@rightkit/document-core 0.1.2 → 0.2.0

package/dist/documents/pdf/pdfSave.js

@@ -1,250 +0,0 @@
-/**
- * PDF save orchestration (`docs/PDF-EDITOR-NOW.md` §9.4, §10).
- *
- * Two independent flows live here, both built against the *injected-bridge* function types frozen
- * in `pdfEditModel.ts` (§10.3's "Frozen cross-lane seams" — the real `tauriBridge.ts` Tauri-backed
- * implementations land later at Dispatch D2, so every backend call here is a parameter, never an
- * import):
- *
- *  1. `recomposePdfPreview` (§9.4) — calls an injected `ComposePdfPreview`-shaped function to get
- *     memory-only candidate bytes for the current source-command history, so the on-page display of
- *     committed-but-unsaved edits always renders from the same composer Save itself will use.
- *
- *  2. `createPdfSavePlan` (§10.1/§10.2/§10.3) — the two-phase Save protocol: prepare (Rust composes
- *     + structurally validates a candidate and returns it plus a short-lived token), isolated
- *     pdf.js render-verify (catch corruption client-side before committing), then commit (Rust
- *     atomically replaces bytes). Returns a `PdfSavePlan` (§10.3's exact shape) rather than bare
- *     bytes, so the generic document saver only updates recents/fingerprints/checkpoints after a
- *     real commit succeeds.
- *
- * `verifyPdfCandidate` (§10.2) loads candidate bytes into a FRESH, isolated pdf.js document — never
- * the active on-screen `PDFDocumentProxy` — and renders affected pages to an offscreen canvas before
- * extracting text for the expected-text checks. This mirrors `pdfPrinting.ts`'s isolated-document
- * pattern (pdf.js caches page proxies per document; reusing the live viewer's document risks a
- * second concurrent `render()` against a `PDFPageProxy` it may already be rendering — a hang, not an
- * error). Every await is time-bounded so a stuck pdf.js worker surfaces as a rejected promise
- * (`VerificationFailed`) instead of silence.
- */
-import * as pdfjs from "pdfjs-dist";
-import { pdfDocumentInit, pdfWasmUrl } from "@rightkit/pdf-ui";
-import { withTimeout } from "./pdfShared.js";
-import { logEvent } from "../../state/diagnostics.js";
-// ---------------------------------------------------------------------------------------------
-// §9.4 in-memory preview composition
-// ---------------------------------------------------------------------------------------------
-/** Compose final replacements against original bytes without deleting undo history. A failed
- * preview leaves its original locator visible; correcting it supersedes earlier replacements
- * of exactly those bytes, rather than retrying an invalid intermediate edit first. */
-function effectiveSourceCommands(commands) {
-    const seen = new Set();
-    const effective = [];
-    for (const command of [...commands].reverse()) {
-        if (command.kind === "replaceSourceText") {
-            const { documentSha256, page, islandDigest, spans } = command.locator;
-            const key = JSON.stringify([
-                documentSha256,
-                page,
-                islandDigest,
-                spans.map((span) => [
-                    span.streamRef,
-                    span.contentsIndex,
-                    span.operatorIndex,
-                    span.operandIndex,
-                    span.glyphStart,
-                    span.glyphEnd,
-                    span.expectedText,
-                    span.expectedDigest,
-                ]),
-            ]);
-            if (seen.has(key))
-                continue;
-            seen.add(key);
-        }
-        effective.push(command);
-    }
-    return effective.reverse();
-}
-/** Recompose the memory-only preview candidate for the given source-command history (§9.4). No
- *  token, no retained candidate, no write — callers re-render affected pages from the returned
- *  bytes inside their existing page slots. Thin by design: the composer decides validity; this
- *  function only shapes the request and forwards the injected call. */
-export async function recomposePdfPreview(deps, sourcePath, expectedSourceSha256, sourceCommands) {
-    const request = {
-        sourcePath,
-        expectedSourceSha256,
-        sourceCommands: effectiveSourceCommands(sourceCommands),
-    };
-    return deps.composePdfPreview(request);
-}
-// ---------------------------------------------------------------------------------------------
-// §10.2 isolated pdf.js render-verify
-// ---------------------------------------------------------------------------------------------
-/** Thrown by `verifyPdfCandidate` on any renderability/content failure. Callers map this to
- *  `PdfEditError` code `"verification_failed"` (§11) rather than letting a raw pdf.js error surface,
- *  since §11 requires logs to carry "code/page/strategy but no document text" — this type carries a
- *  `page` when the failure is page-scoped and never embeds extracted document text in its message. */
-export class PdfCandidateVerificationError extends Error {
-    page;
-    constructor(message, page = null) {
-        super(message);
-        this.name = "PdfCandidateVerificationError";
-        this.page = page;
-    }
-}
-const RENDER_TIMEOUT_MS = 20_000;
-/** WKWebView's ReadableStream lacks async iteration, which makes pdf.js `getTextContent()` throw.
- * Read its public stream with a reader instead; this works in WKWebView & standards browsers. */
-async function extractPageText(page) {
-    const reader = page.streamTextContent().getReader();
-    let extracted = "";
-    try {
-        while (true) {
-            const chunk = await withTimeout(reader.read(), RENDER_TIMEOUT_MS, "pdf.js text stream timed out");
-            if (chunk.done)
-                break;
-            extracted += chunk.value.items.map((item) => item.str ?? "").join("");
-        }
-        return extracted;
-    }
-    finally {
-        reader.releaseLock();
-    }
-}
-/** Render every page in `pages` from a FRESH, isolated pdf.js document built from `bytes` (never the
- *  active on-screen document — see module doc), waiting for each render to fully settle, then check
- *  that every `expectedText` entry's text is present somewhere in that page's extracted text.
- *  Resolves on success; rejects with `PdfCandidateVerificationError` on any load/render/text-check
- *  failure. Always disposes the isolated document via `doc.destroy()`, success or failure, so a
- *  failed verification never leaks a pdf.js worker/document. */
-export async function verifyPdfCandidate({ bytes, pages, expectedText, }) {
-    const data = bytes.slice();
-    const loadingTask = pdfjs.getDocument(pdfDocumentInit(data, { wasmUrl: pdfWasmUrl() }));
-    let doc;
-    try {
-        doc = await withTimeout(loadingTask.promise, RENDER_TIMEOUT_MS, "pdf.js getDocument timed out");
-    }
-    catch (e) {
-        throw new PdfCandidateVerificationError(e instanceof Error ? e.message : "candidate could not be opened");
-    }
-    try {
-        const expectedByPage = new Map();
-        for (const check of expectedText) {
-            const list = expectedByPage.get(check.page);
-            if (list)
-                list.push(check);
-            else
-                expectedByPage.set(check.page, [check]);
-        }
-        for (const pageNumber of pages) {
-            let page;
-            try {
-                page = await withTimeout(doc.getPage(pageNumber), RENDER_TIMEOUT_MS, `pdf.js getPage(${pageNumber}) timed out`);
-            }
-            catch (e) {
-                throw new PdfCandidateVerificationError(e instanceof Error ? e.message : "candidate page could not be opened", pageNumber);
-            }
-            const viewport = page.getViewport({ scale: 1 });
-            const canvas = document.createElement("canvas");
-            canvas.width = Math.max(1, Math.ceil(viewport.width));
-            canvas.height = Math.max(1, Math.ceil(viewport.height));
-            const ctx = canvas.getContext("2d");
-            if (!ctx) {
-                throw new PdfCandidateVerificationError("candidate render context unavailable", pageNumber);
-            }
-            try {
-                await withTimeout(page.render({ canvas, canvasContext: ctx, viewport }).promise, RENDER_TIMEOUT_MS, `pdf.js render(page ${pageNumber}) timed out`);
-            }
-            catch (e) {
-                throw new PdfCandidateVerificationError(e instanceof Error ? e.message : "candidate page failed to render", pageNumber);
-            }
-            const checks = expectedByPage.get(pageNumber);
-            if (checks && checks.length > 0) {
-                let extracted;
-                try {
-                    extracted = await extractPageText(page);
-                }
-                catch (e) {
-                    throw new PdfCandidateVerificationError(e instanceof Error ? e.message : "candidate text could not be extracted", pageNumber);
-                }
-                for (const check of checks) {
-                    if (!extracted.includes(check.expectedText)) {
-                        throw new PdfCandidateVerificationError(`expected text missing after candidate render (island ${check.islandId})`, pageNumber);
-                    }
-                }
-            }
-        }
-    }
-    finally {
-        // `PDFDocumentProxy` has no `destroy()` of its own in this pdfjs-dist version — disposal goes
-        // through the `PDFDocumentLoadingTask` that produced it (same pattern as `pdfPrinting.ts`/
-        // `pdfBenchTransport.ts`).
-        await loadingTask.destroy();
-    }
-}
-/** Builds `PreparePdfSaveRequest` from `PdfSaveInput` — kept as a named step so both the happy path
- *  and tests can assert on the exact shape sent to `preparePdfSave` without duplicating field
- *  wiring. */
-function toPrepareRequest(input) {
-    return {
-        sourcePath: input.sourcePath,
-        expectedSourceSha256: input.expectedSourceSha256,
-        annotationPdfBytes: input.annotationPdfBytes,
-        sourceCommands: effectiveSourceCommands(input.sourceCommands),
-    };
-}
-/**
- * Runs §10.1 prepare against the injected `preparePdfSave`, then §10.2 isolated pdf.js
- * render-verify against the returned candidate. Returns a `PdfSavePlan` (§10.3's exact shape) only
- * once both steps succeed — `commit` is the only thing that talks to `commitPdfSave`, and it does so
- * lazily so a caller that decides not to commit (e.g. user cancels a Save As target picker between
- * prepare and commit) never invalidates the token by calling it.
- *
- * Token lifecycle (§10.3): the token this returns is single-use, expires after five minutes, and is
- * invalidated by any source revision in between — this module does not retry or re-prepare on a
- * stale-token rejection from `commit_pdf_save`; it surfaces the typed `PdfEditError` (expected code
- * `"write_conflict"` or `"stale_source"` per §11) to the caller, which owns the "start over" UX.
- * `didCommit` intentionally does nothing beyond what `commitPdfSave` already returned — §10.3 gives
- * the generic document saver, not this module, ownership of when recents/fingerprints/checkpoints
- * update, so this hook exists only as the seam that saver calls after using `result`.
- */
-export async function prepareAndVerifyPdfSave(deps, input) {
-    let prepared;
-    try {
-        prepared = await deps.preparePdfSave(toPrepareRequest(input));
-        logEvent("pdf-save:prepare:done");
-    }
-    catch (error) {
-        logEvent("pdf-save:prepare:error");
-        throw error;
-    }
-    const verify = deps.verifyCandidate ?? verifyPdfCandidate;
-    try {
-        await verify({
-            bytes: prepared.candidateBytes,
-            pages: prepared.affectedPages,
-            expectedText: prepared.expectedText,
-        });
-        logEvent("pdf-save:verify:done");
-    }
-    catch (error) {
-        logEvent("pdf-save:verify:error");
-        throw error;
-    }
-    let committed = false;
-    return {
-        async commit(targetPath) {
-            if (committed) {
-                throw new PdfCandidateVerificationError("save token already used — prepare a new candidate before committing again");
-            }
-            committed = true;
-            // Any rejection (typed `PdfEditError` per §11, e.g. "write_conflict"/"stale_source", or
-            // otherwise) is surfaced as-is — this module does not retry or reinterpret it (see doc
-            // comment above); the caller owns the "start over" UX.
-            return deps.commitPdfSave(prepared.token, targetPath);
-        },
-        async didCommit(_result) {
-            // Intentionally a no-op seam (see doc comment above): the generic document saver
-            // (`useDocumentSaver.ts`) is the sole owner of post-commit bookkeeping.
-        },
-    };
-}

package/dist/documents/pdf/pdfShared.d.ts

@@ -1,16 +0,0 @@
-export { boundedCanvasDpr, clampScale, viewportRotation, withTimeout, ZOOM_MAX, ZOOM_MIN, type OutlineNode, type Size, } from "@rightkit/pdf-ui";
-export declare const MARK_BG = "rgba(196,106,61,.40)";
-export declare const MARK_ACTIVE_BG = "rgba(240,160,96,.65)";
-export declare const THUMB_W = 130;
-export interface Match {
-    page: number;
-    occ: number;
-}
-/** Wrap every occurrence of `q` (honoring match-case / whole-word) inside a rendered text-layer
- *  span with <mark>, using the package's span highlighter with the host app non-active copper
- *  background. The active match is not drawn here: usePdfSearch emphasises it by page-wide
- *  occurrence index. */
-export declare function highlightSpan(span: HTMLElement, q: string, options?: {
-    matchCase?: boolean;
-    wholeWord?: boolean;
-}): void;

package/dist/documents/pdf/pdfShared.js

@@ -1,20 +0,0 @@
-// Shared PDF viewer primitives: @rightkit/pdf-ui's zoom/DPR/rotation/timeout helpers plus
-// the host app brand find-highlight colors, text-layer highlighter, and thumbnail width.
-import { highlightSpan as highlightSpanWithMarks } from "@rightkit/pdf-ui";
-export { boundedCanvasDpr, clampScale, viewportRotation, withTimeout, ZOOM_MAX, ZOOM_MIN, } from "@rightkit/pdf-ui";
-// Both highlight backgrounds are translucent so the canvas-rendered glyph shows through — the
-// text layer is only an invisible selection overlay, so we never repaint its (transparent) text.
-export const MARK_BG = "rgba(196,106,61,.40)"; // copper @ 40% — non-active matches
-export const MARK_ACTIVE_BG = "rgba(240,160,96,.65)"; // brighter copper @ 65% — active match
-export const THUMB_W = 130; // CSS px width of a thumbnail canvas
-/** Wrap every occurrence of `q` (honoring match-case / whole-word) inside a rendered text-layer
- *  span with <mark>, using the package's span highlighter with the host app non-active copper
- *  background. The active match is not drawn here: usePdfSearch emphasises it by page-wide
- *  occurrence index. */
-export function highlightSpan(span, q, options) {
-    highlightSpanWithMarks(span, q, {
-        matchCase: options?.matchCase,
-        wholeWord: options?.wholeWord,
-        markBg: MARK_BG,
-    });
-}

package/dist/documents/pdf/pdfToMarkdown.d.ts

@@ -1,11 +0,0 @@
-interface Item {
-    str: string;
-    hasEOL?: boolean;
-    height?: number;
-    transform?: number[];
-}
-/** One page's text as paragraphs: lines join on EOL/baseline change; a vertical gap over 1.6 line
- *  heights starts a new paragraph. Plain paragraphs only (no font-size heading inference). */
-export declare function pageItemsToParagraphs(items: Item[]): string[];
-export declare function pdfUrlToMarkdown(url: string): Promise<string>;
-export {};

package/dist/documents/pdf/pdfToMarkdown.js

@@ -1,67 +0,0 @@
-import * as pdfjs from "pdfjs-dist";
-/** One page's text as paragraphs: lines join on EOL/baseline change; a vertical gap over 1.6 line
- *  heights starts a new paragraph. Plain paragraphs only (no font-size heading inference). */
-export function pageItemsToParagraphs(items) {
-    const paras = [];
-    let line = "";
-    let cur = "";
-    let lastY = null;
-    let lastH = 0;
-    const flushLine = () => {
-        const t = line.replace(/\s+/g, " ").trim();
-        if (t)
-            cur += (cur ? " " : "") + t;
-        line = "";
-    };
-    const flushPara = () => {
-        flushLine();
-        if (cur)
-            paras.push(cur);
-        cur = "";
-    };
-    for (const it of items) {
-        const y = it.transform?.[5] ?? null;
-        const h = it.height || lastH || 10;
-        if (y !== null && lastY !== null && Math.abs(y - lastY) > h * 0.5) {
-            const gap = Math.abs(y - lastY);
-            flushLine();
-            if (gap > h * 1.6)
-                flushPara();
-        }
-        if (y !== null)
-            lastY = y;
-        lastH = h;
-        line += it.str;
-        if (it.hasEOL)
-            flushLine();
-    }
-    flushPara();
-    return paras;
-}
-export async function pdfUrlToMarkdown(url) {
-    const task = pdfjs.getDocument({ url });
-    const doc = await task.promise;
-    try {
-        const pages = [];
-        for (let n = 1; n <= doc.numPages; n++) {
-            const page = await doc.getPage(n);
-            const reader = page.streamTextContent().getReader();
-            const items = [];
-            for (;;) {
-                const { done, value } = await reader.read();
-                if (done)
-                    break;
-                for (const i of value.items)
-                    if ("str" in i)
-                        items.push(i);
-            }
-            const paras = pageItemsToParagraphs(items);
-            if (paras.length)
-                pages.push(paras.join("\n\n"));
-        }
-        return `${pages.join("\n\n")}\n`;
-    }
-    finally {
-        void task.destroy();
-    }
-}

package/dist/documents/pdf/pdfjsLibGlobal.d.ts

@@ -1 +0,0 @@
-export {};

package/dist/documents/pdf/pdfjsLibGlobal.js

@@ -1,12 +0,0 @@
-// pdf.js publishes `AnnotationEditorUIManager`'s two viewer-component dependencies (EventBus,
-// GenericL10n) only from "pdfjs-dist/web/pdf_viewer.mjs" — and that module's own top-level code
-// destructures `globalThis.pdfjsLib` (Mozilla's own viewer bootstrap sets this global before
-// loading it as a second <script>; a pure-ESM consumer has no such bootstrap). Importing
-// pdf_viewer.mjs without this throws at *module load*, not at use — "Cannot destructure property
-// 'AbortException' of 'globalThis.pdfjsLib' as it is undefined."
-//
-// Import this module (for its side effect only, before any "pdfjs-dist/web/pdf_viewer.mjs"
-// import) to wire that global. ES module evaluation is depth-first in source order, so as long as
-// this import line comes first, the global is set before pdf_viewer.mjs's body runs.
-import * as pdfjs from "pdfjs-dist";
-globalThis.pdfjsLib = pdfjs;

package/dist/documents/pdf/toolbarIcons.d.ts

@@ -1,18 +0,0 @@
-/** Page thumbnails panel toggle. */
-export declare const IconThumbnails: () => import("react").JSX.Element;
-/** Outline (bookmarks) panel toggle. */
-export declare const IconOutline: () => import("react").JSX.Element;
-/** Fit to width. */
-export declare const IconFitWidth: () => import("react").JSX.Element;
-/** Fit full page. */
-export declare const IconFitPage: () => import("react").JSX.Element;
-/** Toggle existing-text edit mode (edits source text runs, not overlaid annotations). */
-export declare const IconEditText: () => import("react").JSX.Element;
-/** Add-text annotation tool. */
-export declare const IconAnnotateText: () => import("react").JSX.Element;
-/** Ink/draw annotation tool. */
-export declare const IconDraw: () => import("react").JSX.Element;
-/** Highlight annotation tool. */
-export declare const IconHighlight: () => import("react").JSX.Element;
-/** Image stamp annotation tool. */
-export declare const IconImage: () => import("react").JSX.Element;

package/dist/documents/pdf/toolbarIcons.js

@@ -1,31 +0,0 @@
-import { jsx as _jsx } from "react/jsx-runtime";
-// PDF toolbar icons — Phosphor, matching the existing zoom/rotate icons in PdfToolbar.tsx
-// (14px, regular weight, aria-hidden — the accessible name lives on the wrapping button's
-// aria-label/title) and the DOCX toolbar's icon-based convention (src/documents/docx/toolbarIcons.tsx).
-import { SquaresFour } from "@phosphor-icons/react/dist/icons/SquaresFour";
-import { TreeStructure } from "@phosphor-icons/react/dist/icons/TreeStructure";
-import { ArrowsOutLineHorizontal } from "@phosphor-icons/react/dist/icons/ArrowsOutLineHorizontal";
-import { CornersOut } from "@phosphor-icons/react/dist/icons/CornersOut";
-import { TextT } from "@phosphor-icons/react/dist/icons/TextT";
-import { TextAa } from "@phosphor-icons/react/dist/icons/TextAa";
-import { Scribble } from "@phosphor-icons/react/dist/icons/Scribble";
-import { Highlighter } from "@phosphor-icons/react/dist/icons/Highlighter";
-import { ImageSquare } from "@phosphor-icons/react/dist/icons/ImageSquare";
-/** Page thumbnails panel toggle. */
-export const IconThumbnails = () => _jsx(SquaresFour, { size: 14, weight: "regular", "aria-hidden": true });
-/** Outline (bookmarks) panel toggle. */
-export const IconOutline = () => _jsx(TreeStructure, { size: 14, weight: "regular", "aria-hidden": true });
-/** Fit to width. */
-export const IconFitWidth = () => (_jsx(ArrowsOutLineHorizontal, { size: 14, weight: "regular", "aria-hidden": true }));
-/** Fit full page. */
-export const IconFitPage = () => _jsx(CornersOut, { size: 14, weight: "regular", "aria-hidden": true });
-/** Toggle existing-text edit mode (edits source text runs, not overlaid annotations). */
-export const IconEditText = () => _jsx(TextT, { size: 14, weight: "regular", "aria-hidden": true });
-/** Add-text annotation tool. */
-export const IconAnnotateText = () => _jsx(TextAa, { size: 14, weight: "regular", "aria-hidden": true });
-/** Ink/draw annotation tool. */
-export const IconDraw = () => _jsx(Scribble, { size: 14, weight: "regular", "aria-hidden": true });
-/** Highlight annotation tool. */
-export const IconHighlight = () => _jsx(Highlighter, { size: 14, weight: "regular", "aria-hidden": true });
-/** Image stamp annotation tool. */
-export const IconImage = () => _jsx(ImageSquare, { size: 14, weight: "regular", "aria-hidden": true });

package/dist/documents/pdf/usePdfDocumentLoader.d.ts

@@ -1,40 +0,0 @@
-import { type Dispatch, type MutableRefObject, type SetStateAction } from "react";
-import * as pdfjs from "pdfjs-dist";
-import type { PDFDocumentProxy } from "pdfjs-dist";
-import type { ViewMeta } from "../types.js";
-import type { OutlineNode, Size } from "./pdfShared.js";
-/** Benchmark-only password injection; production & manual unlocks always use the prompt. */
-export declare function getPdfBenchPassword(): string | null;
-interface UsePdfDocumentLoaderArgs {
-    documentPath: string;
-    url?: string;
-    initialView?: ViewMeta;
-    pdfRef: MutableRefObject<PDFDocumentProxy | null>;
-    renderedRef: MutableRefObject<Map<number, () => void>>;
-    slotRefs: MutableRefObject<Map<number, HTMLDivElement>>;
-    visibleRef: MutableRefObject<Set<number>>;
-    restoredRef: MutableRefObject<boolean>;
-    reportReadyRef: MutableRefObject<boolean>;
-    setCurrent: Dispatch<SetStateAction<number>>;
-    setScale: Dispatch<SetStateAction<number>>;
-    setRotation: Dispatch<SetStateAction<number>>;
-    setFitMode: Dispatch<SetStateAction<"width" | "page" | null>>;
-}
-export declare function usePdfDocumentLoader({ documentPath, url, initialView, pdfRef, renderedRef, slotRefs, visibleRef, restoredRef, reportReadyRef, setCurrent, setScale, setRotation, setFitMode, }: UsePdfDocumentLoaderArgs): {
-    pdfDoc: import("pdfjs-dist/types/src/display/api.js").PDFDocumentProxy | null;
-    numPages: number;
-    base: Size | null;
-    pageSizes: Map<number, Size>;
-    setPageSizes: Dispatch<SetStateAction<Map<number, Size>>>;
-    error: string | null;
-    renderErrors: Map<number, string>;
-    setRenderErrors: Dispatch<SetStateAction<Map<number, string>>>;
-    pwPrompt: {
-        retry: boolean;
-        onSubmit: (pw: string) => void;
-        onCancel: () => void;
-    } | null;
-    outline: OutlineNode[] | null;
-    replaceDocument: (doc: PDFDocumentProxy, task: ReturnType<typeof pdfjs.getDocument>) => void;
-};
-export {};