I find the docs about the depth-related metrics in Snowflake to be somewhat lacking. Regarding average_overlaps, we can read the following:
Average number of overlapping micro-partitions for each
micro-partition in the table.
What does this mean precisely? Is it something like:
sum_overlaps = 0
let partitions = all micro partitions
for p in partitions:
sum_overlaps += number_of_overlapping_partitions(p)
return sum_overlaps/len(partitions)
where number_overlapping_partitions(p) simply returns the number of partitions that overlap with p in some range of the cluster key values?
