| Guido van Rossum | a48061a | 1995-01-10 00:31:14 +0000 | [diff] [blame] | 1 | """A slow but simple dbm clone for the Mac. |
| 2 | |
| 3 | For database spam, spam.dir contains the index (a text file), |
| 4 | spam.bak *may* contain a backup of the index (also a text file), |
| 5 | while spam.dat contains the data (a binary file). |
| 6 | |
| 7 | XXX TO DO: |
| 8 | |
| 9 | - reclaim free space (currently, space once occupied by deleted or expanded |
| 10 | items is never reused) |
| 11 | |
| 12 | - support concurrent access (currently, if two processes take turns making |
| 13 | updates, they can mess up the index) |
| 14 | |
| 15 | - support efficient access to large databases (currently, the whole index |
| 16 | is read when the database is opened, and some updates rewrite the whole index) |
| 17 | """ |
| 18 | |
| 19 | _os = __import__('os') |
| 20 | import __builtin__ |
| 21 | |
| 22 | _open = __builtin__.open |
| 23 | |
| 24 | _BLOCKSIZE = 512 |
| 25 | |
| 26 | class _Database: |
| 27 | |
| 28 | def __init__(self, file): |
| 29 | self._dirfile = file + '.dir' |
| 30 | self._datfile = file + '.dat' |
| 31 | self._bakfile = file + '.bak' |
| 32 | self._update() |
| 33 | |
| 34 | def _update(self): |
| 35 | self._index = {} |
| 36 | try: |
| 37 | f = _open(self._dirfile) |
| 38 | except IOError: |
| 39 | pass |
| 40 | else: |
| 41 | while 1: |
| 42 | line = f.readline() |
| 43 | if not line: break |
| 44 | key, (pos, siz) = eval(line) |
| 45 | self._index[key] = (pos, siz) |
| 46 | f.close() |
| 47 | |
| 48 | def _commit(self): |
| 49 | try: _os.unlink(self._bakfile) |
| 50 | except _os.error: pass |
| 51 | try: _os.rename(self._dirfile, self._bakfile) |
| 52 | except _os.error: pass |
| 53 | f = _open(self._dirfile, 'w') |
| 54 | for key, (pos, siz) in self._index.items(): |
| 55 | f.write("%s, (%s, %s)\n" % (`key`, `pos`, `siz`)) |
| 56 | f.close() |
| 57 | |
| 58 | def __getitem__(self, key): |
| 59 | pos, siz = self._index[key] # may raise KeyError |
| 60 | f = _open(self._datfile, 'rb') |
| 61 | f.seek(pos) |
| 62 | dat = f.read(siz) |
| 63 | f.close() |
| 64 | return dat |
| 65 | |
| 66 | def _addval(self, val): |
| 67 | f = _open(self._datfile, 'rb+') |
| 68 | f.seek(0, 2) |
| 69 | pos = f.tell() |
| 70 | pos = ((pos + _BLOCKSIZE - 1) / _BLOCKSIZE) * _BLOCKSIZE |
| 71 | f.seek(pos) |
| 72 | f.write(val) |
| 73 | f.close() |
| 74 | return (pos, len(val)) |
| 75 | |
| 76 | def _setval(self, pos, val): |
| 77 | f = _open(self._datfile, 'rb+') |
| 78 | f.seek(pos) |
| 79 | f.write(val) |
| 80 | f.close() |
| 81 | return pos, (val) |
| 82 | |
| 83 | def _addkey(self, key, (pos, siz)): |
| 84 | self._index[key] = (pos, siz) |
| 85 | f = _open(self._dirfile, 'a') |
| 86 | f.write("%s, (%s, %s)\n" % (`key`, `pos`, `siz`)) |
| 87 | f.close() |
| 88 | |
| 89 | def __setitem__(self, key, val): |
| 90 | if not type(key) == type('') == type(val): |
| 91 | raise TypeError, "dbmac keys and values must be strings" |
| 92 | if not self._index.has_key(key): |
| 93 | (pos, siz) = self._addval(val) |
| 94 | self._addkey(key, (pos, siz)) |
| 95 | else: |
| 96 | pos, siz = self._index[key] |
| 97 | oldblocks = (siz + _BLOCKSIZE - 1) / _BLOCKSIZE |
| 98 | newblocks = (len(val) + _BLOCKSIZE - 1) / _BLOCKSIZE |
| 99 | if newblocks <= oldblocks: |
| 100 | pos, siz = self._setval(pos, val) |
| 101 | self._index[key] = pos, siz |
| 102 | else: |
| 103 | pos, siz = self._addval(val) |
| 104 | self._index[key] = pos, siz |
| 105 | self._addkey(key, (pos, siz)) |
| 106 | |
| 107 | def __delitem__(self, key): |
| 108 | del self._index[key] |
| 109 | self._commit() |
| 110 | |
| 111 | def keys(self): |
| 112 | return self._index.keys() |
| 113 | |
| 114 | def has_key(self, key): |
| 115 | return self._index.has_key(key) |
| 116 | |
| 117 | def __len__(self): |
| 118 | return len(self._index) |
| 119 | |
| 120 | def close(self): |
| 121 | self._index = self._datfile = self._dirfile = self._bakfile = None |
| 122 | |
| 123 | |
| 124 | def open(file, mode = None): |
| 125 | return _Database(file) |