virtual icon indicating copy to clipboard operation
virtual copied to clipboard

Virtualized list wrapper keeps stale height after data changes

Open kosovatoolsadmin opened this issue 3 months ago • 0 comments

Describe the bug

Summary

List uses useWindowVirtualizer for row virtualization, but the outer div’s inline height is read directly from rowVirtualizer.getTotalSize() during render. Because the virtualizer instance is stable and the getter mutates internal state without changing referential identity, React never receives a new height value when the rows array is filtered or otherwise resized. The wrapper therefore preserves an outdated, overly tall height, leaving a large blank spacer below the visible rows and causing scroll position jumps.

Impact

Any operation that changes the number of rows (search, filters, pagination reset) leaves the list container at its previous height. Users see excessive empty space and the scroll bar no longer reflects the actual result size.

Root Cause

getTotalSize() is invoked inline inside JSX. The virtualizer recomputes its measurement cache, but React doesn’t detect that the returned string (e.g., "32000px") should change because the component never re-renders with a different computed value tied to the rows count.

Fix in my case

Memoize the total height using useMemo(() => rowVirtualizer.getTotalSize(), [rowVirtualizer, rows.length]) and pass that memoized number to the style prop. This forces a fresh height string whenever the virtualizer’s count changes, keeping the spacer in sync. Implemented in commit

Your minimal, reproducible example

https://github.com/kosovatools/kosovatools.org/commit/da4a20196356fd7d96e0fb9b69be9bd662294eeb

Steps to reproduce

.

Expected behavior

Height react to count change

How often does this bug happen?

Every time

Screenshots or Videos

No response

Platform

Firefox

tanstack-virtual version

3.13.12

TypeScript version

5.7.3

Additional context

No response

Terms & Code of Conduct

  • [x] I agree to follow this project's Code of Conduct
  • [x] I understand that if my bug cannot be reliable reproduced in a debuggable environment, it will probably not be fixed and this issue may even be closed.

kosovatoolsadmin avatar Nov 08 '25 15:11 kosovatoolsadmin