import array
import collections
import itertools
import pprint
# Placeholder constants
FREE = -1
DUMMY = -2
UNUSED = ':UNUSED:' # In C, be sure to use a unique object
def round_upto_powtwo(n):
'Round-up to the next power of two'
result = 1
while result <= n:
result <<= 1
return result
class Dict(collections.MutableMapping):
'''Space efficient dictionary with fast iteration
and cheap resizes (minimum data movement)
'''
@staticmethod
def _gen_probes(hashvalue, n):
'Same sequence of probes used in the current dictionary design'
PERTURB_SHIFT = 5
if hashvalue < 0:
hashvalue = int(2 ** 64 - hashvalue)
i = hashvalue & (n-1)
yield int(i)
perturb = hashvalue
while True:
i = ((i << 2) + i + perturb + 1)
yield int(i & (n-1))
perturb >>= PERTURB_SHIFT
def _lookup(self, key, hashvalue):
'Same lookup logic as currently used in real dicts'
assert len(self.keylist) < len(self.indices) # At least one open slot
freeslot = None
for i in self._gen_probes(hashvalue, len(self.indices)):
if self.indices[i] is FREE:
if freeslot is not None:
i = freeslot
return (False, i)
elif self.indices[i] is DUMMY:
freeslot = i
else:
index = self.indices[i]
if (self.keylist[index] is key or
self.hashlist[index] == hashvalue
and self.keylist[index] == key):
return (True, i)
def _next_open_index(self):
'''New logic: Store hash/key/value entries in a dense list.
If an unused entry is available, re-use it; otherwise,
append a new entry to the end of the list. '''
if self.size == len(self.keylist):
index = len(self.keylist)
self.hashlist.append(None)
self.keylist.append(None)
self.valuelist.append(None)
return index
# Least common case. Can make fast in C with a pointer chain.
return self.keylist.index(UNUSED)
def _make_index(self, n):
'New sequence of indices using the smallest possible datatype'
typecode = 'b' if n <= 128 else 'h' if n <= 65536 else 'l'
return array.array(typecode, itertools.repeat(FREE, n))
def _resize(self, n):
'''Reindex the existing hash/key/value entries
and compact the entry list.
No calls are made to hash() or __eq__().
No hash/key/value entries get moved except
to fill-in an UNUSED slot in the entry sequence.
'''
n = round_upto_powtwo(n)
self.indices = self._make_index(n)
for index, hashvalue in enumerate(self.hashlist):
while hashvalue is UNUSED:
hashvalue = self.hashlist.pop()
key = self.keylist.pop()
value = self.valuelist.pop()
if index >= len(self.hashlist):
return
self.hashlist[index] = hashvalue
self.keylist[index] = key
self.valuelist[index] = value
for i in Dict._gen_probes(hashvalue, n):
if self.indices[i] is FREE:
break
self.indices[i] = index
def __init__(self, *args, **kwds):
self.indices = self._make_index(8)
self.hashlist = []
self.keylist = []
self.valuelist = []
self.size = 0
self.update(*args, **kwds)
def __setitem__(self, key, value):
hashvalue = hash(key)
found, i = self._lookup(key, hashvalue)
if found:
index = self.indices[i]
self.valuelist[index] = value
else:
index = self._next_open_index()
self.indices[i] = index
self.hashlist[index] = hashvalue
self.keylist[index] = key
self.valuelist[index] = value
self.size += 1
if self.indices.count(FREE) * 3 < len(self.indices):
self._resize(2 * len(self))
def __getitem__(self, key):
hashvalue = hash(key)
found, i = self._lookup(key, hashvalue)
if found:
index = self.indices[i]
return self.valuelist[index]
else:
raise KeyError(key)
def __delitem__(self, key, hashvalue=None):
if hashvalue is None:
hashvalue = hash(key)
found, i = self._lookup(key, hashvalue)
if found:
index = self.indices[i]
self.indices[i] = DUMMY
self.hashlist[index] = UNUSED
self.keylist[index] = UNUSED
self.valuelist[index] = UNUSED
self.size -= 1
else:
raise KeyError(key)
def __iter__(self):
for key in self.keylist:
if key is not UNUSED:
yield key
def __len__(self):
return self.size
def __repr__(self):
return 'Dict(%r)' % self.items()
###### Diagnostic methods. Not part of the API ##############
def show_structure(self):
print '=' * 50
print 'Dict:', self
print 'Size in bytes:', self.memory()
print 'Indices:', self.indices
entries = itertools.izip_longest(
xrange(len(self.hashlist)),
self.hashlist, self.keylist, self.valuelist,
fillvalue = ':ERROR:')
pprint.pprint(list(entries))
print '-' * 50
def memory(self):
'Size in bytes assuming a 64-bit build'
return (len(self.indices) * self.indices.itemsize
+ len(self.hashlist) * 8 * 3)
if __name__ == '__main__':
d = Dict([('timmy', 'red'), ('barry', 'green'), ('guido', 'blue')])
d.show_structure()
Diff to Previous Revision
--- revision 3 2012-12-10 04:16:25
+++ revision 4 2012-12-10 04:39:57
@@ -26,12 +26,12 @@
PERTURB_SHIFT = 5
if hashvalue < 0:
hashvalue = int(2 ** 64 - hashvalue)
- i = hashvalue % n # When n is a power-of-two, binary-and is faster
+ i = hashvalue & (n-1)
yield int(i)
perturb = hashvalue
while True:
i = ((i << 2) + i + perturb + 1)
- yield int(i % n)
+ yield int(i & (n-1))
perturb >>= PERTURB_SHIFT
def _lookup(self, key, hashvalue):
@@ -79,6 +79,7 @@
to fill-in an UNUSED slot in the entry sequence.
'''
+ n = round_upto_powtwo(n)
self.indices = self._make_index(n)
for index, hashvalue in enumerate(self.hashlist):
while hashvalue is UNUSED:
@@ -117,7 +118,7 @@
self.valuelist[index] = value
self.size += 1
if self.indices.count(FREE) * 3 < len(self.indices):
- self._resize(round_upto_powtwo(2 * len(self)))
+ self._resize(2 * len(self))
def __getitem__(self, key):
hashvalue = hash(key)
@@ -161,7 +162,8 @@
print 'Size in bytes:', self.memory()
print 'Indices:', self.indices
entries = itertools.izip_longest(
- xrange(len(self.hashlist)), self.hashlist, self.keylist, self.valuelist,
+ xrange(len(self.hashlist)),
+ self.hashlist, self.keylist, self.valuelist,
fillvalue = ':ERROR:')
pprint.pprint(list(entries))
print '-' * 50
