Appendix D — Dependency matrix (full)
Summary
For every live app: what state a document should be in before it (needs), which apps make its result better when run earlier (improves), and what it can break for later steps. Entries come from the app chapters' "Before you start", "Quality & limits" and "Works well with" sections. An entry marked expected, not tested is inferred by combining those chapters and has not been checked on the live app. The general ordering logic is in III.4 ordering rules and the short version of this matrix in III.3.
Details
The golden order in one line
Repair → unlock → fix orientation/skew → OCR → edit (merge/split/delete/resize) → add structure (TOC, tables) → flatten → compress → protect → sign → verify. Everything below is a detail of that line. Source: chapter combos in Organize, Add & Structure, Protect, Shrink & Repair, Extract.
Matrix: needs / improves / breaks
| App | Needs (state before) | Improved by (run earlier) | Breaks or invalidates |
|---|---|---|---|
| Merger | Unencrypted inputs | Rotation, Deskew, Unlock, Converter for odd formats | Nothing from your files; but later page-number-based work (TOC, deletes) refers to the new order |
| Splitter | Unlocked PDF | Unlock | Bookmarks/links pointing outside a piece cannot follow it |
| Delete Page | Unlocked PDF | Search (to find pages) | An existing table of contents (page numbers shift) |
| Rotation | Any PDF | – | Auto Skew re-renders pages; a signature (expected, not tested) |
| Resize | Any PDF | Merger, Rotation | Tile mode turns text into images (searchability lost unless OCR) |
| Booklet | Final page set at half-sheet size | Resize, Delete Page, Add TOC | Reorders pages: a contents page no longer matches reading order |
| Add TOC | Final page order | Merger, Delete Page, Splitter, Flatten | Page numbers wrong if pages change afterwards |
| TOC AI | Real text, unlocked, final order | Searchable (scans), Merger, Delete Page, Resize | Stale outline if pages change afterwards |
| Add Table | PDF + .xlsx | Merger / Delete Page (settle order) | Shifts page numbers of everything after the insert point |
| Flatten | Unlocked PDF | Extract Form Data (keep answers first) | Form interactivity (irreversible on that copy) |
| Collage | Image files | Deskew for scans | – |
| Illustrator | TXT or PDF with title/author up front | Chat, Converter, Add TOC, Merger | Page appearance (adds frames/ornaments) |
| Protector | Unencrypted PDF; all edits done | Delete Page, Flatten, Compressor, Unlock | Any later edit needs Unlock; layer modes rebuild pages |
| Unlock | PDF + known password | – | An existing signature (expected, not tested; the Verify chapter warns it may) |
| Sign | .pfx + password; unencrypted PDF; all edits done | Flatten, Compressor, Protector (see caveat in combos) | Any edit after signing invalidates it |
| Verify Signature | File exactly as received | – | Re-saving or compressing before verifying destroys what is tested |
| PDF/A Validator | Unencrypted PDF | Flatten, Optimize, Compressor (then validate) | Nothing (read-only); any later rewrite can change the verdict |
| Compressor | Finished content | Merger, Deskew, Searchable, Recovery | A signature; PDF/A compliance (validate again); softens pictures (do not run twice) |
| Optimize | Openable PDF | Merger, Recovery, Flatten | A signature (expected, not tested) |
| Deskew | Upright pages | Rotation, Merger (photos to PDF), Recovery | The text layer (pages rebuilt as images); run Searchable afterwards |
| Recovery | Nothing (first step) | – | Metadata/attachments with the default Full policy |
| Unwatermark | Real text; one file | Searchable (scans), Converter | PDF layout (images, tables, fonts, links lost) |
| OCR Image | Upright images, known language | Rotation, Deskew | – |
| Searchable | Upright scans, known language, unlocked | Rotation, Deskew, Unwatermark, Unlock, Merger | Grows file size; Office outputs drop the page picture |
| Extract Text | Text layer | Searchable, Unlock, Splitter | – |
| Extract Image | Unlocked PDF | Unlock, Splitter | – |
| Extract Metadata | Openable PDF | Unlock, Recovery | – |
| Extract Form Data | Filled fillable form, same template | Unlock, Splitter | – (but Flatten before it would empty the fields) |
| Parser | Unlocked PDF | Unlock, Searchable | – |
| Search | Text layer, or ticked scanned | Searchable, Unlock, Rotation | – (adds annotations to a copy) |
| Resume Analyzer | CV (PDF/DOC) + ideally a job description | Searchable (poor scans), Unlock, Merger | – |
| Converter | Unencrypted input matching the chosen type | Unlock, Rotation, Deskew | Pagination (Office → PDF); PDF layout in text-like targets |
| Chat | Optional text PDF | Searchable, Extract Text | – |
| SVG Maker | An idea | – | – |
| Abstracts | Text or scan (slow) | Searchable, Merger | Adds a page: shifts later page numbers; refresh the TOC afterwards |
| Ask PDF | Selectable text | Searchable, Deskew, Merger | – |
| Cheat Sheet | Text PDF or a topic | Searchable, Splitter | – |
| Checklist | Text PDF + topic | Searchable, Splitter, Ask PDF | – |
| PDF KB | Selectable text | Searchable, Merger | – |
| Assembly Chat | .NET assembly or package | NuGet Updater | – |
| Code Graph | GitHub repo with .sln, repo access | – | – |
| Repo Docs | GitHub repo | Code Graph, NuGet Updater | – (opens a PR) |
| Refactor Assistant | GitHub C# repo | Code Graph | Source structure via PR (review) |
| NuGet Updater | GitHub .NET repo | Assembly Chat | Package versions via PR (review) |
| Page SEO Checker | Public URL / sitemap | – | – |
Classic chains (from the chapters)
| Chain | Why this order |
|---|---|
| Rotation → Deskew → Searchable → Search | OCR reads sideways or crooked pages poorly |
| Recovery → anything | Repair first; other apps assume an openable file |
| Unlock → almost everything | Most apps cannot read an encrypted file |
| Searchable → TOC AI / Ask PDF / Abstracts / Checklist | The AI apps read text; scans have none |
| Merger / Delete Page → Add TOC | TOC page numbers must match the final order |
| Extract Form Data → Flatten | Save the answers before freezing the fields |
| Flatten → Compressor → Protector → Sign | Signing last; compressing after signing would change the signed file |
| Compressor → PDF/A Validator | Validate after any rewriting step |
| Searchable → Compressor | The text layer enlarges the file |
Cross-family lookalike rules (expected, not tested)
- A signature is expected to be invalidated by any app that rewrites the file (every app that writes a new PDF). The chapters state this explicitly for Compressor and Sign; for the others it is inferred.
- A text layer is expected to be lost by any app that rebuilds pages as images (Deskew, Resize tile mode without OCR, Converter to image formats).
- Bookmarks and a contents page are expected to go stale after any page-adding or page-removing step (Delete Page, Merger, Splitter, Add Table, Abstracts, Booklet).
Apps that are not part of chains
Global Search (a site search box), Advisor/Document Engineer (chat), and the developer apps work on repositories, packages or URLs, not on documents. Not available: Redactor, Compare PDFs. Retired: Release Notes, Fork Sync.