GraphCompose v2.1.0 — twinned

[DemchaAV]

The banner above is itself a GraphCompose document — rendered by the engine from the same DSL you'd use in your app. View the source deck (PDF).

GraphCompose v2.1.0 — twinned

v2.1.0 is live on Maven Central. The theme: one document, two outputs. The session that prints your PDF now also emits an editable PowerPoint deck — not a rendering of the PDF, and not a second layout to keep in sync. The same resolved layout graph feeds both backends, so a page and its slide agree by construction rather than by effort.

• PowerPoint output (@Beta) — add graph-compose-render-pptx and call buildPptx(path) on the session you already have. One page becomes one identically-sized slide, and every element keeps its coordinates because layout is compiled in core before either backend runs. Text lands as real text frames, panels and vectors as native DrawingML shapes — a reviewer can click a heading and retype it.
• Honest about what a deck cannot carry. Of the 38 capabilities the capability matrix tracks, 24 map to a native PowerPoint equivalent and 4 do not exist at all. The remaining 10 are partial, and not all in the same way: some render natively with an approximated styling detail (distinct per-corner radii collapse to one value, numeric dash arrays map to the nearest preset), while others lose something the format genuinely cannot hold (bookmarks past the first on a slide, parts of an inline SVG). Both kinds are listed, per capability, with the implementing class.
• A failed render no longer destroys the document it was replacing. Rendering to a path used to truncate the target before writing a byte, so an exception mid-render left you with an empty file where a good one had been. Output is now staged and moved into place — the previous file survives a failure intact.
• Headings stay with their content across a page break — in the engine, and in the CV and proposal presets. A section title orphaned at the bottom of a page now moves down with the block it introduces.
• CV entry, subtitle and project-card titles accept inline [text](url) links.

Geometry identity is a design invariant, not a best effort. What a deck cannot guarantee on its own is glyphs: it names its fonts and the viewer draws them. Standard-14 names travel as their metric-compatible viewer fonts — identical widths, different letterforms — while a registered binary family is embedded and renders identically everywhere. Every substitution is announced at WARN, once per family, naming what was asked for and what the deck will reference instead. The module page walks the three cases.

Two things worth knowing before you add it: the backend is @Beta — the API shape (PptxFixedLayoutBackend, its builder, the fragment-handler seam) may still move in a minor while feedback lands — and graph-compose-render-pptx depends on graph-compose-render-pdf at compile scope, because the PDF backend supplies the shared font measurement that keeps glyph widths aligned with the layout graph. A PPTX-only consumer therefore also resolves the PDF stack.

Nothing here breaks 2.0. The PPTX backend is opt-in and additive; the rest are fixes.

Add PowerPoint output:

<dependency>
    <groupId>io.github.demchaav</groupId>
    <artifactId>graph-compose-render-pptx</artifactId>
    <version>2.1.0</version>
</dependency>

Drop-in (PDF included): io.github.demchaav:graph-compose:2.1.0 — core + the PDF backend, unchanged from 2.0.

Release notes — https://github.com/DemchaAV/GraphCompose/releases/tag/v2.1.0
Maven Central — https://central.sonatype.com/artifact/io.github.demchaav/graph-compose/2.1.0
Backend capability matrix — https://github.com/DemchaAV/GraphCompose/blob/v2.1.0/docs/architecture/backend-capability-matrix.md
PPTX module guide — https://github.com/DemchaAV/GraphCompose/blob/v2.1.0/render-pptx/README.md
Live showcase — https://demchaav.github.io/GraphCompose/
Examples gallery — https://github.com/DemchaAV/GraphCompose/blob/v2.1.0/examples/README.md

As always — author intent, not coordinates.