The tag on a page's backdrop copy, `/PCBackdrop << /Original ref /At /Len >>`, kept the page's original content stream reachable. After Add background, then Redact (or Remove Hidden Information, or Edit text) and a full save, the text taken off the page was still in the file, in the original stream the tag pointed at. Text extraction didn't show it; reading the file's bytes did. Restore also gave up after any edit, so the fill stayed off without a word. The tag is now `/PCBackdrop << /Index n /Op (f) /Before <SHA-256> /Was [num gen] >>`: how many of the copy's operators come before the `n`, the fill it stands in for, a hash of those operators by value (redaction writes the stream out again with other spacing), and the original's number as two integers, which keep nothing alive. Remove background turns the fill back on in the copy when the operators before it still hash the same and a scan with the fill restored finds it painted first, covering the page. Edits and redactions after the fill are kept. When the original stream is still there with the same bytes, the page points at it again. Otherwise the copy stays as it is, and Remove background says which pages keep their own backdrop off: `EditNotice::BackdropsLeftOff` (engine), a toast in the app, `warnings` from `doc_remove_marks`. `edit_notices` holds `EditNotice` now, with the undrawn-text notice as one of its kinds. The two new UI strings are in all 33 catalogues. `sha2` (already in the lock file) becomes a dependency of `pdfcraft-edit`. Tests, each seen failing with its piece of the fix reverted: - `the_tag_holds_no_reference_and_a_full_save_keeps_no_second_copy_of_the_page` and `removed_text_leaves_the_file_when_a_background_replaced_the_pages_backdrop` (automation, redaction and Remove Hidden Information through a full save): fail with `/Original` back. - `a_fill_that_would_cover_content_painted_before_it_stays_off`: fails without the re-scan. - `a_copy_written_out_again_with_other_spacing_still_comes_back` and the automation test: fail when the raw bytes are hashed instead of the operators. - `an_edit_after_the_fill_is_kept_and_the_backdrop_still_comes_back`: fails under the old rule (only an unchanged copy comes back). - `remove_background_names_the_pages_whose_own_backdrop_stays_off` (engine), `remove_background_warns_about_pages_whose_own_backdrop_stays_off` (automation), `remove_background_says_which_pages_keep_their_own_backdrop_off` (UI): fail without the notice. - `a_change_before_the_fill_leaves_it_off_and_names_the_page`, `a_tag_that_does_not_add_up_is_left_alone` (hostile and old-format tags, never a panic), `operators_hash_by_value_not_by_spelling`. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
pdfcraft-engine
The façade every frontend talks to. It holds the open documents, their edit history and the tool catalogue; frontends never reach past it into the parsing, rendering or editing crates.
word_count::count provides cancellable, page-at-a-time Unicode word and character statistics
with optional semantic exclusions; see word-count behavior and limits.
- Layer: L6. It depends on most of L1–L4 and on nothing above itself. No UI toolkit —
egui,winit,eframeandrfdare forbidden below L7, andxtask layersenforces it. - The crate denies
unwrap/expect/panic!/todo!. Entry points that run untrusted-document code keep a last-resortcatch_unwindguard so an escaped panic becomes an error and the user keeps their document.
Two frontends exist on top of it today: pdfcraft-ui-egui (desktop and web) and
pdfcraft-automation (headless tools, the CLI and MCP). Anything one can do, the other must be
able to do — see AGENTS.md §3.
The editing model
Each open document holds a pdfcraft_cos::Document: the object graph plus a copy-on-write overlay.
- An edit runs on a clone. On success the previous state is pushed onto the undo stack. Clones share all unchanged data, so a snapshot is cheap — this is what makes 100 levels of undo affordable on a 200 MB file.
- After every edit the working file is produced by an incremental write: the original bytes plus one appended revision.
- The view is refreshed from those bytes, so what you see is exactly what Save will write.
- Saving rebases onto the written bytes, so the next save appends only new edits.
Consequences worth preserving: the original bytes of an opened file are never rewritten in place; a full rewrite happens only where it must (redaction, PDF/A conversion, print imposition, a changed password); and unknown data survives a round trip because nothing is reconstructed that wasn't understood.
API
let mut s = Session::new();
let id = s.open("report.pdf", Some(path), bytes, None)?; // -> DocId
s.apply(id, Edit::RotatePages { pages: vec![0], degrees: 90 })?;
let label = s.undo(id)?; // "Rotate pages"
s.redo(id)?;
let doc = s.get(id).ok_or("gone")?; // &Document: queries, never edits
doc.can_undo(); doc.editable(); doc.allows_printing();
doc.text_blocks(page); doc.page_boxes(page); doc.security_summary();
let bytes = s.save_bytes(id)?; // incremental: keeps signatures over earlier revisions valid
let bytes = s.save_full_bytes(id)?; // full rewrite, garbage-collected
s.mark_saved(id, bytes.clone(), path)?;
Edit is one enum of 75 variants — every undoable change in the product, from
RotatePages to AddAnnotation, SetFieldValue, ApplyRedactions and AddWatermark. Adding a
feature means adding a variant and handling it in apply, which is what keeps undo, the dirty
flag, autosave and the automation tools consistent for free.
Session also owns the operations that produce a new document rather than editing one:
create_blank, create_from_images, create_from_text, combine, extract, split,
split_by_size, reduced_bytes, optimized_bytes, print_pdf, export_data, and the signing
family (sign, sign_with_timestamp, timestamp_document, embed_ltv).
Modules
| Module | What it holds |
|---|---|
commands |
The command registry: id, label, menu, shortcut, and a Needs precondition encoding document security. Menus, the ⌘K palette, shortcuts and the control channel all read it, so every surface agrees on what exists and when it is enabled |
catalog |
Acrobat's "All tools" information architecture as data — groups, sections, icons, hues, and an Availability of Ready, Planned(milestone) or Provider. The UI renders what is not built yet, honestly |
actions |
PDF action dispatch (go-to, URI, named, set-layer-state, JavaScript) |
links |
Link creation, properties, and finding URLs in page text |
js |
The bridge to pdfcraft-js. forms never depends on js directly; the engine injects it, so a build without JavaScript is possible |
xfa, ocr, compare, export |
Façades over the L3/L4 crates of the same names |
Not done here
- Rendering is still borrowed.
pdfcraft-renderwrapshayroand inspects throughlopdf(ADR-0004). The engine only seesinspect() -> DocInfo,RenderPoolandtext::PageText, and that API stays frozen while M2 replaces what is behind it. - Performance budgets (
misc.performance-budgets) are not set or enforced. - Some shipped features are still reachable only from the UI, not through an
automationtool;cargo xtask paritylists them; closing that gap is tracked in the execution plan (D4).
doc.text_edit_preview(page, block) prepares a read-only page background with the selected
paragraph’s glyphs removed, preserving text positioning, images and colours. It shares the
working document’s render configuration and never enters the undo history.