Based on the surrounding code this probably wasn't a bottleneck, but it's nice to avoid in principle anyway.