Conversation
EncoderOrderingTests pins that XMLCoder emits child elements in CodingKeys declaration order and sorts only under the opt-in .sortedKeys, so a future XMLCoder bump that changes default ordering fails loudly. ChoiceEncodingSpike answers the Step 3 (#8) design question: the synthesized decode of an items = "" choice array silently yields an empty array, so Spine must hand-write init(from:) (attributes via the keyed container, items via singleValueContainer choice-array decode); the synthesized encode with items = "" + DynamicNodeEncoding composes and emits children in array order with attributes intact. SpineOrderTests asserts the transitions FeaturePair spine is asset-clip, transition, asset-clip as a raw tree (permanent), and marks the round-trip order assertion XCTExpectFailure until #8 lands ordered Spine.items (XCTSkip on corelibs-xctest platforms). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* Step 3: Ordered Spine.items (#8) * Replace choice if/switch ladders with XMLChoiceField tables Avoid cyclomatic_complexity disables on SpineItem/Spine by walking field-handler and batch-append tables. Document the anti-pattern in agent notes and a Cursor rule. Co-authored-by: Cursor <cursoragent@cursor.com> * Restore XMLChoiceField.encode after helper split The encode method was dropped when OrderedChoiceItems moved to its own file; choice encoding needs it. Co-authored-by: Cursor <cursoragent@cursor.com> * Add XMLChoiceCodable and OrderedChoiceItems payloads/replace Share choice decode/encode via protocol extension and fold typed slice get/set into OrderedChoiceItems so Spine drops isX accessors. Co-authored-by: Cursor <cursoragent@cursor.com> * Move Spine typed item accessors into Spine+Items Keep Spine.swift under the file_length warning limit. Co-authored-by: Cursor <cursoragent@cursor.com> * Fix strict lint violations - Expand one-line conditional returns in SpineItem+Accessors.swift - Split XMLChoiceCodable protocol and extension out of XMLChoiceField.swift into XMLChoiceCodable.swift (one declaration per file, file types order) - Move ChoiceKey typealias above CodingKeys enum in SpineItem.swift - Apply swift-format/swiftlint autofixes (line breaks, whitespace) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Replace OrderedChoiceItems with OrderedChoiceContainer protocol. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
…ngKeys (#23) * feat: implement AnchoredItem choice enum and DTD-ordered CodingKeys for story elements (Issue #9) * Replace choice if/switch ladders with XMLChoiceField tables Avoid cyclomatic_complexity disables on SpineItem/Spine by walking field-handler and batch-append tables. Document the anti-pattern in agent notes and a Cursor rule. Co-authored-by: Cursor <cursoragent@cursor.com> * Apply XMLChoiceField tables to AnchoredItem and seed inits Convert AnchoredItem choice Codable and AssetClip/Storyline/SyncClip/ RefClip/MCClip convenience seed chains to table-driven helpers so large if/switch ladders are gone on Step 4. Co-authored-by: Cursor <cursoragent@cursor.com> * Add XMLChoiceCodable and OrderedChoiceItems payloads/replace Share choice decode/encode via protocol extension and fold typed slice get/set into OrderedChoiceItems so Spine drops isX accessors. Co-authored-by: Cursor <cursoragent@cursor.com> * Clear file_length/line_length via choice protocols and splits Adopt XMLChoiceCodable and AnchoredChoiceContainer, wrap DTD comments, and move decode inits into +Coding files so clip types need no lint disables. Co-authored-by: Cursor <cursoragent@cursor.com> * Fix step4 lint violations: decoder docs, AnchoredItem ordering, auto-format - Document public init(from decoder:) in the five clip +Coding files - Put conditional returns on their own lines in AnchoredItem+Accessors - Move ChoiceKey typealias ahead of CodingKeys in AnchoredItem - Apply swift-format/swiftlint auto-fixes (trailing comma, whitespace) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Remove stray doc comments duplicated from +Coding inits Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Replace OrderedChoiceItems with OrderedChoiceContainer protocol. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* Step 5: ResourceID/ResourceRef and typed resource shells (#10). Adopt phantom-typed refs across resources, document shells, and story refs; keep Asset.audioRate as String Hz while Sequence/Multicam use AudioRate. Co-authored-by: Cursor <cursoragent@cursor.com> * Replace choice if/switch ladders with XMLChoiceField tables Avoid cyclomatic_complexity disables on SpineItem/Spine by walking field-handler and batch-append tables. Document the anti-pattern in agent notes and a Cursor rule. Co-authored-by: Cursor <cursoragent@cursor.com> * Restore XMLChoiceField.encode after helper split The encode method was dropped when OrderedChoiceItems moved to its own file; choice encoding needs it. Co-authored-by: Cursor <cursoragent@cursor.com> * Apply XMLChoiceField tables to AnchoredItem and seed inits Convert AnchoredItem choice Codable and AssetClip/Storyline/SyncClip/ RefClip/MCClip convenience seed chains to table-driven helpers so large if/switch ladders are gone on Step 4. Co-authored-by: Cursor <cursoragent@cursor.com> * Add XMLChoiceCodable and OrderedChoiceItems payloads/replace Share choice decode/encode via protocol extension and fold typed slice get/set into OrderedChoiceItems so Spine drops isX accessors. Co-authored-by: Cursor <cursoragent@cursor.com> * Move Spine typed item accessors into Spine+Items Keep Spine.swift under the file_length warning limit. Co-authored-by: Cursor <cursoragent@cursor.com> * Clear file_length/line_length via choice protocols and splits Adopt AnchoredChoiceContainer on clip hosts, wrap DTD comments, and move decode inits into +Coding extensions while keeping ResourceRef typing. Co-authored-by: Cursor <cursoragent@cursor.com> * Apply lint auto-fixes from swift-format and swiftlint No step5-specific violations found; remaining lint findings are inherited from step3/step4 and resolve when those branches merge up. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * Replace OrderedChoiceItems with OrderedChoiceContainer protocol. Co-authored-by: Cursor <cursoragent@cursor.com> --------- Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
#14). - .gitignore: /transitions.fcpxml and /titles.fcpxml are reproducible via `swift run fcpxml-dsl export`. - typed-generation-gate.md: accept the DSL import gate — transitions and titles imported into Final Cut Pro, dissolve confirmed between clips, anchored title lane confirmed. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
Important Review skippedToo many files! This PR contains 457 files, which is 357 over the limit of 100. To get a review, reduce the PR to 100 files or fewer by splitting it into smaller PRs or changing its base branch. Upgrade to a paid plan to raise the limit. Usage-priced reviews support at most 300 files. ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (457)
You can disable this status message by setting the 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. Comment |
SBObject proxies resolve sdef term names (name, id, duration, file, ...) through value(forKey:), not the sdef cocoa keys the inspector was passing (displayName, uniqueIdentifier, durationDict, URL). The cocoa keys raised NSUnknownKeyException against a running Final Cut Pro, and ObjC exceptions are uncatchable from Swift, so the whole test process died. Every key was probed individually against a live FCP in isolated child processes before adoption: - name/id/file/duration/frameDuration/startTime/timecodeFormat and the children keys all resolve live. - persistent ID is declared in the sdef but unimplemented by FCP (errors -1728 even in AppleScript); persistentID is now optional on the scripted models and read tolerantly. - media time records arrive as value/timescale/epoch/flags dictionaries, which the existing parser already handles. - timecode format arrives as an OSType NSNumber (drop/ndrp/unsp); the parser now decodes four-char codes, with a regression test. Mock tests re-pin the term-name contract so they can no longer pass while the live path is broken, and the live test now asserts non-empty library content when Final Cut is running (verified live: 12 tests, 6 suites pass with FCP open). Docs known-issue callouts removed accordingly. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Apple-platform legs were using the library-only FCPKit scheme (no test action); omit scheme so swift-build derives FCPKit-Package. Stub xmllint Process usage on iOS/tvOS/watchOS/visionOS. Point CodeQL at xcode-27 so autobuild can parse swift-tools-version 6.4. Co-authored-by: Cursor <cursoragent@cursor.com>
|
You are seeing this message because GitHub Code Scanning has recently been set up for this repository, or this pull request contains the workflow file for the Code Scanning tool. What Enabling Code Scanning Means:
For more information about GitHub Code Scanning, check out the documentation. |
This was referenced Aug 2, 2026
…42) * Fix #41: gate match-analysis-type smart collection on FCPXML version `Defaults.smartCollections()` unconditionally emitted a "Missing Analysis" smart collection containing `<match-analysis-type>`. That element does not exist in the FCPXML 1.13 DTD (0 occurrences in FCPXMLv1_13.dtd, 5 in FCPXMLv1_14.dtd), so every DSL document exported at the library default version — `FCPXMLVersion.supportedGeneration`, which is 1.13 — was invalid against Final Cut's own DTD. The `fcpxml-dsl` CLI defaults to 1.14, which is why the common path validated and this went unnoticed. Thread the target version through `ResourceStore`, which is already passed `inout` to every `build` call, and gate the collection on `compatibility(relativeTo:)` rather than string-matching a version. The collection is gated, not dropped: the feature-pair acceptance tests export at 1.14 and their real Final Cut fixtures do contain "Missing Analysis", so removing it outright would fail the symmetric structural diff. Also extracts the previously private `assertDTDValidates` helper into a shared `DTDValidationSupport.swift` for reuse by other test suites. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Namespace soft-promotion behind an enum instead of global functions Addresses review feedback on #42: "no global properties or functions". `softPromote` and its four private helpers become statics on a `SoftPromote` enum. The entry point is renamed to `promote(_:resources:)` so it does not overload the private `library(for:version:)` helpers it calls. These globals predate this branch, but the rule is a general one, so they are converted here rather than left as the only exceptions in the module. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Replace Defaults and SoftPromote with type extensions; adopt typed throws Addresses review feedback on #42: Defaults belongs on the types it concerns, and SoftPromote reads as an anti-pattern rather than a namespace. `Defaults` held two unrelated concerns and is split along that seam: - The version-gated vocabulary moves onto the type that decides it — `FCPXMLVersion.admitsAnalysisMatching` and `.defaultSmartCollections`. Both depended on nothing but the version, so this is a clean move. The documented `.malformed`-admits-the-element behaviour is preserved. - The sequence factory becomes `ModelSequence.init(packing:format:)`, an extension on the FCPXML model type it constructs. The typealias exists because FCPKitDSL declares its own `Sequence`, so the file cannot be named for a qualified `FCPKit.Sequence`. `SoftPromote` becomes `Built.promotedLibrary(resources:)`. The switch is a fold up a fixed ladder — item to spine to sequence to project to event to library — where each case enters at its own rung, which is an operation on `Built` rather than a bag of functions. No protocol here: one implementation, closed case set. Two incidental fixes fall out of the rewrite: the `.event` arm no longer inlines `FCPKit.Library(...)` instead of calling its own helper, and the `.item` arm no longer packs twice (once here, once inside the old `Defaults.sequence`). Adopts typed throws across FCPKitDSL. Every throw site in the module already raised `BuildError` and nothing else, and the module calls no throwing FCPKit API, so `throws(BuildError)` now runs end to end from `DSLNode.build` through `Document.export` — callers get a concrete catch type. Closures need explicit `throws(BuildError)` annotations where they propagate. Note this narrows `DSLNode.build`: an external conformer declaring plain `throws` no longer satisfies the protocol. Pre-1.0 and in-policy for a protocol already documented as evolving, but it is an API narrowing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
* Fix #36: lift .anchor onto StoryItem and make DSLNode public `.anchor(lane:offset:content:)` was declared only on `AssetClip`, so nothing could be anchored onto a generator or a color. A media-free slide deck needs exactly that: text over a solid color background. Introduce `StoryItem` — "content that can sit in a spine", matching the DTD's `%clip_item;` entity — and hang the shared `.anchor` modifier off it. `AssetClip`, `Generator`, `Title`, `Gap`, and `Transition` conform directly. `Color` promotes to `Generator`, since it is a model type in another module and cannot gain stored properties; `Color.build` already desugared that way, so anchoring just performs the desugaring one step earlier. A public protocol requirement cannot mention internal types, so `DSLNode`, `Built`, and `ResourceStore` become public. Every `ResourceStore` member stays internal, so only the type name is exposed and no usable extension point ships. Anchoring onto a `Transition` is accepted but never emitted: the DTD does not admit anchored items there. Pinned by a test so it stays documented behavior rather than something a later change "fixes" into a throw. Also fixes a latent bug: `.anchor` replaced rather than appended, so chained calls silently dropped earlier lanes. The shared modifier now appends, and `replacingAnchors` is documented as the pure setter it is. This was latent because nothing chained; `StoryItem` makes chaining the natural idiom. Adds the missing `.video` and `.generator` cases to `anchoredExtent`, so an anchored generator contributes to its parent's extent. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Cover the new anchoredExtent case, negative lanes, and Title/Gap hosts Follow-up from an adversarial review of e4325f7. The `.video` case added to `anchoredExtent` had no test: the existing "anchored item longer than its background" test used a Title, which the pre-existing `.title` case already handled. The new branch could have been deleted with the suite staying green. Now covered by an anchored *generator* outrunning its background. Also adds the negative-lane case (legal FCPXML — content below the primary storyline; only lane 0 is reserved) and Title/Gap as anchor hosts, both of which gained StoryItem conformance in e4325f7 without direct coverage. Documents why the `.generator` packing branch passes `anchoredExtent: 0`: `FCPKit.Generator` has no `anchoredItems` property at all, because `<generator>` is absent from the DTD's `%anchor_item;` list. Zero is the correct value there, not a stub — the previous commit message implied that call site was also updated, which overstated the change. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Replace global functions with namespaced statics; StoryItem refines DSLNode Addresses review feedback on #43. "As a rule never use global functions or properties": the anchored-item lowering helpers become statics on a new `AnchoredItemBuilder` enum, and the pre-existing `storyItems` global becomes `StoryItemLowering.items`. The latter predates this branch, but the rule is a general one, so leaving it behind would undercut it. `StoryItems.swift` is renamed to match its type. "Are all StoryItem a DSLNode?" — yes. Every conformer is a `DSLNode`, and the two extension-based conformances (`Color`, `Generator`) extend types already declared as such. `StoryItem` now refines `DSLNode` rather than `DocumentContent`, which makes the relationship explicit and lets the four `DSLNode, StoryItem` declarations drop the redundant conformance. This also tightens the protocol: the free `.anchor` extension calls `replacingAnchors` and relies on the conformer's `build` lowering those anchors, which only a `DSLNode` can do. The old declaration did not require that. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Replace lowering namespaces with a protocol and array extensions Addresses review feedback on #43: AnchoredItemBuilder and StoryItemLowering are namespaces holding functions rather than types that mean anything. The anchoring case had a real duplication behind it. `Anchor.applyLaneOffset` and `AnchoredItemBuilder.item` each switched over the same four shapes, and both already omitted `.gap` despite the schema admitting it — two switches, one concept, already drifted. `AnchorableItem` states that set once. Each conformer describes how to re-wrap itself into both ordered-choice containers, so `Built.anchorable` is the only switch and no downcasts are needed. `Anchor.build` drops from a body with its own parallel switch to three lines. `.gap` stays unsupported, so behaviour is unchanged; widening it is a capability change and gets its own issue. The rest become extensions on what they operate on: - `Built.anchoredItem()` and `Built.placed(lane:offset:)` — the mapping is genuinely `Built` to an anchored item, so it lives on `Built`. - `[any DSLNode].anchoredItems(resources:)` and `[any DocumentContent].spineItems(resources:)` — both always took a collection, so they read as collection operations: `content.contents.spineItems(resources: &resources)`. Adopts typed throws across the module, matching #42. Every throw site already raised `BuildError`, so `throws(BuildError)` runs end to end and callers get a concrete catch type. Note this narrows `DSLNode.build`: an external conformer declaring plain `throws` no longer satisfies the protocol. Pre-1.0 and in-policy for a protocol documented as evolving, but it is an API narrowing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Carry the version property through the v0.1.x merge `ResourceStore` needs both this branch's public cascade and #42's `version` property. The merge resolution restored this branch's copy, which predates #42, so `ResourceStore(version:)` went missing and `Document+Export` failed to compile against it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
…yle ids (#44) * Fix #35: allocate unique text-style-def ids per document `Title.build` hardcoded the id "ts1" in both the `TextStyle(ref:)` and the `TextStyleDef(id:)` it emitted, for every title. The DTD declares `text-style-def/@id` as `ID`, and XML requires `ID` values to be unique within a document, so any document containing two titles emitted invalid XML. This was latent only because no shipped document had more than one title. The presentation deck (#38) has seven. Add a `textStyleID()` allocator to `ResourceStore`, modeled on the existing `nextNumber` resource counter but in its own `ts1`, `ts2`, … namespace. Ids are numbered globally across titles, matching what Final Cut itself writes: `Tests/FCPKitTests/TestData/UntitledXML.fcpxml` shows two titles producing ts1..ts4 rather than restarting per title. The counter starts at 1, so single-title documents still emit exactly "ts1" and the feature-pair fixtures need no edits. The DTD regression test exports at 1.14 on purpose: at the 1.13 default the document is also invalid because the default smart collections emit `match-analysis-type`, which 1.13 does not declare (#41, fixed separately). That unrelated failure would otherwise mask the ID regression this guards. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Fix #37: add Title styling and frame positioning modifiers `Title.build` hardcoded Helvetica 63pt Regular white centered for every title. A slide deck needs to distinguish a heading from body text and to place text somewhere other than dead centre. Adds `TitleStyle` plus `.font`, `.fontSize`, `.fontFace`, `.fontColor`, `.alignment`, `.bold`, and `.name` modifiers, and `FramePosition` with an alignment-first `.position(_:inset:)` and an absolute `.position(x:y:)` escape hatch. Positioning emits a sibling `adjust-transform`, which composes over whatever the title preset does internally. The unit is percent of frame HEIGHT on both axes, measured from the centre, Y-up — verified against real Final Cut output (TestData/UntitledXML.fcpxml:448, position="-17.8241 7.77778"), which the tests reproduce exactly rather than asserting a hand-derived number. The frame size rides on `ResourceStore`, which is already threaded through every `build` call, so `DSLNode.build`'s signature is untouched. `Sequence` publishes its format's dimensions before building children and restores the outer value afterwards, so nested sequences do not leak frame sizes. Nothing is emitted unless a modifier is applied: an unstyled, unpositioned title serializes byte-identically to today, `bold` is written only when true, and `fontSize` collapses whole numbers (63, not 63.0). The feature-pair structural diff against a real Final Cut export passes unchanged. Also fills in `AdjustTransform`'s missing DTD attributes — `enabled`, `rotation`, and `anchor` — which were silently dropped. CodingKeys follow the DTD's declaration order, in which `enabled` comes first. Adds post-resolution spine ordering tests: resolution rewrites title elements on the way out, and the diff engine's inventory is order-blind, so ordering needed a guard that runs after resolution rather than only after build. Allows `x` and `y` as identifier names, alongside the existing `id`/`no`. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Fix a crash and two silent-wrong-output bugs in title positioning Follow-up from an adversarial review of c6b74c5. All three share a root cause: unvalidated Double -> String conversion, and an Optional treated as "use a default" rather than "cannot resolve". 1. CRASH. `fontSize` and the position formatter both did an unguarded `Int(value)` after a whole-number check. `1e21` is a whole number, so `Title("Hi").fontSize(1e21)` trapped with "Double value cannot be converted to Int" and took the host process down (signal 5). Infinity and NaN hit the same path. Replaced both with a shared `decimalString(_:)` that falls back to the plain Double description outside Int's range or when non-finite. A library must not crash on user input. 2. `.position(.top, inset: 80)` with no enclosing format silently DROPPED the inset and emitted "0 50". An inset is in points, and converting points to Final Cut's percent-of-height unit requires the frame height. The absolute `.position(x:y:)` path already threw `missingFrameSize` in exactly this situation, so the design already agreed it is unresolvable — the alignment path just failed silently instead. It now throws too. A zero inset still never throws, since plain alignments need no frame size. 3. `.position(.center, inset: 100)` emitted a pointless `adjust-transform` with "0 0", breaking the "centred titles emit nothing" invariant while ignoring the inset. `.center` is the frame centre on both axes, so an inset has no direction to move along; it now always resolves to no transform. Also corrects a misleading test comment: the fixture whose position value the formula test reproduces uses FFVideoFormat3840x2160p24, not 1080p. The math is scale-invariant so the test passed for the right reason, but the stated provenance would have misled the next person deriving from it. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Namespace the decimal formatter; document why not Decimal.FormatStyle Addresses review feedback on #44. "Make it a rule never to have global functions": `decimalString` becomes `AttributeValue.decimal(_:)`, and the file is renamed to match. The trivial `FramePosition.format` wrapper is inlined rather than left as an alias. "Is there something here to use Decimal.FormatStyle?" — I tested it, and no. Three reasons, all verified rather than assumed: - It is locale-aware. `63.5` renders as "63,5" under de_DE and fr_FR, which would emit invalid FCPXML on any non-English machine. This is the decisive one, and it is now pinned by a test so nobody swaps a locale-aware formatter back in. - It rounds to six fractional digits, so a position component like 7.777777777 would be silently truncated. - `Decimal(Double.infinity)` traps, so it does not even solve the crash that motivated this helper. It does collapse whole numbers natively, which is the one thing it would have bought us — not worth the three costs above. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Express the FCPXML decimal formatter as String.init(fcpxmlValue:) Addresses review feedback on #44: make it an extension on Double or an init extension on String rather than a namespaced static. Went with `String.init(fcpxmlValue:)`. Both call sites produce a String from a Double, which is exactly what an init is for, and it keeps the conversion with the type being produced instead of hanging a domain-specific property off `Double`. The argument label matters: a bare `String(_: Double)` already exists via `LosslessStringConvertible`, so an unlabelled init would quietly overload it. Verified `String(63.5)` still routes to the stdlib and returns "63.5" while `String(fcpxmlValue:)` applies the whole-number collapse. Behaviour is unchanged — same guards, same fallbacks, same tests. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Adopt typed throws across FCPKitDSL Matches the typed-throws work on #42 and #43 so the three branches agree. Every throw site in FCPKitDSL already raised `BuildError` and nothing else, and the module calls no throwing FCPKit API, so `throws(BuildError)` runs end to end from `DSLNode.build` through `Document.export`. Callers get a concrete catch type instead of `any Error`. Closures that propagate need explicit `throws(BuildError)` annotations. In `Title.build` the `flatMap`/`map` chain resolving a deferred position is rewritten as a plain `if let`, which types cleanly and reads better than an annotated chain. Note this narrows `DSLNode.build`: an external conformer declaring plain `throws` no longer satisfies the protocol. Pre-1.0 and in-policy for a protocol documented as evolving, but it is an API narrowing. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Merges the v0.1.0 release branch into main: Steps 0–8 of the accepted plan (ADR 0002) plus scaffolding, gates, and docs sync.
Highlights
FCPTimeand strong value types, DTD-orderedCodingKeys, single orderedSpine.items(fixes the transition mis-ordering Final Cut rejected),AnchoredItemchoice arrays, resources and document shells.FCPKitDSL: create-first authoring (Document+ result builders) with transition overlap packing, anchored lanes, and DTD-valid export (media-rep, not pre-1.6src);fcpxml-dslCLI exports smoke-test cuts and drivesverify-import.FCPKitScripting: macOS-only read-only ScriptingBridge inspector.docs/manual/typed-generation-gate.md) — transitions and titles cuts imported, dissolve confirmed between clips.Before merging
v0.1.x) so it is included here, then close Step 8: Docs sync after code lands #13.Known issue shipped with v0.1.0
FCPKitScriptingSBObject bridge crashes against a live running Final Cut Pro (cocoa keys vs sdef term names). Deliberately deferred past v0.1.0.After merging
v0.1.0.🤖 Generated with Claude Code