musicmuster/app/songdb.py

#!/usr/bin/env python

import argparse
import os
import shutil
import tempfile

from config import Config
from log import DEBUG, INFO
from model import Tracks, Session
from mutagen.flac import FLAC
from pydub import AudioSegment, effects
from tinytag import TinyTag


def main():
    "Main loop"

    INFO("Starting")

    # Parse command line
    p = argparse.ArgumentParser()
    p.add_argument('-u', '--update',
                   action="store_true", dest="update",
                   default=True, help="Update database")
    args = p.parse_args()

    # Run as required
    if args.update:
        INFO("Updating database")
        with Session() as session:
            update_db(session)

    INFO("Finished")


def add_path_to_db(session, path):
    "Add passed path to database along with metadata"

    track = Tracks.get_or_create(session, path)
    tag = TinyTag.get(path)
    audio = get_audio_segment(path)

    track.title = tag.title
    track.artist = tag.artist
    track.duration = int(round(tag.duration,
                               Config.MILLISECOND_SIGFIGS) * 1000)
    track.start_gap = leading_silence(audio)
    track.fade_at = round(fade_point(audio) / 1000,
                          Config.MILLISECOND_SIGFIGS) * 1000
    track.silence_at = round(trailing_silence(audio) / 1000,
                             Config.MILLISECOND_SIGFIGS) * 1000
    track.mtime = os.path.getmtime(path)
    session.commit()

    if Config.NORMALISE_ON_IMPORT:
        # Get current file gid, uid and permissions
        stats = os.stat(path)
        try:
            # Copy original file
            fd, temp_path = tempfile.mkstemp()
            shutil.copyfile(path, temp_path)
        except Exception as err:
            DEBUG(f"songdb.add_path_to_db({path}): err1: {str(err)}")
            return

    # Overwrite original file with normalised output
    normalised = effects.normalize(audio)
    try:
        normalised.export(path, format=os.path.splitext(path)[1][1:])
        # Fix up permssions and ownership
        os.chown(path, stats.st_uid, stats.st_gid)
        os.chmod(path, stats.st_mode)
        # Copy tags
        src = FLAC(temp_path)
        dst = FLAC(path)
        for tag in src:
            dst[tag] = src[tag]
        dst.save()
    except Exception as err:
        DEBUG(f"songdb.add_path_to_db({path}): err2: {str(err)}")
        # Restore original file
        shutil.copyfile(path, temp_path)
    finally:
        if os.path.exists(temp_path):
            os.remove(temp_path)

    return track


def get_audio_segment(path):
    try:
        if path.endswith('.mp3'):
            return AudioSegment.from_mp3(path)
        elif path.endswith('.flac'):
            return AudioSegment.from_file(path, "flac")
    except AttributeError:
        return None


def leading_silence(audio_segment, silence_threshold=Config.DBFS_SILENCE,
                    chunk_size=Config.AUDIO_SEGMENT_CHUNK_SIZE):
    """
    Returns the millisecond/index that the leading silence ends.
    audio_segment - the segment to find silence in
    silence_threshold - the upper bound for how quiet is silent in dFBS
    chunk_size - chunk size for interating over the segment in ms

    https://github.com/jiaaro/pydub/blob/master/pydub/silence.py
    """

    trim_ms = 0  # ms
    assert chunk_size > 0  # to avoid infinite loop
    while (
        audio_segment[trim_ms:trim_ms + chunk_size].dBFS <      # noqa W504
            silence_threshold and trim_ms < len(audio_segment)):
        trim_ms += chunk_size

    # if there is no end it should return the length of the segment
    return min(trim_ms, len(audio_segment))


def fade_point(audio_segment, fade_threshold=Config.DBFS_FADE,
               chunk_size=Config.AUDIO_SEGMENT_CHUNK_SIZE):
    """
    Returns the millisecond/index of the point where the fade is down to
    fade_threshold and doesn't get louder again.
    audio_segment - the sdlg_search_database_uiegment to find silence in
    fade_threshold - the upper bound for how quiet is silent in dFBS
    chunk_size - chunk size for interating over the segment in ms
    """

    assert chunk_size > 0  # to avoid infinite loop

    segment_length = audio_segment.duration_seconds * 1000  # ms
    trim_ms = segment_length - chunk_size
    while (
        audio_segment[trim_ms:trim_ms + chunk_size].dBFS < fade_threshold
            and trim_ms > 0):       # noqa W503
        trim_ms -= chunk_size

    # if there is no trailing silence, return lenght of track (it's less
    # the chunk_size, but for chunk_size = 10ms, this may be ignored)
    return int(trim_ms)


# Current unused (1 June 2021)
# def rescan_database(session):
#
#     tracks = Tracks.get_all_tracks(session)
#     total_tracks = len(tracks)
#     track_count = 0
#     for track in tracks:
#         track_count += 1
#         print(f"Track {track_count} of {total_tracks}")
#         audio = get_audio_segment(track.path)
#         track.start_gap = leading_silence(audio)
#         track.fade_at = fade_point(audio)
#         track.silence_at = trailing_silence(audio)
#         session.commit()


def trailing_silence(audio_segment, silence_threshold=-50.0,
                     chunk_size=Config.AUDIO_SEGMENT_CHUNK_SIZE):
    return fade_point(audio_segment, silence_threshold, chunk_size)


def update_db(session):
    """
    Repopulate database
    """

    # Search for tracks in only one of directory and database

    db_paths = set(Tracks.get_all_paths(session))

    os_paths_list = []
    for root, dirs, files in os.walk(Config.ROOT):
        for f in files:
            path = os.path.join(root, f)
            ext = os.path.splitext(f)[1]
            if ext in [".flac", ".mp3"]:
                os_paths_list.append(path)
    os_paths = set(os_paths_list)

    for path in list(db_paths - os_paths):
        # TODO
        INFO(f"To remove from database: {path}")

    for path in list(os_paths - db_paths):
        # TODO
        INFO(f"Adding to dataabase: {path}")
        add_path_to_db(session, path)


if __name__ == '__main__' and '__file__' in globals():
    main()