[db,ffmpeg] calculate and persist audio hash on local files

owntone.conf.in

@@ -238,6 +238,12 @@ library {
 	# the playlist like MPD does. Note that some dacp clients do not show
 	# the playqueue if playback is stopped.
 #	clear_queue_on_stop_disable = false
+
+	# Enable audio hash persistance for local files
+	# Audio streams of files can be hashed to give identifiers to audio files
+	# that are independant of any metadata, but generation is a more expensive
+	# operation
+#	audio_hash = false
 }
 
 # Local audio output

src/conffile.c

@@ -121,6 +121,7 @@ static cfg_opt_t sec_library[] =
     CFG_BOOL("only_first_genre", cfg_false, CFGF_NONE),
     CFG_STR_LIST("decode_audio_filters", NULL, CFGF_NONE),
     CFG_STR_LIST("decode_video_filters", NULL, CFGF_NONE),
+    CFG_BOOL("audio_hash", cfg_false, CFGF_NONE),
     CFG_END()
   };
 

src/db.c

@@ -230,6 +230,7 @@ static const struct col_type_map mfi_cols_map[] =
     { "channels",           mfi_offsetof(channels),           DB_TYPE_INT },
     { "usermark",           mfi_offsetof(usermark),           DB_TYPE_INT },
     { "scan_kind",          mfi_offsetof(scan_kind),          DB_TYPE_INT },
+    { "audio_hash",         mfi_offsetof(audio_hash),         DB_TYPE_STRING },
   };
 
 /* This list must be kept in sync with
@@ -371,6 +372,7 @@ static const ssize_t dbmfi_cols_map[] =
     dbmfi_offsetof(channels),
     dbmfi_offsetof(usermark),
     dbmfi_offsetof(scan_kind),
+    dbmfi_offsetof(audio_hash),
   };
 
 /* This list must be kept in sync with
@@ -777,6 +779,7 @@ free_mfi(struct media_file_info *mfi, int content_only)
   free(mfi->composer_sort);
   free(mfi->album_artist_sort);
   free(mfi->virtual_path);
+  free(mfi->audio_hash);
 
   if (!content_only)
     free(mfi);

src/db.h

@@ -248,6 +248,8 @@ struct media_file_info {
   char *composer_sort;
 
   uint32_t scan_kind; /* Identifies the library_source that created/updates this item */
+
+  char *audio_hash;  /* sha256 - ffmpeg -i foo.mp3 -c:a copy -bsf:a null -f hash - */
 };
 
 #define mfi_offsetof(field) offsetof(struct media_file_info, field)
@@ -422,6 +424,7 @@ struct db_media_file_info {
   char *channels;
   char *usermark;
   char *scan_kind;
+  char *audio_hash;
 };
 
 #define dbmfi_offsetof(field) offsetof(struct db_media_file_info, field)

src/db_init.c

@@ -98,7 +98,8 @@
   "   composer_sort      VARCHAR(1024) DEFAULT NULL COLLATE DAAP,"	\
   "   channels           INTEGER DEFAULT 0,"		\
   "   usermark           INTEGER DEFAULT 0,"		\
-  "   scan_kind          INTEGER DEFAULT 0"		\
+  "   scan_kind          INTEGER DEFAULT 0,"		\
+  "   audio_hash         VARCHAR(128) DEFAULT NULL"	\
   ");"
 
 #define T_PL					\

src/db_init.h

@@ -26,7 +26,7 @@
  * is a major upgrade. In other words minor version upgrades permit downgrading
  * the server after the database was upgraded. */
 #define SCHEMA_VERSION_MAJOR 22
-#define SCHEMA_VERSION_MINOR 0
+#define SCHEMA_VERSION_MINOR 1
 
 int
 db_init_indices(sqlite3 *hdl);

src/db_upgrade.c

@@ -1227,6 +1227,21 @@ static const struct db_upgrade_query db_upgrade_v2200_queries[] =
     { U_v2200_SCVER_MINOR,    "set schema_version_minor to 00" },
   };
 
+/* ---------------------------- 22.00 -> 22.01 ------------------------------ */
+#define U_v2201_ALTER_FILES_AUDIOHASH \
+  "ALTER TABLE files ADD COLUMN audio_hash VARCHAR(128) DEFAULT NULL;"
+#define U_v2201_SCVER_MINOR                    \
+  "UPDATE admin SET value = '01' WHERE key = 'schema_version_minor';"
+
+static const struct db_upgrade_query db_upgrade_v2201_queries[] =
+  {
+    { U_v2201_ALTER_FILES_AUDIOHASH, "update files adding audio_hash" },
+
+    { U_v2201_SCVER_MINOR,    "set schema_version_minor to 01" },
+  };
+
+
+
 /* -------------------------- Main upgrade handler -------------------------- */
 
 int
@@ -1437,6 +1452,12 @@ db_upgrade(sqlite3 *hdl, int db_ver)
       if (ret < 0)
 	return -1;
 
+      /* FALLTHROUGH */
+
+    case 2200:
+      ret = db_generic_upgrade(hdl, db_upgrade_v2201_queries, ARRAY_SIZE(db_upgrade_v2201_queries));
+      if (ret < 0)
+	return -1;
 
       /* Last case statement is the only one that ends with a break statement! */
       break;

src/httpd_jsonapi.c

@@ -312,6 +312,8 @@ track_to_json(struct db_media_file_info *dbmfi)
   safe_json_add_int_from_string(item, "channels", dbmfi->channels);
   safe_json_add_int_from_string(item, "usermark", dbmfi->usermark);
 
+  safe_json_add_string(item, "audio_hash", dbmfi->audio_hash);
+
   ret = safe_atoi32(dbmfi->media_kind, &intval);
   if (ret == 0)
     safe_json_add_string(item, "media_kind", db_media_kind_label(intval));

src/library/filescanner_ffmpeg.c

@@ -29,6 +29,7 @@
 
 #include <libavcodec/avcodec.h>
 #include <libavformat/avformat.h>
+#include <libavutil/hash.h>
 #include <libavutil/opt.h>
 
 #include "db.h"
@@ -373,6 +374,53 @@ extract_metadata(struct media_file_info *mfi, AVFormatContext *ctx, AVStream *au
   return mdcount;
 }
 
+static int
+calc_audio_hash(struct media_file_info *mfi, AVFormatContext *ctx, const int audio_stream_idx)
+{
+    struct AVPacket *pkt = NULL;
+    struct AVHashContext *hash = NULL;
+    char res[2 * AV_HASH_MAX_SIZE + 4] = { 0 };
+    int ret = -1;
+
+    if (!cfg_getbool(cfg_getsec(cfg, "library"), "audio_hash"))
+      return 0;
+
+
+    pkt = av_packet_alloc();
+
+    ret = av_hash_alloc(&hash, "sha256");
+    if (ret < 0)
+      {
+	DPRINTF(E_WARN, L_SCAN, "Failed to find sha256 hash - %s\n", ret == EINVAL ? "unknown hash" : strerror(ret));
+        goto cleanup;
+      }
+
+    av_hash_init(hash);
+
+    while ((ret = av_read_frame(ctx, pkt)) >= 0)
+      {
+	if (pkt->stream_index != audio_stream_idx) {
+	    av_packet_unref(pkt);
+	    continue;
+	}
+
+	av_hash_update(hash, pkt->data, pkt->size);
+	av_packet_unref(pkt);
+      }
+
+    av_hash_final_hex(hash, (uint8_t*)res, sizeof(res));
+
+    mfi->audio_hash = strdup(res);
+
+    ret = 0;
+
+cleanup:
+    if (pkt)   av_packet_free(&pkt);
+    if (hash)  av_hash_freep(&hash);
+
+    return ret;
+}
+
 /*
  * Fills metadata read with ffmpeg/libav from the given path into the given mfi
  *
@@ -513,6 +561,10 @@ scan_metadata_ffmpeg(struct media_file_info *mfi, const char *file)
 		audio_stream = ctx->streams[i];
 		audio_codec_id = codec_id;
 
+		/* only scan for files because we need to read the entire audio */
+	        if (mfi->data_kind == DATA_KIND_FILE)
+		  calc_audio_hash(mfi, ctx, i);
+
 		mfi->samplerate = sample_rate;
 		mfi->bits_per_sample = 8 * av_get_bytes_per_sample(sample_fmt);
 		if (mfi->bits_per_sample == 0)

src/parsers/smartpl_lexer.l

@@ -87,6 +87,7 @@ samplerate     { yylval->str = strdup(yytext); return SMARTPL_T_INTTAG; }
 song_length    { yylval->str = strdup(yytext); return SMARTPL_T_INTTAG; }
 usermark       { yylval->str = strdup(yytext); return SMARTPL_T_INTTAG; }
 file_size      { yylval->str = strdup(yytext); return SMARTPL_T_INTTAG; }
+audio_hash     { yylval->str = strdup(yytext); return SMARTPL_T_STRTAG; }
 
 time_added     { yylval->str = strdup(yytext); return SMARTPL_T_DATETAG; }
 time_modified  { yylval->str = strdup(yytext); return SMARTPL_T_DATETAG; }