Welcome, guest | Sign In | My Account | Store | Cart

Implements the observer design pattern via generator coroutines, wrapped up to use the new 'with' statement of Python 2.5. Enables the loosely-coupled observation of any container implementing the dictionary protocol.

Python, 254 lines
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
"""
observer module

Typical usage is as follows:

from __future__ import with_statement
from observer import consumer, observation

@consumer
def do_something_with_notification():
    while True:
        key, old, new = (yield)
        print "%s: %s -> %s" % (key, old, new)

container = {}

# Any modification to `container`, now called `observed` in the
# body of the with statement, is sent to the coroutine
# do_something_with_notification()

with observation(observe=container,
                 notify=[do_something_with_notification()) as observed:
    modify_observed(observed)

Requires Python 2.5
Author: Jim Baker (jbaker@zyasoft.com)
"""

from __future__ import with_statement
from contextlib import contextmanager
import unittest

@contextmanager
def observation(observe, notify):
    """Simple boilerplate to link to the 'with' statement.

    Contextlib's contextmanager decorator is a very convenient way to
    create simple context managers, specifically the __enter__ and
    __exit__ special methods.
    """

    proxy = Observation(observe, notify)
    try:
        yield proxy
    finally:
        proxy.close()


class NoneSuch(object):
    """A useful alternative to None in the case of a key being deleted or inserted."""
    def __new__(cls, *args, **kwargs):
        if '_inst' not in vars(cls):
            cls._inst = object.__new__(cls, *args, **kwargs)
        return cls._inst
    def __init__(self, *args, **kwargs): pass
    def __repr__(self): return "NoneSuch()"
    def __call__(self, *args, **kwargs): return self
    def __nonzero__(self): return False

NoneSuch = NoneSuch()


class Observation(object):
    """Enables observation of dictionaries.

    Proxies the `observe` dictionary such that any modifications to
    it are sent via `send()` to the notifiers in the `notify`
    sequence.  The sent value is a triple (key, old, new).
    Notifications are sent AFTER the change.

    Other mutable containers, such as sets and lists or your custom
    container, can be readily added by supporting their interface.
    """

    def __init__(self, observe, notify):
        self._obj = observe
        self.notify = notify

    def close(self):
        self._obj = None
        self.notify = None

    def __iter__(self):
        if self._obj is None: raise ValueError("Operation on closed observation")
        return iter(self._obj)

    # all mutating methods go here, this list should be comprehensive as of 2.5
    def __delitem__(self, K):
        if self._obj is None: raise ValueError("Operation on closed observation")
        old = self._obj[K]
        del self._obj[K]
        for notify in self.notify:
            notify.send((K, old, NoneSuch))

    def __setitem__(self, K, V):
        if self._obj is None: raise ValueError("Operation on closed observation")
        old = self._obj.get(K, NoneSuch)
        self._obj[K] = V
        for notify in self.notify:
            notify.send((K, old, V))

    def setdefault(self, K, default):
        if self._obj is None: raise ValueError("Operation on closed observation")
        try:
            return self._obj[K]
        except KeyError:
            self._obj[K] = default
            for notify in self.notify:
                notify.send((K, NoneSuch, default))

    def clear(self):
        if self._obj is None: raise ValueError("Operation on closed observation")
        items = self._obj.items()
        self._obj.clear()
        for K, old in items:
            for notify in self.notify:
                notify.send((K, old, NoneSuch))

    def update(self, *seq_or_map, **kw):
        from itertools import chain

        if self._obj is None: raise ValueError("Operation on closed observation")
        try: seq = seq_or_map[0].iteritems()
        except IndexError: seq = ((K,None) for K in seq_or_map)
        for K, V in chain(seq, kw.iteritems()):
            old = self._obj.get(K, NoneSuch)
            self._obj[K] = V
            for notify in self.notify:
                notify.send((K, old, V))

    def pop(self, K, *default):
        if self._obj is None: raise ValueError("Operation on closed observation")

        # this may be unexpected to have old be the default
        # value. what do you think?
        if default:
            old = self._obj.pop(K, default[0])
        else:
             old = self._obj.pop(K)
        for notify in self.notify:
            notify.send((K, old, NoneSuch))
        return old

    def popitem(self):
        if self._obj is None: raise ValueError("Operation on closed observation")
        K,old = self._obj.popitem()
        for notify in self.notify:
            notify.send((K, old, NoneSuch))
        return old

    def __contains__(self, K):
        if self._obj is None: raise ValueError("Operation on closed observation")
        return K in self._obj
    def __getitem__(self, K):
        if self._obj is None: raise ValueError("Operation on closed observation")
        return self._obj[K]
    def __len__(self):
        if self._obj is None: raise ValueError("Operation on closed observation")
        return len(self._obj)

    # otherwise, just pass through
    def __getattr__(self, attrib):
        if self._obj is None: raise ValueError("Operation on closed observation")
        return getattr(self._obj, attrib)


def consumer(func):
    """A decorator, advances func to its first yield point when called.

    Modifed this original example code from PEP 342 to use the new
    functools.wraps decorator. This convenience function makes it look
    like the original function, which is almost always what we want,
    especially if we designed the original function to be wrapped in
    the first place!

    Maybe `consumer` should go into functools too!
    """

    from functools import wraps

    @wraps(func)
    def wrapper(*args,**kw):
        gen = func(*args, **kw)
        gen.next()
        return gen
    return wrapper



class ObserverTestCase(unittest.TestCase):
    """Tests observer module, special emphasis on dictionary protocol.

    We keep the tests monolithic, just RunTest(), to keep the scope of
    the with statement visible and simple.
    """

    def runTest(self):
        from collections import deque
        changes = deque()

        def consume(X):
            def _consume(X):
                while X:
                    yield X.popleft()
            return list(_consume(X))

        @consumer
        def observe_changes():
            while True:
                change = (yield)
                changes.append(change)

        fruits = dict(apple=1, banana=2, cherry=3)
        with observation(observe=fruits, notify=[observe_changes()]) as observed_fruits:
            # typical mutations
            observed_fruits['cherry'] *= 2
            del observed_fruits['apple']
            self.assertEquals(consume(changes), [('cherry', 3, 6), ('apple', 1, NoneSuch)])

            # .update with keyword args
            observed_fruits.update(durian=4, figs=5)
            self.assertEquals(fruits['durian'], 4)

            # .clear
            observed_fruits.clear()
            self.assertEquals(len(observed_fruits), 0)
            consume(changes) # keep it simple, just throw away

            # .update with map and keyword args, kw should override
            observed_fruits.update({'grapefruit':6, 'jackfruit':7}, jackfruit=8)
            self.assertEquals(observed_fruits['jackfruit'], 8)
            self.assertEquals(consume(changes), [('jackfruit', NoneSuch, 7), ('grapefruit', NoneSuch, 6), ('jackfruit', 7, 8)])

            # .pop, default here may be controversial
            observed_fruits.pop('durian', None)
            self.assertEquals(consume(changes), [('durian', None, NoneSuch)])

            # .setdefault
            observed_fruits.setdefault('jackfruit', -1)
            observed_fruits.setdefault('kiwi', 9)
            self.assertEquals(consume(changes), [('kiwi', NoneSuch, 9)])

            # .popitem
            while observed_fruits:
                observed_fruits.popitem()
            self.assertEquals(fruits, dict())

        # verify that outside of with statement scope, the observation
        # is closed
        self.assertRaises(ValueError, lambda: observed_fruits.update(foo=0, fum=1))


if __name__ == "__main__":
    unittest.main()

Note that we do not provide any mechanism for explicitly unregistering an observer, except through exiting the scope of the with statement, which closes the observation as a whole. This is intentional.

There are other implementations of the observer pattern in the Python cookbook. The comments in this recipe has a list: http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/415310

References:

http://c2.com/cgi/wiki?ObserverPattern (we implement a push model) http://www.suttoncourtenay.org.uk/duncan/accu/pythonpatterns.html

Extensions:

Other mutable containers like lists or sets are a natural extension left to the reader.

If you want to rendezvous with another thread, enqueuing to Queue presumably makes sense; on the other hand, the handoff takes away from some of the nice low-overhead qualities we get with generator coroutines (although wrapping any container protocol incurs its own overhead). This can probably be remedied by using the techniques Goetz Graefe explored in the implementation of the Exchange Iterator, as first seen in his Volcano system and apparently now in MS SQLServer. Start at http://icde2005.is.tsukuba.ac.jp/influential.html.

1 comment

Christopher Dunn 17 years, 2 months ago  # | flag

@consumer. The 'consumer' decorator is useful all by itself.