pandora_cdosea/render.py

#!/usr/bin/python3
from argparse import ArgumentParser
from collections import defaultdict
from glob import glob
import json
import os
import string
import subprocess
import sys

from pi import random
from keywords import KEYWORDS
import ox
import ox.web.auth


base_url = 'http://127.0.0.1:2620'

FRAME_DURATION = 1/60
MAX_DURATION = 40

HIDDEN_TAGS = [
    "women with white males",
    "gene z hanrahan"
]

# items to not use at all
BLACKLIST = [
    'XN'
]

api = None

def get_api():
    global api
    if not api:
        api = ox.API(base_url + '/api/')
        api.signin(**ox.web.auth.get('cdosea'))

if os.path.exists('PATHS.json'):
    PATHS = json.load(open('PATHS.json'))
else:
    PATHS = {}

if os.path.exists('CLIPS.json'):
    CLIPS = json.load(open('CLIPS.json'))
else:
    CLIPS = {}

if not os.path.exists('MUSIC.json'):
    MUSIC = defaultdict(list)
    for letter in os.listdir('music'):
        for d in range(10):
            path = os.path.join('music', letter, '%d.mp3' % d)
            MUSIC[letter].append({
                'path': path,
                'duration': ox.avinfo(path)['duration']
            })
    with open('MUSIC.json', 'w') as fd:
        json.dump(MUSIC, fd, indent=2, sort_keys=True)
else:
    MUSIC = json.load(open('MUSIC.json'))

if not os.path.exists('VOCALS.json'):
    VOCALS = defaultdict(list)
    for letter in os.listdir('vocals'):
        for fn in sorted(os.listdir(os.path.join('vocals', letter))):
            path = os.path.join('vocals', letter, fn)
            VOCALS[letter].append({
                'path': path,
                'duration': ox.avinfo(path)['duration']
            })
        while len(VOCALS[letter]) < 10:
            VOCALS[letter] += VOCALS[letter]
        VOCALS[letter] = VOCALS[letter][:10]
    with open('VOCALS.json', 'w') as fd:
        json.dump(VOCALS, fd, indent=2, sort_keys=True)
else:
    VOCALS = json.load(open('VOCALS.json'))

if not os.path.exists('DRONES.json'):
    DRONES = defaultdict(list)
    prefix = 'drones_wav'
    for letter in os.listdir(prefix):
        for fn in sorted(os.listdir(os.path.join(prefix, letter))):
            path = os.path.join(prefix, letter, fn)
            DRONES[letter[0]].append({
                'path': path,
                'duration': ox.avinfo(path)['duration']
            })
    with open('DRONES.json', 'w') as fd:
        json.dump(DRONES, fd, indent=2, sort_keys=True)
else:
    DRONES = json.load(open('DRONES.json'))

def get_path(id):
    global PATHS
    if id not in PATHS:
        get_api()
        info = api.findMedia({
            'query': {
                'conditions': [
                    {'key': 'id', 'operator': '==', 'value': id}
                ]
            },
            'keys': ['id', 'extension'],
            'range': [0, 1]
        })['data']['items'][0]
        path = os.path.join('cache', '%s.%s' % (info['id'], info['extension']))
        h = info['id']
        source = '/srv/pandora/data/media/%s/%s/%s/%s/data.*' % (
            h[:2], h[2:4], h[4:6], h[6:]
        )
        source = glob(source)[0]
        if not os.path.exists(path):
            if not os.path.exists(source):
                print('WTF', source)
                sys.exit(1)
            os.symlink(source, path)
            '''
            url = '%s/%s/download/source/' % (base_url, id)
            print('get video', url)
            '''
        PATHS[id] = path
        with open('PATHS.json', 'w') as fd:
            json.dump(PATHS, fd, indent=4, sort_keys=True)
    return PATHS[id]

def get_clips(tag):
    global CLIPS
    if tag not in CLIPS:
        get_api()
        clips = api.findAnnotations({
            'query': {
                'conditions': [
                    {'key': 'layer', 'operator': '==', 'value': 'keywords'},
                    {'key': 'value', 'operator': '==', 'value': tag}
                ],
                'operator': '&'
            },
            'keys': ['id', 'in', 'out'],
            'range': [0, 90000]})['data']['items']
        clips = [clip for clip in clips if clip['id'].split('/')[0] not in BLACKLIST]

        for clip in clips:
            clip['path'] = get_path(clip['id'].split('/')[0])
            # or use round?
            clip['in'] = int(clip['in'] / FRAME_DURATION) * FRAME_DURATION
            clip['out'] = int(clip['out'] / FRAME_DURATION) * FRAME_DURATION
            clip['duration'] = clip['out'] - clip['in']
            clip['tag'] = tag
        clips = [clip for clip in clips if clip['duration'] and clip['duration'] <= MAX_DURATION]
        for clip in clips:
            fduration = ox.avinfo(clip['path'])['duration']
            if clip['out'] > fduration:
                if clip['in'] == 0:
                    clip['out'] = fduration
                    clip['duration'] = clip['out'] - clip['in']
                else:
                    print('FAIL', clip, fduration)
                    sys.exit(1)
        CLIPS[tag] = list(sorted(clips, key=lambda c: c['id']))
        with open('CLIPS.json', 'w') as fd:
            json.dump(CLIPS, fd, indent=4, sort_keys=True)
    return CLIPS[tag].copy()

def random_choice(seq, items):
    n = n_ = len(items) - 1
    #print('len', n)
    if n == 0:
        return items[0]
    r = seq()
    base = 10
    while n > 10:
        n /= 10
        #print(r)
        r += seq()
        base += 10
    r = int(n_ * r / base)
    #print('result', r, items)
    return items[r]

def splitint(number, by):
    div = int(number/by)
    mod = number % by
    return [div + 1 if i > (by - 1 - mod) else div for i in range(by)]

def filter_clips(clips, duration, max_duration=0):
    # 1 minute
    blur = 0.5
    low = 1
    high = 10

    # 2 minute
    blur = 1
    low = 2
    high = 20

    buckets = {}
    clips_ = []
    for tag in clips:
        for clip in clips[tag]:
            clip['tag'] = tag
            clips_.append(clip)
    clips_.sort(key=lambda c: c['duration'])
    #print(clips_)
    size = splitint(len(clips_), 10)
    p = 0
    for i in range(10):
        buckets[i+1] = clips_[p:+p+size[i]]
        p += size[i]
    clips_ = {}
    #print(buckets[duration])
    for clip in buckets[duration]:
        if clip['tag'] not in clips_:
            clips_[clip['tag']] = []
        clips_[clip['tag']].append(clip)
    return clips_

def add_blank(track, d):
    if track and track[-1].get('blank'):
        track[-1]['duration'] += d
    else:
        blank = {'blank': True, 'duration': d}
        track.append(blank)
    return d

def sequence(seq, letter):


    tags = KEYWORDS[letter]
    clips = {tag: get_clips(tag) for tag in tags}
    all_clips = clips.copy()
    result = {
        'clips': [],
        'text': [],
        'vocals': [],
        'music': [],
        'drones0': [],
        'drones1': [],
    }
    duration = 0
    MAX_DURATION = 60 * 2 + 5
    MIN_DURATION = 60 * 2 - 4

    # add 1 black frame for sync playback
    duration = 1 * FRAME_DURATION
    result['clips'].append({'black': True, 'duration': duration})

    while duration < MAX_DURATION and not duration >= MIN_DURATION:
        # clip duration: 1-10
        n = seq()
        if n == 0:
            n = 10

        max_duration = MAX_DURATION - duration
        clips_n = filter_clips(clips, n, max_duration)
        tags_n = [tag for tag in tags if tag in clips_n]
        if not tags_n:
            print('MISSING CLIPS, fall back to all clips!', letter, n)
            clips_n = filter_clips(all_clips, n, max_duration)
            tags_n = [tag for tag in tags if tag in clips_n]
        if not tags_n:
            print('NO tags for', letter, n)
            sys.exit(1)
        tag = random_choice(seq, tags_n)
        #if 'tiger' in tags_n:
        #    tag = 'tiger'
        clip = random_choice(seq, clips_n[tag])
        duration += clip['duration']
        result['clips'].append(clip.copy())

        # no reuse
        for t in clips:
            clips[t] = [c for c in clips[t] if not c['id'] == clip['id']]
    '''
    for clip in result['clips']:
        if seq() == 0:
            clip['black'] = True
    '''

    # text overlay
    position = last_text = 0
    tags_text = []

    # no overlay for the first 2 frames
    position = last_text = add_blank(result['text'], 2 * FRAME_DURATION)

    while position < duration:
        n = seq()
        if n == 0:
            blank = {'blank': True, 'duration': position - last_text}
            if blank['duration']:
                result['text'].append(blank)
            n = seq()
            if n == 0:
                n = 10
            n = min(n, duration-position)
            if not tags_text:
                tags_text = list(sorted(set(tags)))
                tags_text = [t for t in tags_text if t not in HIDDEN_TAGS]

            ttag = random_choice(seq, tags_text)
            tags_text.remove(ttag)
            text = {
                'text': ttag,
                'duration': n
            }
            result['text'].append(text)
            position += n
            last_text = position
        else:
            position += n
    if last_text < duration:
        blank = {'blank': True, 'duration': duration - last_text}
        result['text'].append(blank)


    # music
    track = 'music'
    if letter in MUSIC:
        position = 0
        while position < duration:
            n = seq()
            if n < 5:
                n = seq()
                position += add_blank(result[track], min(n, duration-position))
            else:
                n = seq()
                clip = MUSIC[letter][n]
                position += clip['duration']
                if result[track] and position > duration \
                        and result[track][-1].get('blank') \
                        and result[track][-1]['duration'] > clip['duration']:
                    result[track][-1]['duration'] -= (position-duration)
                    position = duration
                    if not result[track][-1]['duration']:
                        result[track].pop(-1)
                if position <= duration:
                    result[track].append(clip)
                else:
                    position -= clip['duration']
                    break
        if position < duration:
            position += add_blank(result[track], duration - position)
    # vocals
    track = 'vocals'
    if letter in VOCALS:
        position = 0
        loop = 0
        while position < duration:
            n = seq()
            # vocals should start after one blank
            if len(result[track])  and result[track][-1].get('blank'):
                n = 10
            # 50 % chance of a silence of up to 5 seconds
            if n < 5:
                n = seq() / 2  # (0 - 5 seconds)
                position += add_blank(result[track], min(n, duration-position))
            else:
                n = seq()
                clip = VOCALS[letter][n]
                position += clip['duration']
                if result[track] and position > duration \
                        and result[track][-1].get('blank') \
                        and result[track][-1]['duration'] > clip['duration']:
                    result[track][-1]['duration'] -= (position-duration)
                    position = duration
                    if not result[track][-1]['duration']:
                        result[track].pop(-1)
                if position <= duration:
                    result[track].append(clip)
                else:
                    position -= clip['duration']
                    if duration - position < 10 or loop > 10:
                        break
                    loop += 1
        if position < duration:
            position += add_blank(result[track], duration - position)
    '''
    if letter in VOCALS:
        n = seq()
        clip = VOCALS[letter][n]

        n = 1.0 / (seq() + 1)  # 0.1 - 1

        silence = duration - clip['duration']
        silence_start = n * silence
        blank = {'blank': True, 'duration': silence_start}
        if n != 0:
            result['vocals'].append(blank)
        result['vocals'].append(clip)
        if n != 1:
            blank = {'blank': True, 'duration': silence - silence_start}
            result['vocals'].append(blank)
    '''
    # drones
    if letter in DRONES:
        for track in ('drones0', 'drones1'):
            position = 0
            while position < duration:
                n = seq()
                if n == 9:
                    position += add_blank(result[track], min(3.141, duration - position))
                else:
                    clip = DRONES[letter][n]
                    if position + clip['duration'] < duration:
                        position += clip['duration']
                        result[track].append(clip)
                    else:
                        c = clip.copy()
                        c['duration'] = duration - position
                        result[track].append(c)
                        position += c['duration']
            if position < duration:
                position += add_blank(result[track], duration - position)

    for track in result:
        if result[track]:
            tduration = sum([c['duration'] for c in result[track]])
            if not abs(tduration - duration) < 0.000001:
                raise Exception('invalid duration on track: %s %s vs %s %s' % (track, tduration, duration, result[track]))
    result['gongs'] = {
        'tracks': 100 + seq() * 23,
        'offset': seq.position,
    }
    return result


if __name__ == '__main__':
    usage = "usage: %(prog)s [options] xml"
    parser = ArgumentParser(usage=usage)
    parser.add_argument('-p', '--prefix', dest='prefix',
                    help='version prefix', default='.')
    parser.add_argument('files', metavar='path', type=str, nargs='*', help='json files')
    opts = parser.parse_args()

    render_xml = opts.files != ['json']
    for n in range(10):
        seq = random(n * 1000)
        #for letter in ('T', 'W'):
        for letter in string.ascii_uppercase:
            r = sequence(seq, letter)
            tjson = os.path.join(opts.prefix, 'output/%02d/%s.json' % (n, letter))
            folder = os.path.dirname(tjson)
            if not os.path.exists(folder):
                ox.makedirs(folder)
            if os.path.exists(tjson):
                with open(tjson, 'r') as fd:
                    old = fd.read()
            else:
                old = None
            current = json.dumps(r, indent=4, sort_keys=True)
            #print(current)
            #print(sum([c['duration'] for c in r['clips']]))
            if current != old:
                with open(tjson, 'w') as fd:
                    fd.write(current)
            if render_xml:
                if current != old or os.path.getmtime(tjson) < os.path.getmtime('render_mlt.py'):
                    subprocess.call(['./render_mlt.py', '--prefix', prefix, tjson])
                    #subprocess.call(['./render_audio.py', tjson])