DidlLiteCodec.kt
244 lines · 12.1 KB
Changed in: 0.5.0
| 1 | package dev.stave.upnp.renderer |
| 2 | |
| 3 | import dev.stave.domain.model.DidlItem |
| 4 | import dev.stave.net.UrlPolicy |
| 5 | import org.jupnp.support.contentdirectory.DIDLParser |
| 6 | import org.jupnp.support.model.DIDLObject |
| 7 | import org.jupnp.support.model.item.Item |
| 8 | import org.jupnp.support.model.item.MusicTrack |
| 9 | import kotlin.time.Duration |
| 10 | import kotlin.time.Duration.Companion.milliseconds |
| 11 | |
| 12 | /** |
| 13 | * DIDL-Lite codec between [DidlItem.Track] and the XML payload UPnP control points |
| 14 | * exchange with us via AVTransport. |
| 15 | * |
| 16 | * Two halves: |
| 17 | * |
| 18 | * - [encodeTrack] generates DIDL-Lite XML by hand-assembled string-builder. We don't |
| 19 | * use jUPnP's [MusicTrack] constructor here because its overload signatures vary |
| 20 | * between jUPnP versions, and the Java-style builder-returning setters trip up |
| 21 | * Kotlin's property-style assignment. Hand-rolling the XML keeps the output stable |
| 22 | * against jUPnP API changes and produces a known-good spec-compliant document. |
| 23 | * |
| 24 | * - [decodeFirstTrack] uses [DIDLParser] (which is stable across jUPnP versions and |
| 25 | * the same parser the existing [dev.stave.upnp.DidlParser] is built on top of) |
| 26 | * to parse incoming setAVTransportURI metadata. Property navigation mirrors the |
| 27 | * existing parser — convenience methods like `firstArtist` don't exist on jUPnP's |
| 28 | * [Item], so we walk `properties` by descriptor name for fields beyond what |
| 29 | * [MusicTrack] exposes directly. |
| 30 | * |
| 31 | * Both directions tolerate sparse input. Many control points send minimal metadata |
| 32 | * (`<item><res>$url</res></item>` and nothing else) and rely on the renderer to fill |
| 33 | * blanks at play time from id3/Vorbis tags. We do the same — empty / null fields are |
| 34 | * normal and don't break the flow. |
| 35 | */ |
| 36 | internal object DidlLiteCodec { |
| 37 | |
| 38 | private val parser = DIDLParser() |
| 39 | |
| 40 | // ---- Encode ------------------------------------------------------------- |
| 41 | |
| 42 | /** |
| 43 | * Generate a DIDL-Lite XML document for [track]. Returns the empty string when |
| 44 | * [track] is null — a valid AVTransport metadata value meaning "no media". |
| 45 | * |
| 46 | * Hand-rolled XML rather than going through jUPnP's [MusicTrack] builder for the |
| 47 | * reasons in the class kdoc. Produces a single-item document with the standard |
| 48 | * upnp:musicTrack class and one `<res>` element pointing at [DidlItem.Track.resourceUrl]. |
| 49 | */ |
| 50 | fun encodeTrack(track: DidlItem.Track?): String { |
| 51 | if (track == null) return "" |
| 52 | return buildString { |
| 53 | append("""<?xml version="1.0" encoding="UTF-8"?>""") |
| 54 | append("""<DIDL-Lite xmlns="urn:schemas-upnp-org:metadata-1-0/DIDL-Lite/" """) |
| 55 | append("""xmlns:upnp="urn:schemas-upnp-org:metadata-1-0/upnp/" """) |
| 56 | append("""xmlns:dc="http://purl.org/dc/elements/1.1/">""") |
| 57 | |
| 58 | val itemId = track.id.ifBlank { "earhorn-current" } |
| 59 | val parentId = track.parentId.ifBlank { "0" } |
| 60 | append("""<item id="${esc(itemId)}" parentID="${esc(parentId)}" restricted="1">""") |
| 61 | |
| 62 | append("""<dc:title>${esc(track.title)}</dc:title>""") |
| 63 | (track.albumArtist ?: track.artist)?.takeIf { it.isNotBlank() }?.let { |
| 64 | append("""<dc:creator>${esc(it)}</dc:creator>""") |
| 65 | } |
| 66 | track.artist?.takeIf { it.isNotBlank() }?.let { |
| 67 | append("""<upnp:artist>${esc(it)}</upnp:artist>""") |
| 68 | } |
| 69 | track.albumArtist?.takeIf { it.isNotBlank() }?.let { |
| 70 | append("""<upnp:artist role="AlbumArtist">${esc(it)}</upnp:artist>""") |
| 71 | } |
| 72 | track.album?.takeIf { it.isNotBlank() }?.let { |
| 73 | append("""<upnp:album>${esc(it)}</upnp:album>""") |
| 74 | } |
| 75 | track.albumArtUri?.takeIf { it.isNotBlank() }?.let { |
| 76 | append("""<upnp:albumArtURI>${esc(it)}</upnp:albumArtURI>""") |
| 77 | } |
| 78 | track.genre?.takeIf { it.isNotBlank() }?.let { |
| 79 | append("""<upnp:genre>${esc(it)}</upnp:genre>""") |
| 80 | } |
| 81 | track.trackNumber?.let { |
| 82 | append("""<upnp:originalTrackNumber>$it</upnp:originalTrackNumber>""") |
| 83 | } |
| 84 | |
| 85 | append("""<upnp:class>object.item.audioItem.musicTrack</upnp:class>""") |
| 86 | |
| 87 | // Resource element. Required `protocolInfo` attribute carries the MIME |
| 88 | // for sink-matching by control points. Optional attributes are emitted |
| 89 | // only when we have a non-null value, since jUPnP rejects empty quoted |
| 90 | // values for some of them. |
| 91 | val mime = track.mimeType?.takeIf { it.isNotBlank() } ?: "audio/mpeg" |
| 92 | append("""<res protocolInfo="http-get:*:${esc(mime)}:*"""") |
| 93 | track.duration?.let { append(""" duration="${formatDurationAsUpnpClock(it)}"""") } |
| 94 | track.size?.let { append(""" size="$it"""") } |
| 95 | track.bitrate?.let { append(""" bitrate="$it"""") } |
| 96 | track.sampleFrequency?.let { append(""" sampleFrequency="$it"""") } |
| 97 | track.bitsPerSample?.let { append(""" bitsPerSample="$it"""") } |
| 98 | track.channels?.let { append(""" nrAudioChannels="$it"""") } |
| 99 | append(">") |
| 100 | append(esc(track.resourceUrl)) |
| 101 | append("</res>") |
| 102 | |
| 103 | append("</item>") |
| 104 | append("</DIDL-Lite>") |
| 105 | } |
| 106 | } |
| 107 | |
| 108 | // ---- Decode ------------------------------------------------------------- |
| 109 | |
| 110 | /** |
| 111 | * Parse a single DIDL-Lite item out of [didl] (the `CurrentURIMetaData` argument |
| 112 | * to setAVTransportURI) and return a [DidlItem.Track] populated from it. |
| 113 | * |
| 114 | * [fallbackUri] is the `CurrentURI` argument the control point also sent — used |
| 115 | * if the metadata lacks a `<res>` element or to anchor a Track when metadata is |
| 116 | * completely empty. We favor metadata-provided URLs when present (they may |
| 117 | * differ from the play-target URI for transcoding-proxy setups). |
| 118 | * |
| 119 | * Returns null when no parseable item is found AND no fallback URI is provided. |
| 120 | * |
| 121 | * Safety (the payload comes from any device on the LAN, unauthenticated): |
| 122 | * - A document carrying a `<!DOCTYPE` or `<!ENTITY` is not parsed at all. |
| 123 | * DIDL-Lite never needs a DTD, and refusing it here closes XXE / entity- |
| 124 | * expansion regardless of how the underlying SAX factory is configured. The |
| 125 | * bare [fallbackUri] is still considered. |
| 126 | * - With [requirePlayableUrl] (the default), only http(s) URLs to a |
| 127 | * non-loopback host are accepted ([UrlPolicy.isPlayableRemoteUrl]). Otherwise |
| 128 | * a peer could make libvlc open `file://`, `smb://`, `alsa://`, `v4l2://` or a |
| 129 | * service on this machine. When the metadata's `<res>` is refused but the |
| 130 | * `CurrentURI` is acceptable, the latter is used; if neither is, null. |
| 131 | * - `albumArtURI` passes the same check, or is dropped. |
| 132 | */ |
| 133 | fun decodeFirstTrack( |
| 134 | didl: String?, |
| 135 | fallbackUri: String, |
| 136 | requirePlayableUrl: Boolean = true, |
| 137 | ): DidlItem.Track? { |
| 138 | val safeDidl = didl?.takeUnless { |
| 139 | it.contains("<!DOCTYPE", ignoreCase = true) || it.contains("<!ENTITY", ignoreCase = true) |
| 140 | } |
| 141 | if (safeDidl == null && !didl.isNullOrBlank()) { |
| 142 | System.err.println("[EarhornRenderer/DIDL] ignored metadata carrying a DOCTYPE/ENTITY declaration") |
| 143 | } |
| 144 | val item: Item? = if (!safeDidl.isNullOrBlank()) { |
| 145 | try { |
| 146 | parser.parse(safeDidl).items?.firstOrNull() |
| 147 | } catch (_: Exception) { |
| 148 | null |
| 149 | } |
| 150 | } else null |
| 151 | |
| 152 | if (item == null && fallbackUri.isBlank()) return null |
| 153 | |
| 154 | val res = item?.resources?.firstOrNull() |
| 155 | val candidates = listOfNotNull(res?.value?.takeIf { it.isNotBlank() }, fallbackUri.takeIf { it.isNotBlank() }) |
| 156 | if (candidates.isEmpty()) return null |
| 157 | val resolvedUri = if (requirePlayableUrl) { |
| 158 | candidates.firstOrNull { UrlPolicy.isPlayableRemoteUrl(it) } ?: run { |
| 159 | System.err.println("[EarhornRenderer/DIDL] rejected URI: only http(s) to a non-loopback host is accepted") |
| 160 | return null |
| 161 | } |
| 162 | } else candidates.first() |
| 163 | |
| 164 | // Cast to MusicTrack for the music-specific accessors. Non-music items |
| 165 | // (generic audioItem with no album / artist / trackNumber) fall through |
| 166 | // to the property-based extraction below. |
| 167 | val musicTrack = item as? MusicTrack |
| 168 | val mime = res?.protocolInfo?.toString()?.split(":")?.getOrNull(2) |
| 169 | ?.takeIf { it.isNotBlank() } |
| 170 | |
| 171 | return DidlItem.Track( |
| 172 | id = item?.id?.takeIf { it.isNotBlank() } ?: "remote:${resolvedUri.hashCode()}", |
| 173 | parentId = item?.parentID?.takeIf { it.isNotBlank() } ?: "0", |
| 174 | title = item?.title?.takeIf { it.isNotBlank() } |
| 175 | ?: resolvedUri.substringAfterLast('/').substringBeforeLast('.'), |
| 176 | artist = musicTrack?.artists?.firstOrNull()?.name?.takeIf { it.isNotBlank() } |
| 177 | ?: item?.creator?.takeIf { it.isNotBlank() }, |
| 178 | albumArtist = musicTrack?.artists |
| 179 | ?.firstOrNull { it.role?.equals("AlbumArtist", ignoreCase = true) == true } |
| 180 | ?.name, |
| 181 | album = musicTrack?.album?.takeIf { it.isNotBlank() }, |
| 182 | composer = item?.firstPropertyByName("composer"), |
| 183 | genre = item?.firstPropertyByName("genre"), |
| 184 | date = item?.firstPropertyByName("date"), |
| 185 | trackNumber = musicTrack?.originalTrackNumber, |
| 186 | discNumber = item?.firstPropertyByName("discNumber")?.toIntOrNull(), |
| 187 | duration = parseDurationFromUpnpClock(res?.duration), |
| 188 | albumArtUri = item?.firstPropertyByName("albumArtURI") |
| 189 | ?.takeIf { !requirePlayableUrl || UrlPolicy.isPlayableRemoteUrl(it) }, |
| 190 | resourceUrl = resolvedUri, |
| 191 | mimeType = mime, |
| 192 | sampleFrequency = res?.sampleFrequency?.toInt(), |
| 193 | bitsPerSample = res?.bitsPerSample?.toInt(), |
| 194 | channels = res?.nrAudioChannels?.toInt(), |
| 195 | bitrate = res?.bitrate?.toInt(), |
| 196 | size = res?.size, |
| 197 | extras = emptyMap(), |
| 198 | ) |
| 199 | } |
| 200 | |
| 201 | // ---- Helpers ------------------------------------------------------------ |
| 202 | |
| 203 | /** Generic property lookup by descriptor name. Same pattern the existing |
| 204 | * [dev.stave.upnp.DidlParser] uses — convenience methods like `firstArtist` |
| 205 | * don't exist on jUPnP 3.0.2 [Item]. */ |
| 206 | private fun DIDLObject.firstPropertyByName(name: String): String? = |
| 207 | properties?.firstOrNull { it.descriptorName?.equals(name, ignoreCase = true) == true } |
| 208 | ?.value?.toString() |
| 209 | ?.takeIf { it.isNotBlank() } |
| 210 | |
| 211 | /** XML attribute / text escaping. Covers the five mandatory characters; we |
| 212 | * don't worry about other Unicode since the XML doc is UTF-8 and jUPnP's |
| 213 | * parser handles entities natively. */ |
| 214 | private fun esc(s: String): String = s |
| 215 | .replace("&", "&") |
| 216 | .replace("<", "<") |
| 217 | .replace(">", ">") |
| 218 | .replace("\"", """) |
| 219 | .replace("'", "'") |
| 220 | |
| 221 | /** "H:MM:SS.mmm" — the UPnP clock-value format AVTransport uses for durations. */ |
| 222 | private fun formatDurationAsUpnpClock(duration: Duration): String { |
| 223 | val totalMs = duration.inWholeMilliseconds.coerceAtLeast(0) |
| 224 | val ms = totalMs % 1000 |
| 225 | val total = totalMs / 1000 |
| 226 | val h = total / 3600 |
| 227 | val m = (total % 3600) / 60 |
| 228 | val s = total % 60 |
| 229 | return "%d:%02d:%02d.%03d".format(h, m, s, ms) |
| 230 | } |
| 231 | |
| 232 | /** Inverse of [formatDurationAsUpnpClock]. Returns null on parse failure rather |
| 233 | * than throwing — sparse / malformed metadata is normal. */ |
| 234 | private fun parseDurationFromUpnpClock(clock: String?): Duration? { |
| 235 | if (clock.isNullOrBlank() || clock == "NOT_IMPLEMENTED") return null |
| 236 | val parts = clock.split(":") |
| 237 | if (parts.size != 3) return null |
| 238 | val h = parts[0].toLongOrNull() ?: return null |
| 239 | val m = parts[1].toLongOrNull() ?: return null |
| 240 | val s = parts[2].toDoubleOrNull() ?: return null |
| 241 | val ms = ((h * 3600 + m * 60) * 1000 + (s * 1000).toLong()) |
| 242 | return ms.milliseconds |
| 243 | } |
| 244 | } |