Skip to content

Documentation update #92

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Cadons merged 3 commits into from
Aug 30, 2026
Merged

Documentation update #92

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
dae0985
docs(claude): reflect completed legacy-to-loom migration
Cadons Aug 30, 2026
058186b
fix(backend): drop constexpr on kDefaultFont to fix MSVC build
Cadons Aug 30, 2026
e5b75ab
docs(claude): clarify loom pipeline architecture after migration
Cadons Aug 30, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
88 changes: 37 additions & 51 deletions CLAUDE.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,69 +51,58 @@ add a font file, register it in `fonts.json` — the CMake step generates the em

Docs (Sphinx + Doxygen) live in `doc/source/`; build with `cd doc && make html`.

## Two parallel architectures — know which one you're in
## Architecture: the loom pipeline

The codebase currently contains **two layout/rendering pipelines** side by side. This is
intentional, active migration work (see branch names like
`find-the-best-pattern-for-layout-engine`) — do not assume one is dead code without checking
recent commits/CMakeLists first.
There is a single layout/render pipeline **loom**, built around a pipeline-of-visitors over a plain node tree:

### 1. Legacy pipeline (`docraft/{include,src}/docraft/{model,layout,renderer,backend,services}`)

```
Craft XML → CraftLanguageParser (pugixml) → DocraftNode AST
→ Template Engine (${var} substitution, <Foreach> expansion)
→ Layout Engine: Chain-of-Responsibility handlers (layout/handler/*) compute x/y/width/height + pagination
→ Renderer: Visitor pattern walks the AST, dispatching to Painters (renderer/painter/*)
→ Painters call capability-provider Backend interfaces (backend/docraft_*_rendering_backend.h)
→ DocraftHaruBackend (backend/pdf/*) implements those interfaces on libharu
```

- `DocraftDocument` (include/docraft/docraft_document.h) is the entry point; orchestrates
`configure_settings → template_document → layout → render → save_to_file`.
- `DocraftDocumentContext` holds shared render state; note it's a known Service-Locator-style
god object (20+ getters) — treat as legacy, avoid extending its surface.
- Backend interfaces are split by capability domain (Text/Line/Shape/Image/Page/Output/Font/Metadata
rendering backends), not one aggregated facade — this split is deliberate (a prior aggregated
facade was removed for SRP reasons).
- Node types (Text, Table, List, Image, Shape, Layout, Paragraph, ...) each have: a parser
registration, a layout handler, and a paint/render path. Adding a node type touches all three
plus tests.
- Known architectural pain points are catalogued in `.local/ARCHITETTURA_CRITICITA.md` (circular
deps between PageHaruBackend/DocraftHaruBackend, scattered ownership of cursor/color state, no
fallback strategy for unsupported backend capabilities, etc.) — useful context before touching
`backend/` or `docraft_document_context`.

### 2. New "loom" pipeline (`docraft/{include,src}/docraft/loom/`)

A from-scratch redesign of the layout/render path, built around a **pipeline-of-visitors** over a
plain node tree rather than chain-of-responsibility handlers + a document-context god object:

```
DocraftLoomNode tree (nodes/: Text, Paragraph, Rectangle, VStack, HStack, ...)
Craft XML (pugixml)
→ docraft::craft front-end (craft/, craft/parser/): tokenizes tags and validates attributes into
typed "parsed element" data (ParsedElement, Parsed*Data, shared line-style/shape-style parsing)
→ docraft::craft::DocraftLoomCraftLanguageParser: top-level driver for <Document>/<Settings>/
<Metadata>/<Header>/<Body>/<Footer>/<Foreach>/<NewPage>
→ DocraftLoomTreeBuilder + per-tag handlers (loom/craft/handlers/) build the DocraftLoomNode
tree (nodes/: Text, Paragraph, Rectangle, Canvas, VStack, HStack, Table, ...)
→ MeasureProcessor (visitor: computes LayoutBox.measured_size per node)
→ LayoutProcessor (visitor: walks tree with a DocraftLoomCursor, fills LayoutBox.frame)
→ PaginationProcessor(visitor: fills LayoutBox.page_index)
→ RenderingProcessor (visitor: paints using the final frame/page_index)
→ docraft::backend::pdf (DocraftHaruBackend + capability interfaces) writes to libharu
```

- Each processor implements `interfaces::DocraftLoomIVisitor` (double-dispatch visitor, one
`visit(NodeType*)` overload per node type) — adding a node type means adding a `visit` overload
to every processor, not editing a big switch.
- `docraft::craft` (`docraft/{include,src}/docraft/craft/`) is the shared XML front-end: it parses
Craft-language tags into generic, typed "parsed element" data — attribute validation, enum
parsing (e.g. `border_style="solid"/"dashed"`), no knowledge of loom's node types or layout.
Despite the name, `DocraftCraftLanguageParser` here is this low-level tag/attribute parser, not
a document-level entry point.
- `docraft::craft::DocraftLoomCraftLanguageParser` (`docraft_loom_craft_language_parser.h`) is the
actual top-level orchestrator — the one that knows about `<Document>` structure and walks it
into a fully wired `DocraftLoomPdfCreator`. `docraft_tool` (`docraft/src/docraft/main.cpp`)
exercises it end to end.
- `docraft::loom::craft` (`docraft/{include,src}/docraft/loom/craft/`) holds the per-tag handlers
(`handlers/docraft_loom_*_handler.h`) and `DocraftLoomTreeBuilder` that turn parsed elements into
the actual `DocraftLoomNode` tree.
- Each pipeline processor implements `interfaces::DocraftLoomIVisitor` (double-dispatch visitor,
one `visit(NodeType*)` overload per node type) — adding a node type means adding a `visit`
overload to every processor, not editing a big switch.
- `DocraftLoomNode::layout_box()` accumulates the results of each pipeline stage
(`measured_size` → `page_index` → `frame`) — later stages read fields earlier stages wrote, so
pipeline order matters and stages must not be skipped for a node type.
- `DocraftLoomCursor` tracks the "pen" position during layout (origin top-left, y grows downward),
separate from any node — this replaces the old handler-local cursor bookkeeping.
- `DocraftLoomPdfCreator` is the loom-side equivalent of `DocraftDocument` (top-level orchestrator);
`docraft_tool` (`docraft/src/docraft/main.cpp`) exercises it end to end via
`DocraftLoomCraftLanguageParser`.
separate from any node.
- `docraft::backend` (`docraft/{include,src}/docraft/backend/`) is the one piece of the old
architecture that survived: capability-split rendering interfaces
(Text/Line/Shape/Image/Page/Output/Font/Metadata), implemented on libharu under `backend/pdf/`
(`DocraftHaruBackend` and friends). This split by capability domain, not one aggregated facade,
is deliberate (SRP). It's low-level PDF-writing infrastructure used directly by loom's
`RenderingProcessor`/`DocraftLoomPdfCreator` — not itself legacy, just the layer that talks to
libharu.
- Charts (`<Chart>`/`<Series>`) are native to loom, not a legacy leftover: `docraft/{include,src}/
docraft/loom/charts/` has builders for scatter/spline/line/histogram/pie, synthesized as a
`<Canvas>` full of primitives (Line/Circle/Rectangle/Text) via `docraft_loom_chart_handler`.
- Tests live in `docraft/test/docraft/loom/`; `docraft/test/CMakeLists.txt` is not glob-based — if
you add a new loom test file, you must add it to TEST_SOURCES there or it silently won't run.

When asked to work on layout/rendering, check whether the task is about the legacy pipeline or
`loom` — they don't share types, and "the layout engine" is ambiguous in this repo right now.

## Coding conventions (enforced by `.clang-tidy`, warnings are errors)

- Namespaces: `lower_case` (e.g. `docraft::loom::pipeline`)
Expand All @@ -139,7 +128,4 @@ Format enforced by convention (`.github/git-commit-instructions.md`): first line
- Templating: `${variable}` substitution and `<Foreach model="${items}">` expansion happen before
layout, driven by JSON data (`data("key")` inside a `Foreach`).
- Common attributes across nodes: `name`, `x`, `y`, `width`, `height`, `padding`,
`z_index`, `position`, `visible`. Colors are `#RRGGBB` or named (`red`, `blue`, ...).

Loom architecture is the newer architecture, follow it as main architecture, and use the old one only to understand
which items must be port into the new architecture
`z_index`, `position`, `visible`. Colors are `#RRGGBB` or named (`red`, `blue`, ...).
2 changes: 1 addition & 1 deletion docraft/include/docraft/backend/pdf/docraft_haru_text_backend.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -110,7 +110,7 @@ namespace docraft::backend::pdf {
*/
HPDF_Font resolve_font(const std::string& font_name, bool* needs_win_ansi_transcode = nullptr) const;
std::shared_ptr<DocraftHaruSharedState> state_;
static constexpr std::string kDefaultFont = "Helvetica";
static inline const std::string kDefaultFont = "Helvetica";
// Sticky across draw_text()/draw_text_matrix() calls that don't take a font_name
// of their own -- set by the preceding set_font() call, which does know which
// encoding resolve_font() picked for the font it just activated.
Expand Down
Loading