Sheetwrite — Spreadsheet, data grid & multi-sheet workbook engine

How it works

Three layers. One clear line of ownership.

The Grid handles input and painting. The engine stores cells and recalculates. Your application decides what to save and where.

Read the runtime-ownership model →

What you can build

Real workbooks, not a table with formulas on top.

Every card opens a live example that runs the real engine in your browser. See all showcases →

Performance & scaleAddress 1,000,000 rows × 1,000 columns. The Grid requests only the tiles on screen and keeps memory inside a fixed cache.Open the live example →
Formula analysisGROUPBY, PIVOTBY, FILTER, LAMBDA, statistics and finance. Results spill, and other formulas read the spill with A2#.Open the live example →
CollaborationClients commit in order through your server. Duplicates, gaps, and conflicts recover with explicit protocol steps.Open the live example →
Database & documentsSnapshots and append-only commits in your storage. Pending edits survive a reload and drain in order.Open the live example →
XLSX & CSVImport and export workbooks with formulas and styles. Every lossy case returns a named warning, not a silent change.Open the live example →
Host-owned rowsYour store keeps the records. Sort, filter, and edit the Grid, and every delta names the same record ID.Open the live example →

See it happen

The hard parts of a spreadsheet, handled.

Three short stories about work that a table component leaves to you. Each one links to the live example that runs it for real.

One formula. The answer takes the space it needs.

FILTER spills its result into the cells below. When a source value changes, the spill grows, and every formula that reads E2# follows.

Try the formula showcase →

Start in minutes

Mount a Grid in the framework you already use.

One engine, four entry points. The adapters share one lifecycle: create on mount, apply live option changes, and destroy on unmount.

Build your first grid →

import { createGrid, initSheetwrite } from "@sheetwrite/core";
import "@sheetwrite/core/styles.css";

await initSheetwrite();

const grid = createGrid(host, {
  workbook: {
    activeSheet: "sales",
    sheets: [
      {
        id: "sales",
        name: "Sales",
        rowCount: 3,
        columns: [
          { key: "product", header: "Product", width: 180, type: "text" },
          { key: "amount", header: "Amount", width: 100, type: "number" },
        ],
      },
    ],
  },
  data: {
    rowCount: 3,
    columns: {
      product: ["Notebook", "Pen", "Folder"],
      amount: new Float64Array([12, 4, 9]),
    },
  },
});

$ npm install @sheetwrite/core

New in 0.5

Pick the formula engine that fits your download budget.

Default@sheetwrite/wasm

Included by the core package. The standard function set and the smaller download.

await initSheetwrite();

Full@sheetwrite/formulas

The same engine with the analysis families: grouping, pivots, LAMBDA, statistics, regression, finance, regex, and matrices.

await initSheetwrite(undefined, formulas);

Array arithmetic and spill references in both engines
Large undo through compressed restores and atomic sync batches
Host-owned rows with stable IDs and per-record deltas
Faster formula kernels, bulk loading, and streamed CSV and XLSX imports
Read what is new and the upgrade checklist →

Results you can check

Measured, not promised.

Median interaction ratios against Handsontable 18.1.2 in one saved browser capture. Every interaction must return the correct result; failed runs stay failed.

1k rows6.5×median of 14 interactions
10k rows6.4×median of 14 interactions
100k rows8.2×median of 14 interactions
1M rows7.1×median of 14 interactions

Bar length uses a log scale from 1× (as fast as Handsontable) to 10×.

Browser: Chromium 153.0.8010.12
Rounds: 10 counterbalanced
Commit: 49e05e2
Captured: 2026-10-17

See the full method and every result →

The spreadsheet engine for your web app.