid of excessive buffering in the fileinput module.
+ The bufsize parameter is no longer used.
+
- Issue #2202: Fix UnboundLocalError in
AbstractDigestAuthHandler.get_algorithm_impls. Initial patch by Mathieu Dupuy.
--
Repository URL: https://hg.python.org/cpython
More information about the Python-checkins
mailing list