allow_session_lost documentation for ChildWatchers seems

[jimfulton]

The documentation says:

"If allow_session_lost is set to True, then the function will no longer be called if the session is lost."

But:

>>> import kazoo.client
>>> c = kazoo.client.KazooClient('localhost:2182')
>>> c2 = kazoo.client.KazooClient('localhost:2183')

I have 2 connections against 2 tunnels.

>>> @c.add_listener
... def _(state):
...     print state
... 
>>> c.start()
CONNECTED
>>> c2.start()
>>> @c.ChildrenWatch('/test', allow_session_lost=True)
... def _(data):
...     print True, data
... 
True [u'b', u'a', u'aa', u'z']
>>> @c.ChildrenWatch('/test', allow_session_lost=False)
... def _(data):
...     print False, data
... 
False [u'b', u'a', u'aa', u'z']

I break the first tunnel

SUSPENDED

Wait a while and recreate it:

LOST
CONNECTED

>>> _ = c2.delete('test/a')
True [u'b', u'aa', u'z']

>>> _ = c2.create('test/a')
True [u'b', u'a', u'aa', u'z']

Note that the session is lost when allow_session_lost is false, not when is True. Is this a bug? doc bug?

[bbangert]

I'm going to guess at the moment that its a bug, as I didn't go through it again when refactoring DataWatch recently and it did several things like DataWatch did (incorrectly) at first.

[jimfulton]

So, from #157 it appears that allow_session_lost should be treated as deprecated and ignored. The current default behavior for both DataWatch and ChildrenWatch is to restore watches to new sessions. This is what I want. I was thrown by the documentation, which suggests otherwise.

For my use case, I should simply not pass allow_session_lost.

WRT documentation, I suggest removing mention of allow_session_lost, and saying that watches are restored.

I also suggest that deprecation warnings be issued if allow_session_lost is used.

If you agree, I'll make a pull request.

[bbangert]

Yep, that sounds good to me.

[bbangert]

Was there a pull request ever made? ;)

[jimfulton]

I don't remember. I guess not. Sorry.

[bbangert]

Closing this as I'm not sure if it's still an issue. Changes have been made, etc. Watchers now restore on reconnect as well which could alter this.

[l2dy]

Closing this as I'm not sure if it's still an issue. Changes have been made, etc. Watchers now restore on reconnect as well which could alter this.

The documentation suggests that if allow_session_lost is set to True, the function will no longer be called if the session is lost. However, the test_child_stop_on_session_loss test in test_watchers.py expects that it won't be called when allow_session_lost is set to False after session loss. Which is the expected behavior?