cpython/Lib/test/test_bz2.py

1002 lines
37 KiB
Python

from test import support
from test.support import bigmemtest, _4G
import unittest
from io import BytesIO, DEFAULT_BUFFER_SIZE
import os
import pickle
import glob
import random
import subprocess
import sys
from test.support import unlink
import _compression
try:
import threading
except ImportError:
threading = None
# Skip tests if the bz2 module doesn't exist.
bz2 = support.import_module('bz2')
from bz2 import BZ2File, BZ2Compressor, BZ2Decompressor
class BaseTest(unittest.TestCase):
"Base for other testcases."
TEXT_LINES = [
b'root:x:0:0:root:/root:/bin/bash\n',
b'bin:x:1:1:bin:/bin:\n',
b'daemon:x:2:2:daemon:/sbin:\n',
b'adm:x:3:4:adm:/var/adm:\n',
b'lp:x:4:7:lp:/var/spool/lpd:\n',
b'sync:x:5:0:sync:/sbin:/bin/sync\n',
b'shutdown:x:6:0:shutdown:/sbin:/sbin/shutdown\n',
b'halt:x:7:0:halt:/sbin:/sbin/halt\n',
b'mail:x:8:12:mail:/var/spool/mail:\n',
b'news:x:9:13:news:/var/spool/news:\n',
b'uucp:x:10:14:uucp:/var/spool/uucp:\n',
b'operator:x:11:0:operator:/root:\n',
b'games:x:12:100:games:/usr/games:\n',
b'gopher:x:13:30:gopher:/usr/lib/gopher-data:\n',
b'ftp:x:14:50:FTP User:/var/ftp:/bin/bash\n',
b'nobody:x:65534:65534:Nobody:/home:\n',
b'postfix:x:100:101:postfix:/var/spool/postfix:\n',
b'niemeyer:x:500:500::/home/niemeyer:/bin/bash\n',
b'postgres:x:101:102:PostgreSQL Server:/var/lib/pgsql:/bin/bash\n',
b'mysql:x:102:103:MySQL server:/var/lib/mysql:/bin/bash\n',
b'www:x:103:104::/var/www:/bin/false\n',
]
TEXT = b''.join(TEXT_LINES)
DATA = b'BZh91AY&SY.\xc8N\x18\x00\x01>_\x80\x00\x10@\x02\xff\xf0\x01\x07n\x00?\xe7\xff\xe00\x01\x99\xaa\x00\xc0\x03F\x86\x8c#&\x83F\x9a\x03\x06\xa6\xd0\xa6\x93M\x0fQ\xa7\xa8\x06\x804hh\x12$\x11\xa4i4\xf14S\xd2<Q\xb5\x0fH\xd3\xd4\xdd\xd5\x87\xbb\xf8\x94\r\x8f\xafI\x12\xe1\xc9\xf8/E\x00pu\x89\x12]\xc9\xbbDL\nQ\x0e\t1\x12\xdf\xa0\xc0\x97\xac2O9\x89\x13\x94\x0e\x1c7\x0ed\x95I\x0c\xaaJ\xa4\x18L\x10\x05#\x9c\xaf\xba\xbc/\x97\x8a#C\xc8\xe1\x8cW\xf9\xe2\xd0\xd6M\xa7\x8bXa<e\x84t\xcbL\xb3\xa7\xd9\xcd\xd1\xcb\x84.\xaf\xb3\xab\xab\xad`n}\xa0lh\tE,\x8eZ\x15\x17VH>\x88\xe5\xcd9gd6\x0b\n\xe9\x9b\xd5\x8a\x99\xf7\x08.K\x8ev\xfb\xf7xw\xbb\xdf\xa1\x92\xf1\xdd|/";\xa2\xba\x9f\xd5\xb1#A\xb6\xf6\xb3o\xc9\xc5y\\\xebO\xe7\x85\x9a\xbc\xb6f8\x952\xd5\xd7"%\x89>V,\xf7\xa6z\xe2\x9f\xa3\xdf\x11\x11"\xd6E)I\xa9\x13^\xca\xf3r\xd0\x03U\x922\xf26\xec\xb6\xed\x8b\xc3U\x13\x9d\xc5\x170\xa4\xfa^\x92\xacDF\x8a\x97\xd6\x19\xfe\xdd\xb8\xbd\x1a\x9a\x19\xa3\x80ankR\x8b\xe5\xd83]\xa9\xc6\x08\x82f\xf6\xb9"6l$\xb8j@\xc0\x8a\xb0l1..\xbak\x83ls\x15\xbc\xf4\xc1\x13\xbe\xf8E\xb8\x9d\r\xa8\x9dk\x84\xd3n\xfa\xacQ\x07\xb1%y\xaav\xb4\x08\xe0z\x1b\x16\xf5\x04\xe9\xcc\xb9\x08z\x1en7.G\xfc]\xc9\x14\xe1B@\xbb!8`'
EMPTY_DATA = b'BZh9\x17rE8P\x90\x00\x00\x00\x00'
BAD_DATA = b'this is not a valid bzip2 file'
# Some tests need more than one block of uncompressed data. Since one block
# is at least 100 kB, we gather some data dynamically and compress it.
# Note that this assumes that compression works correctly, so we cannot
# simply use the bigger test data for all tests.
test_size = 0
BIG_TEXT = bytearray(128*1024)
for fname in glob.glob(os.path.join(os.path.dirname(__file__), '*.py')):
with open(fname, 'rb') as fh:
test_size += fh.readinto(memoryview(BIG_TEXT)[test_size:])
if test_size > 128*1024:
break
BIG_DATA = bz2.compress(BIG_TEXT, compresslevel=1)
def setUp(self):
self.filename = support.TESTFN
def tearDown(self):
if os.path.isfile(self.filename):
os.unlink(self.filename)
if sys.platform == "win32":
# bunzip2 isn't available to run on Windows.
def decompress(self, data):
return bz2.decompress(data)
else:
def decompress(self, data):
pop = subprocess.Popen("bunzip2", shell=True,
stdin=subprocess.PIPE,
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT)
pop.stdin.write(data)
pop.stdin.close()
ret = pop.stdout.read()
pop.stdout.close()
if pop.wait() != 0:
ret = bz2.decompress(data)
return ret
class BZ2FileTest(BaseTest):
"Test the BZ2File class."
def createTempFile(self, streams=1, suffix=b""):
with open(self.filename, "wb") as f:
f.write(self.DATA * streams)
f.write(suffix)
def testBadArgs(self):
self.assertRaises(TypeError, BZ2File, 123.456)
self.assertRaises(ValueError, BZ2File, os.devnull, "z")
self.assertRaises(ValueError, BZ2File, os.devnull, "rx")
self.assertRaises(ValueError, BZ2File, os.devnull, "rbt")
self.assertRaises(ValueError, BZ2File, os.devnull, compresslevel=0)
self.assertRaises(ValueError, BZ2File, os.devnull, compresslevel=10)
def testRead(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
self.assertRaises(TypeError, bz2f.read, float())
self.assertEqual(bz2f.read(), self.TEXT)
def testReadBadFile(self):
self.createTempFile(streams=0, suffix=self.BAD_DATA)
with BZ2File(self.filename) as bz2f:
self.assertRaises(OSError, bz2f.read)
def testReadMultiStream(self):
self.createTempFile(streams=5)
with BZ2File(self.filename) as bz2f:
self.assertRaises(TypeError, bz2f.read, float())
self.assertEqual(bz2f.read(), self.TEXT * 5)
def testReadMonkeyMultiStream(self):
# Test BZ2File.read() on a multi-stream archive where a stream
# boundary coincides with the end of the raw read buffer.
buffer_size = _compression.BUFFER_SIZE
_compression.BUFFER_SIZE = len(self.DATA)
try:
self.createTempFile(streams=5)
with BZ2File(self.filename) as bz2f:
self.assertRaises(TypeError, bz2f.read, float())
self.assertEqual(bz2f.read(), self.TEXT * 5)
finally:
_compression.BUFFER_SIZE = buffer_size
def testReadTrailingJunk(self):
self.createTempFile(suffix=self.BAD_DATA)
with BZ2File(self.filename) as bz2f:
self.assertEqual(bz2f.read(), self.TEXT)
def testReadMultiStreamTrailingJunk(self):
self.createTempFile(streams=5, suffix=self.BAD_DATA)
with BZ2File(self.filename) as bz2f:
self.assertEqual(bz2f.read(), self.TEXT * 5)
def testRead0(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
self.assertRaises(TypeError, bz2f.read, float())
self.assertEqual(bz2f.read(0), b"")
def testReadChunk10(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
text = b''
while True:
str = bz2f.read(10)
if not str:
break
text += str
self.assertEqual(text, self.TEXT)
def testReadChunk10MultiStream(self):
self.createTempFile(streams=5)
with BZ2File(self.filename) as bz2f:
text = b''
while True:
str = bz2f.read(10)
if not str:
break
text += str
self.assertEqual(text, self.TEXT * 5)
def testRead100(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
self.assertEqual(bz2f.read(100), self.TEXT[:100])
def testPeek(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
pdata = bz2f.peek()
self.assertNotEqual(len(pdata), 0)
self.assertTrue(self.TEXT.startswith(pdata))
self.assertEqual(bz2f.read(), self.TEXT)
def testReadInto(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
n = 128
b = bytearray(n)
self.assertEqual(bz2f.readinto(b), n)
self.assertEqual(b, self.TEXT[:n])
n = len(self.TEXT) - n
b = bytearray(len(self.TEXT))
self.assertEqual(bz2f.readinto(b), n)
self.assertEqual(b[:n], self.TEXT[-n:])
def testReadLine(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
self.assertRaises(TypeError, bz2f.readline, None)
for line in self.TEXT_LINES:
self.assertEqual(bz2f.readline(), line)
def testReadLineMultiStream(self):
self.createTempFile(streams=5)
with BZ2File(self.filename) as bz2f:
self.assertRaises(TypeError, bz2f.readline, None)
for line in self.TEXT_LINES * 5:
self.assertEqual(bz2f.readline(), line)
def testReadLines(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
self.assertRaises(TypeError, bz2f.readlines, None)
self.assertEqual(bz2f.readlines(), self.TEXT_LINES)
def testReadLinesMultiStream(self):
self.createTempFile(streams=5)
with BZ2File(self.filename) as bz2f:
self.assertRaises(TypeError, bz2f.readlines, None)
self.assertEqual(bz2f.readlines(), self.TEXT_LINES * 5)
def testIterator(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
self.assertEqual(list(iter(bz2f)), self.TEXT_LINES)
def testIteratorMultiStream(self):
self.createTempFile(streams=5)
with BZ2File(self.filename) as bz2f:
self.assertEqual(list(iter(bz2f)), self.TEXT_LINES * 5)
def testClosedIteratorDeadlock(self):
# Issue #3309: Iteration on a closed BZ2File should release the lock.
self.createTempFile()
bz2f = BZ2File(self.filename)
bz2f.close()
self.assertRaises(ValueError, next, bz2f)
# This call will deadlock if the above call failed to release the lock.
self.assertRaises(ValueError, bz2f.readlines)
def testWrite(self):
with BZ2File(self.filename, "w") as bz2f:
self.assertRaises(TypeError, bz2f.write)
bz2f.write(self.TEXT)
with open(self.filename, 'rb') as f:
self.assertEqual(self.decompress(f.read()), self.TEXT)
def testWriteChunks10(self):
with BZ2File(self.filename, "w") as bz2f:
n = 0
while True:
str = self.TEXT[n*10:(n+1)*10]
if not str:
break
bz2f.write(str)
n += 1
with open(self.filename, 'rb') as f:
self.assertEqual(self.decompress(f.read()), self.TEXT)
def testWriteNonDefaultCompressLevel(self):
expected = bz2.compress(self.TEXT, compresslevel=5)
with BZ2File(self.filename, "w", compresslevel=5) as bz2f:
bz2f.write(self.TEXT)
with open(self.filename, "rb") as f:
self.assertEqual(f.read(), expected)
def testWriteLines(self):
with BZ2File(self.filename, "w") as bz2f:
self.assertRaises(TypeError, bz2f.writelines)
bz2f.writelines(self.TEXT_LINES)
# Issue #1535500: Calling writelines() on a closed BZ2File
# should raise an exception.
self.assertRaises(ValueError, bz2f.writelines, ["a"])
with open(self.filename, 'rb') as f:
self.assertEqual(self.decompress(f.read()), self.TEXT)
def testWriteMethodsOnReadOnlyFile(self):
with BZ2File(self.filename, "w") as bz2f:
bz2f.write(b"abc")
with BZ2File(self.filename, "r") as bz2f:
self.assertRaises(OSError, bz2f.write, b"a")
self.assertRaises(OSError, bz2f.writelines, [b"a"])
def testAppend(self):
with BZ2File(self.filename, "w") as bz2f:
self.assertRaises(TypeError, bz2f.write)
bz2f.write(self.TEXT)
with BZ2File(self.filename, "a") as bz2f:
self.assertRaises(TypeError, bz2f.write)
bz2f.write(self.TEXT)
with open(self.filename, 'rb') as f:
self.assertEqual(self.decompress(f.read()), self.TEXT * 2)
def testSeekForward(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
self.assertRaises(TypeError, bz2f.seek)
bz2f.seek(150)
self.assertEqual(bz2f.read(), self.TEXT[150:])
def testSeekForwardAcrossStreams(self):
self.createTempFile(streams=2)
with BZ2File(self.filename) as bz2f:
self.assertRaises(TypeError, bz2f.seek)
bz2f.seek(len(self.TEXT) + 150)
self.assertEqual(bz2f.read(), self.TEXT[150:])
def testSeekBackwards(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
bz2f.read(500)
bz2f.seek(-150, 1)
self.assertEqual(bz2f.read(), self.TEXT[500-150:])
def testSeekBackwardsAcrossStreams(self):
self.createTempFile(streams=2)
with BZ2File(self.filename) as bz2f:
readto = len(self.TEXT) + 100
while readto > 0:
readto -= len(bz2f.read(readto))
bz2f.seek(-150, 1)
self.assertEqual(bz2f.read(), self.TEXT[100-150:] + self.TEXT)
def testSeekBackwardsFromEnd(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
bz2f.seek(-150, 2)
self.assertEqual(bz2f.read(), self.TEXT[len(self.TEXT)-150:])
def testSeekBackwardsFromEndAcrossStreams(self):
self.createTempFile(streams=2)
with BZ2File(self.filename) as bz2f:
bz2f.seek(-1000, 2)
self.assertEqual(bz2f.read(), (self.TEXT * 2)[-1000:])
def testSeekPostEnd(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
bz2f.seek(150000)
self.assertEqual(bz2f.tell(), len(self.TEXT))
self.assertEqual(bz2f.read(), b"")
def testSeekPostEndMultiStream(self):
self.createTempFile(streams=5)
with BZ2File(self.filename) as bz2f:
bz2f.seek(150000)
self.assertEqual(bz2f.tell(), len(self.TEXT) * 5)
self.assertEqual(bz2f.read(), b"")
def testSeekPostEndTwice(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
bz2f.seek(150000)
bz2f.seek(150000)
self.assertEqual(bz2f.tell(), len(self.TEXT))
self.assertEqual(bz2f.read(), b"")
def testSeekPostEndTwiceMultiStream(self):
self.createTempFile(streams=5)
with BZ2File(self.filename) as bz2f:
bz2f.seek(150000)
bz2f.seek(150000)
self.assertEqual(bz2f.tell(), len(self.TEXT) * 5)
self.assertEqual(bz2f.read(), b"")
def testSeekPreStart(self):
self.createTempFile()
with BZ2File(self.filename) as bz2f:
bz2f.seek(-150)
self.assertEqual(bz2f.tell(), 0)
self.assertEqual(bz2f.read(), self.TEXT)
def testSeekPreStartMultiStream(self):
self.createTempFile(streams=2)
with BZ2File(self.filename) as bz2f:
bz2f.seek(-150)
self.assertEqual(bz2f.tell(), 0)
self.assertEqual(bz2f.read(), self.TEXT * 2)
def testFileno(self):
self.createTempFile()
with open(self.filename, 'rb') as rawf:
bz2f = BZ2File(rawf)
try:
self.assertEqual(bz2f.fileno(), rawf.fileno())
finally:
bz2f.close()
self.assertRaises(ValueError, bz2f.fileno)
def testSeekable(self):
bz2f = BZ2File(BytesIO(self.DATA))
try:
self.assertTrue(bz2f.seekable())
bz2f.read()
self.assertTrue(bz2f.seekable())
finally:
bz2f.close()
self.assertRaises(ValueError, bz2f.seekable)
bz2f = BZ2File(BytesIO(), "w")
try:
self.assertFalse(bz2f.seekable())
finally:
bz2f.close()
self.assertRaises(ValueError, bz2f.seekable)
src = BytesIO(self.DATA)
src.seekable = lambda: False
bz2f = BZ2File(src)
try:
self.assertFalse(bz2f.seekable())
finally:
bz2f.close()
self.assertRaises(ValueError, bz2f.seekable)
def testReadable(self):
bz2f = BZ2File(BytesIO(self.DATA))
try:
self.assertTrue(bz2f.readable())
bz2f.read()
self.assertTrue(bz2f.readable())
finally:
bz2f.close()
self.assertRaises(ValueError, bz2f.readable)
bz2f = BZ2File(BytesIO(), "w")
try:
self.assertFalse(bz2f.readable())
finally:
bz2f.close()
self.assertRaises(ValueError, bz2f.readable)
def testWritable(self):
bz2f = BZ2File(BytesIO(self.DATA))
try:
self.assertFalse(bz2f.writable())
bz2f.read()
self.assertFalse(bz2f.writable())
finally:
bz2f.close()
self.assertRaises(ValueError, bz2f.writable)
bz2f = BZ2File(BytesIO(), "w")
try:
self.assertTrue(bz2f.writable())
finally:
bz2f.close()
self.assertRaises(ValueError, bz2f.writable)
def testOpenDel(self):
self.createTempFile()
for i in range(10000):
o = BZ2File(self.filename)
del o
def testOpenNonexistent(self):
self.assertRaises(OSError, BZ2File, "/non/existent")
def testReadlinesNoNewline(self):
# Issue #1191043: readlines() fails on a file containing no newline.
data = b'BZh91AY&SY\xd9b\x89]\x00\x00\x00\x03\x80\x04\x00\x02\x00\x0c\x00 \x00!\x9ah3M\x13<]\xc9\x14\xe1BCe\x8a%t'
with open(self.filename, "wb") as f:
f.write(data)
with BZ2File(self.filename) as bz2f:
lines = bz2f.readlines()
self.assertEqual(lines, [b'Test'])
with BZ2File(self.filename) as bz2f:
xlines = list(bz2f.readlines())
self.assertEqual(xlines, [b'Test'])
def testContextProtocol(self):
f = None
with BZ2File(self.filename, "wb") as f:
f.write(b"xxx")
f = BZ2File(self.filename, "rb")
f.close()
try:
with f:
pass
except ValueError:
pass
else:
self.fail("__enter__ on a closed file didn't raise an exception")
try:
with BZ2File(self.filename, "wb") as f:
1/0
except ZeroDivisionError:
pass
else:
self.fail("1/0 didn't raise an exception")
@unittest.skipUnless(threading, 'Threading required for this test.')
def testThreading(self):
# Issue #7205: Using a BZ2File from several threads shouldn't deadlock.
data = b"1" * 2**20
nthreads = 10
with BZ2File(self.filename, 'wb') as f:
def comp():
for i in range(5):
f.write(data)
threads = [threading.Thread(target=comp) for i in range(nthreads)]
with support.start_threads(threads):
pass
def testWithoutThreading(self):
module = support.import_fresh_module("bz2", blocked=("threading",))
with module.BZ2File(self.filename, "wb") as f:
f.write(b"abc")
with module.BZ2File(self.filename, "rb") as f:
self.assertEqual(f.read(), b"abc")
def testMixedIterationAndReads(self):
self.createTempFile()
linelen = len(self.TEXT_LINES[0])
halflen = linelen // 2
with BZ2File(self.filename) as bz2f:
bz2f.read(halflen)
self.assertEqual(next(bz2f), self.TEXT_LINES[0][halflen:])
self.assertEqual(bz2f.read(), self.TEXT[linelen:])
with BZ2File(self.filename) as bz2f:
bz2f.readline()
self.assertEqual(next(bz2f), self.TEXT_LINES[1])
self.assertEqual(bz2f.readline(), self.TEXT_LINES[2])
with BZ2File(self.filename) as bz2f:
bz2f.readlines()
self.assertRaises(StopIteration, next, bz2f)
self.assertEqual(bz2f.readlines(), [])
def testMultiStreamOrdering(self):
# Test the ordering of streams when reading a multi-stream archive.
data1 = b"foo" * 1000
data2 = b"bar" * 1000
with BZ2File(self.filename, "w") as bz2f:
bz2f.write(data1)
with BZ2File(self.filename, "a") as bz2f:
bz2f.write(data2)
with BZ2File(self.filename) as bz2f:
self.assertEqual(bz2f.read(), data1 + data2)
def testOpenBytesFilename(self):
str_filename = self.filename
try:
bytes_filename = str_filename.encode("ascii")
except UnicodeEncodeError:
self.skipTest("Temporary file name needs to be ASCII")
with BZ2File(bytes_filename, "wb") as f:
f.write(self.DATA)
with BZ2File(bytes_filename, "rb") as f:
self.assertEqual(f.read(), self.DATA)
# Sanity check that we are actually operating on the right file.
with BZ2File(str_filename, "rb") as f:
self.assertEqual(f.read(), self.DATA)
def testDecompressLimited(self):
"""Decompressed data buffering should be limited"""
bomb = bz2.compress(bytes(int(2e6)), compresslevel=9)
self.assertLess(len(bomb), _compression.BUFFER_SIZE)
decomp = BZ2File(BytesIO(bomb))
self.assertEqual(bytes(1), decomp.read(1))
max_decomp = 1 + DEFAULT_BUFFER_SIZE
self.assertLessEqual(decomp._buffer.raw.tell(), max_decomp,
"Excessive amount of data was decompressed")
# Tests for a BZ2File wrapping another file object:
def testReadBytesIO(self):
with BytesIO(self.DATA) as bio:
with BZ2File(bio) as bz2f:
self.assertRaises(TypeError, bz2f.read, float())
self.assertEqual(bz2f.read(), self.TEXT)
self.assertFalse(bio.closed)
def testPeekBytesIO(self):
with BytesIO(self.DATA) as bio:
with BZ2File(bio) as bz2f:
pdata = bz2f.peek()
self.assertNotEqual(len(pdata), 0)
self.assertTrue(self.TEXT.startswith(pdata))
self.assertEqual(bz2f.read(), self.TEXT)
def testWriteBytesIO(self):
with BytesIO() as bio:
with BZ2File(bio, "w") as bz2f:
self.assertRaises(TypeError, bz2f.write)
bz2f.write(self.TEXT)
self.assertEqual(self.decompress(bio.getvalue()), self.TEXT)
self.assertFalse(bio.closed)
def testSeekForwardBytesIO(self):
with BytesIO(self.DATA) as bio:
with BZ2File(bio) as bz2f:
self.assertRaises(TypeError, bz2f.seek)
bz2f.seek(150)
self.assertEqual(bz2f.read(), self.TEXT[150:])
def testSeekBackwardsBytesIO(self):
with BytesIO(self.DATA) as bio:
with BZ2File(bio) as bz2f:
bz2f.read(500)
bz2f.seek(-150, 1)
self.assertEqual(bz2f.read(), self.TEXT[500-150:])
def test_read_truncated(self):
# Drop the eos_magic field (6 bytes) and CRC (4 bytes).
truncated = self.DATA[:-10]
with BZ2File(BytesIO(truncated)) as f:
self.assertRaises(EOFError, f.read)
with BZ2File(BytesIO(truncated)) as f:
self.assertEqual(f.read(len(self.TEXT)), self.TEXT)
self.assertRaises(EOFError, f.read, 1)
# Incomplete 4-byte file header, and block header of at least 146 bits.
for i in range(22):
with BZ2File(BytesIO(truncated[:i])) as f:
self.assertRaises(EOFError, f.read, 1)
class BZ2CompressorTest(BaseTest):
def testCompress(self):
bz2c = BZ2Compressor()
self.assertRaises(TypeError, bz2c.compress)
data = bz2c.compress(self.TEXT)
data += bz2c.flush()
self.assertEqual(self.decompress(data), self.TEXT)
def testCompressEmptyString(self):
bz2c = BZ2Compressor()
data = bz2c.compress(b'')
data += bz2c.flush()
self.assertEqual(data, self.EMPTY_DATA)
def testCompressChunks10(self):
bz2c = BZ2Compressor()
n = 0
data = b''
while True:
str = self.TEXT[n*10:(n+1)*10]
if not str:
break
data += bz2c.compress(str)
n += 1
data += bz2c.flush()
self.assertEqual(self.decompress(data), self.TEXT)
@bigmemtest(size=_4G + 100, memuse=2)
def testCompress4G(self, size):
# "Test BZ2Compressor.compress()/flush() with >4GiB input"
bz2c = BZ2Compressor()
data = b"x" * size
try:
compressed = bz2c.compress(data)
compressed += bz2c.flush()
finally:
data = None # Release memory
data = bz2.decompress(compressed)
try:
self.assertEqual(len(data), size)
self.assertEqual(len(data.strip(b"x")), 0)
finally:
data = None
def testPickle(self):
for proto in range(pickle.HIGHEST_PROTOCOL + 1):
with self.assertRaises(TypeError):
pickle.dumps(BZ2Compressor(), proto)
class BZ2DecompressorTest(BaseTest):
def test_Constructor(self):
self.assertRaises(TypeError, BZ2Decompressor, 42)
def testDecompress(self):
bz2d = BZ2Decompressor()
self.assertRaises(TypeError, bz2d.decompress)
text = bz2d.decompress(self.DATA)
self.assertEqual(text, self.TEXT)
def testDecompressChunks10(self):
bz2d = BZ2Decompressor()
text = b''
n = 0
while True:
str = self.DATA[n*10:(n+1)*10]
if not str:
break
text += bz2d.decompress(str)
n += 1
self.assertEqual(text, self.TEXT)
def testDecompressUnusedData(self):
bz2d = BZ2Decompressor()
unused_data = b"this is unused data"
text = bz2d.decompress(self.DATA+unused_data)
self.assertEqual(text, self.TEXT)
self.assertEqual(bz2d.unused_data, unused_data)
def testEOFError(self):
bz2d = BZ2Decompressor()
text = bz2d.decompress(self.DATA)
self.assertRaises(EOFError, bz2d.decompress, b"anything")
self.assertRaises(EOFError, bz2d.decompress, b"")
@bigmemtest(size=_4G + 100, memuse=3.3)
def testDecompress4G(self, size):
# "Test BZ2Decompressor.decompress() with >4GiB input"
blocksize = 10 * 1024 * 1024
block = random.getrandbits(blocksize * 8).to_bytes(blocksize, 'little')
try:
data = block * (size // blocksize + 1)
compressed = bz2.compress(data)
bz2d = BZ2Decompressor()
decompressed = bz2d.decompress(compressed)
self.assertTrue(decompressed == data)
finally:
data = None
compressed = None
decompressed = None
def testPickle(self):
for proto in range(pickle.HIGHEST_PROTOCOL + 1):
with self.assertRaises(TypeError):
pickle.dumps(BZ2Decompressor(), proto)
def testDecompressorChunksMaxsize(self):
bzd = BZ2Decompressor()
max_length = 100
out = []
# Feed some input
len_ = len(self.BIG_DATA) - 64
out.append(bzd.decompress(self.BIG_DATA[:len_],
max_length=max_length))
self.assertFalse(bzd.needs_input)
self.assertEqual(len(out[-1]), max_length)
# Retrieve more data without providing more input
out.append(bzd.decompress(b'', max_length=max_length))
self.assertFalse(bzd.needs_input)
self.assertEqual(len(out[-1]), max_length)
# Retrieve more data while providing more input
out.append(bzd.decompress(self.BIG_DATA[len_:],
max_length=max_length))
self.assertLessEqual(len(out[-1]), max_length)
# Retrieve remaining uncompressed data
while not bzd.eof:
out.append(bzd.decompress(b'', max_length=max_length))
self.assertLessEqual(len(out[-1]), max_length)
out = b"".join(out)
self.assertEqual(out, self.BIG_TEXT)
self.assertEqual(bzd.unused_data, b"")
def test_decompressor_inputbuf_1(self):
# Test reusing input buffer after moving existing
# contents to beginning
bzd = BZ2Decompressor()
out = []
# Create input buffer and fill it
self.assertEqual(bzd.decompress(self.DATA[:100],
max_length=0), b'')
# Retrieve some results, freeing capacity at beginning
# of input buffer
out.append(bzd.decompress(b'', 2))
# Add more data that fits into input buffer after
# moving existing data to beginning
out.append(bzd.decompress(self.DATA[100:105], 15))
# Decompress rest of data
out.append(bzd.decompress(self.DATA[105:]))
self.assertEqual(b''.join(out), self.TEXT)
def test_decompressor_inputbuf_2(self):
# Test reusing input buffer by appending data at the
# end right away
bzd = BZ2Decompressor()
out = []
# Create input buffer and empty it
self.assertEqual(bzd.decompress(self.DATA[:200],
max_length=0), b'')
out.append(bzd.decompress(b''))
# Fill buffer with new data
out.append(bzd.decompress(self.DATA[200:280], 2))
# Append some more data, not enough to require resize
out.append(bzd.decompress(self.DATA[280:300], 2))
# Decompress rest of data
out.append(bzd.decompress(self.DATA[300:]))
self.assertEqual(b''.join(out), self.TEXT)
def test_decompressor_inputbuf_3(self):
# Test reusing input buffer after extending it
bzd = BZ2Decompressor()
out = []
# Create almost full input buffer
out.append(bzd.decompress(self.DATA[:200], 5))
# Add even more data to it, requiring resize
out.append(bzd.decompress(self.DATA[200:300], 5))
# Decompress rest of data
out.append(bzd.decompress(self.DATA[300:]))
self.assertEqual(b''.join(out), self.TEXT)
def test_failure(self):
bzd = BZ2Decompressor()
self.assertRaises(Exception, bzd.decompress, self.BAD_DATA * 30)
# Previously, a second call could crash due to internal inconsistency
self.assertRaises(Exception, bzd.decompress, self.BAD_DATA * 30)
class CompressDecompressTest(BaseTest):
def testCompress(self):
data = bz2.compress(self.TEXT)
self.assertEqual(self.decompress(data), self.TEXT)
def testCompressEmptyString(self):
text = bz2.compress(b'')
self.assertEqual(text, self.EMPTY_DATA)
def testDecompress(self):
text = bz2.decompress(self.DATA)
self.assertEqual(text, self.TEXT)
def testDecompressEmpty(self):
text = bz2.decompress(b"")
self.assertEqual(text, b"")
def testDecompressToEmptyString(self):
text = bz2.decompress(self.EMPTY_DATA)
self.assertEqual(text, b'')
def testDecompressIncomplete(self):
self.assertRaises(ValueError, bz2.decompress, self.DATA[:-10])
def testDecompressBadData(self):
self.assertRaises(OSError, bz2.decompress, self.BAD_DATA)
def testDecompressMultiStream(self):
text = bz2.decompress(self.DATA * 5)
self.assertEqual(text, self.TEXT * 5)
def testDecompressTrailingJunk(self):
text = bz2.decompress(self.DATA + self.BAD_DATA)
self.assertEqual(text, self.TEXT)
def testDecompressMultiStreamTrailingJunk(self):
text = bz2.decompress(self.DATA * 5 + self.BAD_DATA)
self.assertEqual(text, self.TEXT * 5)
class OpenTest(BaseTest):
"Test the open function."
def open(self, *args, **kwargs):
return bz2.open(*args, **kwargs)
def test_binary_modes(self):
for mode in ("wb", "xb"):
if mode == "xb":
unlink(self.filename)
with self.open(self.filename, mode) as f:
f.write(self.TEXT)
with open(self.filename, "rb") as f:
file_data = self.decompress(f.read())
self.assertEqual(file_data, self.TEXT)
with self.open(self.filename, "rb") as f:
self.assertEqual(f.read(), self.TEXT)
with self.open(self.filename, "ab") as f:
f.write(self.TEXT)
with open(self.filename, "rb") as f:
file_data = self.decompress(f.read())
self.assertEqual(file_data, self.TEXT * 2)
def test_implicit_binary_modes(self):
# Test implicit binary modes (no "b" or "t" in mode string).
for mode in ("w", "x"):
if mode == "x":
unlink(self.filename)
with self.open(self.filename, mode) as f:
f.write(self.TEXT)
with open(self.filename, "rb") as f:
file_data = self.decompress(f.read())
self.assertEqual(file_data, self.TEXT)
with self.open(self.filename, "r") as f:
self.assertEqual(f.read(), self.TEXT)
with self.open(self.filename, "a") as f:
f.write(self.TEXT)
with open(self.filename, "rb") as f:
file_data = self.decompress(f.read())
self.assertEqual(file_data, self.TEXT * 2)
def test_text_modes(self):
text = self.TEXT.decode("ascii")
text_native_eol = text.replace("\n", os.linesep)
for mode in ("wt", "xt"):
if mode == "xt":
unlink(self.filename)
with self.open(self.filename, mode) as f:
f.write(text)
with open(self.filename, "rb") as f:
file_data = self.decompress(f.read()).decode("ascii")
self.assertEqual(file_data, text_native_eol)
with self.open(self.filename, "rt") as f:
self.assertEqual(f.read(), text)
with self.open(self.filename, "at") as f:
f.write(text)
with open(self.filename, "rb") as f:
file_data = self.decompress(f.read()).decode("ascii")
self.assertEqual(file_data, text_native_eol * 2)
def test_x_mode(self):
for mode in ("x", "xb", "xt"):
unlink(self.filename)
with self.open(self.filename, mode) as f:
pass
with self.assertRaises(FileExistsError):
with self.open(self.filename, mode) as f:
pass
def test_fileobj(self):
with self.open(BytesIO(self.DATA), "r") as f:
self.assertEqual(f.read(), self.TEXT)
with self.open(BytesIO(self.DATA), "rb") as f:
self.assertEqual(f.read(), self.TEXT)
text = self.TEXT.decode("ascii")
with self.open(BytesIO(self.DATA), "rt") as f:
self.assertEqual(f.read(), text)
def test_bad_params(self):
# Test invalid parameter combinations.
self.assertRaises(ValueError,
self.open, self.filename, "wbt")
self.assertRaises(ValueError,
self.open, self.filename, "xbt")
self.assertRaises(ValueError,
self.open, self.filename, "rb", encoding="utf-8")
self.assertRaises(ValueError,
self.open, self.filename, "rb", errors="ignore")
self.assertRaises(ValueError,
self.open, self.filename, "rb", newline="\n")
def test_encoding(self):
# Test non-default encoding.
text = self.TEXT.decode("ascii")
text_native_eol = text.replace("\n", os.linesep)
with self.open(self.filename, "wt", encoding="utf-16-le") as f:
f.write(text)
with open(self.filename, "rb") as f:
file_data = self.decompress(f.read()).decode("utf-16-le")
self.assertEqual(file_data, text_native_eol)
with self.open(self.filename, "rt", encoding="utf-16-le") as f:
self.assertEqual(f.read(), text)
def test_encoding_error_handler(self):
# Test with non-default encoding error handler.
with self.open(self.filename, "wb") as f:
f.write(b"foo\xffbar")
with self.open(self.filename, "rt", encoding="ascii", errors="ignore") \
as f:
self.assertEqual(f.read(), "foobar")
def test_newline(self):
# Test with explicit newline (universal newline mode disabled).
text = self.TEXT.decode("ascii")
with self.open(self.filename, "wt", newline="\n") as f:
f.write(text)
with self.open(self.filename, "rt", newline="\r") as f:
self.assertEqual(f.readlines(), [text])
def test_main():
support.run_unittest(
BZ2FileTest,
BZ2CompressorTest,
BZ2DecompressorTest,
CompressDecompressTest,
OpenTest,
)
support.reap_children()
if __name__ == '__main__':
test_main()