text: Finer-grained types transport-side

Author: peaBerberian

src/main_thread/text_displayer/html/html_text_displayer.ts

@@ -3,7 +3,10 @@ import { onEnded, onSeeked, onSeeking } from "../../../compat/event_listeners";
 import onHeightWidthChange from "../../../compat/on_height_width_change";
 import config from "../../../config";
 import log from "../../../log";
-import type { ITextTrackSegmentData } from "../../../transports";
+import type {
+  ISupportedTextTrackFormat,
+  ITextTrackSegmentData,
+} from "../../../transports";
 import type { IRange } from "../../../utils/ranges";
 import { convertToRanges } from "../../../utils/ranges";
 import type { CancellationSignal } from "../../../utils/task_canceller";
@@ -409,11 +412,13 @@ export default class HTMLTextDisplayer implements ITextDisplayer {
 }
 
 /** Data of chunks that should be pushed to the `HTMLTextDisplayer`. */
-export interface ITextTracksBufferSegmentData {
+export interface ITextTracksBufferSegmentData<
+  TDataFormatName extends ISupportedTextTrackFormat = ISupportedTextTrackFormat,
+> {
   /** The text track data, in the format indicated in `type`. */
-  data: string | BufferSource;
+  data: TDataFormatName extends "mp4vtt" ? BufferSource : string;
   /** The format of `data` (examples: "ttml", "srt" or "vtt") */
-  type: string;
+  type: TDataFormatName;
   /**
    * Language in which the text track is, as a language code.
    * This is mostly needed for "sami" subtitles, to know which cues can / should
@@ -437,9 +442,7 @@ export interface ITextTracksBufferSegmentData {
  */
 if ((__ENVIRONMENT__.CURRENT_ENV as number) === (__ENVIRONMENT__.DEV as number)) {
   // @ts-expect-error: uncalled function just for type-checking
-  function _checkType<T extends string | BufferSource>(
-    input: ITextTrackSegmentData<T>,
-  ): void {
+  function _checkType(input: ITextTrackSegmentData): void {
     function checkEqual(_arg: ITextTracksBufferSegmentData): void {
       /* nothing */
     }

src/transports/types.ts

@@ -435,12 +435,12 @@ export type ISupportedTextTrackFormat = "ttml" | "vtt" | "mp4vtt" | "srt" | "sam
 
 /** Text track segment data, once parsed. */
 export interface ITextTrackSegmentData<
-  T extends string | BufferSource = string | BufferSource,
+  TDataFormatName extends ISupportedTextTrackFormat = ISupportedTextTrackFormat,
 > {
   /** The text track data, in the format indicated in `type`. */
-  data: T;
+  data: TDataFormatName extends "mp4vtt" ? BufferSource : string;
   /** The format of `data`. */
-  type: ISupportedTextTrackFormat;
+  type: TDataFormatName;
   /**
    * Language in which the text track is, as a language code.
    * This is mostly needed for "sami" subtitles, to know which cues can / should
@@ -491,10 +491,7 @@ export interface ITransportAudioVideoSegmentPipeline {
 
 export interface ITransportTextSegmentPipeline {
   loadSegment: ISegmentLoader<ILoadedTextSegmentFormat>;
-  parseSegment: ISegmentParser<
-    ILoadedTextSegmentFormat,
-    ITextTrackSegmentData<Uint8Array | string> | null
-  >;
+  parseSegment: ISegmentParser<ILoadedTextSegmentFormat, ITextTrackSegmentData | null>;
 }
 
 export type ITransportSegmentPipeline =

src/transports/utils/parse_text_track.ts

@@ -19,12 +19,7 @@ import type { ISegment } from "../../manifest";
 import { getMDAT } from "../../parsers/containers/isobmff";
 import startsWith from "../../utils/starts_with";
 import { utf8ToStr } from "../../utils/string_parsing";
-import type {
-  IChunkTimeInfo,
-  ISegmentContext,
-  ISupportedTextTrackFormat,
-  ITextTrackSegmentData,
-} from "../types";
+import type { IChunkTimeInfo, ISegmentContext, ITextTrackSegmentData } from "../types";
 
 /**
  * Returns the a string expliciting the format of a text track when that text
@@ -122,24 +117,27 @@ export function getISOBMFFEmbeddedTextTrackData(
     }
   }
 
-  let type: ISupportedTextTrackFormat = getISOBMFFTextTrackFormat(codecs);
-  let textData: string | BufferSource;
+  const type: "ttml" | "vtt" = getISOBMFFTextTrackFormat(codecs);
   const mdat = getMDAT(chunkBytes);
   const mdatStr = mdat !== null ? utf8ToStr(mdat) : "";
-  if (codecs === "wvtt") {
-    if (!startsWith(mdatStr, "WEBVTT") && !startsWith(mdatStr, "\xfe\xffWEBVTT")) {
-      // From how I understand it, we're here in a special WebVTT format  embedded
-      // in an MP4 where the whole chunk contains information, now just the MDAT
-      textData = chunkBytes;
-      type = "mp4vtt";
-    } else {
-      textData = mdatStr;
-    }
-  } else {
-    textData = mdatStr;
+  if (
+    codecs === "wvtt" &&
+    !startsWith(mdatStr, "WEBVTT") &&
+    !startsWith(mdatStr, "\xfe\xffWEBVTT")
+  ) {
+    // From how I understand it, we're here in a special WebVTT format  embedded
+    // in an MP4 where the whole chunk contains information, now just the MDAT
+    return {
+      data: chunkBytes,
+      type: "mp4vtt",
+      language,
+      start: startTime,
+      end: endTime,
+      timescale: initTimescale ?? null,
+    };
   }
   return {
-    data: textData,
+    data: mdatStr,
     type,
     language,
     start: startTime,