csv-xlsx-import-export-stinger

# csv-xlsx-import-export Stinger Procedural arsenal for `csv-xlsx-import-export-wasp-drone`. Encodes every decision the Drone needs to make when helping a team add "upload your spreadsheet" to a React/Next.js product. This stinger is deliberately opinionated: it picks winners from the research and explains the trade-offs rather than presenting false symmetry. When the research surfaces a gotcha (SheetJS cannot stream XLSX, ExcelJS WorkbookWriter has a memory leak), the guides encode the workaround, not the happy fiction. ## When this stinger applies Load when the Drone is invoked on any of: - "Build a CSV import feature" - "Add XLSX file upload" - "Column-mapping wizard for my importer" - "Parse a 100MB spreadsheet without freezing the browser" - "Export data to Excel with styled headers" - "Is my CSV export safe from formula injection?" - "Compare OneSchema vs Flatfile vs dromo vs hand-rolled" - "Why is SheetJS not streaming my XLSX?" Do NOT load for: - Generic file upload drop-zone UX (ux-ui-svelte-wasp-drone owns the react-dropzone + progress indicator surface). - Database upsert performance after import (db-wasp-drone). - Server-side upload endpoint hardening (security-wasp-drone must audit the endpoint before production). ## First action when this stinger is loaded Read these in order before doing anything else: 1. **`guides/00-library-decision-tree.md`** -- pick the right library stack for the user's constraints before writing a line of code. 2. **`guides/04-csv-injection-prevention.md`** -- the sanitization rule is non-negotiable; read it early so it is never forgotten. 3. **`research/research-summary.md`** -- the five headline findings from scripture-historian, including the SheetJS streaming gotcha and ExcelJS memory leak. Then walk the remaining guides in order of the task at hand. Each guide is self-contained. ## Folder layout ```text csv-xlsx-import-export-stinger/ +- SKILL.md (this file) +- README.md (one-page human overview) +- guides/ | +- 00-library-decision-tree.md (papaparse vs SheetJS vs exceljs vs csv-parse) | +- 01-streaming-parse-large-files.md (Web Worker pattern, ExcelJS ReadStream, chunked CSV) | +- 02-column-mapping-ux.md (5-stage wizard, managed vs self-hosted, RSI) | +- 03-validation-rules.md (Zod row schema, error shape, abort vs collect) | +- 04-csv-injection-prevention.md (dangerous prefixes, sanitize function, CWE-1236) | +- 05-encoding-edge-cases.md (UTF-8 BOM, CP1252, line-ending normalization) | +- 06-export-xlsx.md (exceljs WorkbookWriter, header style, freeze pane) | +- 07-export-csv.md (streaming response, MIME type, BOM for Excel) | +- 08-error-reporting-ux.md (row-level errors, partial import, download error CSV) +- examples/ | +- papaparse-chunked-worker.tsx (large CSV via Web Worker + chunk callback) | +- sheetjs-webworker-xlsx.ts (large XLSX via Web Worker isolation) | +- exceljs-workbook-builder.ts (styled export with WorkbookWriter) | +- column-mapping-wizard.tsx (minimal hand-rolled mapping component) | +- csv-injection-sanitize.ts (canonical sanitize function) +- templates/ | +- row-error-object.ts (TypeScript interface for error shape) | +- import-result.ts (TypeScript interface for full import result) | +- import-report.md (markdown report skeleton for findings) +- reports/ | +- README.md (how past-run reports accumulate) +- research/ (DO NOT MODIFY -- owned by scripture-historian) ``` ## Critical directives (from the Command Brief) - **Never skip CSV injection sanitization even if the import target is a database, not Excel.** Why: exported data may be re-downloaded as CSV and opened in Excel later, creating a deferred injection surface. See `guides/04-csv-injection-prevention.md`. - **Always specify the streaming strategy for files over 5 MB.** Why: synchronous `readAsArrayBuffer` on a 100 MB XLSX freezes the browser main thread and crashes low-RAM devices. See `guides/01-streaming-parse-large-files.md`. - **Distinguish browser-side vs. server-side parse.** Why: SheetJS and exceljs have different APIs, memory models, and security profiles. See `guides/00-library-decision-tree.md`. - **Do not recommend a managed importer without noting the pricing model.** Why: OneSchema has no free tier and a ~$38K/year median contract; Flatfile has 50 files/month free; Dromo starts at $499/month. See `guides/02-column-mapping-ux.md`. - **Always validate and report errors at the row level, not just file level.** Why: row-level errors let users fix and re-upload rather than starting over. See `guides/08-error-reporting-ux.md`. - **Hand off to `security-wasp-drone` before any upload endpoint reaches production.** Why: file parsing is a classic attack surface (zip bombs, billion-laughs via XLSX shared strings, path traversal). The stinger covers sanitization but not endpoint hardening. ## Key gotchas from research These are not in the official docs but emerged from the 2026 literature sweep: 1. **SheetJS cannot stream-read XLSX.** The ZIP container requires buffering the full file. Use a Web Worker to keep the main thread responsive. `xlsx-stream-rows` (maintenance status unverified -- check before recommending as primary solution) offers true bounded-memory iteration. See `guides/01-streaming-parse-large-files.md` and `research/external/2026-05-20-sheetjs-streaming-api.md` (source-repository research, not shipped with the plugin). 2. **ExcelJS WorkbookWriter memory leak.** Issue #2916, PR #2558 pending for v5.0. Workaround: call `row.commit()` on every row immediately after writing. Do NOT accumulate rows in memory. See `guides/06-export-xlsx.md` and `research/external/2026-05-20-exceljs-github-readme.md` (source-repository research, not shipped with the plugin). 3. **Tab-prefix CSV injection survives Excel save/reopen ~90% of the time; single-quote does not.** Layer all three defenses (cell sanitization + correct Content-Type + Content-Disposition). See `guides/04-csv-injection-prevention.md` and `research/external/2026-05-20-csv-injection-dead-ends.md` (source-repository research, not shipped with the plugin). 4. **react-spreadsheet-import (RSI) adds Chakra UI.** For shadcn/ui stacks, this means ~200KB+ bundle addition. Prefer the hand-rolled wizard from `examples/column-mapping-wizard.tsx` or the `bow-react-spreadsheet-import` fork for Chakra-free teams. See `guides/02-column-mapping-ux.md`. 5. **Dromo processes data client-side (Private Mode) on all plans.** OneSchema and Flatfile process server-side. This matters for GDPR/HIPAA -- only Dromo guarantees PII never leaves the browser. See `guides/02-column-mapping-ux.md` and `research/external/2026-05-20-oneschema-vs-dromo-comparison.md` (source-repository research, not shipped with the plugin). --- *Forged by `stinger-forge` from `command-briefs/csv-xlsx-import-export-wasp-drone-command-brief.md` and research gathered by `scripture-historian`. Part of The Wasp Nest by [Mario Aldayuz a.k.a @thenotoriousllama](https://github.com/thenotoriousllama).*