refactor: two names that meant different things in different files

src/ods/stream.ts

@@ -11,7 +11,19 @@ import { MAX_COL_INDEX, MAX_REPEAT_COUNT, MAX_ROW_INDEX } from "../limits"
 
 // ── Helpers ──────────────────────────────────────────────────────────
 
-function decodeUtf8(data: Uint8Array): string {
+/**
+ * Decode a part with no string-length ceiling — see the fuller note on
+ * the same function in `xlsx/stream-reader.ts`. `ods/reader.ts` has a
+ * `decodeUtf8` that does check and raises the #514 `ParseError`; this
+ * one does not, and the name says so rather than leaving two functions
+ * with one name and two contracts.
+ *
+ * `streamOdsRows` streams the *rows* out of `content.xml` but still
+ * builds the whole part as a string to do it, so an ODS over the ceiling
+ * fails here with V8's raw error — unlike `streamXlsxRows`, which SAX-parses
+ * the worksheet off the decompression stream and never builds one.
+ */
+function decodeUtf8Unchecked(data: Uint8Array): string {
   return new TextDecoder("utf-8").decode(data)
 }
 
@@ -331,7 +343,7 @@ export async function* streamOdsRows(
   if (!zip.has("content.xml")) {
     throw new ParseError("Invalid ODS: missing content.xml")
   }
-  const contentXml = decodeUtf8(await zip.extract("content.xml"))
+  const contentXml = decodeUtf8Unchecked(await zip.extract("content.xml"))
 
   // 4. Yield rows via SAX
   // 4. Yield rows via SAX, applying the filters

src/xlsx/pivot-writer.ts

@@ -623,7 +623,7 @@ function computeLocation(
   colFieldCount: number,
   meta: PivotFieldsMeta,
 ): Record<string, string | number> {
-  const { col, row } = parseCellRef(targetCell)
+  const { col, row } = parseCellRefStrict(targetCell)
 
   // Header rows: 1 (page filters) + 1 (column field header per col field) + 1 (data field header)
   const firstHeaderRow = 0
@@ -644,8 +644,17 @@ function computeLocation(
   }
 }
 
-/** Parse `"B3"` → `{col: 1, row: 2}` (0-based). */
-function parseCellRef(cell: string): { col: number; row: number } {
+/**
+ * Parse `"B3"` → `{col: 1, row: 2}` (0-based), rejecting anything else.
+ *
+ * Deliberately not `xlsx/worksheet.ts`'s `parseCellRef`, which is lenient
+ * because a cell that mis-parses in a worksheet should cost one cell, not
+ * the read. This one takes a `targetCell` the caller wrote by hand, where
+ * a typo has no salvageable reading and silently pivoting at A1 would be
+ * worse than the throw. The name says `Strict` so the two are not
+ * mistaken for each other.
+ */
+function parseCellRefStrict(cell: string): { col: number; row: number } {
   const m = /^([A-Z]+)(\d+)$/i.exec(cell.trim())
   if (!m) {
     throw new Error(`Invalid pivot targetCell "${cell}" — expected an A1-style reference`)

src/xlsx/stream-reader.ts

@@ -68,7 +68,21 @@ const REL_STYLES = "styles"
 
 // ── Helpers ──────────────────────────────────────────────────────────
 
-function decodeUtf8(data: Uint8Array): string {
+/**
+ * Decode a package part with no string-length ceiling.
+ *
+ * The buffered readers route the same job through `_decode.decodePart`,
+ * which catches V8's `MAX_STRING_LENGTH` and reports it as the #514
+ * `ParseError` naming the part. This path does not, and that is what
+ * `docs/PARITY.md` records — `streamXlsxRows` is the answer *to* the
+ * ceiling for worksheets, since it never builds one string for them.
+ *
+ * The parts decoded here are the small ones (content types, rels, the
+ * workbook, styles); a `sharedStrings.xml` over the ceiling would still
+ * throw V8's raw error. The name is `Unchecked` so a line moved between
+ * the two readers cannot quietly drop the check.
+ */
+function decodeUtf8Unchecked(data: Uint8Array): string {
   return new TextDecoder("utf-8").decode(data)
 }
 
@@ -645,9 +659,9 @@ function resolveFromParts(
   const ct = parts.get("[Content_Types].xml")
   const rootRelsBytes = parts.get("_rels/.rels")
   if (!ct || !rootRelsBytes) return null
-  parseContentTypes(decodeUtf8(ct))
+  parseContentTypes(decodeUtf8Unchecked(ct))
 
-  const rootRels = parseRelationships(decodeUtf8(rootRelsBytes))
+  const rootRels = parseRelationships(decodeUtf8Unchecked(rootRelsBytes))
   const workbookRel = rootRels.find((r) => matchesRelType(r.type, REL_WORKBOOK))
   if (!workbookRel) return null
   const workbookPath = workbookRel.target.startsWith("/")
@@ -662,9 +676,9 @@ function resolveFromParts(
     ? `${workbookDir}/_rels/${workbookPath.slice(workbookDir.length + 1)}.rels`
     : `_rels/${workbookPath}.rels`
   const wbRelsBytes = parts.get(workbookRelsPath)
-  const workbookRels = wbRelsBytes ? parseRelationships(decodeUtf8(wbRelsBytes)) : []
+  const workbookRels = wbRelsBytes ? parseRelationships(decodeUtf8Unchecked(wbRelsBytes)) : []
 
-  const { sheets: sheetInfos, dateSystem } = parseWorkbookXml(decodeUtf8(wbBytes), options)
+  const { sheets: sheetInfos, dateSystem } = parseWorkbookXml(decodeUtf8Unchecked(wbBytes), options)
   const targetSheet = resolveTargetSheet(sheetInfos, options?.sheet)
   if (!targetSheet) return null
 
@@ -692,7 +706,7 @@ function resolveFromParts(
     const ssPath = resolvePath(workbookDir, ssRel.target)
     const ssBytes = parts.get(ssPath)
     if (!ssBytes) return null
-    sharedStrings = parseSharedStrings(decodeUtf8(ssBytes))
+    sharedStrings = parseSharedStrings(decodeUtf8Unchecked(ssBytes))
   }
 
   let parsedStyles: ParsedStyles | null = null
@@ -701,7 +715,7 @@ function resolveFromParts(
     const stylesPath = resolvePath(workbookDir, stylesRel.target)
     const stylesBytes = parts.get(stylesPath)
     if (!stylesBytes) return null
-    parsedStyles = parseStyles(decodeUtf8(stylesBytes))
+    parsedStyles = parseStyles(decodeUtf8Unchecked(stylesBytes))
   }
 
   return { wsPath, sharedStrings, parsedStyles, dateSystem }
@@ -829,14 +843,14 @@ export async function* streamXlsxRows(
   if (!zip.has("[Content_Types].xml")) {
     throw new ParseError("Invalid XLSX: missing [Content_Types].xml")
   }
-  const contentTypesXml = decodeUtf8(await zip.extract("[Content_Types].xml"))
+  const contentTypesXml = decodeUtf8Unchecked(await zip.extract("[Content_Types].xml"))
   parseContentTypes(contentTypesXml)
 
   // 3. Parse _rels/.rels to find the workbook path
   if (!zip.has("_rels/.rels")) {
     throw new ParseError("Invalid XLSX: missing _rels/.rels")
   }
-  const rootRelsXml = decodeUtf8(await zip.extract("_rels/.rels"))
+  const rootRelsXml = decodeUtf8Unchecked(await zip.extract("_rels/.rels"))
   const rootRels = parseRelationships(rootRelsXml)
   const workbookRel = rootRels.find((r) => matchesRelType(r.type, REL_WORKBOOK))
   if (!workbookRel) {
@@ -855,15 +869,15 @@ export async function* streamXlsxRows(
 
   let workbookRels: Relationship[] = []
   if (zip.has(workbookRelsPath)) {
-    const wbRelsXml = decodeUtf8(await zip.extract(workbookRelsPath))
+    const wbRelsXml = decodeUtf8Unchecked(await zip.extract(workbookRelsPath))
     workbookRels = parseRelationships(wbRelsXml)
   }
 
   // 5. Parse workbook XML for sheet names and date system
   if (!zip.has(workbookPath)) {
     throw new ParseError(`Invalid XLSX: missing workbook at ${workbookPath}`)
   }
-  const workbookXml = decodeUtf8(await zip.extract(workbookPath))
+  const workbookXml = decodeUtf8Unchecked(await zip.extract(workbookPath))
   const { sheets: sheetInfos, dateSystem } = parseWorkbookXml(workbookXml, options)
 
   // 6. Parse shared strings (small, needed for cell resolution)
@@ -872,7 +886,7 @@ export async function* streamXlsxRows(
   if (ssRel) {
     const ssPath = resolvePath(workbookDir, ssRel.target)
     if (zip.has(ssPath)) {
-      const ssXml = decodeUtf8(await zip.extract(ssPath))
+      const ssXml = decodeUtf8Unchecked(await zip.extract(ssPath))
       sharedStrings = parseSharedStrings(ssXml)
     }
   }
@@ -883,7 +897,7 @@ export async function* streamXlsxRows(
   if (stylesRel) {
     const stylesPath = resolvePath(workbookDir, stylesRel.target)
     if (zip.has(stylesPath)) {
-      const stylesXml = decodeUtf8(await zip.extract(stylesPath))
+      const stylesXml = decodeUtf8Unchecked(await zip.extract(stylesPath))
       parsedStyles = parseStyles(stylesXml)
     }
   }