1
# Copyright (C) 2006, 2008 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
"""A simple least-recently-used (LRU) cache."""
26
class _LRUNode(object):
27
"""This maintains the linked-list which is the lru internals."""
29
__slots__ = ('prev', 'next_key', 'key', 'value', 'cleanup', 'size')
31
def __init__(self, key, value, cleanup=None):
33
self.next_key = _null_key
36
self.cleanup = cleanup
37
# TODO: We could compute this 'on-the-fly' like we used to, and remove
38
# one pointer from this object, we just need to decide if it
39
# actually costs us much of anything in normal usage
46
prev_key = self.prev.key
47
return '%s(%r n:%r p:%r)' % (self.__class__.__name__, self.key,
48
self.next_key, prev_key)
50
def run_cleanup(self):
51
if self.cleanup is not None:
52
self.cleanup(self.key, self.value)
54
# Just make sure to break any refcycles, etc
58
class LRUCache(object):
59
"""A class which manages a cache of entries, removing unused ones."""
61
def __init__(self, max_cache=100, after_cleanup_count=None,
62
after_cleanup_size=symbol_versioning.DEPRECATED_PARAMETER):
63
if symbol_versioning.deprecated_passed(after_cleanup_size):
64
symbol_versioning.warn('LRUCache.__init__(after_cleanup_size) was'
65
' deprecated in 1.11. Use'
66
' after_cleanup_count instead.',
68
after_cleanup_count = after_cleanup_size
70
# The "HEAD" of the lru linked list
71
self._most_recently_used = None
72
# The "TAIL" of the lru linked list
73
self._least_recently_used = None
74
self._update_max_cache(max_cache, after_cleanup_count)
76
def __contains__(self, key):
77
return key in self._cache
79
def __getitem__(self, key):
82
# Inlined from _record_access to decrease the overhead of __getitem__
83
# We also have more knowledge about structure if __getitem__ is
84
# succeeding, then we know that self._most_recently_used must not be
86
mru = self._most_recently_used
88
# Nothing to do, this node is already at the head of the queue
90
# Remove this node from the old location
92
next_key = node.next_key
93
# benchmarking shows that the lookup of _null_key in globals is faster
94
# than the attribute lookup for (node is self._least_recently_used)
95
if next_key is _null_key:
96
# 'node' is the _least_recently_used, because it doesn't have a
97
# 'next' item. So move the current lru to the previous node.
98
self._least_recently_used = node_prev
100
node_next = cache[next_key]
101
node_next.prev = node_prev
102
node_prev.next_key = next_key
103
# Insert this node at the front of the list
104
node.next_key = mru.key
106
self._most_recently_used = node
111
return len(self._cache)
114
"""Walk the LRU list, only meant to be used in tests."""
115
node = self._most_recently_used
117
if node.prev is not None:
118
raise AssertionError('the _most_recently_used entry is not'
119
' supposed to have a previous entry'
121
while node is not None:
122
if node.next_key is _null_key:
123
if node is not self._least_recently_used:
124
raise AssertionError('only the last node should have'
125
' no next value: %s' % (node,))
128
node_next = self._cache[node.next_key]
129
if node_next.prev is not node:
130
raise AssertionError('inconsistency found, node.next.prev'
131
' != node: %s' % (node,))
132
if node.prev is None:
133
if node is not self._most_recently_used:
134
raise AssertionError('only the _most_recently_used should'
135
' not have a previous node: %s'
138
if node.prev.next_key != node.key:
139
raise AssertionError('inconsistency found, node.prev.next'
140
' != node: %s' % (node,))
144
def add(self, key, value, cleanup=None):
145
"""Add a new value to the cache.
147
Also, if the entry is ever removed from the cache, call
150
:param key: The key to store it under
151
:param value: The object to store
152
:param cleanup: None or a function taking (key, value) to indicate
153
'value' should be cleaned up.
156
raise ValueError('cannot use _null_key as a key')
157
if key in self._cache:
158
node = self._cache[key]
161
node.cleanup = cleanup
163
node = _LRUNode(key, value, cleanup=cleanup)
164
self._cache[key] = node
165
self._record_access(node)
167
if len(self._cache) > self._max_cache:
168
# Trigger the cleanup
171
def cache_size(self):
172
"""Get the number of entries we will cache."""
173
return self._max_cache
175
def get(self, key, default=None):
176
node = self._cache.get(key, None)
179
self._record_access(node)
183
"""Get the list of keys currently cached.
185
Note that values returned here may not be available by the time you
186
request them later. This is simply meant as a peak into the current
189
:return: An unordered list of keys that are currently cached.
191
return self._cache.keys()
194
"""Get the key:value pairs as a dict."""
195
return dict((k, n.value) for k, n in self._cache.iteritems())
198
"""Clear the cache until it shrinks to the requested size.
200
This does not completely wipe the cache, just makes sure it is under
201
the after_cleanup_count.
203
# Make sure the cache is shrunk to the correct size
204
while len(self._cache) > self._after_cleanup_count:
207
def __setitem__(self, key, value):
208
"""Add a value to the cache, there will be no cleanup function."""
209
self.add(key, value, cleanup=None)
211
def _record_access(self, node):
212
"""Record that key was accessed."""
213
# Move 'node' to the front of the queue
214
if self._most_recently_used is None:
215
self._most_recently_used = node
216
self._least_recently_used = node
218
elif node is self._most_recently_used:
219
# Nothing to do, this node is already at the head of the queue
221
# We've taken care of the tail pointer, remove the node, and insert it
224
if node is self._least_recently_used:
225
self._least_recently_used = node.prev
226
if node.prev is not None:
227
node.prev.next_key = node.next_key
228
if node.next_key is not _null_key:
229
node_next = self._cache[node.next_key]
230
node_next.prev = node.prev
232
node.next_key = self._most_recently_used.key
233
self._most_recently_used.prev = node
234
self._most_recently_used = node
237
def _remove_node(self, node):
238
if node is self._least_recently_used:
239
self._least_recently_used = node.prev
240
self._cache.pop(node.key)
241
# If we have removed all entries, remove the head pointer as well
242
if self._least_recently_used is None:
243
self._most_recently_used = None
245
# Now remove this node from the linked list
246
if node.prev is not None:
247
node.prev.next_key = node.next_key
248
if node.next_key is not _null_key:
249
node_next = self._cache[node.next_key]
250
node_next.prev = node.prev
251
# And remove this node's pointers
253
node.next_key = _null_key
255
def _remove_lru(self):
256
"""Remove one entry from the lru, and handle consequences.
258
If there are no more references to the lru, then this entry should be
259
removed from the cache.
261
self._remove_node(self._least_recently_used)
264
"""Clear out all of the cache."""
265
# Clean up in LRU order
269
def resize(self, max_cache, after_cleanup_count=None):
270
"""Change the number of entries that will be cached."""
271
self._update_max_cache(max_cache,
272
after_cleanup_count=after_cleanup_count)
274
def _update_max_cache(self, max_cache, after_cleanup_count=None):
275
self._max_cache = max_cache
276
if after_cleanup_count is None:
277
self._after_cleanup_count = self._max_cache * 8 / 10
279
self._after_cleanup_count = min(after_cleanup_count,
284
class LRUSizeCache(LRUCache):
285
"""An LRUCache that removes things based on the size of the values.
287
This differs in that it doesn't care how many actual items there are,
288
it just restricts the cache to be cleaned up after so much data is stored.
290
The size of items added will be computed using compute_size(value), which
291
defaults to len() if not supplied.
294
def __init__(self, max_size=1024*1024, after_cleanup_size=None,
296
"""Create a new LRUSizeCache.
298
:param max_size: The max number of bytes to store before we start
299
clearing out entries.
300
:param after_cleanup_size: After cleaning up, shrink everything to this
302
:param compute_size: A function to compute the size of the values. We
303
use a function here, so that you can pass 'len' if you are just
304
using simple strings, or a more complex function if you are using
305
something like a list of strings, or even a custom object.
306
The function should take the form "compute_size(value) => integer".
307
If not supplied, it defaults to 'len()'
310
self._compute_size = compute_size
311
if compute_size is None:
312
self._compute_size = len
313
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
314
LRUCache.__init__(self, max_cache=max(int(max_size/512), 1))
316
def add(self, key, value, cleanup=None):
317
"""Add a new value to the cache.
319
Also, if the entry is ever removed from the cache, call
322
:param key: The key to store it under
323
:param value: The object to store
324
:param cleanup: None or a function taking (key, value) to indicate
325
'value' should be cleaned up.
328
raise ValueError('cannot use _null_key as a key')
329
node = self._cache.get(key, None)
330
value_len = self._compute_size(value)
331
if value_len >= self._after_cleanup_size:
332
# The new value is 'too big to fit', as it would fill up/overflow
333
# the cache all by itself
334
trace.mutter('Adding the key %r to an LRUSizeCache failed.'
335
' value %d is too big to fit in a the cache'
336
' with size %d %d', key, value_len,
337
self._after_cleanup_size, self._max_size)
339
# We won't be replacing the old node, so just remove it
340
self._remove_node(node)
341
if cleanup is not None:
345
node = _LRUNode(key, value, cleanup=cleanup)
346
self._cache[key] = node
348
self._value_size -= node.size
349
node.size = value_len
350
self._value_size += value_len
351
self._record_access(node)
353
if self._value_size > self._max_size:
358
"""Clear the cache until it shrinks to the requested size.
360
This does not completely wipe the cache, just makes sure it is under
361
the after_cleanup_size.
363
# Make sure the cache is shrunk to the correct size
364
while self._value_size > self._after_cleanup_size:
367
def _remove_node(self, node):
368
self._value_size -= node.size
369
LRUCache._remove_node(self, node)
371
def resize(self, max_size, after_cleanup_size=None):
372
"""Change the number of bytes that will be cached."""
373
self._update_max_size(max_size, after_cleanup_size=after_cleanup_size)
374
max_cache = max(int(max_size/512), 1)
375
self._update_max_cache(max_cache)
377
def _update_max_size(self, max_size, after_cleanup_size=None):
378
self._max_size = max_size
379
if after_cleanup_size is None:
380
self._after_cleanup_size = self._max_size * 8 / 10
382
self._after_cleanup_size = min(after_cleanup_size, self._max_size)
added
removed
bzrlib/lru_cache.py
