In the current version, when zooWatchGetChildren is triggered, the watcherCtxType, whose type is ZooEvent is always set to ZooChildEvent. However, this is not detailed enough in some situations.

To be more detailed, when zooWatchGetChildren is triggered, we want to know if the event is one of the following:

• ZooCreateEvent
• ZooDeleteEvent
• ZooChangedEvent
• zooNoWatchingEvent

Hello- I attempted to use this library for a CLI tool and had to make some questionable moves in order to reduce log pollution, as zoo_set_log_stream is not exposed by the API, and my naive-to-Haskell attempts to bind it to /dev/null led to segfaults.

My workaround was to set the log-level to (unsafeCoerce @ZooKeeper.ZooLogLevel (0 :: CInt)) during initial connection attempts¹, and run the connection attempt in a thread to allow manual timeout handling, then - upon seeing the state was ZCONNECTED ² or timeout being reached, return or fail.

Perhaps exposing the zoo_set_log_stream API, or maybe providing an easy way to pipe them to a TQueue or similar would be helpful? Similarly, connect timeouts or retry attempt counts before throwing some sort of exception or return state would be nice.

P.S. I'm not clear of the exception handling model, but it seemed very unclear how to actually handle the results of a call without writing an FFI wrapper instead of doing the exception conversions on the fly. I didn't see any examples in the docs or tests that cleared this up, but it didn't seem worth opening another issue for, without further clarification.

• zoo_age_acl: #11
• zoo_aset_acl

ZOOAPI int zoo_state(zhandle_t *zh);

ZOOAPI int is_unrecoverable(zhandle_t *zh);

Just like https://hackage.haskell.org/package/directory-1.3.7.0/docs/System-Directory.html#v:createDirectoryIfMissing

zooCreateIfMissing : Bool -> Path -> IO ()

ZOOAPI: int zoo_recv_timeout(zhandle_t *zh);