Performance live demo

How this demo works

Each column uses the expression v[i] + v[i+1]: column 0 computes v0 + v1, column 1 computes v1 + v2, and so on. Each row model holds numeric fields v0 through v[columns].

rs-x parses each unique column expression once, caches the tree, then clones it for every row binding. So parsing is columns operations regardless of row count, and binding is rows × columns operations.

Rows and columns change in steps ( +100 rows, +5 columns) up to 2,000 rows × 20 columns. After the run all rows of computed expression values are shown in a scrollable table so you can verify the bindings are live.

Run benchmark

// 10 column expressions (field names are namespaced per run)
col 0: v0 + v1
col 1: v1 + v2
col 2: v2 + v3
col 3: v3 + v4
col 4: v4 + v5
col 5: v5 + v6
col 6: v6 + v7
col 7: v7 + v8
col 8: v8 + v9
col 9: v9 + v10

// Each row model has fields v0..v10
// Row 0: { v0: 0, v1: 1, ..., v10: 10 }
// Row 1: { v0: 1, v1: 2, ..., v10: 11 }