@torrent-tv/proxy 2.89.1 → 2.89.3

package/services/encode/output/cut-grid.js

@@ -1,107 +1,107 @@
-/**
- * @file Where a file is cut, and which keyframe each cut is.
- *
- * Two answers, not one, and that is the whole point of this file existing.
- *
- * A cut has a time on the PLAYER's clock — 0-based, because a playlist is —
- * and it has a time on the FILE's clock, which is the keyframe the muxer will
- * actually seek to. The two differ by the container's own start time, and a
- * copied picture cannot be cut anywhere but at a real keyframe, so both are
- * needed and neither can be derived from the other after the fact.
- *
- * **Deriving one from the other after the fact is what broke a viewing on
- * 2026-09-05.** The 0-based time was stored, rounded to six places, and the
- * seek then added the start time back and looked the result up in the file's
- * keyframe list by value. That round trip is lossy: a keyframe at 26.234 s in a
- * container starting at 0.083 s comes back as 26.233999999999998, the lookup
- * takes "the keyframe at or before" that, and answers the PREVIOUS one — 8.717 s
- * earlier. Two parts in a quadrillion became one keyframe interval, that
- * interval became a trim, the trim moved every cut of the run backwards by
- * another interval, and the run's files were numbered from #36 while carrying
- * film 17.4 s before what the playlist says #36 holds. The player's video buffer
- * covered the playhead, its audio buffer had a 17.4 s hole across it, the
- * intersection was empty, and the viewer waited two minutes and was told the
- * proxy had sent no video.
- *
- * So the keyframe is carried, by index, from the one place that knows it.
- */
-
-/**
- * @typedef {object} CutGrid
- * @property {number[]} boundaries - Cut times on the player's clock, 0-based,
- *   ascending, one more than there are segments.
- * @property {number[]} sourceTimes - The same cuts on the FILE's clock: for a
- *   keyframe grid, the keyframe itself, exactly as the container stated it.
- *   Same length as `boundaries`, so an index names both.
- */
-
-/**
- * Cut a file into segments.
- *
- * On a keyframe grid the cuts are the container's own keyframes, thinned so
- * that no segment is much shorter than asked for; the first cut is the start of
- * the file and the last is its end, neither of which need be a keyframe.
- *
- * On a uniform grid the cuts are multiples of the segment length, and the two
- * clocks are the same clock: a re-encode places its own keyframes and owes the
- * container's start time nothing.
- *
- * @param {object} params
- * @param {boolean} params.useKeyframeGrid
- * @param {number} params.durationSeconds
- * @param {number} params.segDur - The segment length asked for.
- * @param {number[]} [params.keyframeTimes] - On the file's own clock.
- * @param {number} [params.startTime] - The container's start time.
- * @returns {CutGrid}
- */
-export function computeCutGrid({ useKeyframeGrid, durationSeconds, segDur, keyframeTimes, startTime }) {
-  const total = Number.isFinite(durationSeconds) && durationSeconds > 0 ? durationSeconds : 0;
-  const step = Number.isFinite(segDur) && segDur > 0 ? segDur : 4;
-  const base = Number.isFinite(startTime) ? startTime : 0;
-  const uniform = () => {
-    const boundaries = [];
-    for (let cutAt = 0; cutAt < total - 0.001; cutAt += step) {
-      boundaries.push(Number(cutAt.toFixed(6)));
-    }
-    boundaries.push(total);
-    // One clock: nothing here is a keyframe of the source, so nothing is owed
-    // the container's start time either.
-    return { boundaries, sourceTimes: [...boundaries] };
-  };
-  if (!useKeyframeGrid || !Array.isArray(keyframeTimes) || keyframeTimes.length === 0 || total <= 0) {
-    return uniform();
-  }
-  const kept = keyframeTimes
-    .filter((time) => Number.isFinite(time))
-    .map((time) => ({ source: time, published: time - base }))
-    // THE END IS MEASURED THE SAME WAY AS EVERY OTHER CUT. A keyframe nearer to
-    // the end than one segment leaves a tail too short to be a segment, and the
-    // rule below — no cut closer than a step to the one before it — never looks
-    // at the end at all. It used to be guarded by fifty milliseconds, a number
-    // from nowhere: field 2026-09-11, a keyframe 160 ms before the end of a
-    // 54-minute film passed it and left segment #541 lasting 0.16 s. The sound
-    // has no data in such a tail, so its run made 541 segments where the picture
-    // made 542, was marked short, and the repair that followed was handed a
-    // start later than its own end — 190 bytes that are not a fragment, and a
-    // viewer held 23 s at the last minute of the film for a 404.
-    .filter((cut) => cut.published >= -0.001 && cut.published < total - step)
-    .sort((left, right) => left.published - right.published);
-  // The first cut is the start of the file, whatever the container's own clock
-  // says that is; the last is its end. Neither is a keyframe, and a run never
-  // seeks to either — index 0 is served without a seek at all.
-  const boundaries = [0];
-  const sourceTimes = [base];
-  for (const cut of kept) {
-    if (cut.published >= boundaries[boundaries.length - 1] + step - 0.05) {
-      boundaries.push(Number(cut.published.toFixed(6)));
-      // NOT the rounded value with the base added back. This is the number the
-      // container stated, and it is what the seek must ask for.
-      sourceTimes.push(cut.source);
-    }
-  }
-  boundaries.push(total);
-  sourceTimes.push(total + base);
-  // A degenerate index — one keyframe, or none inside the film — is no grid at
-  // all, and an even one serves better than a table with a single entry.
-  return boundaries.length >= 2 ? { boundaries, sourceTimes } : uniform();
-}
+/**
+ * @file Where a file is cut, and which keyframe each cut is.
+ *
+ * Two answers, not one, and that is the whole point of this file existing.
+ *
+ * A cut has a time on the PLAYER's clock — 0-based, because a playlist is —
+ * and it has a time on the FILE's clock, which is the keyframe the muxer will
+ * actually seek to. The two differ by the container's own start time, and a
+ * copied picture cannot be cut anywhere but at a real keyframe, so both are
+ * needed and neither can be derived from the other after the fact.
+ *
+ * **Deriving one from the other after the fact is what broke a viewing on
+ * 2026-09-05.** The 0-based time was stored, rounded to six places, and the
+ * seek then added the start time back and looked the result up in the file's
+ * keyframe list by value. That round trip is lossy: a keyframe at 26.234 s in a
+ * container starting at 0.083 s comes back as 26.233999999999998, the lookup
+ * takes "the keyframe at or before" that, and answers the PREVIOUS one — 8.717 s
+ * earlier. Two parts in a quadrillion became one keyframe interval, that
+ * interval became a trim, the trim moved every cut of the run backwards by
+ * another interval, and the run's files were numbered from #36 while carrying
+ * film 17.4 s before what the playlist says #36 holds. The player's video buffer
+ * covered the playhead, its audio buffer had a 17.4 s hole across it, the
+ * intersection was empty, and the viewer waited two minutes and was told the
+ * proxy had sent no video.
+ *
+ * So the keyframe is carried, by index, from the one place that knows it.
+ */
+
+/**
+ * @typedef {object} CutGrid
+ * @property {number[]} boundaries - Cut times on the player's clock, 0-based,
+ *   ascending, one more than there are segments.
+ * @property {number[]} sourceTimes - The same cuts on the FILE's clock: for a
+ *   keyframe grid, the keyframe itself, exactly as the container stated it.
+ *   Same length as `boundaries`, so an index names both.
+ */
+
+/**
+ * Cut a file into segments.
+ *
+ * On a keyframe grid the cuts are the container's own keyframes, thinned so
+ * that no segment is much shorter than asked for; the first cut is the start of
+ * the file and the last is its end, neither of which need be a keyframe.
+ *
+ * On a uniform grid the cuts are multiples of the segment length, and the two
+ * clocks are the same clock: a re-encode places its own keyframes and owes the
+ * container's start time nothing.
+ *
+ * @param {object} params
+ * @param {boolean} params.useKeyframeGrid
+ * @param {number} params.durationSeconds
+ * @param {number} params.segDur - The segment length asked for.
+ * @param {number[]} [params.keyframeTimes] - On the file's own clock.
+ * @param {number} [params.startTime] - The container's start time.
+ * @returns {CutGrid}
+ */
+export function computeCutGrid({ useKeyframeGrid, durationSeconds, segDur, keyframeTimes, startTime }) {
+  const total = Number.isFinite(durationSeconds) && durationSeconds > 0 ? durationSeconds : 0;
+  const step = Number.isFinite(segDur) && segDur > 0 ? segDur : 4;
+  const base = Number.isFinite(startTime) ? startTime : 0;
+  const uniform = () => {
+    const boundaries = [];
+    for (let cutAt = 0; cutAt < total - 0.001; cutAt += step) {
+      boundaries.push(Number(cutAt.toFixed(6)));
+    }
+    boundaries.push(total);
+    // One clock: nothing here is a keyframe of the source, so nothing is owed
+    // the container's start time either.
+    return { boundaries, sourceTimes: [...boundaries] };
+  };
+  if (!useKeyframeGrid || !Array.isArray(keyframeTimes) || keyframeTimes.length === 0 || total <= 0) {
+    return uniform();
+  }
+  const kept = keyframeTimes
+    .filter((time) => Number.isFinite(time))
+    .map((time) => ({ source: time, published: time - base }))
+    // THE END IS MEASURED THE SAME WAY AS EVERY OTHER CUT. A keyframe nearer to
+    // the end than one segment leaves a tail too short to be a segment, and the
+    // rule below — no cut closer than a step to the one before it — never looks
+    // at the end at all. It used to be guarded by fifty milliseconds, a number
+    // from nowhere: field 2026-09-11, a keyframe 160 ms before the end of a
+    // 54-minute film passed it and left segment #541 lasting 0.16 s. The sound
+    // has no data in such a tail, so its run made 541 segments where the picture
+    // made 542, was marked short, and the repair that followed was handed a
+    // start later than its own end — 190 bytes that are not a fragment, and a
+    // viewer held 23 s at the last minute of the film for a 404.
+    .filter((cut) => cut.published >= -0.001 && cut.published < total - step)
+    .sort((left, right) => left.published - right.published);
+  // The first cut is the start of the file, whatever the container's own clock
+  // says that is; the last is its end. Neither is a keyframe, and a run never
+  // seeks to either — index 0 is served without a seek at all.
+  const boundaries = [0];
+  const sourceTimes = [base];
+  for (const cut of kept) {
+    if (cut.published >= boundaries[boundaries.length - 1] + step - 0.05) {
+      boundaries.push(Number(cut.published.toFixed(6)));
+      // NOT the rounded value with the base added back. This is the number the
+      // container stated, and it is what the seek must ask for.
+      sourceTimes.push(cut.source);
+    }
+  }
+  boundaries.push(total);
+  sourceTimes.push(total + base);
+  // A degenerate index — one keyframe, or none inside the film — is no grid at
+  // all, and an even one serves better than a table with a single entry.
+  return boundaries.length >= 2 ? { boundaries, sourceTimes } : uniform();
+}

package/services/encode/output/ladder.js

@@ -1,32 +1,32 @@
-/**
- * @file The heights a picture may be offered at.
- *
- * A fact about an output and about nothing else: it takes a source height and
- * answers with heights. What the machine can afford of them is the quality
- * budget's question, and what a player may splice between is `OutputCatalog`.
- */
-
-/**
- * The standard heights, high to low. One list for both uses: the heights the
- * menu offers, and the rungs the realtime budget steps down through below a
- * viewer's own screen size. They were two lists with the same values, and two
- * owners of one fact drift apart.
- *
- * Only rungs below the source are used: upscaling invents detail and costs the
- * encoder more than the source itself.
- */
-export const LADDER_HEIGHTS = Object.freeze([2160, 1440, 1080, 720, 540, 480, 360, 240]);
-
-/**
- * The heights offered for a source of this height, largest first.
- *
- * @param {number} sourceHeight
- * @returns {number[]}
- */
-export function variantHeightsFor(sourceHeight) {
-  if (!Number.isFinite(sourceHeight) || sourceHeight <= 0) {
-    return [];
-  }
-  const rungs = LADDER_HEIGHTS.filter((height) => height < sourceHeight);
-  return [Math.round(sourceHeight), ...rungs];
-}
+/**
+ * @file The heights a picture may be offered at.
+ *
+ * A fact about an output and about nothing else: it takes a source height and
+ * answers with heights. What the machine can afford of them is the quality
+ * budget's question, and what a player may splice between is `OutputCatalog`.
+ */
+
+/**
+ * The standard heights, high to low. One list for both uses: the heights the
+ * menu offers, and the rungs the realtime budget steps down through below a
+ * viewer's own screen size. They were two lists with the same values, and two
+ * owners of one fact drift apart.
+ *
+ * Only rungs below the source are used: upscaling invents detail and costs the
+ * encoder more than the source itself.
+ */
+export const LADDER_HEIGHTS = Object.freeze([2160, 1440, 1080, 720, 540, 480, 360, 240]);
+
+/**
+ * The heights offered for a source of this height, largest first.
+ *
+ * @param {number} sourceHeight
+ * @returns {number[]}
+ */
+export function variantHeightsFor(sourceHeight) {
+  if (!Number.isFinite(sourceHeight) || sourceHeight <= 0) {
+    return [];
+  }
+  const rungs = LADDER_HEIGHTS.filter((height) => height < sourceHeight);
+  return [Math.round(sourceHeight), ...rungs];
+}