Excel File Repair (Recover Data from a Corrupt XLSX or XLS)

About the Excel File Repair (Recover Data from a Corrupt XLSX or XLS)

An XLSX workbook is a ZIP archive of XML files: one per sheet, one for the text the cells share, one for the styles, and a few that tie them together (ECMA-376 Office Open XML). Excel reads it strictly — one cut-off file inside, one character XML does not allow, one link to a part that is not there — and it says it found a problem with some content, or that it cannot open the file because the file format or file extension is not valid.

This page takes such a workbook apart and rebuilds it: it finds every part of the archive even when its end is missing, repairs each XML part, keeps every complete row of a sheet that was cut short, leaves out what is beyond saving (a damaged chart, a stale calculation chain), ties what is left back together, and reads the result back with a second, independent spreadsheet reader to check it. It handles old XLS workbooks too — it cleans their record stream and reads it again — and CSV files whose encoding or quoting went wrong. You see the report and the first rows of every sheet before anything is saved, and nothing is uploaded.

Common uses

• Workbooks that will not open after a broken download, an interrupted copy or a full disk while saving.
• Workbooks recovered by a deep scan of a disk image and marked partly damaged.
• Exports from other programs that Excel refuses or only opens after its own repair message.
• Old XLS workbooks that Excel calls corrupt.
• CSV files that open as one column, with strange characters, or with half their rows missing.

What the repair does to an XLSX

• Finds the parts. The archive is read from the front, so parts are found even when its end is missing (see ZIP Archive Repair for how). A part cut short is decoded as far as it goes.
• Repairs the XML. Characters XML forbids are removed, bare “&” signs escaped, elements left open closed. A sheet cut short keeps every complete row (<row>), the shared text every complete entry; nothing half-written is kept.
• Leaves out what cannot be saved. The calculation chain is always left out — it is only a cache, and Excel rebuilds it. A damaged styles part is left out with every reference to it (the values stay, the formatting goes). A damaged chart, drawing, comment or pivot table is left out with every link to it.
• Ties it back together. Links to parts that are gone are taken out, sheets without a part leave the workbook’s list, the list of content types is written again, and a workbook part that is missing is written again from the sheets that survived.
• Checks it. The repaired workbook is read back by a second, independent reader (SheetJS). When it reads, the page says so; when it does not, the values come from this page’s own reader and the values-and-formulas workbook is the one to use.

Repaired workbook, or values and formulas?

The repaired workbook is your own file with its damage taken out: column widths, formatting, charts, tables, pictures and macros that survived are still in it. It is the one to try first.

The values and formulas workbook is written new from what was read back: every value, every formula as text, number formats, sheet names. It has none of the damaged file’s structure, so it is the safest to open, and it is the result for XLS, XLSB, ODS and CSV files.

Old XLS workbooks

An Excel 97–2003 workbook is a compound file — a small file system inside one file — holding a stream of records: the workbook’s settings and shared text first, then each sheet (MS-XLS, MS-CFB). The page follows the compound file’s chains as far as they hold, then cleans the record stream: whole records are kept, damaged stretches skipped, a stream cut short closed off, sheets whose records are gone dropped from the list, and the positions of the others fixed. A full reader then reads it, formulas included; when even that fails, the values are read directly from the records.

What the repair is measured against

The page’s automated tests write workbooks with SheetJS, lay them out the way Excel orders its parts, and break them: cut short in the middle of a sheet, with characters XML forbids in the shared text, with the styles part cut in half and a stale calculation chain added, with most of the shared text missing, and with the workbook part gone. Old XLS workbooks are cut short and have a stretch zeroed in the middle; CSV files get double-encoded text, a quote that is never closed and zero bytes. Every repaired workbook must open in SheetJS, and every recovered row must be exactly the original row.

No success rate is published for other people’s files: what comes back depends on how much of the workbook reached the disk.

Limitations

• Missing data stays missing: rows after the point where a sheet was cut off, and cells whose shared text did not survive, cannot be recreated (those cells come back empty and are counted).
• Charts, pivot tables, comments and drawings come back only when their own parts survived whole; a damaged one is left out, and the report says so.
• A workbook whose styles part is damaged comes back without its formatting (fonts, colours, borders, number formats); its values are kept.
• Password-protected workbooks are encrypted: they cannot be repaired without their password, and this page never tries to guess one.
• Macros are kept in the repaired workbook only when the macro project survived whole, and never in the values-and-formulas workbook.
• XLSB, ODS and old XLS files are rebuilt as XLSX values and formulas, not in their own format.
• A damaged record can decode as a value in an impossible place — outside an .xls sheet’s 256 columns, or far beyond every other cell. Such values are left out of the values-and-formulas workbook, and the report says how many.
• Files of up to 400 MB: the repair works on the whole file in memory.
• Excel and Microsoft are trademarks of Microsoft; this page is not made by or affiliated with Microsoft.

Privacy

The workbook is read by your browser on your device; nothing from it is sent anywhere, and the page works offline once it has loaded. The spreadsheet reader used to check the result is part of this site and runs on your device too.