sumatrapdfreader/sumatrapdf

Purpose and entry points

SumatraPDF is a native, multi-format document reader primarily for Windows: PDF, EPUB/MOBI/FB2, comics, CHM, XPS, DjVu, and images. The desktop process starts through src/SumatraPDF.cpp and startup orchestration in src/SumatraStartup.cpp; application/window state is centered on src/MainWindow.cpp. Separate entry points build shell preview and Windows Search integration: src/previewer/PdfPreviewDll.cpp and src/ifilter/SearchFilterDll.cpp.

Application shell and UI

The application is a Win32 desktop client, with its own UI abstractions layered over Windows controls. src/MainWindow.cpp coordinates windows, commands, document tabs, and active views; src/WindowTab.cpp and src/Tabs.cpp represent per-document and tab-group behavior. UI rendering/control wrappers live behind src/wingui/WinGui.h, src/wingui/Wnd.cpp, and src/mui/Mui.cpp. WebView2 is vendored under packages/Microsoft.Web.WebView2.1.0.4022.49 for browser-backed UI surfaces.

Document engine boundary

src/EngineBase.h defines the document-engine abstraction consumed by the viewer. src/EngineCreate.cpp selects an implementation based on input type; engines include MuPDF-backed documents (src/EngineMupdf.cpp), e-books (src/EngineEbook.cpp), images (src/EngineImages.cpp), DjVu, PostScript, and CHM. This boundary keeps format parsing/page access separate from UI and display logic.

Rendering, layout, and interaction flow

Opening a file flows from engine creation to a document controller/view model, then layout and cached page rendering. src/DisplayModel.cpp and src/DocumentLayout.cpp compute visible-page presentation; src/RenderCache.cpp manages rendered-page reuse; src/Canvas.cpp draws the document surface. Search and text selection use engine-provided text data through src/TextSearch.cpp, src/TextSelection.cpp, and src/Selection.cpp. PDF annotations/forms/printing are application-level extensions in src/Annotation.cpp, src/FormFields.cpp, and src/Print.cpp.

MuPDF and format-processing core

The vendored mupdf/ tree is the principal parsing and rasterization runtime. Its generic graphics/document foundation is mupdf/source/fitz/; PDF interpretation is mupdf/source/pdf/; HTML/EPUB and related reflow formats are handled in mupdf/source/html/. Sumatra’s adapter, src/EngineMupdf.cpp, bridges this C library into its engine interface. MuPDF brings embedded resources, fonts, compression, OCR-related facilities, color handling, and support for several document/image formats.

Persistence, OS integration, and services

User preferences, file history, favorites, themes, and settings serialization are handled by src/AppSettings.cpp, src/GlobalPrefs.cpp, and src/FileHistory.cpp. Windows-specific capabilities include registry installation (src/RegistryInstaller.cpp), DDE/search integration (src/SearchAndDDE.cpp), UI Automation providers (src/uia/Provider.cpp), crash handling (src/CrashHandler.cpp), and update checks (src/UpdateCheck.cpp). Network access is abstracted in src/base/Http.h with Windows and macOS implementations.

Build and delivery architecture

The production code is C/C++ with TypeScript/Bun scripts driving builds and generation. cmd/build.ts is the general build driver; premake5.lua generates project definitions, including vs2022/SumatraPDF.sln. Dependencies are largely vendored under ext/ and mupdf/, including HarfBuzz, FreeType, image codecs, compression libraries, archive readers, and MuPDF. CI runs through GitHub Actions (.github/workflows/build.yml) across Windows-oriented builds plus Linux/macOS build paths; macOS-specific application work is isolated under src/mac/.