Created attachment 165612 [details]
test ODS file with autofilter

Assume a column contains n items. When we open the autofilter and un-selected one, then we have n-1 items selected. Doing this filter with AutoFilter is slow because it seems to be using the "x in n-1" logic. However, doing this filter with Standard Filter is fast if we use "Not Equal" logic.

Steps to Reproduce:

1. Open the attached test ODS file

2. Click the autofilter in column A, and de-select the 1st item (i.e., #N/A), then click OK.
--> Slow, it takes more than 8s.

3. Use Standard Filter, and set condition "POMatch <> #N/A", then click OK.
--> Fast, it takes 1s.

Apparently, if we deselect only one item, the condition is the same as <> that item. It should be more expensive to compare the value to see whether it is in the n-1 list.

Actually, if we de-select several items (e.g., 3 items), it would still be faster to use "(x <> a) and (x <> b) and (x <> c)". However, if the deselected items are quite a lot, the "x in n-1" logic should be faster.