page-up/down doesn’t take into account the number of lines of items in mult-line mode

[vejkse]

Checklist

• I have read through the manual page (man fzf)
• I have searched through the existing issues
• For bug reports, I have checked if the bug is reproducible in the latest version of fzf

Output of fzf --version

0.56 (devel)

OS

• Linux
• macOS
• Windows
• Etc.

Shell

• bash
• zsh
• fish

Problem / Steps to reproduce

With one-line items, when doing PgUp (bound by default to page-up) and PgDn (bound by default to page-down), if 21 lines (and so 21 items) are visible as in the following example:

for i in {01..99}; do echo -ne "item $i\0"; done | fzf --read0 --reverse --height=23

each movement one page up or down goes 20 lines (and so 20 items) up or down the list, as one would expect.

But in the following example with 2-line items, where 21 lines (and so 10.5 items) are visible:

for i in {01..99}; do echo -ne "item $i, line 1\nitem $i, line 2\0"; done | fzf --read0 --reverse --height=23

each movement one page up or down doesn’t go about 20 lines up or down the list, as one would expect, but 20 items (so 40 lines) up or down the list, so that each time we go up or down, we skip about 10 items.

[junegunn]

Thanks for the report. It should be fixed now on master.

[vejkse]

Thank you!

But I think it doesn’t work in all cases. I should have made an example with variable length, like this one where each item n > 1 has ⌈n/2⌉ + 1 lines:

for i in {1..99}; do printf "item $i, line %d\n" $(seq 1 $((i/2))); echo -ne "item $i, line $((i/2+1))\0"; done | fzf --read0 --reverse --height=23

Here is how it goes when repeating page-down:

1. Items 1-7 are fully visible and 1 line of item 8.
2. Items 5-9 are fully visible.
3. Items 11-13 are fully visible and 1 line of item 14 — so item 10 was jumped over.
4. Items 14-15 are fully visible and 5 lines of item 16.
5. Items 18-19 are fully visible and 1 line of item 20 — so item 17 was jumped over.

After that, we see all items but some only partially even though they could be fully visible if from that point on, page-down became equivalent to down.

Of course, once we get to items that have too many lines to be shown in that limited height, they will never be fully visible.

[vejkse]

Even a uniform 3-line case has a little problem, like this one:

for i in {01..99}; do echo -ne "item $i, line 1\nitem $i, line 2\nitem $i, line 3\0"; done | fzf --read0 --reverse --height=22

When repeating page-down, after the initial steps, only two lines of the last item on each page are visible, but the full item is not visible on the next page. One could argue that this is acceptable, I suppose.

[junegunn]

Thanks for the repro.

Turns out it's not quite trivial to get it right, especially when considering the different values of --scroll-off. Will revisit this later.

[junegunn]

See 4a85843.

The behavior of the actions could still be improved, but at least they don't skip items.

[vejkse]

Thank you! But I’m afraid there are still cases where items are skipped… This looks very tricky.

Here is one:

{ for i in {1..30}; do printf "item $i, line 1\0"; done; for i in {31..34}; do printf "item $i, line 1\n         line 2\n         line 3\0"; done; for i in {35..90}; do printf "item $i, line 1\0"; done; } | fzf --read0 --reverse --height=28

When iterating page-down, we get:

1. items 1-26;
2. items 4-29;
3. items 29-46 (the only few multi-line items appear here);
4. items 51-76 (so 47-50 were skipped).

I’m wondering if the problem doesn’t come from the fact that when moving down one item, the number of lines by which the ‘window’ is moved depends not on the number of lines of the current or next item, but on the number of lines of the first item in the ‘window’?

[junegunn]

Thanks for finding the edge cases and sharing the insight. I think I'm going to have to take a different approach to the problem. I'll let you know of the progress.