[IMPROVED] Fetch and FetchBatch for draining and closed subscriptions #1582
Conversation
piotrpio
force-pushed
the
js-fetch-drain
branch
from
7458bb1 to
4daf90f
Compare
Signed-off-by: Piotr Piotrowski <piotr@synadia.com>
piotrpio
force-pushed
the
js-fetch-drain
branch
from
4daf90f to
f069202
Compare
| @@ -2861,7 +2861,13 @@ func (sub *Subscription) Fetch(batch int, opts ...PullOpt) ([]*Msg, error) { | |||
| } | |||
| var hbTimer *time.Timer | |||
| var hbErr error | |||
| if err == nil && len(msgs) < batch { | |||
| sub.mu.Lock() | |||
There was a problem hiding this comment.
Maybe extract following logic
sub.mu.Lock()
subClosed := sub.closed || sub.draining
sub.mu.Unlock()in a new function inside Subscription e.g. isClosedOrDraining() or similar
and reuse it in both places: Fetch and FetchBatch?
There was a problem hiding this comment.
That's such a tiny piece of logic only used in 2 places, I would honestly just leave it inlined for clarity :)
With those kinds of checks I usually wait until I have to repeat it at least 3 times before extracting to a separate method, but that's just preference.
|
|
||
| // now drain the subscription, messages should be in the buffer | ||
| sub.Drain() | ||
| msgs, err := sub.Fetch(100) |
There was a problem hiding this comment.
Calling Fetch immediately after Drain returns messages instead of an error. Is this normal behaviour and why?
There was a problem hiding this comment.
Drain is not blocking, so calling sub.Drain() essentially just starts the draining process without waiting for completion. Fetch() first checks if there are any messages in the buffer (those messages should be accessible while draining) and returns if there are any - it will never send a new request when draining though.
Only when we're fully drained/closed and there are no more messages waiting in the buffer we're returning an error. This tests simulates this behavior.
…#1582) Signed-off-by: Piotr Piotrowski <piotr@synadia.com>
…#1582) Signed-off-by: Piotr Piotrowski <piotr@synadia.com>
…#1582) Signed-off-by: Piotr Piotrowski <piotr@synadia.com>
…#1582) Signed-off-by: Piotr Piotrowski <piotr@synadia.com>
Signed-off-by: Piotr Piotrowski piotr@synadia.com