#Reading and writing files

def track as mp3.Track init mp3.open("song.mp3");

$track.path;      # where it was read from
$track.tag;       # the unified view, ID3v2 merged over ID3v1
$track.audio;     # AudioInfo for the stream
$track.id3v2;     # the raw ID3v2 tag
$track.id3v1;     # the raw ID3v1 tag

parse($data) does the same on a buffer and leaves path empty. A Track does not hold the audio bytes, so copying one is cheap; save reads the file again.

#Where the audio is

def r as mp3.AudioRange init mp3.audioRange($data);
$r.start;   # first byte after the ID3v2 tag
$r.end;     # one past the last byte before the ID3v1 tag

Only the ID3v2 and ID3v1 tags are excluded. An APE tag, a Lyrics3 block or a TAG+ tag counts as audio. strip($data) returns the bytes in that range.

fs.writeBytes("bare.mp3", mp3.strip(fs.readBytes("tagged.mp3")));

#Writing a buffer

def out as bytes init mp3.write($data, $tag, mp3.defaultOptions());

write copies the audio byte for byte and writes new tags from tag as options says. The old tags are discarded, including frames Tag does not name; ID3v2 shows how to keep them.

optionsFor($track) keeps what the file had: its ID3v2 version (or DEFAULT_MAJOR if it had none) and an ID3v1 tag only if it carried one.

def out as bytes init mp3.write($data, $tag, mp3.optionsFor($track));

#Writing to disk

mp3.save(mp3.withTag($track, $edited), $track.path);

save(track, target) reads the audio from track.path, writes with optionsFor($track), and may overwrite its own source. It throws when track.path is empty. writeFile(source, target, tag, options) does the same with explicit options.

The write is atomic. The result is built in a temporary file beside the target, named with a random v4 UUID so nothing can create it in advance, and renamed over the target; on failure the temporary file is removed. Because it is a rename, the target is replaced, not written through: a symlink becomes a regular file and a hard link stops sharing. The audio is streamed from the source buffer to the file, so a write holds one copy of the audio in memory, not two.

#A directory

for (def name in fs.list($dir)) {
    if (not strings.endsWith($name, ".mp3")) {
        continue;
    }
    def track as mp3.Track init mp3.open(path.join($dir, $name));
    io.printf("%s\n", mp3.describe($track));
}

describe returns one line, fields separated by two spaces: artist and title, duration, bitrate (with VBR when variable), sampling rate, channel mode. Without audio it ends in no audio.

mp3.describe($track);
# Maike Vandervaart - Ich war müde leer nicht mehr da  4:18  160 kbit/s VBR  44100 Hz  stereo

On a large library, parseWith($data, false) skips the frame walk; see The audio stream.