PdfSpace.Storage 0.5.5-alpha.1

<div align="center">

PdfSpace

Read, edit and review PDFs — on your device.

Open the browser app · Components · Architecture · Compatibility

</div>

PdfSpace is an independent, local-first PDF application built in C# with Uno Platform and SkiaSharp. The browser runs a real Uno WebAssembly application with the same document, editing and rendering libraries as the desktop host. An Acrobat-inspired shell brings together document tabs, floating quick tools, comments, form preparation, original-text editing, protection, redaction and page organization.

0.5.5-alpha.1: functional PDF workflows, not complete or pixel-identical Adobe Acrobat compatibility. PdfSpace has original branding and icons and is not affiliated with Adobe. Review the save and security boundaries before processing important documents.

Native layout and batch efficiency in 0.5.5

Edit → Edit objects → Object layout adds visible-page alignment, explicit reference-object alignment, width/height/size matching and equal-gap distribution. Repeated exact layout operations retain document identity rather than generating redundant native PDF rewrites. Native selection validation replaces quadratic pair checks with sorted occurrence intervals and ancestor lookup. Workflow, APIs, limits and validation.

Resize and vector-point alignment in 0.5.4

Edit → Edit objects now offers independent Snap resizing objects and Snap vector points controls. Resize guides preserve the fixed edge/corner and Shift proportions; point guides preserve axis locks and handle grab offsets. Alt bypasses snapping. A completed gesture makes one native PDF edit; stationary, returned and cancelled gestures leave the source unchanged.

The object inspector reuses a pool of at most 64 rows, with Previous/Next ranges and Go to object reaching every indexed occurrence. Dense selection and range navigation do not reparse PDF objects or add history. Compacted snap anchors and allocation-free point/resize queries reduce geometry overhead. See editing behavior and measurement scope.

Precision object gestures in 0.5.2

Shift constrains movement and aspect ratio; Alt resizes or creates native shapes from their center. Native images and mixed-object selections share the same allocation-free geometry. Repeated selection avoids inspector churn, inline editor handoff retains keyboard focus, and acceptance tooling observes fresh compositor geometry. Interaction and integration guide.

Object appearance in 0.5.1

Native opacity/blending, stroke caps/joins/miter and dash editing are available under Edit → Edit objects → Object appearance. Dense pages use a reusable ordered spatial index for picking and visible selection outlines. The source-pinned Apache-2.0 renderer fixes image alpha, odd dash patterns and miter rendering. See appearance, architecture and precise compatibility limits.

Native object editing in 0.5

Edit objects selects native text, paths, images, shading and Form groups directly on the page. Mixed selections support drag/resize/nudge, align/distribute, crop, stacking, grouping, copy/paste across documents, and one-transaction undo. Path controls edit real geometry and fill/stroke; explicit replacement text blocks wrap into a chosen box with overflow checks. Source-run text and native image editing remain available. Object editing guide and exact limits.

Photo compatibility and memory in 0.4.1

Native image import/replacement now honors all EXIF orientations, preserves eligible JPEG compression and streams PNG alpha/RGB compression without full-image staging arrays. Image duplication shares source data; Restore image proportions works after rotation. Undo history now has a distinct-source byte budget alongside its entry cap, with explicit history release in Properties. Photograph workflows · Performance and memory.

Existing-PDF editing and performance in 0.4

Edit a specific native image placement: select, move, resize, rotate, flip, replace, insert or delete without modifying its other shared uses. Supported original text inside nested Form XObjects can now be changed, with explicit Unicode font replacement for independent horizontal runs. No white-cover or page-rasterization editing simulation is used. Native object workflows and limits.

The viewer now uses indexed page geometry instead of repeatedly arranging the entire document. Per-document bounded search indexes, constant-time picture LRU touches, bounded parsed-source caches and one-parse-per-source page assembly reduce repeated work. Performance design and reproducible measurements.

Scan & OCR in 0.3

Recognize image-only PDFs locally in English, Polish or German, review and correct words against the original scan, and export a real searchable PDF with an invisible Unicode text layer. Import PNG/JPEG scans, select page ranges, cancel a batch without partial edits, and undo recognition or corrections. The browser self-hosts its pinned Tesseract.js/WASM/model assets; native hosts use an installed Tesseract 5 executable. No document is sent to an OCR service. OCR workflow, integration and limitations.

What you can do

Read and navigate. Open PDFs and workspaces, use multiple tabs, search and copy text, zoom/pan, switch continuous/single/two-page layouts, browse virtualized thumbnails, navigate original PDF outlines and create workspace bookmarks. Local named destinations and internal links are supported.

Review without flattening everything. Add text, highlights, underlines, strikethrough, notes, replies, ink, shapes, arrows, stamps and links. Move, resize and recolor annotations; resolve comments; undo and redo changes. Structured PDF export writes supported annotations and replies as native PDF objects.

Prepare and fill forms. Create text, check-box and dropdown fields directly on a page. Fill imported text, check boxes, radio groups and single-select choices. Edit tooltips, defaults, read-only/required/multiline flags, font size, maximum length, geometry and choice labels. Delete imported widgets with undo and native PDF persistence. Tab/Shift+Tab move between fields, Space selects a button, and arrows change choices.

Edit supported original text. Replace actual PDF text-showing operands rather than covering them with a white rectangle. Unsupported encodings, absent glyphs and stale edits are rejected. Explicit replacement text blocks provide bounded word wrapping; automatic reconstruction of arbitrary source paragraphs and complex-script shaping remain outside this release.

Organize and export. Reorder, rotate, duplicate, delete, crop, combine, extract and split pages. Download native PDF, a separately labeled flattened visual copy, PNG, text, form data or an editable workspace. Reorganized existing form documents are blocked from structured saving when the field tree cannot be preserved safely.

Protect or redact a copy. Export AES-256 PDFs with distinct opening and owner passwords. Browser protection uses an isolated QPDF WASM worker; desktop uses PDFsharp. Explicit owner authentication is required before reopening for editing. Redaction creates a separately confirmed, image-only reconstruction of all pages with burned-in marks; original PDFs and workspaces remain unredacted.

The included six-page Circular futures report and interactive review form are original synthetic documents for exercising the tools without uploading private data.

Choose the right save operation

Cropping is not redaction. Field/object deletion is not confidential-data erasure. A drawn signature is not a certificate signature. XFA, form scripts/calculations, certificate signatures, full tagging/compliance, Office conversion and cloud collaboration remain unimplemented.

Form data and navigation

Prepare a form → Import form data accepts plain field values from XFDF or JSON. The complete batch is validated, unmatched names are reported, and confirmation applies one undoable transaction. Export either interoperable XFDF or versioned JSON. Check required fields takes you to the first missing supported value. Read-only changes, invalid choices, unsupported multi-values and overlong input reject the whole import.

The Bookmarks panel separates original PDF outlines from editable workspace bookmarks. Supported internal links follow destination page identities through insertion, duplication, reordering, extraction and combination. Links to removed destinations are removed rather than redirected.

Form-data files are unencrypted. Document URLs and scripts are never followed or executed. See the form-data and navigation guide for formats, defensive limits and embedding examples.

Download

Every release ships a self-contained, single-file desktop app — no .NET install needed:

Extract and run PdfSpace (PdfSpace.exe on Windows). Builds are not code-signed yet: on macOS clear the quarantine flag with xattr -d com.apple.quarantine PdfSpace; on Windows choose More info → Run anyway in SmartScreen. Verify downloads against SHA256SUMS.txt.

The libraries below are published to NuGet.org, e.g. dotnet add package PdfSpace.Core --prerelease.

Twelve reusable libraries

The application is a thin browser/desktop host. All twelve libraries are packed with symbols and published to NuGet.org for version tags. See embedding examples.

Toolchain

.NET SDK 10.0.401, Uno SDK 6.7.30 / WinUI 6.7.135, matched managed/native SkiaSharp 3.119.4, PdfPig 0.1.16, the source-pinned PdfSpace.Rendering.Skia adaptation of PdfPig.Rendering.Skia 0.1.16.4, PDFsharp 6.2.4. Browser security uses hash-pinned @neslinesli93/qpdf-wasm 0.3.0 / QPDF 12.2.0. Rendering is still Uno/Skia, not a third-party web PDF viewer.

Build and run

Install the .NET SDK, Python 3, Node.js 22, and the Uno desktop prerequisites for your operating system.

git clone https://github.com/wieslawsoltes/PdfSpace.git
cd PdfSpace
python3 scripts/fetch-assets.py
dotnet workload install wasm-tools

# Native engine and PDF round-trip tests.
dotnet run --project tests/PdfSpace.Tests -c Release

# Desktop host.
dotnet run --project src/PdfSpace.App -f net10.0-desktop -p:PdfSpaceDesktopOnly=true

# Browser publish and static-root collection.
dotnet publish src/PdfSpace.App -f net10.0-browserwasm -c Release \
  -o artifacts/publish -p:WasmShellWebAppBasePath=/PdfSpace/
python3 scripts/collect-site.py artifacts/publish artifacts/site
python3 scripts/serve-site.py --directory artifacts/site --port 4173

Open http://127.0.0.1:4173/PdfSpace/. The asset scripts verify upstream font/runtime hashes. Static collection includes the self-hosted security worker, WASM and dependency notices. No PDF/password processing service is required.

Browser acceptance tests

mkdir -p artifacts/fixtures
cp artifacts/engine/*.pdf artifacts/engine/*.png artifacts/engine/*.jpg artifacts/fixtures/
npm ci --ignore-scripts
npx playwright install chromium
npm run test:browser
bash scripts/verify-browser-exports.sh artifacts/browser-exports

UI tests use real pointer, keyboard, file-picker and download/reopen interactions. ?test=1 enables read-only diagnostics, not a document-mutation API. The security-provider contract also has dedicated boundary tests. CI retains screenshots, reports and failure traces.

Delivery and documentation

Build tests the native engines, publishes and tests Uno WebAssembly, independently verifies the browser-encrypted PDF, and packages the libraries. Desktop compiles on Windows, macOS and Linux. Pages deploys a successful main-branch build and validates its commit identity and public behavior. Release runs for v* tags or a supplied manual version. It reruns the engine and browser gates, publishes self-contained single-file desktop executables for Windows, macOS and Linux (x64 and arm64), packs the libraries with symbols and emits SHA256SUMS.txt. Tags attach all assets to a GitHub Release and publish the packages to NuGet.org with Trusted Publishing (OIDC, no stored API key) from the protected nuget environment. Manual runs are dry runs: they build and upload every asset as workflow artifacts but publish nothing.

Architecture · Components · Compatibility · Shortcuts · Security · Contributing · Changelog · Third-party notices

License

Original PdfSpace source is MIT licensed. The source-pinned src/PdfSpace.Rendering.Skia derivative is Apache-2.0 licensed, with original copyrights, notices and modification records retained. Dependencies and fonts retain their upstream licenses. Adobe and Acrobat are trademarks of their respective owners; Adobe proprietary code, branding and assets are not included.