Document the behaviour of ResponseCache

it looks like everything that uses ResponseCache expects to have to
`make_deferred_yieldable` its results. It's debatable whether that is the best
approach, but let's document it for now to avoid further confusion.
pull/3059/head
Richard van der Hoff 2018-04-04 08:58:53 +01:00
parent 88cc9cc69e
commit a9a74101a4
1 changed files with 32 additions and 0 deletions

View File

@ -31,6 +31,18 @@ class ResponseCache(object):
self.timeout_sec = timeout_ms / 1000.
def get(self, key):
"""Look up the given key.
Returns a deferred which doesn't follow the synapse logcontext rules,
so you'll probably want to make_deferred_yieldable it.
Args:
key (str):
Returns:
twisted.internet.defer.Deferred|None: None if there is no entry
for this key; otherwise a deferred result.
"""
result = self.pending_result_cache.get(key)
if result is not None:
return result.observe()
@ -38,6 +50,26 @@ class ResponseCache(object):
return None
def set(self, key, deferred):
"""Set the entry for the given key to the given deferred.
*deferred* should run its callbacks in the sentinel logcontext (ie,
you should wrap normal synapse deferreds with
logcontext.run_in_background).
Returns a new Deferred which also doesn't follow the synapse logcontext
rules, so you will want to make_deferred_yieldable it
(TODO: before using this more widely, it might make sense to refactor
it and get() so that they do the necessary wrapping rather than having
to do it everywhere ResponseCache is used.)
Args:
key (str):
deferred (twisted.internet.defer.Deferred):
Returns:
twisted.internet.defer.Deferred
"""
result = ObservableDeferred(deferred, consumeErrors=True)
self.pending_result_cache[key] = result