Changelog

What changed, when, and why it matters.

Reference

Release history

Parse The Data is early software. This log separates shipped behavior from planned work and avoids version claims that the product cannot substantiate.

August 23, 2026 — RFC 9535 JSONPath queries

  • Added standards-based JSONPath queries for child selectors, wildcards, array slices, descendants, and filters.
  • Added original-document and working-record targets, normalized match paths, bounded previews, JSON-array copy, and an undoable Use matches as dataset action.
  • Kept query evaluation inside the browser worker with 1,000-character, 10,000-match, and 5 MB output limits.

August 23, 2026 — Name formatting

  • Added a Name option to the Format tab with the same preview-first, undoable workflow as other column formats.
  • Added surrounding-whitespace trimming and name-aware title case for apostrophes, hyphenated names, Unicode letters, and Mc prefixes.
  • Added conservative missing-apostrophe restoration so compact forms such as osullivan, ohenry, and oconnor become O'Sullivan, O'Henry, and O'Connor without changing names such as Oscar, Oliver, or Owen.

August 23, 2026 — Compare and Join

  • Added a dedicated two-source workspace for CSV, TSV, JSON, and JSONL files up to 10 MB each.
  • Added one- or two-column key mapping, optional case-sensitive matching, and explicit added, removed, changed, unchanged, missing-key, and duplicate-key results.
  • Added left, inner, and full joins with controlled collision suffixes plus spreadsheet-safe CSV and JSON exports.
  • Kept parsing, comparison, conflict detection, join preparation, and export inside a dedicated browser worker with bounded previews.
  • Added desktop, mobile, accessibility, security, and deterministic comparison regression coverage.

August 22, 2026 — Natural Voice

  • Added Natural Voice as an additional writing tool without replacing the local data workbench.
  • Added four editing tones, optional writing-sample guidance, an editable result, change summaries, and verbatim checks for structured details.
  • Added a deterministic cadence audit and an independent line-edit pass for substantial or visibly formulaic drafts; a revision is accepted only when protected details remain and the audit does not regress.
  • Added a same-origin Workers AI endpoint with streamed body limits, strict request validation, per-inference rate limiting, fail-closed limiter handling, randomized detail locks, and a route-specific Content Security Policy with Trusted Types.
  • Hardened local JSON and JSON Schema complexity limits, prototype-like field handling, and spreadsheet formula escaping for invisible Unicode prefixes.
  • Documented the AI processing boundary and the lack of detector, score, or watermark-removal guarantees.

August 22, 2026 — Search, answers, and trust

  • Added route-specific entity, breadcrumb, application, article, and FAQ structured data.
  • Added answer-first page summaries, accurate sitemap modification dates, AI crawler rules, an AI-readable site guide, and IndexNow release notifications.
  • Published About, Privacy Policy and GDPR information, Security, Methodology, and Changelog references.
  • Added canonical privacy-policy and GDPR aliases plus private privacy and security email routes.

August 22, 2026 — Expanded local workbench

  • Added preview-first date, phone-number, number, text, email, and boolean formatting with invalid-value policies and undo.
  • Added CSV filtering, stable sorting, column selection, renaming, missing-value filling, and multi-column deduplication.
  • Added JSON repair, duplicate-key diagnostics, RFC 6901 path browsing, Draft 2020-12 schema validation, and TSV export.
  • Versioned browser assets so each page and parser worker belong to the same release.

August 22, 2026 — Privacy and interface refresh

  • Rebuilt the landing page around the functional data workspace.
  • Clarified the server-to-browser processing boundary and improved its dark-mode contrast.
  • Added mobile layout regression coverage.

August 21, 2026 — Initial public tools

  • Added exact-row duplicate removal and the initial structured-data sampling work.
  • Hardened response headers and prevented injected browser analytics from changing the no-network boundary.