09 — Minimal bundled template (KeynoteKit resource)

[leogdion]

Closes #21. Part of #12, see also #7.

Ships the hand-authored blank-theme blank.key (98,240 bytes) as a SwiftPM resource on the KeynoteKit target, and establishes the write entry point with its basedOn: parameter.

.copy, not .process — verified empirically

The ticket and PLAN both say .copy "because a .key is a directory-shaped bundle." That rationale does not hold for this file — it is a flat single-file zip (16 stored entries), not a directory-shaped bundle. So I tested both rules rather than inheriting the reasoning.

Both build, and both preserve the bytes on macOS today:

Rule	Result	Bundle layout
.process	builds, SHA-256 unchanged	Contents/Resources/blank.key
.copy	builds, SHA-256 unchanged	Contents/Resources/blank.key

Chose .copy. .process passing today is incidental, not a guarantee: it licenses SwiftPM to transform or rename resources, and the no-op here is just the absence of a known rule for the .key extension — a toolchain or platform that adds one would silently corrupt the archive. .copy states the actual requirement (bytes survive verbatim) and keeps the path stable across the Darwin/Linux/Windows/Android legs. The conclusion matches the ticket; the stated reason is different, and the comment in Package.swift records the real one.

write(to:basedOn:)

try Deck().write(to: url)                                          // bundled template
try Deck().write(to: url, basedOn: KeynoteTemplate(contentsOf: myTheme))

basedOn: defaults to KeynoteTemplate.bundled. It exists from day one so the primary API does not change shape once slide content lands. With no content modelled yet, write emits the template verbatim.

KeynoteTemplate is a value type over either the bundled resource or a caller-supplied URL; TemplateError.bundledResourceMissing covers Bundle.module failing at runtime (stripped or relocated resource bundle).

Tests — verified load-bearing

7 tests: resource resolves from Bundle.module, SHA-256 + byte count pinned, zip local-header signature intact, and the three write paths.

The byte-identity test was checked for false confidence. My first tamper attempt passed — because Bundle.module in a test context resolves to the copy nested inside KeynoteKitTests.xctest, not the top-level product bundle, and I had corrupted the latter. After tampering the copy the test actually reads, a single flipped byte fails the suite. The assertion is real.

SHA-256 is implemented in plain Swift (Tests/KeynoteKitTests/SHA256Digest.swift) rather than CryptoKit, which is Darwin-only, so the Linux/Windows/Android legs keep running. Its output was cross-checked against shasum and Python's hashlib.

Redistribution caveat (#7)

The template derives from Keynote's 21_basicwhite, so a small amount of Apple-authored theme content ships inside this package:

• a 2.5 KB Data/st-*.jpg
• a 50 KB DocumentStylesheet.iwa of theme styling
• three documentResourceLocator entries resolving against Keynote's installed theme bundle at open time

Reducing the template from 458 KB (by removing the theme's slide layouts through Keynote's UI) minimized this exposure but did not eliminate it, exactly as #7 anticipated. Callers who must ship no Apple-authored theme content can supply their own base via KeynoteTemplate(contentsOf:). Documented on the KeynoteTemplate type and at the resource declaration in Package.swift.

Scope

Deliberately minimal per the lane note — resource plus the default basedOn: path only. No authoring DSL, no slide/text-item supply, no SlideContent surface; those stay with #20/#22/#23. Deck is an empty value type whose only job today is to own the write entry point.

Touched only Sources/KeynoteKit/**, Tests/KeynoteKitTests/**, and the KeynoteKit target stanza in Package.swift. No changes to Snappy, IWAFraming, KeynoteKitProtobuf, KeynoteKitScripting, research/, or PLAN.md.

Verification

• swift build / swift test green under Swift 6.4 — 7 tests pass
• LINT_MODE=STRICT ./Scripts/lint.sh exits 0, 0 violations

Two strict-mode violations came up and were fixed by renaming and reordering, not by disabling rules: identifier_name (SHA-256's s0/s1 → sigma0/sigma1) and type_contents_order (static helpers moved above the instance @Test methods). New types went into separate files to satisfy one_declaration_per_file.

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: e7e1f5c1-d9b7-45e7-88f4-b398c562c1da

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}