#
# this is a rather strict implementation of a bit vector class
# it is accessed the same way as an array of python-ints, except
# the value must be 0 or 1
#
import sys; rprt = sys.stderr.write #for debugging
error = 'bitvec.error'
def _check_value(value):
if type(value) != type(0) or not 0 <= value < 2:
raise error, 'bitvec() items must have int value 0 or 1'
import math
def _compute_len(param):
mant, l = math.frexp(float(param))
bitmask = 1L << l
if bitmask <= param:
raise 'FATAL', '(param, l) = ' + `param, l`
while l:
bitmask = bitmask >> 1
if param & bitmask:
break
l = l - 1
return l
def _check_key(len, key):
if type(key) != type(0):
raise TypeError, 'sequence subscript not int'
if key < 0:
key = key + len
if not 0 <= key < len:
raise IndexError, 'list index out of range'
return key
def _check_slice(len, i, j):
#the type is ok, Python already checked that
i, j = max(i, 0), min(len, j)
if i > j:
i = j
return i, j
class BitVec:
def __init__(self, *params):
self._data = 0L
self._len = 0
if not len(params):
pass
elif len(params) == 1:
param, = params
if type(param) == type([]):
value = 0L
bit_mask = 1L
for item in param:
# strict check
#_check_value(item)
if item:
value = value | bit_mask
bit_mask = bit_mask << 1
self._data = value
self._len = len(param)
elif type(param) == type(0L):
if param < 0:
raise error, 'bitvec() can\'t handle negative longs'
self._data = param
self._len = _compute_len(param)
else:
raise error, 'bitvec() requires array or long parameter'
elif len(params) == 2:
param, length = params
if type(param) == type(0L):
if param < 0:
raise error, \
'can\'t handle negative longs'
self._data = param
if type(length) != type(0):
raise error, 'bitvec()\'s 2nd parameter must be int'
computed_length = _compute_len(param)
if computed_length > length:
print 'warning: bitvec() value is longer than the length indicates, truncating value'
self._data = self._data & \
((1L << length) - 1)
self._len = length
else:
raise error, 'bitvec() requires array or long parameter'
else:
raise error, 'bitvec() requires 0 -- 2 parameter(s)'
def append(self, item):
#_check_value(item)
#self[self._len:self._len] = [item]
self[self._len:self._len] = \
BitVec(long(not not item), 1)
def count(self, value):
#_check_value(value)
if value:
data = self._data
else:
data = (~self)._data
count = 0
while data:
data, count = data >> 1, count + (data & 1 != 0)
return count
def index(self, value):
#_check_value(value):
if value:
data = self._data
else:
data = (~self)._data
index = 0
if not data:
raise ValueError, 'list.index(x): x not in list'
while not (data & 1):
data, index = data >> 1, index + 1
return index
def insert(self, index, item):
#_check_value(item)
#self[index:index] = [item]
self[index:index] = BitVec(long(not not item), 1)
def remove(self, value):
del self[self.index(value)]
def reverse(self):
#ouch, this one is expensive!
#for i in self._len>>1: self[i], self[l-i] = self[l-i], self[i]
data, result = self._data, 0L
for i in range(self._len):
if not data:
result = result << (self._len - i)
break
result, data = (result << 1) | (data & 1), data >> 1
self._data = result
def sort(self):
c = self.count(1)
self._data = ((1L << c) - 1) << (self._len - c)
def copy(self):
return BitVec(self._data, self._len)
def seq(self):
result = []
for i in self:
result.append(i)
return result
def __repr__(self):
##rprt('<bitvec class instance object>.' + '__repr__()\n')
return 'bitvec' + `self._data, self._len`
def __cmp__(self, other, *rest):
#rprt(`self`+'.__cmp__'+`(other, ) + rest`+'\n')
if type(other) != type(self):
other = apply(bitvec, (other, ) + rest)
#expensive solution... recursive binary, with slicing
length = self._len
if length == 0 or other._len == 0:
return cmp(length, other._len)
if length != other._len:
min_length = min(length, other._len)
return cmp(self[:min_length], other[:min_length]) or \
cmp(self[min_length:], other[min_length:])
#the lengths are the same now...
if self._data == other._data:
return 0
if length == 1:
return cmp(self[0], other[0])
else:
length = length >> 1
return cmp(self[:length], other[:length]) or \
cmp(self[length:], other[length:])
def __len__(self):
#rprt(`self`+'.__len__()\n')
return self._len
def __getitem__(self, key):
#rprt(`self`+'.__getitem__('+`key`+')\n')
key = _check_key(self._len, key)
return self._data & (1L << key) != 0
def __setitem__(self, key, value):
#rprt(`self`+'.__setitem__'+`key, value`+'\n')
key = _check_key(self._len, key)
#_check_value(value)
if value:
self._data = self._data | (1L << key)
else:
self._data = self._data & ~(1L << key)
def __delitem__(self, key):
#rprt(`self`+'.__delitem__('+`key`+')\n')
key = _check_key(self._len, key)
#el cheapo solution...
self._data = self[:key]._data | self[key+1:]._data >> key
self._len = self._len - 1
def __getslice__(self, i, j):
#rprt(`self`+'.__getslice__'+`i, j`+'\n')
i, j = _check_slice(self._len, i, j)
if i >= j:
return BitVec(0L, 0)
if i:
ndata = self._data >> i
else:
ndata = self._data
nlength = j - i
if j != self._len:
#we'll have to invent faster variants here
#e.g. mod_2exp
ndata = ndata & ((1L << nlength) - 1)
return BitVec(ndata, nlength)
def __setslice__(self, i, j, sequence, *rest):
#rprt(`self`+'.__setslice__'+`(i, j, sequence) + rest`+'\n')
i, j = _check_slice(self._len, i, j)
if type(sequence) != type(self):
sequence = apply(bitvec, (sequence, ) + rest)
#sequence is now of our own type
ls_part = self[:i]
ms_part = self[j:]
self._data = ls_part._data | \
((sequence._data | \
(ms_part._data << sequence._len)) << ls_part._len)
self._len = self._len - j + i + sequence._len
def __delslice__(self, i, j):
#rprt(`self`+'.__delslice__'+`i, j`+'\n')
i, j = _check_slice(self._len, i, j)
if i == 0 and j == self._len:
self._data, self._len = 0L, 0
elif i < j:
self._data = self[:i]._data | (self[j:]._data >> i)
self._len = self._len - j + i
def __add__(self, other):
#rprt(`self`+'.__add__('+`other`+')\n')
retval = self.copy()
retval[self._len:self._len] = other
return retval
def __mul__(self, multiplier):
#rprt(`self`+'.__mul__('+`multiplier`+')\n')
if type(multiplier) != type(0):
raise TypeError, 'sequence subscript not int'
if multiplier <= 0:
return BitVec(0L, 0)
elif multiplier == 1:
return self.copy()
#handle special cases all 0 or all 1...
if self._data == 0L:
return BitVec(0L, self._len * multiplier)
elif (~self)._data == 0L:
return ~BitVec(0L, self._len * multiplier)
#otherwise el cheapo again...
retval = BitVec(0L, 0)
while multiplier:
retval, multiplier = retval + self, multiplier - 1
return retval
def __and__(self, otherseq, *rest):
#rprt(`self`+'.__and__'+`(otherseq, ) + rest`+'\n')
if type(otherseq) != type(self):
otherseq = apply(bitvec, (otherseq, ) + rest)
#sequence is now of our own type
return BitVec(self._data & otherseq._data, \
min(self._len, otherseq._len))
def __xor__(self, otherseq, *rest):
#rprt(`self`+'.__xor__'+`(otherseq, ) + rest`+'\n')
if type(otherseq) != type(self):
otherseq = apply(bitvec, (otherseq, ) + rest)
#sequence is now of our own type
return BitVec(self._data ^ otherseq._data, \
max(self._len, otherseq._len))
def __or__(self, otherseq, *rest):
#rprt(`self`+'.__or__'+`(otherseq, ) + rest`+'\n')
if type(otherseq) != type(self):
otherseq = apply(bitvec, (otherseq, ) + rest)
#sequence is now of our own type
return BitVec(self._data | otherseq._data, \
max(self._len, otherseq._len))
def __invert__(self):
#rprt(`self`+'.__invert__()\n')
return BitVec(~self._data & ((1L << self._len) - 1), \
self._len)
def __coerce__(self, otherseq, *rest):
#needed for *some* of the arithmetic operations
#rprt(`self`+'.__coerce__'+`(otherseq, ) + rest`+'\n')
if type(otherseq) != type(self):
otherseq = apply(bitvec, (otherseq, ) + rest)
return self, otherseq
def __int__(self):
return int(self._data)
def __long__(self):
return long(self._data)
def __float__(self):
return float(self._data)
bitvec = BitVec
|