Merge pull request #32 from eshaz/ogg-sample-count

Fix samples counts

Author: eshaz

package-lock.json

@@ -1,6 +1,6 @@
 {
   "name": "codec-parser",
-  "version": "2.4.3",
+  "version": "2.5.0",
   "description": "Library that parses raw data from audio codecs into frames containing data, header values, duration, and other information.",
   "main": "index.js",
   "types": "index.d.ts",
@@ -37,8 +37,8 @@
   "sideEffects": false,
   "homepage": "https://github.com/eshaz/codec-parser#readme",
   "devDependencies": {
-    "@types/jest": "^29.5.3",
-    "jest": "^29.6.2",
-    "prettier": "^3.0.1"
+    "@types/jest": "^29.5.13",
+    "jest": "^29.7.0",
+    "prettier": "^3.3.3"
   }
 }

package.json

@@ -42,6 +42,7 @@ import {
   parseFrame,
   checkCodecUpdate,
   reset,
+  isLastPage,
 } from "./constants.js";
 import HeaderCache from "./codecs/HeaderCache.js";
 import MPEGParser from "./codecs/mpeg/MPEGParser.js";
@@ -229,12 +230,37 @@ export default class CodecParser {
   [mapFrameStats](frame) {
     if (frame[codecFrames]) {
       // Ogg container
-      frame[codecFrames].forEach((codecFrame) => {
-        frame[duration] += codecFrame[duration];
-        frame[samples] += codecFrame[samples];
-        this[mapCodecFrameStats](codecFrame);
-      });
+      if (frame[isLastPage]) {
+        // cut any excess samples that fall outside of the absolute granule position
+        // some streams put invalid data in absolute granule position, so only do this
+        // for the end of the stream
+        let absoluteGranulePositionSamples = frame[samples];
+
+        frame[codecFrames].forEach((codecFrame) => {
+          const untrimmedCodecSamples = codecFrame[samples];
+
+          if (absoluteGranulePositionSamples < untrimmedCodecSamples) {
+            codecFrame[samples] =
+              absoluteGranulePositionSamples > 0
+                ? absoluteGranulePositionSamples
+                : 0;
+            codecFrame[duration] =
+              (codecFrame[samples] / codecFrame[header][sampleRate]) * 1000;
+          }
+
+          absoluteGranulePositionSamples -= untrimmedCodecSamples;
+
+          this[mapCodecFrameStats](codecFrame);
+        });
+      } else {
+        frame[samples] = 0;
+        frame[codecFrames].forEach((codecFrame) => {
+          frame[samples] += codecFrame[samples];
+          this[mapCodecFrameStats](codecFrame);
+        });
+      }
 
+      frame[duration] = (frame[samples] / this._sampleRate) * 1000 || 0;
       frame[totalSamples] = this._totalSamples;
       frame[totalDuration] =
         (this._totalSamples / this._sampleRate) * 1000 || 0;

src/CodecParser.js

@@ -100,9 +100,8 @@ export default class FLACParser extends Parser {
             ))
           ) {
             // found a valid next frame header
-            let frameData = yield* this._codecParser[readRawData](
-              nextHeaderOffset,
-            );
+            let frameData =
+              yield* this._codecParser[readRawData](nextHeaderOffset);
 
             if (!this._codecParser._flushing)
               frameData = frameData[subarray](0, nextHeaderOffset);

src/codecs/flac/FLACParser.js

@@ -16,15 +16,10 @@
     along with this program.  If not, see <https://www.gnu.org/licenses/>
 */
 
-import { sampleRate, frameCount, frameSize } from "../../constants.js";
 import CodecFrame from "../CodecFrame.js";
 
 export default class OpusFrame extends CodecFrame {
-  constructor(data, header) {
-    super(
-      header,
-      data,
-      ((header[frameSize] * header[frameCount]) / 1000) * header[sampleRate],
-    );
+  constructor(data, header, samples) {
+    super(header, data, samples);
   }
 }

src/codecs/opus/OpusFrame.js

@@ -27,6 +27,10 @@ import {
   parseOggPage,
   enable,
   getHeaderFromUint8Array,
+  preSkip,
+  frameSize,
+  frameCount,
+  sampleRate,
 } from "../../constants.js";
 import Parser from "../Parser.js";
 import OpusFrame from "./OpusFrame.js";
@@ -40,6 +44,7 @@ export default class OpusParser extends Parser {
 
     onCodec(this[codec]);
     this._identificationHeader = null;
+    this._preSkipRemaining = null;
   }
 
   get [codec]() {
@@ -67,7 +72,22 @@ export default class OpusParser extends Parser {
             this._headerCache,
           );
 
-          if (header) return new OpusFrame(segment, header);
+          if (header) {
+            if (this._preSkipRemaining === null)
+              this._preSkipRemaining = header[preSkip];
+
+            let samples =
+              ((header[frameSize] * header[frameCount]) / 1000) *
+              header[sampleRate];
+
+            if (this._preSkipRemaining > 0) {
+              this._preSkipRemaining -= samples;
+              samples =
+                this._preSkipRemaining < 0 ? -this._preSkipRemaining : 0;
+            }
+
+            return new OpusFrame(segment, header, samples);
+          }
 
           this._codecParser[logError](
             "Failed to parse Ogg Opus Header",

src/codecs/opus/OpusParser.js

@@ -70,6 +70,7 @@ import {
   uint8Array,
   dataView,
 } from "../../constants.js";
+import { readInt64le } from "../../utilities.js";
 
 export default class OggPageHeader {
   static *[getHeader](codecParser, headerCache, readOffset) {
@@ -111,13 +112,7 @@ export default class OggPageHeader {
     // Byte (7-14 of 28)
     // * `FFFFFFFF|FFFFFFFF|FFFFFFFF|FFFFFFFF|FFFFFFFF|FFFFFFFF|FFFFFFFF|FFFFFFFF`
     // * Absolute Granule Position
-
-    /**
-     * @todo Safari does not support getBigInt64, but it also doesn't support Ogg
-     */
-    try {
-      header[absoluteGranulePosition] = view.getBigInt64(6, true);
-    } catch {}
+    header[absoluteGranulePosition] = readInt64le(view, 6);
 
     // Byte (15-18 of 28)
     // * `GGGGGGGG|GGGGGGGG|GGGGGGGG|GGGGGGGG`

src/containers/ogg/OggPageHeader.js

@@ -26,6 +26,7 @@ import {
   codec,
   data,
   length,
+  samples,
   segments,
   subarray,
   vorbis,
@@ -38,6 +39,7 @@ import {
   uint8Array,
   isLastPage,
   streamSerialNumber,
+  absoluteGranulePosition,
 } from "../../constants.js";
 
 import Parser from "../../codecs/Parser.js";
@@ -57,6 +59,7 @@ class OggStream {
     this._continuedPacket = new uint8Array();
     this._codec = null;
     this._isSupported = null;
+    this._previousAbsoluteGranulePosition = null;
   }
 
   get [codec]() {
@@ -149,6 +152,16 @@ class OggStream {
       );
     }
 
+    // set total samples in this ogg page
+    if (this._previousAbsoluteGranulePosition !== null) {
+      oggPage[samples] = Number(
+        oggPage[absoluteGranulePosition] -
+          this._previousAbsoluteGranulePosition,
+      );
+    }
+
+    this._previousAbsoluteGranulePosition = oggPage[absoluteGranulePosition];
+
     if (this._isSupported) {
       const frame = this._parser[parseOggPage](oggPage);
       this._codecParser[mapFrameStats](frame);

src/containers/ogg/OggParser.js

@@ -185,12 +185,37 @@ class BitReader {
   }
 }
 
+/**
+ * @todo Old versions of Safari do not support BigInt
+ */
+const readInt64le = (view, offset) => {
+  try {
+    return view.getBigInt64(offset, true);
+  } catch {
+    const sign = view.getUint8(offset + 7) & 0x80 ? -1 : 1;
+    let firstPart = view.getUint32(offset, true);
+    let secondPart = view.getUint32(offset + 4, true);
+
+    if (sign === -1) {
+      firstPart = ~firstPart + 1;
+      secondPart = ~secondPart + 1;
+    }
+
+    if (secondPart > 0x000fffff) {
+      console.warn("This platform does not support BigInt");
+    }
+
+    return sign * (firstPart + secondPart * 2 ** 32);
+  }
+};
+
 export {
   crc8,
   flacCrc16,
   crc32Function,
   reverse,
   concatBuffers,
   bytesToString,
+  readInt64le,
   BitReader,
 };

src/utilities.js

@@ -311,6 +311,7 @@ describe("CodecParser", () => {
       testParser("ogg.opus.framesize_60", mimeType, "opus", 251, 0);
       testParser("ogg.opus.surround", mimeType, "opus", 737, 0);
       testParser("ogg.opus.channel_family_255", mimeType, "opus", 284, 0);
+      testParser("ogg.opus.last_page_trimming", mimeType, "opus", 51, 0);
     });
 
     describe("Ogg Vorbis", () => {
@@ -322,7 +323,7 @@ describe("CodecParser", () => {
         "ogg.vorbis.setup_packets_separate_pages",
         mimeType,
         "vorbis",
-        118,
+        119,
       );
       testParser("metronome2.vorbis", mimeType, "vorbis", 3);
     });