Excel to JSON Converter

Modern .xlsx workbooks are SpreadsheetML files defined by ECMA-376 / ISO/IEC 29500 (the Office Open XML standard): the .xlsx extension is a ZIP container whose entries are XML documents under xl/worksheets/sheetN.xml, with strings stored in a shared-string table xl/sharedStrings.xml and cell types declared by the t attribute (s = sharedString, n = number, b = boolean, str = inline string, d = date in ISO 8601). The page reads the uploaded File with FileReader.readAsArrayBuffer, then hands the ArrayBuffer to SheetJS (xlsx). XLSX.read(buffer, {type: 'array'}) returns a workbook object whose SheetNames array drives the sheet selector, and XLSX.utils.sheet_to_json(sheet, options) walks the cell grid to produce either an array-of-objects (with the header row as keys) or an array-of-arrays when header: 1 is passed. Legacy .xls is the binary BIFF8 format (Excel 97-2003), which SheetJS also decodes through the same read() entry point but with type: 'binary' for FileReader.readAsBinaryString or 'array' when reading as bytes. Cell date values in both formats are stored as serial numbers (days since the 1900 epoch in Excel, with the historical 1900-02-29 leap-year bug that shifts dates before 1900-03-01 by one day, or days since 1904-01-01 when the workbook's date1904 flag is set on Excel for Mac). Conversion to a JavaScript Date is (serial - 25569) * 86400 * 1000 milliseconds after Unix epoch for the 1900 system once the leap-year bug is accounted for. Formulas live in the f attribute and SheetJS reports their last cached result in v; if the workbook was saved by a tool that did not recalculate (e.g. some scripted exports), v can be missing or stale. When the first row is used as headers, duplicate or empty header cells force SheetJS to either skip or auto-rename keys (e.g. __EMPTY, __EMPTY_1), which breaks downstream JSON contracts. Merged-cell ranges (declared in worksheet/mergeCells) expose the value only at the top-left anchor and emit undefined in the merged tail cells, so a merged title row produces one populated record followed by N - 1 sparse ones. For workbooks above tens of megabytes, the synchronous XLSX.read() path can block the UI thread; the streaming alternative is to move parsing into a Web Worker so the main thread stays responsive.

• .xlsx is a ZIP container of XML (ECMA-376 / ISO/IEC 29500); .xls is the binary BIFF8 format; SheetJS XLSX.read(buffer, {type: 'array'}) handles both via FileReader.readAsArrayBuffer.
• sheet_to_json options: header: 1 returns array-of-arrays; header: 'A' uses spreadsheet column letters; defval: null fills sparse cells with null instead of dropping the key; raw: false applies Excel display formatting.
• Excel date serial is days since 1900-01-01 (with the 1900-02-29 leap-year bug; for the Mac 1904 date system, days since 1904-01-01); JS Date ms = (serial - 25569) * 86400000 after Unix epoch.
• Duplicate or empty header cells become auto-renamed keys (__EMPTY, __EMPTY_1) or collide silently; validate the JSON for these strings before relying on the schema.
• Merged-cell ranges (worksheet.mergeCells) expose the value only at the top-left anchor; tail cells in the merge are undefined, so a merged title row yields one record with the value and N-1 records with empty fields.
• Formula cells store the expression in f and the last cached result in v; tools that save without recalculating leave v stale or missing, and SheetJS returns v as-is.
• Move XLSX.read into a Web Worker for workbooks above ~10 MB; the synchronous parse path blocks requestAnimationFrame and can freeze the tab for several seconds on large sheets.