Course contentsShow
← Web Development
Lesson 2103 of 2,98238. JavaScript Performance in the BrowserFree lesson

Virtual Grid Layouts

Extending virtualization to two-dimensional grids with rows and columns for image galleries and data tables.

Virtual Grid Layouts

You'll learn how to extend virtualization techniques to two-dimensional grids, rendering only visible rows and columns for efficient image galleries and data tables.

Beyond Lists: Working in Two Dimensions

So far you've seen virtual scrolling work vertically with lists. But what about grids—like photo galleries or spreadsheet-style data tables? These need both row and column awareness. A virtual grid renders only the cells that fit in the current viewport, dramatically reducing DOM nodes when dealing with thousands or millions of items.

The Core Approach

Think of a virtual grid as nested virtualization: you calculate which rows are visible (just like vertical scrolling), then for each visible row, calculate which columns are visible horizontally.

The calculation looks like:

  1. Divide scrollTop by row height to find the first visible row index
  2. Divide scrollLeft by column width to find the first visible column index
  3. Calculate how many rows and columns fit in the viewport dimensions
  4. Render only those cells: grid[row][col] for each visible position

For a 10,000×1,000 grid where only 20 rows and 5 columns are visible, you render just 100 DOM elements instead of 10 million.

Practical Considerations

Positioning: Each cell needs position: absolute with calculated top and left based on its row/column indices and your item dimensions.

Overscan: Buffer a few extra rows/columns beyond the viewport edges (concept you learned earlier) to hide rendering during fast diagonal scrolling.

Variable sizes: Like variable-height lists, variable column widths require tracking cumulative widths for accurate positioning.

Key Takeaway: Virtual grids apply list virtualization principles in two dimensions simultaneously, calculating visible row and column ranges to render only what users see, making massive datasets interactive without performance penalties.