1"""Manage shelves of pickled objects. 2 3A "shelf" is a persistent, dictionary-like object. The difference 4with dbm databases is that the values (not the keys!) in a shelf can 5be essentially arbitrary Python objects -- anything that the "pickle" 6module can handle. This includes most class instances, recursive data 7types, and objects containing lots of shared sub-objects. The keys 8are ordinary strings. 9 10To summarize the interface (key is a string, data is an arbitrary 11object): 12 13 import shelve 14 d = shelve.open(filename) # open, with (g)dbm filename -- no suffix 15 16 d[key] = data # store data at key (overwrites old data if 17 # using an existing key) 18 data = d[key] # retrieve a COPY of the data at key (raise 19 # KeyError if no such key) -- NOTE that this 20 # access returns a *copy* of the entry! 21 del d[key] # delete data stored at key (raises KeyError 22 # if no such key) 23 flag = d.has_key(key) # true if the key exists; same as "key in d" 24 list = d.keys() # a list of all existing keys (slow!) 25 26 d.close() # close it 27 28Dependent on the implementation, closing a persistent dictionary may 29or may not be necessary to flush changes to disk. 30 31Normally, d[key] returns a COPY of the entry. This needs care when 32mutable entries are mutated: for example, if d[key] is a list, 33 d[key].append(anitem) 34does NOT modify the entry d[key] itself, as stored in the persistent 35mapping -- it only modifies the copy, which is then immediately 36discarded, so that the append has NO effect whatsoever. To append an 37item to d[key] in a way that will affect the persistent mapping, use: 38 data = d[key] 39 data.append(anitem) 40 d[key] = data 41 42To avoid the problem with mutable entries, you may pass the keyword 43argument writeback=True in the call to shelve.open. When you use: 44 d = shelve.open(filename, writeback=True) 45then d keeps a cache of all entries you access, and writes them all back 46to the persistent mapping when you call d.close(). This ensures that 47such usage as d[key].append(anitem) works as intended. 48 49However, using keyword argument writeback=True may consume vast amount 50of memory for the cache, and it may make d.close() very slow, if you 51access many of d's entries after opening it in this way: d has no way to 52check which of the entries you access are mutable and/or which ones you 53actually mutate, so it must cache, and write back at close, all of the 54entries that you access. You can call d.sync() to write back all the 55entries in the cache, and empty the cache (d.sync() also synchronizes 56the persistent dictionary on disk, if feasible). 57""" 58 59# Try using cPickle and cStringIO if available. 60 61try: 62 from cPickle import Pickler, Unpickler 63except ImportError: 64 from pickle import Pickler, Unpickler 65 66try: 67 from cStringIO import StringIO 68except ImportError: 69 from StringIO import StringIO 70 71import UserDict 72 73__all__ = ["Shelf","BsdDbShelf","DbfilenameShelf","open"] 74 75class _ClosedDict(UserDict.DictMixin): 76 'Marker for a closed dict. Access attempts raise a ValueError.' 77 78 def closed(self, *args): 79 raise ValueError('invalid operation on closed shelf') 80 __getitem__ = __setitem__ = __delitem__ = keys = closed 81 82 def __repr__(self): 83 return '<Closed Dictionary>' 84 85class Shelf(UserDict.DictMixin): 86 """Base class for shelf implementations. 87 88 This is initialized with a dictionary-like object. 89 See the module's __doc__ string for an overview of the interface. 90 """ 91 92 def __init__(self, dict, protocol=None, writeback=False): 93 self.dict = dict 94 if protocol is None: 95 protocol = 0 96 self._protocol = protocol 97 self.writeback = writeback 98 self.cache = {} 99 100 def keys(self): 101 return self.dict.keys() 102 103 def __len__(self): 104 return len(self.dict) 105 106 def has_key(self, key): 107 return key in self.dict 108 109 def __contains__(self, key): 110 return key in self.dict 111 112 def get(self, key, default=None): 113 if key in self.dict: 114 return self[key] 115 return default 116 117 def __getitem__(self, key): 118 try: 119 value = self.cache[key] 120 except KeyError: 121 f = StringIO(self.dict[key]) 122 value = Unpickler(f).load() 123 if self.writeback: 124 self.cache[key] = value 125 return value 126 127 def __setitem__(self, key, value): 128 if self.writeback: 129 self.cache[key] = value 130 f = StringIO() 131 p = Pickler(f, self._protocol) 132 p.dump(value) 133 self.dict[key] = f.getvalue() 134 135 def __delitem__(self, key): 136 del self.dict[key] 137 try: 138 del self.cache[key] 139 except KeyError: 140 pass 141 142 def close(self): 143 if self.dict is None: 144 return 145 try: 146 self.sync() 147 try: 148 self.dict.close() 149 except AttributeError: 150 pass 151 finally: 152 # Catch errors that may happen when close is called from __del__ 153 # because CPython is in interpreter shutdown. 154 try: 155 self.dict = _ClosedDict() 156 except: 157 self.dict = None 158 159 def __del__(self): 160 if not hasattr(self, 'writeback'): 161 # __init__ didn't succeed, so don't bother closing 162 return 163 self.close() 164 165 def sync(self): 166 if self.writeback and self.cache: 167 self.writeback = False 168 for key, entry in self.cache.iteritems(): 169 self[key] = entry 170 self.writeback = True 171 self.cache = {} 172 if hasattr(self.dict, 'sync'): 173 self.dict.sync() 174 175 176class BsdDbShelf(Shelf): 177 """Shelf implementation using the "BSD" db interface. 178 179 This adds methods first(), next(), previous(), last() and 180 set_location() that have no counterpart in [g]dbm databases. 181 182 The actual database must be opened using one of the "bsddb" 183 modules "open" routines (i.e. bsddb.hashopen, bsddb.btopen or 184 bsddb.rnopen) and passed to the constructor. 185 186 See the module's __doc__ string for an overview of the interface. 187 """ 188 189 def __init__(self, dict, protocol=None, writeback=False): 190 Shelf.__init__(self, dict, protocol, writeback) 191 192 def set_location(self, key): 193 (key, value) = self.dict.set_location(key) 194 f = StringIO(value) 195 return (key, Unpickler(f).load()) 196 197 def next(self): 198 (key, value) = self.dict.next() 199 f = StringIO(value) 200 return (key, Unpickler(f).load()) 201 202 def previous(self): 203 (key, value) = self.dict.previous() 204 f = StringIO(value) 205 return (key, Unpickler(f).load()) 206 207 def first(self): 208 (key, value) = self.dict.first() 209 f = StringIO(value) 210 return (key, Unpickler(f).load()) 211 212 def last(self): 213 (key, value) = self.dict.last() 214 f = StringIO(value) 215 return (key, Unpickler(f).load()) 216 217 218class DbfilenameShelf(Shelf): 219 """Shelf implementation using the "anydbm" generic dbm interface. 220 221 This is initialized with the filename for the dbm database. 222 See the module's __doc__ string for an overview of the interface. 223 """ 224 225 def __init__(self, filename, flag='c', protocol=None, writeback=False): 226 import anydbm 227 Shelf.__init__(self, anydbm.open(filename, flag), protocol, writeback) 228 229 230def open(filename, flag='c', protocol=None, writeback=False): 231 """Open a persistent dictionary for reading and writing. 232 233 The filename parameter is the base filename for the underlying 234 database. As a side-effect, an extension may be added to the 235 filename and more than one file may be created. The optional flag 236 parameter has the same interpretation as the flag parameter of 237 anydbm.open(). The optional protocol parameter specifies the 238 version of the pickle protocol (0, 1, or 2). 239 240 See the module's __doc__ string for an overview of the interface. 241 """ 242 243 return DbfilenameShelf(filename, flag, protocol, writeback) 244