[postgis-users] Slow construction of GiST index, but better with smaller # of big rows

[Paul Norman]

On 2019-01-12 8:28 a.m., Wenbo Tao wrote:
>  So essentially, the index construction is slow for too many rows, but 
> much faster for a smaller # of bigger rows. Any intuition why this is 
> the case would be greatly appreciated!


GiST indexes work on bounding boxes only, so the portion of the row that 
matters to the index is the same size per row. I would expect the two 
factors that matter most for index size and generation speed to be 
number of rows and number of pages. You've reduced the number of rows 
and possibly the number of pages, so it's expected the index will be 
smaller and faster to generate.