@mengine/medeo-client 2.1.1-dsl.2 → 2.1.1-dsl.21

package/dist/legacy.js

@@ -1,6 +1,6 @@
 import { n as bytesToBase64, t as base64ToBytes } from "./base64-D-jd-dr5.js";
 import { c as readMengineEventStream, d as MenginePushRejectedError, i as openManualDocument, l as MengineHttpClient, n as MedeoHttpDocStorage, o as decodeDocVersionMark, r as ManualSyncTransport, s as encodeDocVersionMark, t as MemoryDocStorage, u as MengineHttpRequestError } from "./storage-BW3tI_ER.js";
-import { a as documentFormat, r as UnsupportedDocumentFormatError, t as DEFAULT_UNIT_TIME_MS } from "./video-draft-types-TXRv3wS6.js";
+import { a as documentFormat, r as UnsupportedDocumentFormatError, t as DEFAULT_UNIT_TIME_MS } from "./video-draft-types-BfIXM8q3.js";
 import { u as volumeSchema } from "./shared-iWnU2osE.js";
 import { sha256 } from "@noble/hashes/sha2.js";
 import { LoroDoc, LoroMap, LoroMovableList, LoroText, UndoManager } from "loro-crdt";

package/dist/testing.d.ts

@@ -1,5 +1,5 @@
-import { i as VideoDraft } from "./video-draft-types-c_shaRyq.js";
-import { _ as MedeoDsl } from "./index-Cu7SdKN5.js";
+import { i as VideoDraft } from "./video-draft-types-0J2vfXBu.js";
+import { b as MedeoDsl } from "./index-DVmvFhEn.js";
 import { LoroDoc } from "loro-crdt";
 
 //#region src/testing/in-memory-mengine-server.d.ts

package/dist/testing.js

@@ -1,6 +1,6 @@
 import { n as bytesToBase64, t as base64ToBytes } from "./base64-D-jd-dr5.js";
-import { a as documentFormat, i as assertDslDocument } from "./video-draft-types-TXRv3wS6.js";
-import { a as MedeoDsl, i as DslEditor, n as toVideoDraft, r as createInitialMedeoDsl } from "./dsl-EV6kCdr6.js";
+import { a as documentFormat, i as assertDslDocument } from "./video-draft-types-BfIXM8q3.js";
+import { i as DslEditor, n as toVideoDraft, o as MedeoDsl, r as createInitialMedeoDsl } from "./dsl-B7-4dMjr.js";
 import { t as classifyUpdate } from "./loro-relay-doc-Bi6JKqtQ.js";
 import { LoroDoc, VersionVector, encodeFrontiers } from "loro-crdt";
 //#region src/testing/in-memory-mengine-server.ts
@@ -450,7 +450,7 @@ function createTestDsl(options = {}) {
 			durationMs: 1e3,
 			storageKey: "speech.mp3"
 		},
-		script: { body: "hello" },
+		script: { segments: ["hello"] },
 		volume: 1,
 		voice: {
 			system: "voice-library",
@@ -466,10 +466,7 @@ function createTestDsl(options = {}) {
 			clipId: "caption",
 			startMs: 0,
 			durationMs: 1e3,
-			range: {
-				start: 0,
-				end: 5
-			},
+			segment: 0,
 			style: {}
 		}]
 	});

package/dist/{video-draft-types-c_shaRyq.d.ts → video-draft-types-0J2vfXBu.d.ts}

@@ -17,9 +17,24 @@ interface SeqMarker {
   rate?: number;
   fillTimeline?: true;
 }
+/**
+ * Where a Clip starts: at an absolute time on the Timeline, or following another Clip. No anchor
+ * means it plays in Tree order. One value, so concurrent placements keep one whole placement.
+ */
+interface TimeAnchor {
+  /** The Timeline (absolute) or another Clip (following). */
+  to: string;
+  offsetMs: number;
+  /** The absolute start recorded by the placement edit; used when the anchor cannot be solved. */
+  fallback_abs_ms: number;
+}
+/** A Clip's single-valued relations are fields of its own row: its content and its time anchor. */
 interface Clip {
   seqMarker: SeqMarker;
   volume?: number;
+  /** The media or Caption this use plays. */
+  source: string;
+  anchor?: TimeAnchor;
 }
 interface MediaLocator {
   system: ResourceSystem;
@@ -38,8 +53,23 @@ type Media = MediaLocator & ({
   kind: 'image';
   storageKey?: string;
 });
-interface AudioScript {
-  body: string;
+/** One segment of a script: its own text, and the id a Caption or PhoneticScript names it by. */
+interface AudioScriptSegment {
+  id: string;
+  text: string;
+  /**
+   * Where these words were spoken in the medium the script was read off, as the transcription
+   * file it names gave them (D186). Only a transcript's segments state it, both ends together.
+   */
+  startMs?: number;
+  endMs?: number;
+}
+/**
+ * A script is segmented text: its segments in order (D181). A transcript names the transcription
+ * file its words and times were read from (D186); a written script names none.
+ */
+interface AudioScript extends Partial<StoredFile> {
+  segments: AudioScriptSegment[];
   defaultLanguage?: string;
 }
 interface CaptionStyle$1 {
@@ -54,24 +84,54 @@ interface CaptionStyle$1 {
   entranceAnimationDurationMs?: number;
   strokeColor?: string;
   strokeWidth?: number;
+  /** The caption's center across the frame, from 0 at the left edge to 1 at the right. */
   positionX?: number;
+  /** The caption's center down the frame, from 0 at the top edge to 1 at the bottom. */
   positionY?: number;
 }
+/**
+ * What a Caption shows or a PhoneticScript is written for: one segment of a script, or a run inside
+ * it (D181). The run's ends are Cursors on that segment's Text, so they follow its edits; without
+ * them the whole segment is meant, whatever it comes to say. One value: taken or replaced whole.
+ */
+interface SegmentSelection {
+  audioScriptId: string;
+  segmentId: string;
+  start?: Cursor;
+  end?: Cursor;
+}
+type CaptionContent = SegmentSelection;
+/** Words come from a selected run of a script or from the Caption's own text, never both. */
 interface Caption {
+  selection?: CaptionContent;
   text?: string;
   style?: CaptionStyle$1;
 }
+/** The segment a PhoneticScript is written for, selected as a Caption selects one, and its pronunciation. */
+interface Phoneme extends SegmentSelection {
+  /** What the voice synthesizes for that selection, written in what the voice accepts. */
+  phonemeScript?: string;
+}
+/** A file Memota stored from raw content: how a Media File entity names it, and its digest. */
+interface StoredFile {
+  system: 'memota';
+  key: string;
+  storageKey?: string;
+  digest: string;
+}
+/**
+ * A variant of an AudioScript: the segment it is written for, plus what to synthesize. It has the Media File
+ * trait (D178): its `phonemeScript` is raw content, stored in Memota as a text file whenever it is
+ * written, and the file's facts are kept beside it. Without a `phonemeScript` it has no file.
+ */
+interface PhoneticScript extends Partial<StoredFile> {
+  phoneme: Phoneme;
+}
 interface Voice {
   system: 'voice-library';
   key: string;
   name?: string;
 }
-interface Asset {
-  system: ResourceSystem;
-  key: string;
-  storageKey?: string;
-  digest?: string;
-}
 /** Table keys carry entity identity; values never duplicate an id or a single-table kind. */
 interface EntityValues {
   'timeline-entities': Record<string, never>;
@@ -79,52 +139,32 @@ interface EntityValues {
   'clip-entities': Clip;
   'media-entities': Media;
   'audio-script-entities': AudioScript;
+  'phonetic-script-entities': PhoneticScript;
   'caption-entities': Caption;
   'voice-entities': Voice;
-  'asset-entities': Asset;
-}
-interface TimeAnchor {
-  from: string;
-  to: string;
-  offsetMs: number;
-  fallback_abs_ms: number;
-}
-interface CaptionContent {
-  captionId: string;
-  audioScriptId: string;
-  start: Cursor;
-  end: Cursor;
 }
+/** Only relations an entity may hold many of have tables; single-valued ones are fields of their owner. */
 interface RelationValues {
-  'clip-source-relations': {
-    clipId: string;
-    sourceId: string;
-  };
-  'time-anchor-relations': TimeAnchor;
-  'caption-content-relations': CaptionContent;
-  'caption-alignment-relations': {
-    captionId: string;
-    mediaId: string;
-    startMs: number;
-    endMs: number;
-  };
   'voice-timbre-relations': {
     recordingId: string;
     voiceId: string;
   };
+  /**
+   * `sourceDigest`: the source's digest when the product was made from it (D177). A product whose
+   * source's digest differs now was made from content the source no longer has.
+   */
   'derived-from-relations': {
     entityId: string;
     sourceId: string;
-  };
-  'from-asset-relations': {
-    audioScriptId: string;
-    assetId: string;
-  };
-  'to-asset-relations': {
-    audioScriptId: string;
-    assetId: string;
+    sourceDigest?: string;
   };
 }
+/** What a create states: a PhoneticScript's phoneme may follow by set-phoneme in the same write. */
+type EntityInput = Omit<EntityValues, 'phonetic-script-entities'> & {
+  'phonetic-script-entities': Omit<PhoneticScript, 'phoneme'> & {
+    phoneme?: Phoneme;
+  };
+};
 type EntityTable = keyof EntityValues;
 type RelationTable = keyof RelationValues;
 type EntityTables = { [K in EntityTable]: Record<string, EntityValues[K]> };
@@ -134,7 +174,7 @@ type EntityMutation = { [K in EntityTable]: {
   type: 'create-entity';
   table: K;
   id: string;
-  value: EntityValues[K];
+  value: EntityInput[K];
 } | {
   type: 'patch-entity';
   table: K;
@@ -168,12 +208,39 @@ type ContainMutation = {
   type: 'delete-node';
   entityId: string;
 };
-interface CaptionSelection {
-  captionId: string;
+/** A segment, or a run of it in UTF-16 offsets of its current text, as a write states it. */
+interface SegmentRange {
   audioScriptId: string;
-  start: number;
-  end: number;
+  segmentId: string;
+  start?: number;
+  end?: number;
+}
+interface CaptionSelection extends SegmentRange {
+  captionId: string;
+}
+interface PhonemeSelection extends SegmentRange {
+  phoneticScriptId: string;
+  phonemeScript?: string;
 }
+/** A script's segments are added, moved and removed by id; their text is edited in place. */
+type SegmentMutation = {
+  type: 'insert-segment';
+  audioScriptId: string;
+  segmentId: string;
+  index?: number;
+  text: string;
+  startMs?: number;
+  endMs?: number;
+} | {
+  type: 'move-segment';
+  audioScriptId: string;
+  segmentId: string;
+  index: number;
+} | {
+  type: 'delete-segment';
+  audioScriptId: string;
+  segmentId: string;
+};
 type TextMutation = {
   type: 'edit-caption-text';
   captionId: string;
@@ -182,12 +249,15 @@ type TextMutation = {
 } | {
   type: 'edit-text';
   audioScriptId: string;
+  segmentId: string;
   expectedBodyId?: string;
   edits: readonly TextEdit[];
 } | ({
   type: 'set-caption-content';
-} & CaptionSelection);
-type DslMutation = EntityMutation | RelationMutation | ContainMutation | TextMutation;
+} & CaptionSelection) | ({
+  type: 'set-phoneme';
+} & PhonemeSelection);
+type DslMutation = EntityMutation | RelationMutation | ContainMutation | SegmentMutation | TextMutation;
 interface ContainNode {
   treeId: TreeID;
   entityId: string;
@@ -207,6 +277,7 @@ interface RichTextSnapshot {
   plainText: string;
   delta: TextRun[];
 }
+/** Selected text, with its UTF-16 range in the Text it lies in: the segment's, or a Caption's own. */
 interface CaptionText {
   text: string;
   start: number;
@@ -215,12 +286,13 @@ interface CaptionText {
 }
 type DeepReadonly<T> = T extends Cursor ? T : T extends object ? { readonly [K in keyof T]: DeepReadonly<T[K]> } : T;
 type MedeoDslSnapshot = DeepReadonly<EntityTables & RelationTables & {
-  'contain-relations': ContainNode[];
+  'contain-relations': ContainNode[]; /** A Caption's own text by the Caption's id, and each script segment's text by the segment's id. */
   richText: Record<string, RichTextSnapshot>;
-  captionText: Record<string, CaptionText>;
+  captionText: Record<string, CaptionText>; /** What each PhoneticScript's selection holds now, resolved as a Caption's is. */
+  phoneticText: Record<string, CaptionText>;
   diagnostics: DslDiagnostic[];
 }>;
-type TextMark = 'language' | 'pauseMs';
+type TextMark = 'language';
 type TextEdit = {
   type: 'insert';
   at: number;
@@ -234,7 +306,7 @@ type TextEdit = {
   start: number;
   end: number;
   key: TextMark;
-  value: string | number;
+  value: string;
 } | {
   type: 'unmark';
   start: number;
@@ -256,9 +328,10 @@ type TextDelta = {
 //#endregion
 //#region ../medeo-dsl/src/ids.d.ts
 declare function tupleHash(tuple: readonly string[]): string;
-declare function resourceId(kind: 'media' | 'voice' | 'asset', system: string, key: string): string;
+declare function resourceId(kind: 'media' | 'voice', system: string, key: string): string;
 declare function relationId<K extends RelationTable>(table: K, value: RelationValues[K]): string;
-type EntityIdKind = 'timeline' | 'track' | 'clip' | 'audio-script' | 'caption';
+/** Kinds of id a document mints: its entities', and a script segment's. */
+type EntityIdKind = 'timeline' | 'track' | 'clip' | 'audio-script' | 'segment' | 'phonetic-script' | 'caption';
 type DslIdFactory = (kind: EntityIdKind) => string;
 declare const defaultIdFactory: DslIdFactory;
 //#endregion
@@ -806,4 +879,4 @@ type VideoDraftContent = Pick<VideoDraft, 'timeline' | 'main_track' | 'above_mai
 type VideoDraftInput = Pick<VideoDraft, keyof VideoDraftContent>;
 declare const DEFAULT_UNIT_TIME_MS = 33.333;
 //#endregion
-export { Track$1 as $, ContainNode as A, Media as B, Caption as C, CaptionText as D, CaptionStyle$1 as E, EntityTable as F, ResourceSystem as G, RelationTable as H, EntityTables as I, TextEdit as J, RichTextSnapshot as K, EntityValues as L, DslDiagnostic as M, DslMutation as N, Clip as O, EntityMutation as P, TimeAnchor as Q, FieldPatch as R, AudioScript as S, CaptionSelection as T, RelationTables as U, RelationMutation as V, RelationValues as W, TextMutation as X, TextMark as Y, TextRun as Z, defaultIdFactory as _, VideoDraftCaptionStyle as a, tupleHash as b, VideoDraftInput as c, VideoDraftTrack as d, TrackRole as et, VideoDraftTrackItem as f, EntityIdKind as g, DslIdFactory as h, VideoDraft as i, DeepReadonly as j, ContainMutation as k, VideoDraftPartUnion as l, OpActor as m, DEFAULT_UNIT_TIME_MS as n, VideoDraftContent as o, CommitOptions as p, TextDelta as q, VideoCreationSettings as r, VideoDraftContentPartUnion as s, CaptionDisplayCue as t, Voice as tt, VideoDraftTimeline as u, relationId as v, CaptionContent as w, Asset as x, resourceId as y, MedeoDslSnapshot as z };
+export { TextMutation as $, DeepReadonly as A, PhonemeSelection as B, CaptionContent as C, Clip as D, CaptionText as E, EntityTables as F, ResourceSystem as G, RelationTable as H, EntityValues as I, SegmentSelection as J, RichTextSnapshot as K, FieldPatch as L, DslMutation as M, EntityMutation as N, ContainMutation as O, EntityTable as P, TextMark as Q, MedeoDslSnapshot as R, Caption as S, CaptionStyle$1 as T, RelationTables as U, RelationMutation as V, RelationValues as W, TextDelta as X, StoredFile as Y, TextEdit as Z, defaultIdFactory as _, VideoDraftCaptionStyle as a, tupleHash as b, VideoDraftInput as c, VideoDraftTrack as d, TextRun as et, VideoDraftTrackItem as f, EntityIdKind as g, DslIdFactory as h, VideoDraft as i, Voice as it, DslDiagnostic as j, ContainNode as k, VideoDraftPartUnion as l, OpActor as m, DEFAULT_UNIT_TIME_MS as n, Track$1 as nt, VideoDraftContent as o, CommitOptions as p, SegmentRange as q, VideoCreationSettings as r, TrackRole as rt, VideoDraftContentPartUnion as s, CaptionDisplayCue as t, TimeAnchor as tt, VideoDraftTimeline as u, relationId as v, CaptionSelection as w, AudioScript as x, resourceId as y, Media as z };