Skip to content
RevoGrid [Pro]

Performance And Large Projects

RevoGrid virtualizes the task table and timeline projection. Preserve that advantage by updating project data through public collections rather than rendering parallel task DOM or reading state back from cells.

Recommendations

Section titled “Recommendations”
  • Use stable task and linked-entity IDs.
  • Replace only collections that changed and keep wrapper inputs memoized or reactive at the correct boundary.
  • Batch related task, dependency, calendar, resource, and assignment changes before triggering a new schedule pass.
  • Keep custom cell and task-bar renderers pure and inexpensive.
  • Limit visible marker and time-range overlays to information users need in the active view.
  • Prefer column presets over rendering every analytical field at once.
  • Measure scheduling separately from browser rendering with the Gantt scheduler benchmark, and use the browser benchmark for end-to-end interaction evidence.
  • Test representative hierarchy depth, dependency density, resource assignments, and timeline span—not only row count.

Current Boundary

Section titled “Current Boundary”

Virtual rendering is implemented. A server-side, date-windowed Gantt project model is not currently part of the public contract, so the complete linked project graph must be available to the client scheduler. See Limitations before designing very large remote projects.

Two benchmark layers

Section titled “Two benchmark layers”

The Gantt scheduler benchmark isolates scheduling-engine recalculation in Node.js. It does not measure browser rendering, scrolling, pointer interactions, hierarchy projection, memory, or mounted DOM.

Run the scheduler benchmark with:

Terminal window
pnpm --filter @revolist/gantt bench

For the reproducible browser suite, raw JSON/CSV, exact environment, screenshot, video, and live workload, open the RevoGrid Gantt browser benchmark. Its published matrix covers 100, 1,000, 5,000, and 10,000 tasks across sparse, normal, and highly connected dependency graphs, with one warm-up and five measured runs per isolated scenario.

The browser suite leads with the measurable capacity claim “10,000 editable tasks and 19,796 dependencies in a live browser demo.” It does not make a “fastest” claim or compare numbers collected with a different harness.