mirror of
https://github.com/python/cpython.git
synced 2025-10-06 23:21:06 +00:00
Issue #17812: Fixed quadratic complexity of base64.b32encode().
Optimize base64.b32encode() and base64.b32decode() (speed up to 3x).
This commit is contained in:
commit
87aa7dc211
2 changed files with 51 additions and 77 deletions
125
Lib/base64.py
125
Lib/base64.py
|
@ -138,21 +138,10 @@ def urlsafe_b64decode(s):
|
||||||
|
|
||||||
|
|
||||||
# Base32 encoding/decoding must be done in Python
|
# Base32 encoding/decoding must be done in Python
|
||||||
_b32alphabet = {
|
_b32alphabet = b'ABCDEFGHIJKLMNOPQRSTUVWXYZ234567'
|
||||||
0: b'A', 9: b'J', 18: b'S', 27: b'3',
|
_b32tab = [bytes([i]) for i in _b32alphabet]
|
||||||
1: b'B', 10: b'K', 19: b'T', 28: b'4',
|
_b32tab2 = [a + b for a in _b32tab for b in _b32tab]
|
||||||
2: b'C', 11: b'L', 20: b'U', 29: b'5',
|
_b32rev = {v: k for k, v in enumerate(_b32alphabet)}
|
||||||
3: b'D', 12: b'M', 21: b'V', 30: b'6',
|
|
||||||
4: b'E', 13: b'N', 22: b'W', 31: b'7',
|
|
||||||
5: b'F', 14: b'O', 23: b'X',
|
|
||||||
6: b'G', 15: b'P', 24: b'Y',
|
|
||||||
7: b'H', 16: b'Q', 25: b'Z',
|
|
||||||
8: b'I', 17: b'R', 26: b'2',
|
|
||||||
}
|
|
||||||
|
|
||||||
_b32tab = [v[0] for k, v in sorted(_b32alphabet.items())]
|
|
||||||
_b32rev = dict([(v[0], k) for k, v in _b32alphabet.items()])
|
|
||||||
|
|
||||||
|
|
||||||
def b32encode(s):
|
def b32encode(s):
|
||||||
"""Encode a byte string using Base32.
|
"""Encode a byte string using Base32.
|
||||||
|
@ -161,41 +150,30 @@ def b32encode(s):
|
||||||
"""
|
"""
|
||||||
if not isinstance(s, bytes_types):
|
if not isinstance(s, bytes_types):
|
||||||
raise TypeError("expected bytes, not %s" % s.__class__.__name__)
|
raise TypeError("expected bytes, not %s" % s.__class__.__name__)
|
||||||
quanta, leftover = divmod(len(s), 5)
|
leftover = len(s) % 5
|
||||||
# Pad the last quantum with zero bits if necessary
|
# Pad the last quantum with zero bits if necessary
|
||||||
if leftover:
|
if leftover:
|
||||||
s = s + bytes(5 - leftover) # Don't use += !
|
s = s + bytes(5 - leftover) # Don't use += !
|
||||||
quanta += 1
|
encoded = bytearray()
|
||||||
encoded = bytes()
|
from_bytes = int.from_bytes
|
||||||
for i in range(quanta):
|
b32tab2 = _b32tab2
|
||||||
# c1 and c2 are 16 bits wide, c3 is 8 bits wide. The intent of this
|
for i in range(0, len(s), 5):
|
||||||
# code is to process the 40 bits in units of 5 bits. So we take the 1
|
c = from_bytes(s[i: i + 5], 'big')
|
||||||
# leftover bit of c1 and tack it onto c2. Then we take the 2 leftover
|
encoded += (b32tab2[c >> 30] + # bits 1 - 10
|
||||||
# bits of c2 and tack them onto c3. The shifts and masks are intended
|
b32tab2[(c >> 20) & 0x3ff] + # bits 11 - 20
|
||||||
# to give us values of exactly 5 bits in width.
|
b32tab2[(c >> 10) & 0x3ff] + # bits 21 - 30
|
||||||
c1, c2, c3 = struct.unpack('!HHB', s[i*5:(i+1)*5])
|
b32tab2[c & 0x3ff] # bits 31 - 40
|
||||||
c2 += (c1 & 1) << 16 # 17 bits wide
|
)
|
||||||
c3 += (c2 & 3) << 8 # 10 bits wide
|
|
||||||
encoded += bytes([_b32tab[c1 >> 11], # bits 1 - 5
|
|
||||||
_b32tab[(c1 >> 6) & 0x1f], # bits 6 - 10
|
|
||||||
_b32tab[(c1 >> 1) & 0x1f], # bits 11 - 15
|
|
||||||
_b32tab[c2 >> 12], # bits 16 - 20 (1 - 5)
|
|
||||||
_b32tab[(c2 >> 7) & 0x1f], # bits 21 - 25 (6 - 10)
|
|
||||||
_b32tab[(c2 >> 2) & 0x1f], # bits 26 - 30 (11 - 15)
|
|
||||||
_b32tab[c3 >> 5], # bits 31 - 35 (1 - 5)
|
|
||||||
_b32tab[c3 & 0x1f], # bits 36 - 40 (1 - 5)
|
|
||||||
])
|
|
||||||
# Adjust for any leftover partial quanta
|
# Adjust for any leftover partial quanta
|
||||||
if leftover == 1:
|
if leftover == 1:
|
||||||
return encoded[:-6] + b'======'
|
encoded[-6:] = b'======'
|
||||||
elif leftover == 2:
|
elif leftover == 2:
|
||||||
return encoded[:-4] + b'===='
|
encoded[-4:] = b'===='
|
||||||
elif leftover == 3:
|
elif leftover == 3:
|
||||||
return encoded[:-3] + b'==='
|
encoded[-3:] = b'==='
|
||||||
elif leftover == 4:
|
elif leftover == 4:
|
||||||
return encoded[:-1] + b'='
|
encoded[-1:] = b'='
|
||||||
return encoded
|
return bytes(encoded)
|
||||||
|
|
||||||
|
|
||||||
def b32decode(s, casefold=False, map01=None):
|
def b32decode(s, casefold=False, map01=None):
|
||||||
"""Decode a Base32 encoded byte string.
|
"""Decode a Base32 encoded byte string.
|
||||||
|
@ -217,8 +195,7 @@ def b32decode(s, casefold=False, map01=None):
|
||||||
characters present in the input.
|
characters present in the input.
|
||||||
"""
|
"""
|
||||||
s = _bytes_from_decode_data(s)
|
s = _bytes_from_decode_data(s)
|
||||||
quanta, leftover = divmod(len(s), 8)
|
if len(s) % 8:
|
||||||
if leftover:
|
|
||||||
raise binascii.Error('Incorrect padding')
|
raise binascii.Error('Incorrect padding')
|
||||||
# Handle section 2.4 zero and one mapping. The flag map01 will be either
|
# Handle section 2.4 zero and one mapping. The flag map01 will be either
|
||||||
# False, or the character to map the digit 1 (one) to. It should be
|
# False, or the character to map the digit 1 (one) to. It should be
|
||||||
|
@ -232,42 +209,36 @@ def b32decode(s, casefold=False, map01=None):
|
||||||
# Strip off pad characters from the right. We need to count the pad
|
# Strip off pad characters from the right. We need to count the pad
|
||||||
# characters because this will tell us how many null bytes to remove from
|
# characters because this will tell us how many null bytes to remove from
|
||||||
# the end of the decoded string.
|
# the end of the decoded string.
|
||||||
padchars = 0
|
l = len(s)
|
||||||
mo = re.search(b'(?P<pad>[=]*)$', s)
|
s = s.rstrip(b'=')
|
||||||
if mo:
|
padchars = l - len(s)
|
||||||
padchars = len(mo.group('pad'))
|
|
||||||
if padchars > 0:
|
|
||||||
s = s[:-padchars]
|
|
||||||
# Now decode the full quanta
|
# Now decode the full quanta
|
||||||
parts = []
|
decoded = bytearray()
|
||||||
acc = 0
|
b32rev = _b32rev
|
||||||
shift = 35
|
for i in range(0, len(s), 8):
|
||||||
for c in s:
|
quanta = s[i: i + 8]
|
||||||
val = _b32rev.get(c)
|
acc = 0
|
||||||
if val is None:
|
try:
|
||||||
|
for c in quanta:
|
||||||
|
acc = (acc << 5) + b32rev[c]
|
||||||
|
except KeyError:
|
||||||
raise TypeError('Non-base32 digit found')
|
raise TypeError('Non-base32 digit found')
|
||||||
acc += _b32rev[c] << shift
|
decoded += acc.to_bytes(5, 'big')
|
||||||
shift -= 5
|
|
||||||
if shift < 0:
|
|
||||||
parts.append(binascii.unhexlify(bytes('%010x' % acc, "ascii")))
|
|
||||||
acc = 0
|
|
||||||
shift = 35
|
|
||||||
# Process the last, partial quanta
|
# Process the last, partial quanta
|
||||||
last = binascii.unhexlify(bytes('%010x' % acc, "ascii"))
|
if padchars:
|
||||||
if padchars == 0:
|
acc <<= 5 * padchars
|
||||||
last = b'' # No characters
|
last = acc.to_bytes(5, 'big')
|
||||||
elif padchars == 1:
|
if padchars == 1:
|
||||||
last = last[:-1]
|
decoded[-5:] = last[:-1]
|
||||||
elif padchars == 3:
|
elif padchars == 3:
|
||||||
last = last[:-2]
|
decoded[-5:] = last[:-2]
|
||||||
elif padchars == 4:
|
elif padchars == 4:
|
||||||
last = last[:-3]
|
decoded[-5:] = last[:-3]
|
||||||
elif padchars == 6:
|
elif padchars == 6:
|
||||||
last = last[:-4]
|
decoded[-5:] = last[:-4]
|
||||||
else:
|
else:
|
||||||
raise binascii.Error('Incorrect padding')
|
raise binascii.Error('Incorrect padding')
|
||||||
parts.append(last)
|
return bytes(decoded)
|
||||||
return b''.join(parts)
|
|
||||||
|
|
||||||
|
|
||||||
|
|
||||||
|
|
|
@ -10,6 +10,9 @@ What's New in Python 3.4.0 Alpha 1?
|
||||||
Core and Builtins
|
Core and Builtins
|
||||||
-----------------
|
-----------------
|
||||||
|
|
||||||
|
- Issue #17812: Fixed quadratic complexity of base64.b32encode().
|
||||||
|
Optimize base64.b32encode() and base64.b32decode() (speed up to 3x).
|
||||||
|
|
||||||
- Issue #17937: Try harder to collect cyclic garbage at shutdown.
|
- Issue #17937: Try harder to collect cyclic garbage at shutdown.
|
||||||
|
|
||||||
- Issue #12370: Prevent class bodies from interfering with the __class__
|
- Issue #12370: Prevent class bodies from interfering with the __class__
|
||||||
|
|
Loading…
Add table
Add a link
Reference in a new issue