Files
deckcraft/docs/file-format-parity.md
T
aiongg f001f37906 CI: one test binary per crate, nextest, line tables only (#1213)
* Build: line tables only for workspace crates, no debug info for dependencies

* Tests: one integration-test binary per crate

* xtask ci: nextest, doctests, test build without examples, one wasm check
2026-10-23 12:43:34 -04:00

18 KiB
Raw Blame History

File-format parity with PowerPoint

Last reviewed: 2026-10-14 · Last updated: 2026-10-18 · Change: minor (native chart payloads stored in bounded ZIP members; previous: major, community PR wave 2026-10-14: seven owned PPTX corpus files, embedded fonts, EMF/WMF clip regions, threaded comments (#70), chart trendlines and workbooks (#335), SmartArt written as a diagram, tags and custom XML kept, picture export DPI; duplicated rows and headers folded; previous: trivial, lines in scaled groups, #233) · Target: Microsoft PowerPoint for Mac 16.113 (Microsoft 365)

Every format PowerPoint reads or writes, DeckCraft's support, fidelity and how it is tested. Gaps are itemized in gaps.md. Dimension score: ~58% ready (estimated), 60–110 h to parity.

Weights inside this dimension: PPTX 70%, PDF export 10%, image export 5%, legacy binary formats 5%, everything else (templates, shows, macros, ODP, video, GIF, outline) 10%.

Documents

From /Applications/Microsoft PowerPoint.app/Contents/Info.plist CFBundleDocumentTypes (Editor = PowerPoint opens and saves; Viewer = opens only).

Format PowerPoint DeckCraft read DeckCraft write Fidelity / notes Tested by
.pptx presentation Editor yes yes Broad (see below); an owned corpus of decks we author round-trips tables, charts, animations, media, sections, masters and layouts, comments, links and actions, custom shows, rotated groups, inherited placeholders, text effects and body layout, theme effects, fills, bevels, connectors, 3-D models and animation bounce end. Not measured against PowerPoint renders or decks from outside this repo crates/pptx/tests/it/corpus.rs, corpus_kinds.rs, corpus_elements.rs, corpus_style.rs, corpus_themes.rs, corpus_charts.rs, corpus_media.rs, roundtrip.rs, sample.rs, malformed.rs, sniff.rs; opened in PowerPoint by hand 2026-10-06
.potx template Editor yes (as a presentation) no: written with the .pptx content type file.saveTemplate is Save As under another name none
.ppsx slide show Editor yes (as a presentation) no: written as .pptx content Handed over by the system (command line, file association) it starts the slide show, as in PowerPoint; File › Open opens it for editing. The macOS Finder event arrives with #432 crates/ui-egui/src/opening.rs tests
.pptm / .potm / .ppsm macro-enabled Editor no no Should open and keep vbaProject.bin without running it none
.ppt / .pot / .pps 97–2003 binary Editor no no Owner decision; import only would be enough none
.odp OpenDocument Editor no no none
.thmx Office theme Viewer no no Browse for themes is a P2 row none
.deckcraft native (zip + JSON) n/a yes (also .slidecraft) yes Lossless for the DeckCraft model, forward compatible crates/format tests
PDF export no yes Slides, notes pages, handouts (1–9); real-text layer with embedded subsets, links, bookmarks. Slide artwork is a raster image at print resolution, not vector crates/pdf tests
Outline .rtf / .txt export (RTF), import (outline) .txt outline; .md as Markdown (headings are slides, list items bullets) (slide.fromOutline) .txt outline No RTF engine tests
PNG / JPEG slide images export n/a yes (file.export, deckcraft-cli render) Matches the editor's renderer. PNG pHYs and JPEG JFIF density are scale × 72 dpi (default scale 2 is 144 dpi), so Preview shows the slide's real size. PNG can leave the slide background out (transparentBackground, Export… “Transparent background”, deckcraft-cli render --transparent), so only the slide's content is opaque (#784) render, engine, UI and CLI tests
TIFF / BMP / GIF slide images export n/a no Easy (the image crate writes them) none
Animated GIF of the deck export n/a yes (file.export format gif, Export…, deckcraft-cli, MCP export) One frame per slide from the slide renderer; size small/medium/large/extra large or 16–3840 px wide, 0.1–600 s per slide, loop forever or play once, slide range, hidden slides skipped; a 256-colour palette per frame (exact, or NeuQuant above 256 colours). Transitions and build animations are not animated crates/engine/src/gif_export.rs tests (decoded frames, size, delay, loop, hidden slides, hostile params), MCP, CLI and export dialog tests
MP4 / MOV video of the deck export n/a no Needs timeline render + H.264 encode none
Print yes n/a no PR #66 (browser) open none

Pictures and media inside decks

Native files retain a nonempty chart kept sidecar: relationships, part names, content types and nativeEntry references live in presentation.json; exact binary payloads live in distinct chart-parts/<index> ZIP members. This keeps workbooks and drawings through a native save before exporting back to PPTX without expanding workbook bytes into JSON numbers. Chart parts and media share the 4 GiB inflated-byte and 100,000-item budgets; each member is read through the existing 2 GiB entry limit. Missing chart members are reported by load_checked and left empty. A present chart member that fails decoding, CRC or its entry limit refuses the file immediately, so later corrupt members cannot reuse the remaining inflation budget. Invalid, reused or ambiguous references are refused. A synthetic 3 MiB payload test checks the ZIP layout and bounded reads, alongside shared-budget and all-owner round trips.

Older native files without a sidecar still load. Earlier inline chart byte arrays are read under the existing manifest-size limit, then charged to the shared budget; saving migrates them to ZIP members. Builds predating chart-sidecar support ignore it and lose it on resave. The intermediate build that required inline bytes cannot read the new referenced representation.

Kind PowerPoint DeckCraft Notes
PNG, JPEG, GIF (animated), BMP, TIFF, WebP yes yes image 0.25; animated GIFs play since #39
SVG yes (with PNG fallback) no The PNG fallback shows; SVG files can't be inserted
EMF / WMF yes partial Drawn as vectors (paths, pens, pattern and hatch brushes, bitmaps, text in the document's fonts with the file's character advances). Rectangle clip regions (INTERSECTCLIPRECT, EXCLUDECLIPRECT, EXTSELECTCLIPRGN) limit later drawing. EMF+ only pictures play every EMF+ drawing record (brushes incl. gradients and textures, pens, paths, curves, images, strings, transforms, clipping; see crates/metafile/src/emfplus.rs for the few exceptions). Legacy OLE objects (VML preview) and ActiveX controls show their pictures
HEIC yes (macOS) no
Audio: MP3, AAC/M4A, ALAC, WAV, AIFF, CAF, FLAC, Ogg Vorbis, Opus yes (Ogg/Opus partly) yes Pure-Rust decoders in crates/media, crates/opus
WMA yes (Windows) embedded, not playable
Video: H.264, HEVC in MP4/MOV yes yes Pure-Rust, software decode
Video: VP9, AV1 in WebM/MKV partly (OS codecs) yes Ahead of PowerPoint for Mac
WMV yes (Windows) embedded, not playable
Online video (YouTube embeds) yes no Cloud feature
Embedded OLE objects yes kept on round trip (opaque), not insertable or editable ShapeKind::Opaque
Embedded fonts yes yes p:embeddedFontLst read and written. TrueType, OpenType and Embedded OpenType faces are used to draw; MicroType Express subsets are kept and written back, not drawn
3-D models (.glb) yes kept The model part round-trips (p:contentPart, including an mc:AlternateContent choice and its fallback picture). Not drawn as a mesh and not inserted from the ribbon

PPTX coverage (PresentationML / DrawingML)

What crates/pptx reads and writes, from ECMA-376. "Kept" means the XML is preserved untouched for the round trip even though DeckCraft doesn't model it.

Part / feature Read Write Notes
OPC package, content types, relationships yes yes Content types found anywhere in the zip (#19); zip bombs capped; relationship cycles bounded
Presentation, slide size, sections (p14), custom shows yes yes
Slide masters, layouts, placeholders and inheritance yes yes
Themes: colour, font (Latin/EA/CS), effect schemes, background styles yes yes Unknown theme parts kept
Notes slides and notes master yes yes Header, footer, date and slide-number text on the notes master is kept. Edited in View ▸ Notes Master. Notes keep their run and paragraph formatting (bold, italic, underline, size, levels, indents, bullets), shown and kept when edited in the notes pane and Notes Page (crates/pptx/tests/it/notes.rs). PDF notes pages print them as plain text; levels and bullets inherited from a custom notes master aren't shown
Handout master yes yes Header, footer, date and slide-number text is kept. Edited in View ▸ Handout Master
Shapes: preset geometry (~180 presets), custom geometry, groups, connectors yes yes Presets re-derived in crates/geom/src/preset.rs (and preset/extra.rs). Outlines, dashes, arrowheads and ink inside a group, and text, keep their size in points whatever the group's child-coordinate scale (#233).
Fills: solid, gradient, picture (stretch/tile/crop), pattern, group yes yes
Lines: dash, compound, caps, joins, arrowheads yes yes
Text: runs, paragraphs, bullets, numbering, levels, autofit, columns, vertical, RTL, script fonts, fields, hyperlinks yes yes Round-trip tests incl. justLow, date fields and paragraph defRPr (the defaults a placeholder's runs inherit, including a link). A click or hover action keeps its sound, endSnd, and a custom show's "Show and return". Gradient and outline on a run, autofit, columns, vertical and RTL text, tab stops and line spacing round-trip (corpus_style.rs)
Effects: shadow, glow, soft edges, reflection yes yes On a run, inner shadow, reflection and soft edges are kept along with the outer shadow and glow
3-D: scene3d, sp3d kept kept A bevel preset with no sp3d of its own is written back on shapes and theme effect styles. Bevels, cameras, face lighting and extrusion sides draw approximately
Text warp (prstTxWarp) yes yes All 40 presets rendered (crates/geom/src/warp.rs); adjust values (avLst) round-trip
Pictures with crop, recolour, duotone, alpha, luminance yes yes
Tables with styles yes yes Merges and both diagonal borders round-trip
Charts (c:chartSpace) yes yes Original XML kept in Chart.raw. Trendlines (type, order or period, forecast, intercept, equation and R²) are read, written and drawn. A combo chart's secondary value axis is read and is not written back when the chart is rebuilt. An unchanged chart also keeps its embedded workbook and the other parts its relationships point at, under fresh keptN names so they never replace a part DeckCraft writes (#335); editing the chart drops that workbook instead of writing a stale one
Charts (cx: chartex: waterfall, funnel, treemap, sunburst, histogram, Pareto, box) yes yes Unchanged parts kept, with their embedded workbook (#335); edited and new ones written from the model with a plain-chart fallback. Treemap and sunburst hierarchy levels kept; region maps kept, drawn as tiles (no map geometry)
SmartArt (dgm) the drawing part as a group (text laid out on its dsp:txXfrm rectangle), and the outline from the data model (Text Pane) as a diagram: data model from the outline; layout, quick style and colour parts (the file's own kept until the layout is switched); the group as the cached drawing DeckCraft's layout definitions are simple; a diagram without a cached drawing isn't drawn. SmartArt kept only as XML, without its relationship parts, still loses the drawing and data parts (#203)
Equations (OMML, a14:m) no no PR #55 open
Audio / video (p14:media, trim, fades, bookmarks) yes partly Playback options lost on save (#74)
Ink (contentPart) yes as a group of freeform lines PowerPoint sees shapes, not ink
Transitions incl. p14/p15 (Vortex, Morph…) all 48 all 48 every_transition_kind_round_trips
Animations: timing tree, build paragraphs, triggers, motion paths yes (flattened to the Animation Pane list) yes (rebuilt) every_animation_effect_survives; repeat, triggers and bounce end kept; rewind lost (#77)
Comments: legacy yes yes
Comments: modern threaded (p188) yes yes Written next to the flattened legacy part; replies and resolved state round-trip (threads_and_resolved_round_trip)
Unknown slide-level elements (graphicFrame OLE, extensions, mc:AlternateContent) kept kept Rendered as a labelled placeholder with the preview picture when present
Embedded fonts yes yes MicroType Express subsets kept, not drawn
Tags (p:custDataLst) on the presentation, slides and shapes yes yes Add-ins such as think-cell keep their identity; tags_survive_round_trip
Custom XML data, sensitivity label (docMetadata/LabelInfo.xml), custom document properties kept verbatim yes SharePoint metadata and the label survive a save; custom_xml_label_and_custom_properties_survive_save
VBA no no
Encryption (password to open) no no

What would move the PPTX number

  1. A larger owned corpus (corporate templates, SmartArt, chartex charts, EMF, embedded fonts, every transition and animation) in a corpus repo. crates/pptx/tests/it/corpus.rs and corpus_kinds.rs already check open → save → reopen (parts and text) for decks we author: tables, charts, animations, media, sections and masters/layouts among them. Still needed: render checks and a PowerPoint reference render for each (owner runs PowerPoint).
  2. The round-trip losses above (#70, #74, #77).
  3. SVG rendering. EMF+ records inside EMF/WMF are still not drawn.

Revision history

Date Change Summary
2026-10-18 minor Native chart payloads use referenced ZIP members and share media byte/item limits; old inline sidecars remain readable. Synthetic tests cover multi-megabyte layout, all shape owners, shared budgets, missing/corrupt members and invalid references. Scores and milestone scope unchanged
2026-10-15 trivial Native .deckcraft files serialize a nonempty chart kept sidecar (relationships, part names, content types and bytes). Older files without it still load; empty sidecars are omitted. Headless PPTX → edit → native save/reopen → PPTX checks keep the workbook and drawing bytes; editing chart data still drops the stale workbook
2026-10-15 trivial PPTX interactive timing sequences keep later On Click steps and paragraph click builds (trigger_steps.rs); mixed starts and independent buttons survive import/edit/save/reopen
2026-10-15 minor A .pptx slide that can't be read is reported instead of dropped in silence: file.open and file.openBytes return lostSlides, the app shows a message and opens a repaired copy (Save asks for a new name), and deckcraft-cli warns on stderr (#203)
2026-10-16 minor PNG slide export with a transparent background (#784)
2026-10-16 trivial A .ppsx opened from the command line or the file association starts the slide show (#862)
2026-10-14 minor Community PR wave folded: one row each for .pptx (seven owned corpus files), outline, 3-D models, embedded fonts, text, effects, 3-D (extrusion sides draw now), SmartArt (written as a diagram) and VBA (tags and custom XML are kept)
2026-10-12 minor SmartArt import lays each shape's text out on its dsp:txXfrm text rectangle instead of the shape box
2026-10-13 minor EMF/WMF rectangle clip regions limit later drawing; EMF+ records remain
2026-10-13 minor Embedded fonts read and written; TrueType, OpenType and Embedded OpenType faces drawn
2026-10-13 minor Owned corpus round-trip (parts and text). Notes-master header, footer, date and slide-number text is written back
2026-10-13 minor Owned corpus for comments, links and actions, custom shows, rotated groups and inherited placeholders. Paragraph defRPr and action sounds (including custom-show return) survive a save
2026-10-13 minor Owned corpus for text effects, body layout, theme effects, fills, bevels and connectors. A run's inner shadow, reflection and soft edge, and a bevel preset with no kept 3-D XML, survive a save
2026-10-14 minor 3-D models (.glb) round-trip, including an alternate-content fallback picture; animation bounce end is read and written (corpus_media.rs)
2026-10-14 minor Chart trendlines are drawn (linear, exponential, logarithmic, polynomial, power, moving average, forecast, equation and R²)
2026-10-14 minor Chart trendlines round-trip (corpus_charts.rs); merges, diagonal borders, columns, autofit, vertical text and WordArt warp are in that slice
2026-10-13 minor Add-in tags on the presentation, slides and shapes read and written
2026-10-13 minor Custom XML parts, the sensitivity label and custom properties kept on save
2026-10-13 minor Owned corpus round-trip for tables, charts, animations, media, sections and masters. Handout-master header, footer, date and slide-number text is written back
2026-10-14 minor PNG and JPEG slide export record their resolution (scale × 72 dpi; the default 2× picture is 144 dpi)
2026-10-13 trivial Unchanged charts keep the embedded workbook (ppt/embeddings/) and chart relationships (#335)
2026-10-12 minor EMF/WMF text and pattern brushes drawn; OLE (VML) and ActiveX previews read
2026-10-12 minor Text warps rendered; their adjust values read and written
2026-10-12 trivial Groups with a small legacy child coordinate space draw outlines, dashes, arrowheads and ink at their own width (#233)
2026-10-11 minor Speaker notes: formatting shown in the notes pane and Notes Page and kept when they are edited
2026-10-11 minor Animated GIF of the deck: export yes (#171)
2026-10-10 major Created from PowerPoint 16.113.4's CFBundleDocumentTypes, crates/pptx reader/writer and tests, and the open round-trip issues
2026-10-13 minor Comments: the writer emits the modern threaded part (p188) next to the flattened legacy one; the reader prefers the modern part and scales its EMU positions (#70)