Comparison:
flat_map 2.41 K
map.flatten 1.29 K - 1.87x slower +362.05 μs
flat_map (first-try) 2.39 K map.flatten (first-try) 1.26 K - 1.91x slower +378.52 μs
Extended statistics:
Name minimum maximum sample size mode
flat_map 345.18 μs 1195.24 μs 12.00 K 405.28 μs, 405.28 μs
map.flatten 494.08 μs 1479.42 μs 6.42 K 739.22 μs
flat_map (first-try) 345.21 μs 1273.69 μs 11.94 K 405.52 μs, 405.46 μs
map.flatten (first-try) 493.44 μs 1614.85 μs 6.27 K 741.30 μs
See our sample for that.
I'm guessing it is because we didn't compute relative statistics for it... which is true because they are loaded after statistics calculation. This means we probably should have a separate Relative/Meta Statistics step that is called after loading them and always recomputed.
That probably has other problems again (such as sorting etc...) which I'll then have to figure out tomorrow.
See our sample for that.
I'm guessing it is because we didn't compute relative statistics for it... which is true because they are loaded after statistics calculation. This means we probably should have a separate Relative/Meta Statistics step that is called after loading them and always recomputed.
That probably has other problems again (such as sorting etc...) which I'll then have to figure out tomorrow.