diff --git a/.changeset/pre.json b/.changeset/pre.json
deleted file mode 100644
index f679993..0000000
--- a/.changeset/pre.json
+++ /dev/null
@@ -1,4 +0,0 @@
-{
- "mode": "pre",
- "tag": "beta"
-}
diff --git a/.changeset/pre/builders-shared-styles.md b/.changeset/pre/builders-shared-styles.md
deleted file mode 100644
index b267fde..0000000
--- a/.changeset/pre/builders-shared-styles.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-`fromDataset` shares style objects across rows (one per column, and one per column and `rowStyle` object for banding) instead of building a new one per cell. The writer then resolves each distinct style once from its identity cache. In a local run of 1M rows × 4 styled, banded columns, building went from 2.6 s to 0.5 s and writing from 11.7 s to 7.3 s.
diff --git a/.changeset/pre/builtin-codes-excel.md b/.changeset/pre/builtin-codes-excel.md
deleted file mode 100644
index e95db1f..0000000
--- a/.changeset/pre/builtin-codes-excel.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-Built-in number formats 37–40 and 47 now use the codes Excel itself shows, checked in Excel for Mac: `#,##0_);(#,##0)`, `#,##0_);[Red](#,##0)`, `#,##0.00_);(#,##0.00)`, `#,##0.00_);[Red](#,##0.00)` and `mm:ss.0`. The ECMA-376 table lists other codes. This changes `BUILTIN_NUM_FMTS` and what `formatValue`/`formatCell` show for those ids: positive numbers keep a trailing space where the `)` would be (`1,234.50 `), and 47 shows `31:00.0` instead of `3100.0`. Files are unaffected, because the writer writes only the id.
diff --git a/.changeset/pre/css-color-4-from-table.md b/.changeset/pre/css-color-4-from-table.md
deleted file mode 100644
index 01eb554..0000000
--- a/.changeset/pre/css-color-4-from-table.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-`fromTable` with `styles: 'computed'` now maps text and background colours that browsers report as `lab()`, `lch()`, `oklab()`, `oklch()` or `color()` (for example Tailwind v4 colours), converting them to sRGB. Before, only `rgb()`, hex and named colours were kept.
diff --git a/.changeset/pre/csv-parse-and-serialize.md b/.changeset/pre/csv-parse-and-serialize.md
deleted file mode 100644
index 2e4e001..0000000
--- a/.changeset/pre/csv-parse-and-serialize.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": minor
----
-
-Add `javascript-excel/csv`: `parseCsv`, `readCsv`, `toCsv` and `toCsvBlob`. RFC 4180 parsing with delimiter detection, BOM and CRLF/LF/CR handling, lenient or strict modes, optional type inference, and an async `readCsv` with encoding detection and AbortSignal support. Serialization with quoting, BOM, ISO dates and a formula-injection guard that is on by default.
diff --git a/.changeset/pre/dataset-builders.md b/.changeset/pre/dataset-builders.md
deleted file mode 100644
index e680c48..0000000
--- a/.changeset/pre/dataset-builders.md
+++ /dev/null
@@ -1,9 +0,0 @@
----
-"javascript-excel": minor
----
-
-Add the dataset builders:
-- `fromDataset({ columns, data }, options)`: `key`, `path` and function accessors; header groups as merged header cells; falsy (conditional) columns skipped; per-column `type` coercion, `numFmt`, `width` (including `'auto'`), `hidden` and styles.
-- Freeze and autofilter on the header rows, and a totals row with `SUBTOTAL` formulas whose cached values are computed from the data.
-- `fromDatasets` stacks several blocks on one sheet.
-- `toDataset(sheet)` reads a sheet back into `{ columns, data }` with null-prototype rows.
diff --git a/.changeset/pre/datasets-placement.md b/.changeset/pre/datasets-placement.md
deleted file mode 100644
index 655bc2c..0000000
--- a/.changeset/pre/datasets-placement.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": minor
----
-
-`fromDatasets` blocks take an optional `placement: 'below' | 'right'`. `'below'` (the default) is the existing behaviour. `'right'` puts a block on the previous block's top row, `gap` empty columns after its last column, with `offset.cols` and `offset.rows` adding to that. A `'below'` block after `'right'` blocks starts below the lowest of them, so blocks do not overlap. Existing calls give the same output as before.
diff --git a/.changeset/pre/date-to-serial-day-zero.md b/.changeset/pre/date-to-serial-day-zero.md
deleted file mode 100644
index ad1938d..0000000
--- a/.changeset/pre/date-to-serial-day-zero.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-`dateToSerial`, and so the writer, now accepts 1899-12-31, Excel's day 0, as serials 0 to below 1. `serialToDate` returns time-only values on that day, so a time-only cell read back by `read()` can now be written again; it used to throw `INVALID_DATE`.
diff --git a/.changeset/pre/dom-from-table-download.md b/.changeset/pre/dom-from-table-download.md
deleted file mode 100644
index 201b881..0000000
--- a/.changeset/pre/dom-from-table-download.md
+++ /dev/null
@@ -1,11 +0,0 @@
----
-"javascript-excel": minor
----
-
-Add `javascript-excel/dom`:
-- `fromTable(table, options)` turns an HTML `
` into a sheet. It reads the rows of ``, then ``, then ``, and turns `colspan`/`rowspan` into merges (including `rowspan="0"`). It understands `data-type`, `data-value` and `data-numfmt` hints, and skips hidden and `data-export-ignore` rows and cells.
-- In a cell, `
` becomes a newline with wrapping, and the first link becomes the cell's hyperlink, limited to `http`, `https` and `mailto` unless `hyperlinkSchemes` allows more.
-- Optional computed-style mapping covers bold, italic, underline, strike, colour, background (including a row's background) and explicit alignment. There are also header, freeze, autofilter and auto-width options, a per-cell hook, and `onWarning` for data that cannot be coerced or for dropped links.
-- `download(data, filename)` saves bytes, a Blob or a string through a temporary `` and an object URL, with no dependencies. The module never touches globals at import.
-
-Also adds the `HYPERLINK_DROPPED` warning code and the `AutoWidthOptions` type.
diff --git a/.changeset/pre/exports-types-conditions.md b/.changeset/pre/exports-types-conditions.md
deleted file mode 100644
index 8714da0..0000000
--- a/.changeset/pre/exports-types-conditions.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-The `exports` map now names the type declarations explicitly: each subpath has `import: { types: "./dist/x.d.ts", default: "./dist/x.js" }` and `require: { types: "./dist/x.d.cts", default: "./dist/x.cjs" }`. TypeScript already found the declarations next to the JavaScript files; the explicit `types` conditions don't depend on that lookup.
diff --git a/.changeset/pre/group-title-fill-right.md b/.changeset/pre/group-title-fill-right.md
deleted file mode 100644
index 8cac300..0000000
--- a/.changeset/pre/group-title-fill-right.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-`toDataset` with `headerRows` (and `importDataset` with a multi-row header) now carries a group title right across the empty upper-row cells that follow it when there are no merges, as in CSV: `Name,Q1,` over `,Jan,Feb` gives `Name`, `Q1 / Jan`, `Q1 / Feb` instead of `Name`, `Q1 / Jan`, `Feb`. Merged cells still decide where they are present, and a column with no text in the lower header rows is not filled.
diff --git a/.changeset/pre/import-pipeline.md b/.changeset/pre/import-pipeline.md
deleted file mode 100644
index 9f0f0f5..0000000
--- a/.changeset/pre/import-pipeline.md
+++ /dev/null
@@ -1,21 +0,0 @@
----
-"javascript-excel": minor
----
-
-New entry `javascript-excel/import` turns a user's `.xlsx` or `.csv` file into `{ columns, data }` with `importDataset`. It handles:
-
-- the format, detected from the bytes
-- the sheet: the first visible one, or one picked by name or index
-- the header row, detected (title rows skipped, multi-row headers flattened)
-- column types, inferred all-or-nothing, with explicit number and date locales
-- column mapping to a schema by name or alias
-- per-row issues with their cell refs, and `maxRows` / `preview`
-
-The building blocks are exported too: `detectHeader`, `inferColumnTypes`, `coerceColumns`, `parseNumber`, `parseDate`, `mapColumns` and `applyMapping`.
-
-Other additions:
-
-- `read` gains `maxRows`. Rows past it are skipped without tokenizing them, while merges and sheet views are still read. It comes with a `ROWS_TRUNCATED` warning.
-- `toDataset` gains `headerRows` and `join` for multi-row headers.
-- The worker client gains `importDataset`.
-- New error codes: `SHEET_NOT_FOUND`, and `VALIDATION_FAILED` (reserved for react-data-import).
diff --git a/.changeset/pre/issue-fields.md b/.changeset/pre/issue-fields.md
deleted file mode 100644
index be1c097..0000000
--- a/.changeset/pre/issue-fields.md
+++ /dev/null
@@ -1,7 +0,0 @@
----
-"javascript-excel": minor
----
-
-`importDataset` issues now name the field as it appears in `data` rows. With a schema, `TYPE`, `ERROR_CELL` and `DATE_ORDER_AMBIGUOUS` issues use the schema field name; they used the source column key. The new optional `column` property on `ImportIssue` holds the source column key (also set by `coerceColumns` and on `applyMapping`'s `REQUIRED`). A value that doesn't convert to a forced type in a required field gives one `TYPE` issue, no longer a `REQUIRED` issue as well. Issues on columns the mapping drops are left out.
-
-Migration: if you grouped issues by the source column key, group by `issue.column`.
diff --git a/.changeset/pre/lazy-shared-strings.md b/.changeset/pre/lazy-shared-strings.md
deleted file mode 100644
index 010247b..0000000
--- a/.changeset/pre/lazy-shared-strings.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-`read` with `maxRows` decodes shared strings lazily. They are read after the sheets, and only up to the highest index the rows read use, so previews of string-heavy files no longer parse the whole shared strings table: 412 → 166 ms on a 100,000 × 10 sheet of unique strings. Without `maxRows` nothing changes.
diff --git a/.changeset/pre/model-types-and-zip.md b/.changeset/pre/model-types-and-zip.md
deleted file mode 100644
index fd9fd4a..0000000
--- a/.changeset/pre/model-types-and-zip.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": minor
----
-
-Add the workbook model types and typed errors to the core entry. The model types (`Workbook`, `Sheet`, `Cell`, `CellInput`, `CellValue`, `Style`, `Font`, `Fill`, `Border`, `Alignment`, `NumFmt`, `Color`, `Theme`, `SheetView`, `PrintSetup`, `ColumnProps`, `RowProps`, `DocProperties`, `DefinedName` and the rest) are type-only, so they add no runtime code. Optional properties accept an explicit `undefined`, for projects using `exactOptionalPropertyTypes`. The errors are `ExcelError`, `ValidationError`, `FormatError`, `LimitError`, `CsvError` and `EnvironmentError`, with the `ExcelErrorCode` union and `isExcelError`, which also recognises errors from another ESM/CJS copy of the package or another realm. Internally, the package now has its own zip reader and writer built on the native `CompressionStream`/`DecompressionStream('deflate-raw')`, with no runtime dependencies; the zip codec is not part of the public API.
diff --git a/.changeset/pre/number-format-renderer.md b/.changeset/pre/number-format-renderer.md
deleted file mode 100644
index f8018d8..0000000
--- a/.changeset/pre/number-format-renderer.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": minor
----
-
-New entry `javascript-excel/format`: `formatValue(value, numFmt, options)` and `formatCell(cell, options)` render values the way Excel shows them in a cell. They cover number-format sections, conditions and colours, digit placeholders, grouping and scaling, percent, scientific, fractions, text, `_x` spacing and `*x` fills, General, dates and times from serials (time-zone free), elapsed time, built-in ids and locale tags. Separators, month and day names and the short date come from `Intl` for the given locale.
diff --git a/.changeset/pre/reader-yields.md b/.changeset/pre/reader-yields.md
deleted file mode 100644
index 3aaf7f6..0000000
--- a/.changeset/pre/reader-yields.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-`read` now returns to the event loop about every 16 ms while it parses, and feeds the decompressor in slices, so reading a large file on the main thread no longer freezes the page: frames render, timers and input run, and an abort from a timer or a click takes effect while the read is in flight. The resulting model is unchanged.
diff --git a/.changeset/pre/readme-and-api-docs.md b/.changeset/pre/readme-and-api-docs.md
deleted file mode 100644
index 97b465c..0000000
--- a/.changeset/pre/readme-and-api-docs.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-A full package README: install, quick starts for writing, datasets, reading, CSV, HTML tables and the worker (each type-checked in CI), what ships, runtime support and links to the new guides. `DetectedFormat` is now spelled out as a union in the published types instead of referring to an internal alias; the type is unchanged.
diff --git a/.changeset/pre/readme-playground-link.md b/.changeset/pre/readme-playground-link.md
deleted file mode 100644
index de15b74..0000000
--- a/.changeset/pre/readme-playground-link.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-README: link the live playground (https://javascript-excel-playground.vercel.app/).
diff --git a/.changeset/pre/render-model.md b/.changeset/pre/render-model.md
deleted file mode 100644
index 9645e5f..0000000
--- a/.changeset/pre/render-model.md
+++ /dev/null
@@ -1,11 +0,0 @@
----
-"javascript-excel": minor
----
-
-New entry `javascript-excel/render`: the render model a sheet viewer draws from. `createRenderSheet(workbook, sheet, options)` returns a handle with:
-- **geometry:** row and column offsets in pixels from prefix sums, plus hit testing;
-- **cells:** the merge map, concrete `#RRGGBB` styles cached per style object, resolved border edges, display text through an optional `format` hook (`formatCell` from `/format` fits), formula-bar text and hyperlink checks;
-- **navigation:** one-cell steps and Excel's Ctrl+arrow data edges;
-- **copy:** TSV/HTML builders and a selection summary.
-
-The core entry gains `resolveColor`, which resolves theme (with tint), indexed, RGB and automatic colours to ARGB.
diff --git a/.changeset/pre/render-row-heights.md b/.changeset/pre/render-row-heights.md
deleted file mode 100644
index 78abf35..0000000
--- a/.changeset/pre/render-row-heights.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": minor
----
-
-`javascript-excel/render`: `setRowHeights(sheet, heights)` sets measured heights for rows without a stored height (a viewer's auto-fit). It grows rows only, clamped to 409 pt, and rebuilds the offsets from the first changed row. `hasStoredRowHeight(sheet, row)` tells which rows the file sized.
diff --git a/.changeset/pre/serial-to-date-day-zero.md b/.changeset/pre/serial-to-date-day-zero.md
deleted file mode 100644
index e222d06..0000000
--- a/.changeset/pre/serial-to-date-day-zero.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-`serialToDate` accepts serials from 0 to below 1 in the 1900 date system (Excel's day 0, used for time-only values such as 0.5 = 12:00) and returns them on 1899-12-31. It used to throw `INVALID_DATE`.
diff --git a/.changeset/pre/sheet-names-code-points.md b/.changeset/pre/sheet-names-code-points.md
deleted file mode 100644
index fd551f1..0000000
--- a/.changeset/pre/sheet-names-code-points.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-Sheet names are limited to 31 code points, as in Excel, instead of 31 UTF-16 units. `validateSheetName`, the writer and `sanitizeSheetName` now accept a name such as 30 letters plus an emoji (Excel for Mac accepts it), and `sanitizeSheetName` cuts long names to 31 code points without splitting a surrogate pair. `History` stays reserved in any case (Excel refuses it), and `sanitizeSheetName` turns it into `Sheet`.
diff --git a/.changeset/pre/style-cache-plain-objects.md b/.changeset/pre/style-cache-plain-objects.md
deleted file mode 100644
index 47bbf11..0000000
--- a/.changeset/pre/style-cache-plain-objects.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-The writer's content-keyed style cache now applies only to styles made of plain objects. A style containing an object with `toJSON` (or another non-plain object) could share a cache key with a valid style and skip validation: an invalid colour `{ toJSON: () => "FF0000" }` was written as red instead of throwing `INVALID_STYLE`. Such styles are now resolved as they are.
diff --git a/.changeset/pre/todataset-skip-totals.md b/.changeset/pre/todataset-skip-totals.md
deleted file mode 100644
index 3074500..0000000
--- a/.changeset/pre/todataset-skip-totals.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": minor
----
-
-`toDataset` gains `skipTotals: 'auto' | boolean`. The default `'auto'` drops a trailing row whose formulas are all `SUBTOTAL(…)`, which is the totals row `fromDataset` writes. So `toDataset(read(write(fromDataset(ds))))` no longer returns the totals as a data row. Pass `skipTotals: false` to keep it.
diff --git a/.changeset/pre/two-row-csv-headers.md b/.changeset/pre/two-row-csv-headers.md
deleted file mode 100644
index 4c567f3..0000000
--- a/.changeset/pre/two-row-csv-headers.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-`detectHeader` (and `importDataset` with `header: 'auto'`) now recognises a two-row header without merges, as CSV files have: a second all-text row whose text cells all sit on columns of typed data joins the header (`Region,Sales,Sales` over `,Q1,Q2` gives `Sales / Q1`, `Sales / Q2`). Two text rows that only fill the table together (`Name,Q1,` over `,Jan,Feb`) are also a two-row header. When the row under the header is all text but only partly fits, it stays data and the confidence is `'low'`.
diff --git a/.changeset/pre/unmapped-fields-null.md b/.changeset/pre/unmapped-fields-null.md
deleted file mode 100644
index b151bf2..0000000
--- a/.changeset/pre/unmapped-fields-null.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": minor
----
-
-`importDataset` and `applyMapping`: a schema field without a column is now `null` in every row, and `columns` lists every schema field. An unmapped field's column has an empty title and letter, `type: 'empty'`, and its `field`. Rows always have the schema's shape, so validators such as Zod's `.nullable()` accept them; before this change the key was missing. A required field without a column still gives one `MAPPING_MISSING` issue, and is `null` in the rows.
diff --git a/.changeset/pre/worker-and-performance.md b/.changeset/pre/worker-and-performance.md
deleted file mode 100644
index b9f06c7..0000000
--- a/.changeset/pre/worker-and-performance.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": minor
----
-
-Add `javascript-excel/worker`: `exposeWorker()` for the app's two-line worker file and `createWorkerClient(worker)` with `write`, `read`, `toCsv` and `parseCsv`; progress, warnings and AbortSignal work across the boundary, results are transferred, errors keep their class and code. Performance: the shared strings table is written in chunks (no long main-thread block for large tables), CRC-32 uses slicing-by-8, and `onProgress` is throttled to one event per 50 ms.
diff --git a/.changeset/pre/writer-yields.md b/.changeset/pre/writer-yields.md
deleted file mode 100644
index a448356..0000000
--- a/.changeset/pre/writer-yields.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": patch
----
-
-`toUint8Array` and `toBlob` now return to the event loop about every 16 ms while they write, so a large export on the main thread no longer freezes the page: frames render, timers and input run, and an abort from a timer or a click takes effect while the write is in flight. The output bytes are unchanged.
diff --git a/.changeset/pre/xlsx-reader.md b/.changeset/pre/xlsx-reader.md
deleted file mode 100644
index 13beca7..0000000
--- a/.changeset/pre/xlsx-reader.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": minor
----
-
-Add `javascript-excel/read`: `read`, `readSheetNames`, `detectFormat` and `DEFAULT_READ_LIMITS`. Reads .xlsx (and .xlsm, and ISO 29500 Strict) into the workbook model: values, shared/inline/rich strings, cached formula values and formula text with shared formulas expanded, styles as shared frozen objects, merges, column widths, row heights, freeze panes, views, print setup, hyperlinks, defined names, document properties and both date systems. Untrusted input is bounded by measured limits (zip sizes and ratio counted on inflated bytes, XML depth, cells, shared strings), and the streaming XML tokenizer rejects DOCTYPE. .xls, .xlsb, .ods and encrypted files are rejected with UNSUPPORTED_FORMAT.
diff --git a/.changeset/pre/xlsx-writer-cells-and-styles.md b/.changeset/pre/xlsx-writer-cells-and-styles.md
deleted file mode 100644
index d88034e..0000000
--- a/.changeset/pre/xlsx-writer-cells-and-styles.md
+++ /dev/null
@@ -1,5 +0,0 @@
----
-"javascript-excel": minor
----
-
-Add the xlsx writer: `toUint8Array` and `toBlob`. It writes strings (shared or inline), numbers, booleans, dates (1900 and 1904 systems), error values, rich text, formulas with cached values and styled empty cells. It also writes the full style model: fonts, pattern and gradient fills, all border styles with a colour per side, alignment, protection, and built-in and custom number formats. Styles are layered sheet → column → row → cell and deduplicated. Sheet order, names, visibility, the active sheet, the default font and document properties are written too, and the model and options are validated with typed errors. Also adds `dateToSerial`, `serialToDate`, `mergeStyles` and the A1 address helpers, and exports the shared `BinaryInput`, `BlobType`, `AbortSignalLike` and `ExcelWarning` types from the core entry.
diff --git a/.changeset/pre/xlsx-writer-sheet-features.md b/.changeset/pre/xlsx-writer-sheet-features.md
deleted file mode 100644
index 93aea40..0000000
--- a/.changeset/pre/xlsx-writer-sheet-features.md
+++ /dev/null
@@ -1,11 +0,0 @@
----
-"javascript-excel": minor
----
-
-The xlsx writer now writes sheet features:
-- column widths (characters, pixels or the new `'auto'`), column and row styles, row heights, hidden rows and columns, outline levels;
-- merges, autofilter, freeze panes and the sheet view (right-to-left, zoom, gridlines, selection), tab colours and default sizes;
-- hyperlinks, external or internal. External URLs are limited to `http`, `https` and `mailto` unless `hyperlinkSchemes` allows more;
-- defined names, and print setup: orientation, paper size, fit to page, scale, margins, print area and titles, headers and footers.
-
-These model fields are no longer ignored, and they are validated with the new `INVALID_SHEET_PROPERTY` and `INVALID_DEFINED_NAME` codes. An invalid sheet `state` now reports `INVALID_SHEET_PROPERTY` instead of `INVALID_SHEET_NAME`. This release also adds `cellText`, `validateSheetName` and `sanitizeSheetName`.
diff --git a/README.md b/README.md
index a346b6e..b9aa3e2 100644
--- a/README.md
+++ b/README.md
@@ -1,8 +1,8 @@
# javascript-excel
-> **2.0 is a from-scratch rewrite, now in beta** (`2.0.0-beta.x` on the `next` dist-tag). The beta builds sheets from `{ columns, data }` datasets (`fromDataset`, `fromDatasets`, and `toDataset` for the way back), writes complete `.xlsx` files (cells, styles and sheet features), reads `.xlsx` files (`javascript-excel/read`), parses and serializes CSV (`javascript-excel/csv`), turns HTML tables into sheets and downloads files (`javascript-excel/dom`), runs big jobs in a Web Worker (`javascript-excel/worker`), renders values with Excel number formats (`javascript-excel/format`), computes what a sheet viewer draws (`javascript-excel/render`), and imports users' `.xlsx`/`.csv` files as datasets (`javascript-excel/import`). The 2.0.0 feature set is complete; changes before 2.0.0 should be limited to fixes found in testing.
+> **2.0 is a from-scratch rewrite** (`2.0.0` on the `latest` dist-tag). It builds sheets from `{ columns, data }` datasets (`fromDataset`, `fromDatasets`, and `toDataset` for the way back), writes complete `.xlsx` files (cells, styles and sheet features), reads `.xlsx` files (`javascript-excel/read`), parses and serializes CSV (`javascript-excel/csv`), turns HTML tables into sheets and downloads files (`javascript-excel/dom`), runs big jobs in a Web Worker (`javascript-excel/worker`), renders values with Excel number formats (`javascript-excel/format`), computes what a sheet viewer draws (`javascript-excel/render`), and imports users' `.xlsx`/`.csv` files as datasets (`javascript-excel/import`).
>
-> The last 1.x release is [`javascript-excel@1.0.3`](https://www.npmjs.com/package/javascript-excel/v/1.0.3) on npm. Keep using it until 2.0 is released.
+> The last 1.x release is [`javascript-excel@1.0.3`](https://www.npmjs.com/package/javascript-excel/v/1.0.3). 2.0 does not share its API; the [2.0.0 changelog](packages/javascript-excel/CHANGELOG.md#200) lists the breaking changes.
## Goal
diff --git a/apps/playground/package.json b/apps/playground/package.json
index a2b81c2..8ac6c67 100644
--- a/apps/playground/package.json
+++ b/apps/playground/package.json
@@ -11,7 +11,7 @@
"test": "vitest run"
},
"dependencies": {
- "javascript-excel": "2.0.0-beta.1"
+ "javascript-excel": "2.0.0"
},
"devDependencies": {
"happy-dom": "21.0.4",
diff --git a/docs/api.md b/docs/api.md
index 2a404d1..424de04 100644
--- a/docs/api.md
+++ b/docs/api.md
@@ -1684,7 +1684,7 @@ The limits are on by default. Raising them is an explicit `limits` option, and t
### 9.1 Versioning
-- Semantic Versioning. 2.0.0 is a breaking rewrite of 1.0.3. Prereleases ship on the `next` dist-tag: `2.0.0-alpha.1` to `2.0.0-alpha.7` while the API was still changing, then `2.0.0-beta.n` from `2.0.0-beta.0`, with the 2.0.0 feature set complete and only fixes expected. `latest` stays on 1.0.3 until 2.0.0 is final.
+- Semantic Versioning. 2.0.0 is a breaking rewrite of 1.0.3. Prereleases ship on the `next` dist-tag: `2.0.0-alpha.1` to `2.0.0-alpha.7` while the API was still changing, then `2.0.0-beta.0` and `2.0.0-beta.1`, with the 2.0.0 feature set complete and only fixes. `2.0.0` is on `latest`; later prereleases (for example of 2.1) ship on `next`.
- **Public API:** the exports of the five entry points, their TypeScript types, the error codes and warning codes, and the behaviour documented here. `ZipCodec` and everything not exported are internal.
- **Breaking (major):** removing or renaming an export, option or code; a narrower accepted input type; a different default that changes output (for example the injection guard or `dateZone`); dropping a runtime from §9.3; required fields added to input types.
- **Not breaking (minor):** new exports, options and optional model fields; the reader populating new optional fields (consumers must tolerate unknown fields); new warning codes; new error codes for inputs that were previously undefined behaviour.
diff --git a/examples/node/package.json b/examples/node/package.json
index 7e2f1a2..50f0562 100644
--- a/examples/node/package.json
+++ b/examples/node/package.json
@@ -8,7 +8,7 @@
"check": "tsc -p tsconfig.json"
},
"dependencies": {
- "javascript-excel": "2.0.0-beta.1"
+ "javascript-excel": "2.0.0"
},
"devDependencies": {
"@types/node": "22.20.5",
diff --git a/examples/react/package.json b/examples/react/package.json
index 8f55698..ae5f78b 100644
--- a/examples/react/package.json
+++ b/examples/react/package.json
@@ -9,7 +9,7 @@
"check": "tsc -p tsconfig.json"
},
"dependencies": {
- "javascript-excel": "2.0.0-beta.1",
+ "javascript-excel": "2.0.0",
"react": "19.3.1",
"react-dom": "19.3.1"
},
diff --git a/examples/snippets/package.json b/examples/snippets/package.json
index eb9bd64..edf34c0 100644
--- a/examples/snippets/package.json
+++ b/examples/snippets/package.json
@@ -7,7 +7,7 @@
"check": "tsc -p tsconfig.json"
},
"dependencies": {
- "javascript-excel": "2.0.0-beta.1"
+ "javascript-excel": "2.0.0"
},
"devDependencies": {
"@types/node": "22.20.5",
diff --git a/examples/vanilla/package.json b/examples/vanilla/package.json
index f090c9a..b64b9f3 100644
--- a/examples/vanilla/package.json
+++ b/examples/vanilla/package.json
@@ -9,7 +9,7 @@
"check": "tsc -p tsconfig.json"
},
"dependencies": {
- "javascript-excel": "2.0.0-beta.1"
+ "javascript-excel": "2.0.0"
},
"devDependencies": {
"typescript": "7.0.2",
diff --git a/examples/vue/package.json b/examples/vue/package.json
index c0bfc74..6ea5a10 100644
--- a/examples/vue/package.json
+++ b/examples/vue/package.json
@@ -9,7 +9,7 @@
"check": "tsc -p tsconfig.json"
},
"dependencies": {
- "javascript-excel": "2.0.0-beta.1",
+ "javascript-excel": "2.0.0",
"vue": "3.5.45"
},
"devDependencies": {
diff --git a/package-lock.json b/package-lock.json
index b24daa7..a2d3366 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -23,7 +23,7 @@
"name": "javascript-excel-playground",
"version": "0.0.0",
"dependencies": {
- "javascript-excel": "2.0.0-beta.1"
+ "javascript-excel": "2.0.0"
},
"devDependencies": {
"happy-dom": "21.0.4",
@@ -35,7 +35,7 @@
"examples/node": {
"name": "example-node",
"dependencies": {
- "javascript-excel": "2.0.0-beta.1"
+ "javascript-excel": "2.0.0"
},
"devDependencies": {
"@types/node": "22.20.5",
@@ -45,7 +45,7 @@
"examples/react": {
"name": "example-react",
"dependencies": {
- "javascript-excel": "2.0.0-beta.1",
+ "javascript-excel": "2.0.0",
"react": "19.3.1",
"react-dom": "19.3.1"
},
@@ -59,7 +59,7 @@
"examples/snippets": {
"name": "example-snippets",
"dependencies": {
- "javascript-excel": "2.0.0-beta.1"
+ "javascript-excel": "2.0.0"
},
"devDependencies": {
"@types/node": "22.20.5",
@@ -69,7 +69,7 @@
"examples/vanilla": {
"name": "example-vanilla",
"dependencies": {
- "javascript-excel": "2.0.0-beta.1"
+ "javascript-excel": "2.0.0"
},
"devDependencies": {
"typescript": "7.0.2",
@@ -79,7 +79,7 @@
"examples/vue": {
"name": "example-vue",
"dependencies": {
- "javascript-excel": "2.0.0-beta.1",
+ "javascript-excel": "2.0.0",
"vue": "3.5.45"
},
"devDependencies": {
@@ -3696,7 +3696,7 @@
}
},
"packages/javascript-excel": {
- "version": "2.0.0-beta.1",
+ "version": "2.0.0",
"license": "MIT",
"devDependencies": {
"@types/node": "22.20.5",
diff --git a/packages/javascript-excel/CHANGELOG.md b/packages/javascript-excel/CHANGELOG.md
index d19f265..6cd4381 100644
--- a/packages/javascript-excel/CHANGELOG.md
+++ b/packages/javascript-excel/CHANGELOG.md
@@ -1,5 +1,99 @@
# javascript-excel
+## 2.0.0
+
+2.0 is a from-scratch rewrite. It reads and writes `.xlsx` and `.csv` in browsers, Web Workers and Node.js, in TypeScript, with no runtime dependencies.
+
+### Breaking changes from 1.x
+
+1.0.3 exported a single default function, `downloadExcel()`, that zipped a placeholder text file with `jszip` and saved it with `file-saver`. None of that API remains:
+
+- **The default export is gone.** There is no default export; use the named exports of the entry points (`javascript-excel`, `/read`, `/csv`, `/dom`, `/worker`, `/format`, `/render`, `/import`). To save a file in the browser, build a workbook, then call `toBlob` or `toUint8Array` and `download` from `javascript-excel/dom`.
+- **No runtime dependencies.** `jszip` and `file-saver` are no longer installed with the package. The zip codec is built in, on the native `CompressionStream`/`DecompressionStream`.
+- **ESM and CommonJS builds with types, from `exports`.** The old `lib/` paths (`lib/javascript-excel.js`, `lib/index.js`) are gone; deep imports outside the documented entry points are not supported.
+- **Runtimes:** Node.js 22 or later, and current browsers with `CompressionStream` (`engines.node` is `>=22`). Consumer TypeScript 5.0 or later.
+- **File I/O is async.** Writers and readers return promises.
+
+The full API is in [docs/api.md](https://github.com/securedeveloper/javascript-excel/blob/main/docs/api.md). Prereleases `2.0.0-alpha.1` to `2.0.0-beta.1` stay on the `next` dist-tag; their entries below are folded into this section.
+
+### Minor Changes
+
+- 6a8ee72: Add `javascript-excel/csv`: `parseCsv`, `readCsv`, `toCsv` and `toCsvBlob`. RFC 4180 parsing with delimiter detection, BOM and CRLF/LF/CR handling, lenient or strict modes, optional type inference, and an async `readCsv` with encoding detection and AbortSignal support. Serialization with quoting, BOM, ISO dates and a formula-injection guard that is on by default.
+- 8e8348f: Add the dataset builders:
+ - `fromDataset({ columns, data }, options)`: `key`, `path` and function accessors; header groups as merged header cells; falsy (conditional) columns skipped; per-column `type` coercion, `numFmt`, `width` (including `'auto'`), `hidden` and styles.
+ - Freeze and autofilter on the header rows, and a totals row with `SUBTOTAL` formulas whose cached values are computed from the data.
+ - `fromDatasets` stacks several blocks on one sheet.
+ - `toDataset(sheet)` reads a sheet back into `{ columns, data }` with null-prototype rows.
+- 252a056: `fromDatasets` blocks take an optional `placement: 'below' | 'right'`. `'below'` (the default) is the existing behaviour. `'right'` puts a block on the previous block's top row, `gap` empty columns after its last column, with `offset.cols` and `offset.rows` adding to that. A `'below'` block after `'right'` blocks starts below the lowest of them, so blocks do not overlap. Existing calls give the same output as before.
+- 9f81c95: Add `javascript-excel/dom`:
+ - `fromTable(table, options)` turns an HTML `` into a sheet. It reads the rows of ``, then ``, then ``, and turns `colspan`/`rowspan` into merges (including `rowspan="0"`). It understands `data-type`, `data-value` and `data-numfmt` hints, and skips hidden and `data-export-ignore` rows and cells.
+ - In a cell, `
` becomes a newline with wrapping, and the first link becomes the cell's hyperlink, limited to `http`, `https` and `mailto` unless `hyperlinkSchemes` allows more.
+ - Optional computed-style mapping covers bold, italic, underline, strike, colour, background (including a row's background) and explicit alignment. There are also header, freeze, autofilter and auto-width options, a per-cell hook, and `onWarning` for data that cannot be coerced or for dropped links.
+ - `download(data, filename)` saves bytes, a Blob or a string through a temporary `` and an object URL, with no dependencies. The module never touches globals at import.
+
+ Also adds the `HYPERLINK_DROPPED` warning code and the `AutoWidthOptions` type.
+- 4669375: New entry `javascript-excel/import` turns a user's `.xlsx` or `.csv` file into `{ columns, data }` with `importDataset`. It handles:
+
+ - the format, detected from the bytes
+ - the sheet: the first visible one, or one picked by name or index
+ - the header row, detected (title rows skipped, multi-row headers flattened)
+ - column types, inferred all-or-nothing, with explicit number and date locales
+ - column mapping to a schema by name or alias
+ - per-row issues with their cell refs, and `maxRows` / `preview`
+
+ The building blocks are exported too: `detectHeader`, `inferColumnTypes`, `coerceColumns`, `parseNumber`, `parseDate`, `mapColumns` and `applyMapping`.
+
+ Other additions:
+
+ - `read` gains `maxRows`. Rows past it are skipped without tokenizing them, while merges and sheet views are still read. It comes with a `ROWS_TRUNCATED` warning.
+ - `toDataset` gains `headerRows` and `join` for multi-row headers.
+ - The worker client gains `importDataset`.
+ - New error codes: `SHEET_NOT_FOUND`, and `VALIDATION_FAILED` (reserved for react-data-import).
+- 07706d3: `importDataset` issues now name the field as it appears in `data` rows. With a schema, `TYPE`, `ERROR_CELL` and `DATE_ORDER_AMBIGUOUS` issues use the schema field name; they used the source column key. The new optional `column` property on `ImportIssue` holds the source column key (also set by `coerceColumns` and on `applyMapping`'s `REQUIRED`). A value that doesn't convert to a forced type in a required field gives one `TYPE` issue, no longer a `REQUIRED` issue as well. Issues on columns the mapping drops are left out.
+
+ Migration: if you grouped issues by the source column key, group by `issue.column`.
+- 6e84545: Add the workbook model types and typed errors to the core entry. The model types (`Workbook`, `Sheet`, `Cell`, `CellInput`, `CellValue`, `Style`, `Font`, `Fill`, `Border`, `Alignment`, `NumFmt`, `Color`, `Theme`, `SheetView`, `PrintSetup`, `ColumnProps`, `RowProps`, `DocProperties`, `DefinedName` and the rest) are type-only, so they add no runtime code. Optional properties accept an explicit `undefined`, for projects using `exactOptionalPropertyTypes`. The errors are `ExcelError`, `ValidationError`, `FormatError`, `LimitError`, `CsvError` and `EnvironmentError`, with the `ExcelErrorCode` union and `isExcelError`, which also recognises errors from another ESM/CJS copy of the package or another realm. Internally, the package now has its own zip reader and writer built on the native `CompressionStream`/`DecompressionStream('deflate-raw')`, with no runtime dependencies; the zip codec is not part of the public API.
+- 68d2df8: New entry `javascript-excel/format`: `formatValue(value, numFmt, options)` and `formatCell(cell, options)` render values the way Excel shows them in a cell. They cover number-format sections, conditions and colours, digit placeholders, grouping and scaling, percent, scientific, fractions, text, `_x` spacing and `*x` fills, General, dates and times from serials (time-zone free), elapsed time, built-in ids and locale tags. Separators, month and day names and the short date come from `Intl` for the given locale.
+- 7f50ecf: New entry `javascript-excel/render`: the render model a sheet viewer draws from. `createRenderSheet(workbook, sheet, options)` returns a handle with:
+ - **geometry:** row and column offsets in pixels from prefix sums, plus hit testing;
+ - **cells:** the merge map, concrete `#RRGGBB` styles cached per style object, resolved border edges, display text through an optional `format` hook (`formatCell` from `/format` fits), formula-bar text and hyperlink checks;
+ - **navigation:** one-cell steps and Excel's Ctrl+arrow data edges;
+ - **copy:** TSV/HTML builders and a selection summary.
+
+ The core entry gains `resolveColor`, which resolves theme (with tint), indexed, RGB and automatic colours to ARGB.
+- 3fcf25b: `javascript-excel/render`: `setRowHeights(sheet, heights)` sets measured heights for rows without a stored height (a viewer's auto-fit). It grows rows only, clamped to 409 pt, and rebuilds the offsets from the first changed row. `hasStoredRowHeight(sheet, row)` tells which rows the file sized.
+- f0c9cf1: `toDataset` gains `skipTotals: 'auto' | boolean`. The default `'auto'` drops a trailing row whose formulas are all `SUBTOTAL(…)`, which is the totals row `fromDataset` writes. So `toDataset(read(write(fromDataset(ds))))` no longer returns the totals as a data row. Pass `skipTotals: false` to keep it.
+- 6ce413a: `importDataset` and `applyMapping`: a schema field without a column is now `null` in every row, and `columns` lists every schema field. An unmapped field's column has an empty title and letter, `type: 'empty'`, and its `field`. Rows always have the schema's shape, so validators such as Zod's `.nullable()` accept them; before this change the key was missing. A required field without a column still gives one `MAPPING_MISSING` issue, and is `null` in the rows.
+- e6a4a94: Add `javascript-excel/worker`: `exposeWorker()` for the app's two-line worker file and `createWorkerClient(worker)` with `write`, `read`, `toCsv` and `parseCsv`; progress, warnings and AbortSignal work across the boundary, results are transferred, errors keep their class and code. Performance: the shared strings table is written in chunks (no long main-thread block for large tables), CRC-32 uses slicing-by-8, and `onProgress` is throttled to one event per 50 ms.
+- bf744b1: Add `javascript-excel/read`: `read`, `readSheetNames`, `detectFormat` and `DEFAULT_READ_LIMITS`. Reads .xlsx (and .xlsm, and ISO 29500 Strict) into the workbook model: values, shared/inline/rich strings, cached formula values and formula text with shared formulas expanded, styles as shared frozen objects, merges, column widths, row heights, freeze panes, views, print setup, hyperlinks, defined names, document properties and both date systems. Untrusted input is bounded by measured limits (zip sizes and ratio counted on inflated bytes, XML depth, cells, shared strings), and the streaming XML tokenizer rejects DOCTYPE. .xls, .xlsb, .ods and encrypted files are rejected with UNSUPPORTED_FORMAT.
+- 54d805b: Add the xlsx writer: `toUint8Array` and `toBlob`. It writes strings (shared or inline), numbers, booleans, dates (1900 and 1904 systems), error values, rich text, formulas with cached values and styled empty cells. It also writes the full style model: fonts, pattern and gradient fills, all border styles with a colour per side, alignment, protection, and built-in and custom number formats. Styles are layered sheet → column → row → cell and deduplicated. Sheet order, names, visibility, the active sheet, the default font and document properties are written too, and the model and options are validated with typed errors. Also adds `dateToSerial`, `serialToDate`, `mergeStyles` and the A1 address helpers, and exports the shared `BinaryInput`, `BlobType`, `AbortSignalLike` and `ExcelWarning` types from the core entry.
+- 8ab2b97: The xlsx writer now writes sheet features:
+ - column widths (characters, pixels or the new `'auto'`), column and row styles, row heights, hidden rows and columns, outline levels;
+ - merges, autofilter, freeze panes and the sheet view (right-to-left, zoom, gridlines, selection), tab colours and default sizes;
+ - hyperlinks, external or internal. External URLs are limited to `http`, `https` and `mailto` unless `hyperlinkSchemes` allows more;
+ - defined names, and print setup: orientation, paper size, fit to page, scale, margins, print area and titles, headers and footers.
+
+ These model fields are no longer ignored, and they are validated with the new `INVALID_SHEET_PROPERTY` and `INVALID_DEFINED_NAME` codes. An invalid sheet `state` now reports `INVALID_SHEET_PROPERTY` instead of `INVALID_SHEET_NAME`. This release also adds `cellText`, `validateSheetName` and `sanitizeSheetName`.
+
+### Patch Changes
+
+- README: drop the beta notice; install from `latest`.
+- 46ad4e1: `fromDataset` shares style objects across rows (one per column, and one per column and `rowStyle` object for banding) instead of building a new one per cell. The writer then resolves each distinct style once from its identity cache. In a local run of 1M rows × 4 styled, banded columns, building went from 2.6 s to 0.5 s and writing from 11.7 s to 7.3 s.
+- c100c46: Built-in number formats 37–40 and 47 now use the codes Excel itself shows, checked in Excel for Mac: `#,##0_);(#,##0)`, `#,##0_);[Red](#,##0)`, `#,##0.00_);(#,##0.00)`, `#,##0.00_);[Red](#,##0.00)` and `mm:ss.0`. The ECMA-376 table lists other codes. This changes `BUILTIN_NUM_FMTS` and what `formatValue`/`formatCell` show for those ids: positive numbers keep a trailing space where the `)` would be (`1,234.50 `), and 47 shows `31:00.0` instead of `3100.0`. Files are unaffected, because the writer writes only the id.
+- 8886d76: `fromTable` with `styles: 'computed'` now maps text and background colours that browsers report as `lab()`, `lch()`, `oklab()`, `oklch()` or `color()` (for example Tailwind v4 colours), converting them to sRGB. Before, only `rgb()`, hex and named colours were kept.
+- 6e37bc8: `dateToSerial`, and so the writer, now accepts 1899-12-31, Excel's day 0, as serials 0 to below 1. `serialToDate` returns time-only values on that day, so a time-only cell read back by `read()` can now be written again; it used to throw `INVALID_DATE`.
+- 3d1b402: The `exports` map now names the type declarations explicitly: each subpath has `import: { types: "./dist/x.d.ts", default: "./dist/x.js" }` and `require: { types: "./dist/x.d.cts", default: "./dist/x.cjs" }`. TypeScript already found the declarations next to the JavaScript files; the explicit `types` conditions don't depend on that lookup.
+- 5020077: `toDataset` with `headerRows` (and `importDataset` with a multi-row header) now carries a group title right across the empty upper-row cells that follow it when there are no merges, as in CSV: `Name,Q1,` over `,Jan,Feb` gives `Name`, `Q1 / Jan`, `Q1 / Feb` instead of `Name`, `Q1 / Jan`, `Feb`. Merged cells still decide where they are present, and a column with no text in the lower header rows is not filled.
+- f26fe0c: `read` with `maxRows` decodes shared strings lazily. They are read after the sheets, and only up to the highest index the rows read use, so previews of string-heavy files no longer parse the whole shared strings table: 412 → 166 ms on a 100,000 × 10 sheet of unique strings. Without `maxRows` nothing changes.
+- 938a6c6: `read` now returns to the event loop about every 16 ms while it parses, and feeds the decompressor in slices, so reading a large file on the main thread no longer freezes the page: frames render, timers and input run, and an abort from a timer or a click takes effect while the read is in flight. The resulting model is unchanged.
+- 21d9d36: A full package README: install, quick starts for writing, datasets, reading, CSV, HTML tables and the worker (each type-checked in CI), what ships, runtime support and links to the new guides. `DetectedFormat` is now spelled out as a union in the published types instead of referring to an internal alias; the type is unchanged.
+- 244577a: README: link the live playground (https://javascript-excel-playground.vercel.app/).
+- bc50cd3: `serialToDate` accepts serials from 0 to below 1 in the 1900 date system (Excel's day 0, used for time-only values such as 0.5 = 12:00) and returns them on 1899-12-31. It used to throw `INVALID_DATE`.
+- 15f8238: Sheet names are limited to 31 code points, as in Excel, instead of 31 UTF-16 units. `validateSheetName`, the writer and `sanitizeSheetName` now accept a name such as 30 letters plus an emoji (Excel for Mac accepts it), and `sanitizeSheetName` cuts long names to 31 code points without splitting a surrogate pair. `History` stays reserved in any case (Excel refuses it), and `sanitizeSheetName` turns it into `Sheet`.
+- 65c5275: The writer's content-keyed style cache now applies only to styles made of plain objects. A style containing an object with `toJSON` (or another non-plain object) could share a cache key with a valid style and skip validation: an invalid colour `{ toJSON: () => "FF0000" }` was written as red instead of throwing `INVALID_STYLE`. Such styles are now resolved as they are.
+- f7e976d: `detectHeader` (and `importDataset` with `header: 'auto'`) now recognises a two-row header without merges, as CSV files have: a second all-text row whose text cells all sit on columns of typed data joins the header (`Region,Sales,Sales` over `,Q1,Q2` gives `Sales / Q1`, `Sales / Q2`). Two text rows that only fill the table together (`Name,Q1,` over `,Jan,Feb`) are also a two-row header. When the row under the header is all text but only partly fits, it stays data and the confidence is `'low'`.
+- 33ba222: `toUint8Array` and `toBlob` now return to the event loop about every 16 ms while they write, so a large export on the main thread no longer freezes the page: frames render, timers and input run, and an abort from a timer or a click takes effect while the write is in flight. The output bytes are unchanged.
+
## 2.0.0-beta.1
The conformance corpus was checked by hand in **Excel for Mac** (2026-11-17): every file opens with no repair prompt. The results are in `docs/conformance.md`. Excel for Windows and Apple Numbers were not checked by hand; for them only the automated LibreOffice and openpyxl checks apply. What changes for users:
diff --git a/packages/javascript-excel/README.md b/packages/javascript-excel/README.md
index cd228c0..2421607 100644
--- a/packages/javascript-excel/README.md
+++ b/packages/javascript-excel/README.md
@@ -2,10 +2,10 @@
Read and write `.xlsx` and `.csv` files in browsers, Web Workers and Node.js. TypeScript, no runtime dependencies, tree-shakable entry points.
-> **Beta.** `2.0.0-beta.x` is a from-scratch rewrite. Everything listed below ships today. The 2.0.0 feature set is complete; changes before 2.0.0 should be limited to fixes found in testing. The last 1.x release is `javascript-excel@1.0.3`.
+> 2.0 is a from-scratch rewrite. It shares nothing with 1.x (`javascript-excel@1.0.3`); see the [2.0.0 changelog](https://github.com/securedeveloper/javascript-excel/blob/main/packages/javascript-excel/CHANGELOG.md#200) for the breaking changes.
```sh
-npm install javascript-excel@next
+npm install javascript-excel
```
**Try it in the browser:** the [playground](https://javascript-excel-playground.vercel.app/) writes, reads and previews workbooks with the published package, and shows the code for what you build.
diff --git a/packages/javascript-excel/package.json b/packages/javascript-excel/package.json
index 15cfa83..3f693c0 100644
--- a/packages/javascript-excel/package.json
+++ b/packages/javascript-excel/package.json
@@ -1,6 +1,6 @@
{
"name": "javascript-excel",
- "version": "2.0.0-beta.1",
+ "version": "2.0.0",
"private": false,
"description": "Read and write .xlsx and .csv files in browsers, Web Workers and Node. 2.0 is a from-scratch rewrite in progress.",
"keywords": [
diff --git a/packages/javascript-excel/src/index.ts b/packages/javascript-excel/src/index.ts
index a97024b..c70cbbe 100644
--- a/packages/javascript-excel/src/index.ts
+++ b/packages/javascript-excel/src/index.ts
@@ -75,4 +75,4 @@ export { decodeCell, decodeCol, decodeRange, encodeCell, encodeCol, encodeRange,
export type { CellAddress, RangeAddress } from "./address";
/** Package version. Set from package.json by `npm run version-packages` (scripts/sync-version.mjs); a unit test checks it. */
-export const VERSION = "2.0.0-beta.1";
+export const VERSION = "2.0.0";