Summary

While benchmarking heavy event scheduling workloads, I noticed that Event.__lt__ currently performs tuple-based comparison:

return (self.time, self.priority, self.unique_id) < (
    other.time,
    other.priority,
    other.unique_id,
)

Since heap operations can trigger a very large number of comparisons, this results in repeated temporary tuple allocations during scheduling. I experimented with replacing the tuple comparison with an explicit field-by-field comparison:

if self.time != other.time:
    return self.time < other.time
if self.priority != other.priority:
    return self.priority < other.priority
return self.unique_id < other.unique_id

This avoids tuple creation while preserving identical ordering semantics

Benchmark Results
Using a recurring-event benchmark with increasing load:

Tuple-based comparison

• 5k events: ~8–9s
• 20k events: ~90–130s
• 50k events: ~210–330s

Manual comparison

• 5k events: ~6–8s
• 20k events: ~69–75s
• 50k events: ~178–190s

The performance difference becomes more noticeable as the number of scheduled events increases. All existing mesa.time tests pass with the manual comparison implementation.