Alex Harri

A deep dive into optimizing a browser-based spreadsheet engine by reducing the fixed cost of cell recalculation. The bottleneck was a method called `_makeCalcCellEvaluationContext` consuming 12.5% of recalculation time by eagerly constructing a large evaluation context object for every cell. The fix involved two techniques: switching eager property computation to lazy getters, and replacing per-cell object creation with a single shared immutable context object (via a new `CellEvaluator` class) whose dynamic parts are updated by mutating private properties. This eliminated repeated object spreading and reduced garbage collector pressure. The result was a ~10% median performance improvement across real-world spreadsheet documents, validated by a dedicated performance and regression testing suite.

Making GRID’s spreadsheet engine 10% faster