Add support for connect_cb #1140
Conversation
| It looks like @hallfox hasn't signed our Contributor License Agreement, yet.
You can read and sign our full Contributor License Agreement here. Once you've signed reply with Appreciation of efforts, clabot |
| | ||
| self->u.Consumer.rebalance_assigned++; | ||
| | ||
| Py_BEGIN_ALLOW_THREADS |
There was a problem hiding this comment.
Looks the indentation doesn't match with the previous code.
| | ||
| Py_BEGIN_ALLOW_THREADS | ||
| err = rd_kafka_assign(self->rk, c_parts); | ||
| Py_END_ALLOW_THREADS |
| PyObject *error_cb; | ||
| PyObject *throttle_cb; | ||
| PyObject *stats_cb; | ||
| PyObject *connect_cb; |
| | ||
| self->u.Consumer.rebalance_assigned++; | ||
| | ||
| Py_BEGIN_ALLOW_THREADS |
|
|
3 participants
This PR adds the ability to set the
connect_cbfunction to a Python function object. Through many iterations it has caused us to run into a few deadlocking situations because of the GIL, which is why severalrdkafkacalls now release it. We are looking for some feedback, but in particular this PR is still missingNotably as discussed in #1097 this patch also releases/acquires the GIL in potentially hot paths, so the overall performance impact here is worth some attention.