Compacting segment refs was slow on a very long queue (thousands of segments). The test had been modified as the new function doesn't support multiple duplicates, but this should never happen.
For 10k segments, this function completes in 0.000005s instead of almost 2 seconds.
Compacting segment refs was slow on a very long queue (thousands of segments). The test had been modified as the new function doesn't support multiple duplicates, but this should never happen.
For 10k segments, this function completes in 0.000005s instead of almost 2 seconds.