fff

android/oral-trainer-sdk/src/main/java/cn/learningpad/oraltrainer/sdk/OralTrainerController.kt

@@ -11,7 +11,6 @@ import androidx.media3.common.MediaItem
 import androidx.media3.common.PlaybackException
 import androidx.media3.common.Player
 import androidx.media3.exoplayer.ExoPlayer
-import androidx.media3.exoplayer.PlayerMessage
 import androidx.media3.exoplayer.source.DefaultMediaSourceFactory
 import java.util.concurrent.CopyOnWriteArraySet
 import kotlin.math.max
@@ -31,8 +30,10 @@ class OralTrainerController internal constructor(
     private var lastSentenceIndex: Int? = null
     private var stopAtSentenceEnd = false
     private var stopAtSentenceEndIndex: Int? = null
-    private var boundaryMessage: PlayerMessage? = null
-    private var boundaryGuardToken = 0
+    private var activeBaseItemIndex: Int = -1
+    private var activeSentenceWindow: SentenceWindow? = null
+    private var sentenceWindows: List<SentenceWindow> = emptyList()
+    private var sentencePlaylistBaseIndex: Int = -1
 
     var config: PlayerConfig = initialConfig
         private set
@@ -47,13 +48,20 @@ class OralTrainerController internal constructor(
             if (released) {
                 return
             }
-            enforceSentenceBoundary()
             notifySentenceIfChanged()
             notifySnapshot()
             mainHandler.postDelayed(this, 250L)
         }
     }
 
+    private data class SentenceWindow(
+        val baseItemIndex: Int,
+        val mediaId: String,
+        val sentenceIndex: Int,
+        val startMs: Long,
+        val endMs: Long,
+    )
+
     init {
         val mediaSourceFactory = DefaultMediaSourceFactory(
             StreamingCache.dataSourceFactory(appContext, sdkConfig)
@@ -66,24 +74,34 @@ class OralTrainerController internal constructor(
                 exoPlayer.playWhenReady = initialConfig.autoPlay
                 exoPlayer.addListener(object : Player.Listener {
                     override fun onPlaybackStateChanged(playbackState: Int) {
+                        if (playbackState == Player.STATE_ENDED) {
+                            applySentenceEndForAnchor()
+                        }
+                        if (playbackState == Player.STATE_READY) {
+                            refreshPlaybackWindow()
+                        }
+                        if (playbackState == Player.STATE_READY ||
+                            playbackState == Player.STATE_ENDED
+                        ) {
+                            refreshSentenceRepeatMode()
+                        }
                         notifySnapshot()
                     }
 
                     override fun onIsPlayingChanged(isPlaying: Boolean) {
                         if (isPlaying) {
-                            refreshSentenceBoundaryGuard()
+                            refreshPlaybackWindow()
                         } else {
-                            cancelSentenceBoundaryGuard()
+                            refreshPlaybackWindow()
                         }
                         notifySnapshot()
                     }
 
                     override fun onMediaItemTransition(mediaItem: MediaItem?, reason: Int) {
-                        lastSentenceIndex = null
-                        cancelSentenceBoundaryGuard()
+                        activeSentenceWindow = currentSentenceWindow()
+                        lastSentenceIndex = activeSentenceWindow?.sentenceIndex
                         listeners.forEach { it.onMediaChanged(currentTrainingItem()) }
                         notifySentenceIfChanged(force = true)
-                        refreshSentenceBoundaryGuard()
                         notifySnapshot()
                     }
 
@@ -93,7 +111,9 @@ class OralTrainerController internal constructor(
                         reason: Int,
                     ) {
                         notifySentenceIfChanged(force = true)
-                        refreshSentenceBoundaryGuard()
+                        activeSentenceWindow = currentSentenceWindow()
+                        lastSentenceIndex = activeSentenceWindow?.sentenceIndex
+                        refreshPlaybackWindow()
                         notifySnapshot()
                     }
 
@@ -140,7 +160,11 @@ class OralTrainerController internal constructor(
         lastSentenceIndex = null
         stopAtSentenceEnd = false
         stopAtSentenceEndIndex = null
-        player.setMediaItems(items.map { it.toMedia3Item() }, safeIndex, max(0L, startPositionMs))
+        activeBaseItemIndex = safeIndex
+        activeSentenceWindow = null
+        sentenceWindows = emptyList()
+        sentencePlaylistBaseIndex = -1
+        player.setMediaItem(items[safeIndex].toMedia3Item(), max(0L, startPositionMs))
         player.prepare()
         player.playWhenReady = config.autoPlay
         listeners.forEach { it.onMediaChanged(currentTrainingItem()) }
@@ -166,10 +190,14 @@ class OralTrainerController internal constructor(
             return true
         }
         mediaItems[index] = updatedItem
-        if (index == player.currentMediaItemIndex) {
+        if (index == activeBaseItemIndex) {
             listeners.forEach { it.onMediaChanged(updatedItem) }
+            if (config.sentenceMode && !config.continuousPlayback && sentences.isNotEmpty()) {
+                installSentencePlaylist(index, activeSentenceWindow?.sentenceIndex ?: 0)
+            } else {
+                refreshPlaybackWindow()
+            }
             notifySentenceIfChanged(force = true)
-            refreshSentenceBoundaryGuard()
             notifySnapshot()
         }
         return true
@@ -177,13 +205,12 @@ class OralTrainerController internal constructor(
 
     fun play() {
         player.play()
-        refreshSentenceBoundaryGuard()
+        refreshPlaybackWindow()
     }
 
     fun pause() {
         stopAtSentenceEnd = false
         stopAtSentenceEndIndex = null
-        cancelSentenceBoundaryGuard()
         player.pause()
     }
 
@@ -198,19 +225,31 @@ class OralTrainerController internal constructor(
     fun stop() {
         stopAtSentenceEnd = false
         stopAtSentenceEndIndex = null
-        cancelSentenceBoundaryGuard()
         player.stop()
     }
 
     fun seekTo(positionMs: Long) {
         stopAtSentenceEnd = false
         stopAtSentenceEndIndex = null
-        player.seekTo(max(0L, positionMs))
-        refreshSentenceBoundaryGuard()
+        val safePositionMs = max(0L, positionMs)
+        val anchor = sentenceAnchorAt(safePositionMs)
+        if (config.sentenceMode && !config.continuousPlayback && anchor != null) {
+            if (ensureSentenceWindow(anchor)) {
+                seekInSentenceWindow(anchor, safePositionMs)
+            } else {
+                player.seekTo(safePositionMs)
+            }
+        } else {
+            restoreFullWindow(safePositionMs)
+            player.seekTo(safePositionMs)
+        }
+        refreshPlaybackWindow()
+        notifySentenceIfChanged(force = true)
+        notifySnapshot()
     }
 
     fun seekRelative(deltaMs: Long) {
-        seekTo(player.currentPosition + deltaMs)
+        seekTo(playbackPositionMs() + deltaMs)
     }
 
     fun rewind(stepMs: Long = config.defaultSeekStepMs) {
@@ -222,11 +261,11 @@ class OralTrainerController internal constructor(
     }
 
     fun previousItem() {
-        player.seekToPreviousMediaItem()
+        switchBaseItem(activeBaseItemIndex - 1)
     }
 
     fun nextItem() {
-        player.seekToNextMediaItem()
+        switchBaseItem(activeBaseItemIndex + 1)
     }
 
     fun setPlaybackSpeed(speed: Float) {
@@ -240,44 +279,50 @@ class OralTrainerController internal constructor(
 
     fun setLoopMode(loopMode: LoopMode) {
         this.loopMode = loopMode
-        player.repeatMode = when (loopMode) {
-            LoopMode.OFF -> Player.REPEAT_MODE_OFF
-            LoopMode.ONE -> Player.REPEAT_MODE_ONE
-            LoopMode.ALL -> Player.REPEAT_MODE_ALL
-        }
+        refreshSentenceRepeatMode()
     }
 
     fun setSentenceMode(enabled: Boolean) {
         config = config.copy(sentenceMode = enabled)
+        refreshPlaybackWindow()
     }
 
     fun setContinuousPlayback(enabled: Boolean) {
         config = config.copy(continuousPlayback = enabled)
-        refreshSentenceBoundaryGuard()
+        refreshPlaybackWindow()
     }
 
     fun playCurrentSentenceAndStop() {
-        val sentence = currentSentence() ?: return
+        val sentence = activeSentenceWindow?.let { window ->
+            currentTrainingItem()?.sentences?.getOrNull(window.sentenceIndex)
+        } ?: currentSentence() ?: return
         playSentenceAndStop(sentence.startMs, sentence.index)
     }
 
     /** Seek to a known sentence and stop automatically when that sentence ends. */
     fun playSentenceAndStop(startMs: Long, sentenceIndex: Int) {
+        val sentence = currentTrainingItem()?.sentences?.getOrNull(sentenceIndex)
         stopAtSentenceEnd = true
         stopAtSentenceEndIndex = sentenceIndex
+        if (sentence != null) {
+            if (!ensureSentenceWindow(sentence)) {
                 player.seekTo(max(0L, startMs))
+            }
+        }
         player.play()
-        refreshSentenceBoundaryGuard()
+        refreshPlaybackWindow()
     }
 
     fun currentTrainingItem(): TrainingMediaItem? {
-        val index = player.currentMediaItemIndex
-        return mediaItems.getOrNull(index)
+        return mediaItems.getOrNull(activeBaseItemIndex)
     }
 
     fun currentSentence(): SentenceBoundary? {
         val item = currentTrainingItem() ?: return null
-        val position = player.currentPosition
+        if (isSentencePlaylistActive()) {
+            return item.sentences.getOrNull(player.currentMediaItemIndex)
+        }
+        val position = playbackPositionMs()
         return item.sentences.lastOrNull { sentence ->
             position >= sentence.startMs && position < sentence.endMs
         }
@@ -288,7 +333,12 @@ class OralTrainerController internal constructor(
         if (item.sentences.isEmpty()) {
             return false
         }
-        val position = player.currentPosition
+        if (isSentencePlaylistActive()) {
+            val index = (player.currentMediaItemIndex - 1).coerceAtLeast(0)
+            seekTo(item.sentences[index].startMs)
+            return true
+        }
+        val position = playbackPositionMs()
         val target = item.sentences
             .asReversed()
             .firstOrNull { it.startMs < position - PREVIOUS_SENTENCE_TOLERANCE_MS }
@@ -302,7 +352,13 @@ class OralTrainerController internal constructor(
         if (item.sentences.isEmpty()) {
             return false
         }
-        val position = player.currentPosition
+        if (isSentencePlaylistActive()) {
+            val index = (player.currentMediaItemIndex + 1)
+                .takeIf { it < item.sentences.size } ?: return false
+            seekTo(item.sentences[index].startMs)
+            return true
+        }
+        val position = playbackPositionMs()
         val target = item.sentences.firstOrNull {
             it.startMs > position + NEXT_SENTENCE_TOLERANCE_MS
         } ?: return false
@@ -362,12 +418,17 @@ class OralTrainerController internal constructor(
 
     fun snapshot(): PlaybackSnapshot {
         val item = currentTrainingItem()
-        val duration = player.duration.takeUnless { it == C.TIME_UNSET } ?: -1L
+        val window = activeSentenceWindow
+        val duration = if (window != null) {
+            window.endMs
+        } else {
+            player.duration.takeUnless { it == C.TIME_UNSET } ?: -1L
+        }
         return PlaybackSnapshot(
             mediaId = item?.id,
-            positionMs = max(0L, player.currentPosition),
+            positionMs = playbackPositionMs(),
             durationMs = duration,
-            bufferedPositionMs = max(0L, player.bufferedPosition),
+            bufferedPositionMs = playbackBufferedPositionMs(),
             isPlaying = player.isPlaying,
             playbackState = player.playbackState.toOralTrainerState(),
             playbackSpeed = player.playbackParameters.speed,
@@ -380,7 +441,6 @@ class OralTrainerController internal constructor(
             return
         }
         released = true
-        cancelSentenceBoundaryGuard()
         mainHandler.removeCallbacks(ticker)
         listeners.clear()
         player.release()
@@ -404,101 +464,42 @@ class OralTrainerController internal constructor(
         lastSentenceIndex = index
         listeners.forEach { it.onSentenceChanged(sentence) }
         if (changed && !config.continuousPlayback && !stopAtSentenceEnd) {
-            refreshSentenceBoundaryGuard()
+            refreshPlaybackWindow()
         }
     }
 
-    private fun enforceSentenceBoundary() {
-        if (!config.sentenceMode || !player.isPlaying) {
-            return
-        }
-        val item = currentTrainingItem() ?: return
-        if (item.sentences.isEmpty()) {
-            return
-        }
-        val position = player.currentPosition
-        if (stopAtSentenceEnd) {
-            val index = stopAtSentenceEndIndex ?: return
-            val sentence = item.sentences.getOrNull(index) ?: return
-            if (position >= sentence.endMs) {
-                applySentenceEndAction(sentence)
-            }
-            return
-        }
-        if (config.continuousPlayback) {
-            return
-        }
-        val anchorIndex = lastSentenceIndex ?: currentSentence()?.index ?: return
-        val anchor = item.sentences.getOrNull(anchorIndex) ?: return
-        if (position >= anchor.endMs) {
-            applySentenceEndAction(anchor)
-        }
-    }
-
-    private fun refreshSentenceBoundaryGuard() {
-        cancelSentenceBoundaryGuard()
+    private fun refreshPlaybackWindow() {
+        val item = currentTrainingItem()
         if (released || !config.sentenceMode) {
-            return
-        }
-        if (!player.isPlaying && !player.playWhenReady) {
-            return
-        }
-        val item = currentTrainingItem() ?: return
-        if (item.sentences.isEmpty()) {
+            restoreFullWindow(playbackPositionMs())
             return
         }
 
-        val sentence = when {
-            stopAtSentenceEnd -> item.sentences.getOrNull(stopAtSentenceEndIndex ?: return)
-            config.continuousPlayback -> null
-            else -> {
-                val index = lastSentenceIndex ?: currentSentence()?.index
-                item.sentences.getOrNull(index ?: return)
-            }
-        } ?: return
-
-        val mediaItemIndex = player.currentMediaItemIndex
-        if (player.currentPosition >= sentence.endMs) {
-            applySentenceEndAction(sentence)
+        if (item == null || item.sentences.isEmpty()) {
+            restoreFullWindow(playbackPositionMs())
             return
         }
 
-        val token = ++boundaryGuardToken
-        boundaryMessage = player.createMessage { _, _ ->
-            if (token != boundaryGuardToken || released) {
-                return@createMessage
-            }
-            val active = currentTrainingItem() ?: return@createMessage
-            val latest = when {
-                stopAtSentenceEnd -> active.sentences.getOrNull(stopAtSentenceEndIndex ?: return@createMessage)
-                else -> active.sentences.getOrNull(sentence.index)
-            } ?: return@createMessage
-            if (player.currentMediaItemIndex != mediaItemIndex) {
-                return@createMessage
-            }
-            applySentenceEndAction(latest)
-        }
-            .setLooper(mainHandler.looper)
-            .setPosition(mediaItemIndex, sentence.endMs)
-            .setDeleteAfterDelivery(true)
-            .send()
+        if (config.continuousPlayback && !stopAtSentenceEnd) {
+            restoreFullWindow(playbackPositionMs())
+            return
         }
 
-    private fun cancelSentenceBoundaryGuard() {
-        boundaryGuardToken++
-        boundaryMessage?.let { message ->
-            runCatching { message.cancel() }
+        val sentence = sentenceForPlaybackWindow(item) ?: return
+        if (!player.isPlaying && !player.playWhenReady && activeSentenceWindow == null) {
+            return
         }
-        boundaryMessage = null
+
+        ensureSentenceWindow(sentence)
+        refreshSentenceRepeatMode()
     }
 
     private fun applySentenceEndAction(sentence: SentenceBoundary) {
         if (stopAtSentenceEnd && stopAtSentenceEndIndex == sentence.index) {
             stopAtSentenceEnd = false
             stopAtSentenceEndIndex = null
-            cancelSentenceBoundaryGuard()
             player.pause()
-            player.seekTo(sentence.startMs)
+            player.seekTo(0L)
             lastSentenceIndex = sentence.index
             notifySentenceIfChanged(force = true)
             notifySnapshot()
@@ -507,20 +508,232 @@ class OralTrainerController internal constructor(
         if (config.continuousPlayback) {
             return
         }
-        // Loop the current sentence without leaking into the next one.
-        player.seekTo(sentence.startMs)
+        player.pause()
+        // 当前 MediaItem 已 clipped 到 [startMs, endMs)，句首就是窗口内的 0ms。
+        player.seekTo(0L)
+        if (player.playbackState == Player.STATE_ENDED) {
+            player.prepare()
+        }
+        player.play()
         lastSentenceIndex = sentence.index
         notifySentenceIfChanged(force = true)
-        refreshSentenceBoundaryGuard()
+        refreshPlaybackWindow()
         notifySnapshot()
     }
 
-    private fun TrainingMediaItem.toMedia3Item(): MediaItem {
+    private fun applySentenceEndForAnchor() {
+        if (!config.sentenceMode || config.continuousPlayback) {
+            return
+        }
+        val item = currentTrainingItem()
+        val index = currentSentenceIndex() ?: lastSentenceIndex ?: return
+        val sentence = item?.sentences?.getOrNull(index) ?: return
+        applySentenceEndAction(sentence)
+    }
+
+    private fun sentenceAnchorAt(positionMs: Long): SentenceBoundary? {
+        val item = currentTrainingItem() ?: return null
+        return item.sentences.lastOrNull { positionMs >= it.startMs && positionMs < it.endMs }
+            ?: item.sentences.lastOrNull { it.startMs <= positionMs }
+    }
+
+    private fun sentenceForPlaybackWindow(item: TrainingMediaItem?): SentenceBoundary? {
+        if (item == null || item.sentences.isEmpty()) {
+            return null
+        }
+        return when {
+            !config.sentenceMode -> null
+            stopAtSentenceEnd -> item.sentences.getOrNull(stopAtSentenceEndIndex ?: return null)
+            config.continuousPlayback -> null
+            else -> {
+                val index = currentSentenceIndex()
+                    ?: lastSentenceIndex
+                    ?: currentSentence()?.index
+                    ?: sentenceAnchorAt(playbackPositionMs())?.index
+                item.sentences.getOrNull(index ?: return null)
+            }
+        }
+    }
+
+    private fun playbackPositionMs(): Long {
+        val window = activeSentenceWindow ?: return max(0L, player.currentPosition)
+        val lengthMs = (window.endMs - window.startMs).coerceAtLeast(0L)
+        val clippedPositionMs = player.currentPosition.coerceIn(0L, lengthMs)
+        return window.startMs + clippedPositionMs
+    }
+
+    private fun playbackBufferedPositionMs(): Long {
+        val window = activeSentenceWindow ?: return max(0L, player.bufferedPosition)
+        val lengthMs = (window.endMs - window.startMs).coerceAtLeast(0L)
+        val clippedBufferedMs = player.bufferedPosition.coerceIn(0L, lengthMs)
+        return window.startMs + clippedBufferedMs
+    }
+
+    private fun ensureSentenceWindow(target: SentenceBoundary): Boolean {
+        if (released) {
+            return false
+        }
+        if (sentencePlaylistBaseIndex != activeBaseItemIndex &&
+            player.playbackState in listOf(Player.STATE_READY, Player.STATE_ENDED)
+        ) {
+            installSentencePlaylist(activeBaseItemIndex, target.index)
+            return true
+        }
+        if (!isSentencePlaylistActive()) {
+            return false
+        }
+
+        val desired = sentenceWindows.getOrNull(target.index) ?: return false
+        if (player.currentMediaItemIndex != target.index) {
+            player.seekTo(target.index, 0L)
+        }
+        activeSentenceWindow = desired
+        return true
+    }
+
+    private fun installSentencePlaylist(baseIndex: Int, startIndex: Int) {
+        if (released) return
+        val baseItem = mediaItems.getOrNull(baseIndex) ?: return
+        val sentences = baseItem.sentences
+        if (sentences.isEmpty()) return
+
+        sentenceWindows = sentences.mapIndexed { index, sentence ->
+            SentenceWindow(
+                baseItemIndex = baseIndex,
+                mediaId = baseItem.id,
+                sentenceIndex = index,
+                startMs = sentence.startMs,
+                endMs = sentence.endMs,
+            )
+        }
+        sentencePlaylistBaseIndex = baseIndex
+        activeBaseItemIndex = baseIndex
+        val safeStartIndex = startIndex.coerceIn(0, sentenceWindows.lastIndex)
+        activeSentenceWindow = sentenceWindows[safeStartIndex]
+        lastSentenceIndex = safeStartIndex
+
+        val playWhenReady = player.playWhenReady
+        player.setMediaItems(
+            sentenceWindows.map { window ->
+                baseItem.toMedia3Item(window.startMs, window.endMs)
+            },
+            safeStartIndex,
+            0L,
+        )
+        player.prepare()
+        player.playWhenReady = playWhenReady
+        refreshSentenceRepeatMode()
+    }
+
+    private fun currentSentenceWindow(): SentenceWindow? {
+        if (!isSentencePlaylistActive()) return null
+        return sentenceWindows.getOrNull(player.currentMediaItemIndex)
+    }
+
+    private fun currentSentenceIndex(): Int? {
+        if (!isSentencePlaylistActive()) return null
+        return player.currentMediaItemIndex.takeIf { it in sentenceWindows.indices }
+    }
+
+    private fun isSentencePlaylistActive(): Boolean {
+        return sentencePlaylistBaseIndex == activeBaseItemIndex &&
+            sentenceWindows.isNotEmpty() &&
+            player.currentMediaItemIndex in sentenceWindows.indices
+    }
+
+    private fun refreshSentenceRepeatMode() {
+        if (!isSentencePlaylistActive()) {
+            player.repeatMode = when (loopMode) {
+                LoopMode.OFF -> Player.REPEAT_MODE_OFF
+                LoopMode.ONE -> Player.REPEAT_MODE_ONE
+                LoopMode.ALL -> Player.REPEAT_MODE_ALL
+            }
+            return
+        }
+        player.repeatMode = when {
+            stopAtSentenceEnd || !config.continuousPlayback -> Player.REPEAT_MODE_ONE
+            loopMode == LoopMode.OFF -> Player.REPEAT_MODE_OFF
+            else -> Player.REPEAT_MODE_ALL
+        }
+    }
+
+    private fun seekInSentenceWindow(target: SentenceBoundary, originalPositionMs: Long) {
+        val relativePositionMs = (originalPositionMs - target.startMs).coerceIn(
+            0L,
+            (target.endMs - target.startMs).coerceAtLeast(0L),
+        )
+        player.seekTo(relativePositionMs)
+    }
+
+    private fun restoreFullWindow(positionMs: Long) {
+        val baseIndex = activeBaseItemIndex
+        val baseItem = mediaItems.getOrNull(baseIndex) ?: return
+        if (sentencePlaylistBaseIndex == -1 &&
+            activeSentenceWindow == null &&
+            baseItem.id == player.currentMediaItem?.mediaId
+        ) {
+            return
+        }
+
+        val playWhenReady = player.playWhenReady
+        activeSentenceWindow = null
+        sentenceWindows = emptyList()
+        sentencePlaylistBaseIndex = -1
+        player.setMediaItem(baseItem.toMedia3Item(), max(0L, positionMs))
+        player.prepare()
+        player.playWhenReady = playWhenReady
+        player.repeatMode = when (loopMode) {
+            LoopMode.OFF -> Player.REPEAT_MODE_OFF
+            LoopMode.ONE -> Player.REPEAT_MODE_ONE
+            LoopMode.ALL -> Player.REPEAT_MODE_ALL
+        }
+    }
+
+    private fun switchBaseItem(targetIndex: Int) {
+        if (targetIndex !in mediaItems.indices) {
+            return
+        }
+        activeBaseItemIndex = targetIndex
+        activeSentenceWindow = null
+        sentenceWindows = emptyList()
+        sentencePlaylistBaseIndex = -1
+        lastSentenceIndex = null
+        stopAtSentenceEnd = false
+        stopAtSentenceEndIndex = null
+        if (config.sentenceMode && mediaItems[targetIndex].sentences.isNotEmpty()) {
+            installSentencePlaylist(targetIndex, 0)
+        } else {
+            player.setMediaItem(mediaItems[targetIndex].toMedia3Item(), 0L)
+            player.prepare()
+            player.repeatMode = when (loopMode) {
+                LoopMode.OFF -> Player.REPEAT_MODE_OFF
+                LoopMode.ONE -> Player.REPEAT_MODE_ONE
+                LoopMode.ALL -> Player.REPEAT_MODE_ALL
+            }
+        }
+        listeners.forEach { it.onMediaChanged(currentTrainingItem()) }
+        notifySentenceIfChanged(force = true)
+        notifySnapshot()
+    }
+
+    private fun TrainingMediaItem.toMedia3Item(
+        clippingStartPositionMs: Long = 0L,
+        clippingEndPositionMs: Long? = null,
+    ): MediaItem {
         val builder = MediaItem.Builder()
             .setMediaId(id)
             .setUri(uri)
         mimeType?.let(builder::setMimeType)
         customCacheKey?.let(builder::setCustomCacheKey)
+        if (clippingEndPositionMs != null) {
+            builder.setClippingConfiguration(
+                MediaItem.ClippingConfiguration.Builder()
+                    .setStartPositionMs(max(0L, clippingStartPositionMs))
+                    .setEndPositionMs(max(0L, clippingEndPositionMs))
+                    .setStartsAtKeyFrame(false)
+                    .build()
+            )
+        }
         subtitleUri?.let { uri ->
             builder.setSubtitleConfigurations(
                 listOf(

android/sample-app/src/main/java/cn/learningpad/oraltrainer/sample/MainActivity.kt

@@ -21,6 +21,7 @@ import cn.learningpad.oraltrainer.sdk.SentenceBoundaryApiCallback
 import cn.learningpad.oraltrainer.sdk.SentenceBoundaryApiResult
 import cn.learningpad.oraltrainer.sdk.TrainingMediaItem
 import cn.learningpad.oraltrainer.sdk.TrainingVideoSummary
+import kotlin.math.roundToInt
 
 class MainActivity : Activity() {
     private lateinit var sdk: OralTrainerSdk
@@ -362,12 +363,43 @@ class MainActivity : Activity() {
         if (!progressBarDragging) {
             bindings.timeText.text =
                 "${PlaybackFormat.time(snapshot.positionMs)} / ${PlaybackFormat.time(snapshot.durationMs)}"
-            bindings.progressBar.progress = PlaybackFormat.progress(snapshot)
+            bindings.progressBar.progress = if (controller.config.continuousPlayback) {
+                PlaybackFormat.progress(snapshot)
+            } else {
+                sentenceProgress(snapshot)
+            }
         }
         bindings.speedText.text = PlaybackFormat.speed(snapshot.playbackSpeed)
         bindings.statusText.text = PlaybackFormat.status(snapshot)
     }
 
+    private fun sentenceCount(): Int {
+        return controller.currentTrainingItem()?.sentences?.size?.takeIf { it > 0 }
+            ?: screenState.currentSentenceCount
+    }
+
+    private fun sentenceProgress(snapshot: PlaybackSnapshot): Int {
+        val count = sentenceCount()
+        val index = snapshot.sentenceIndex ?: return 0
+        if (count <= 1) return 0
+        return (index * PROGRESS_MAX.toFloat() / (count - 1)).roundToInt()
+            .coerceIn(0, PROGRESS_MAX)
+    }
+
+    private fun sentenceForProgress(progress: Int): Triple<Int, SentenceBoundary?, Int> {
+        val sentences = controller.currentTrainingItem()?.sentences.orEmpty()
+        if (sentences.isEmpty()) {
+            return Triple(0, null, 0)
+        }
+        val index = if (sentences.size == 1) {
+            0
+        } else {
+            (progress.toFloat() / PROGRESS_MAX * (sentences.size - 1)).roundToInt()
+                .coerceIn(0, sentences.lastIndex)
+        }
+        return Triple(index, sentences.getOrNull(index), sentences.size)
+    }
+
     private fun applySentence(sentence: SentenceBoundary?) {
         if (!bindings.hasPlayerChrome) return
         if (sentence == null) {
@@ -399,9 +431,21 @@ class MainActivity : Activity() {
 
         override fun onSeekProgressPreview(progress: Int) {
             if (!bindings.hasPlayerChrome) return
+            if (controller.config.continuousPlayback) {
                 val duration = controller.snapshot().durationMs
                 bindings.timeText.text =
-                "${PlaybackFormat.time(PlaybackFormat.seekPosition(progress, duration))} / ${PlaybackFormat.time(duration)}"
+                    "${PlaybackFormat.time(PlaybackFormat.seekPosition(progress, duration))} / " +
+                        PlaybackFormat.time(duration)
+                return
+            }
+            val (index, sentence, count) = sentenceForProgress(progress)
+            bindings.timeText.text =
+                if (sentence == null) {
+                    "暂无句子边界"
+                } else {
+                    "第 ${index + 1}/$count 句  " +
+                        "${PlaybackFormat.time(sentence.startMs)}-${PlaybackFormat.time(sentence.endMs)}"
+                }
         }
 
         override fun onSeekStart() {
@@ -410,9 +454,13 @@ class MainActivity : Activity() {
 
         override fun onSeekStop(progress: Int) {
             progressBarDragging = false
-            controller.seekTo(
-                PlaybackFormat.seekPosition(progress, controller.snapshot().durationMs),
-            )
+            if (controller.config.continuousPlayback) {
+                val duration = controller.snapshot().durationMs
+                controller.seekTo(PlaybackFormat.seekPosition(progress, duration))
+                return
+            }
+            val (_, sentence, _) = sentenceForProgress(progress)
+            sentence?.let { controller.seekTo(it.startMs) }
         }
 
         override fun onContinuousPlaybackChanged(enabled: Boolean) {

ios/OralTrainer/Views/PlayerSectionView.swift

@@ -233,6 +233,7 @@ struct SentenceExplanationSheet: View {
     let courseTitle: String
     let sentence: SentenceBoundary
 
+    @Environment(\.dismiss) private var dismiss
     @State private var explanation = ""
     @State private var errorMessage: String?
 
@@ -264,6 +265,18 @@ struct SentenceExplanationSheet: View {
             .background(Color.white.ignoresSafeArea())
             .navigationTitle("句子讲解")
             .navigationBarTitleDisplayMode(.inline)
+            .toolbar {
+                ToolbarItem(placement: .navigationBarTrailing) {
+                    Button {
+                        dismiss()
+                    } label: {
+                        Image(systemName: "xmark.circle.fill")
+                            .font(.system(size: 22, weight: .semibold))
+                            .foregroundColor(Theme.textMuted)
+                    }
+                    .accessibilityLabel("关闭句子讲解")
+                }
+            }
         }
         .task {
             guard explanation.isEmpty, errorMessage == nil else { return }