Enterprise agent platform for building, deploying, and operating AI teammates.
import * as XLSX from "xlsx"
import DOMPurify from "dompurify"
export interface ExcelSheetPreview {
html: string
missingFormulaResults: number
}
const INLINE_TAGS = ["a", "b", "strong", "i", "em", "u", "s", "sub", "sup", "span", "br"]
function escapeHtml(text: string): string {
return text.replace(/&/g, "&").replace(/</g, "<").replace(/>/g, ">")
.replace(/"/g, """).replace(/'/g, "'")
}
/** Render a preview copy without treating missing formula caches as blank results. */
export function createExcelSheetPreview(
sheet: XLSX.WorkSheet,
missingResultText: string,
): ExcelSheetPreview {
const preview = { ...sheet }
let missingFormulaResults = 0
for (const [address, cell] of Object.entries(sheet)) {
if (address.startsWith("!")) continue
// Format a copy before dropping raw v/z attributes, preserving cell text
// and table structure. Link targets follow the export escaping path below.
let html = ""
if (cell.v != null) {
// Rich-text runs can contain table markup. Keep them inline before
// insertion so they cannot close a cell or inject rows into the sheet.
html = cell.h ? DOMPurify.sanitize(cell.h, {
ALLOWED_TAGS: INLINE_TAGS,
ALLOWED_ATTR: ["href", "title"],
ALLOW_DATA_ATTR: false,
ALLOW_ARIA_ATTR: false,
}) : escapeHtml(XLSX.utils.format_cell({ ...cell })).replace(/\r?\n/g, "<br/>")
}
// With sheetStubs enabled, SheetJS represents an uncached XLSX formula as
// a type-z cell with v: 0. That is not a calculated zero. Empty strings,
// false, zero and cached errors on other cell types are actual results.
if ((cell.f || cell.F) && (cell.t === "z" || cell.v == null)) {
missingFormulaResults++
// Array-formula followers can have F (the array range) but no formula
// text of their own. Show a placeholder rather than inventing a formula.
const text = cell.f ? `=${cell.f}` : missingResultText
html = escapeHtml(text)
}
preview[address] = {
t: cell.t === "n" ? "n" : "s",
v: "",
h: html,
}
if (cell.l?.Target) {
// SheetJS 0.20.3 decodes XML entities on read and escapes HTML attributes
// on export. Keep the original target: pre-escaping would double-encode
// literal quotes/entities. DOMPurify below still rejects unsafe schemes.
preview[address].l = { Target: cell.l.Target }
}
}
// Sanitize the complete output, including rich text and link protocols. Only
// keep spreadsheet presentation markup; no embedded media or workbook CSS.
const html = DOMPurify.sanitize(XLSX.utils.sheet_to_html(preview), {
ALLOWED_TAGS: [
"table", "thead", "tbody", "tfoot", "tr", "th", "td",
...INLINE_TAGS,
],
ALLOWED_ATTR: ["id", "colspan", "rowspan", "href", "title", "data-t"],
ALLOW_DATA_ATTR: false,
ALLOW_ARIA_ATTR: false,
})
return { html, missingFormulaResults }
}