def scrap_naver_data(self): logging.info("########## Scrapper[START] - scrap_naver_data ##########") assert self.src_path, "src_path 가 없음" assert self.tgt_path, "tgt_path 가 없음" movie_df = load_csv(self.src_path) movie_df['score'], movie_df['participa'], movie_df['nation'], movie_df['genre'], movie_df['showtime'], movie_df['exist'] = zip( *movie_df['TITLE'].apply(get_more_info)) movie_df.to_csv(self.tgt_path, encoding=ENCODING) logging.info("########## Scrapper[END] - scrap_naver_data ##########")
import vlc import pafy from flask import Flask, render_template from flask_socketio import SocketIO from tools import load_csv app = Flask(__name__) socketio = SocketIO(app) player = vlc.MediaPlayer(vlc.Instance()) songs = load_csv() current_song = '' @app.route('/') def sessions(): return render_template('index.html', buttons=songs.values()) @socketio.on('my event') def handle_my_custom_event(json): global player, current_song print('received my event:', json['data']) print(player.get_state(), str(player.get_state())) if json['data'] == current_song: if str(player.get_state()) == 'State.Stopped': player.play() else: player.stop() else: player.stop()
def loadNd(dataID): if dataID == 1: # Number 1 dataset fields = [ 'duration', 'width', 'height', 'bitrate', 'framerate', 'i', 'p', 'b', 'frames', 'i_size', 'p_size', 'b_size', 'size', 'o_bitrate', 'o_framerate', 'o_width', 'o_height', 'umem', 'utime' ] # fields = ['i_size', 'umem','utime'] fields = ['duration', 'bitrate', 'framerate', 'size', 'umem', 'utime'] y_column = 4 # should be the order in the input file, not in the "fields" order. data = tools.load_csv(file1, fields, y_column) # "../data/1online_video_dataset/1transcoding_mesurment.csv" if dataID == 2: # Number 2 dataset fields = ["RMSD", "F1", "F2", "F3", "F4", "F5", "F6", "F7", "F8", "F9"] # fields = ["RMSD", "F3", "F5"] # fields = ["RMSD", "F2", "F7"] fields = ["RMSD", "F4", "F5"] y_column = 0 # should be the order in the input file, not in the "fields" order. data = tools.load_csv(file2, fields, y_column) # "../data/2CASP.csv" if dataID == 3: # Number 3 dataset fields = [ 'year', 'month', 'day', 'hour', 'pm2.5', 'DEWP', 'TEMP', 'PRES', 'Iws', 'Is', 'Ir' ] y_column = 4 # fields = ['pm2.5','PRES'] # y_column = 0 # should be the order in the input file, not in the "fields" order. # fields = ['pm2.5','TEMP', 'PRES'] # fields = ['pm2.5', 'TEMP', 'Iws'] # good vision # fields = ['pm2.5', 'PRES', 'Iws'] # y_column = 0 # should be the order in the input file, not in the "fields" order. data = tools.load_csv(file3, fields, y_column) if dataID == 4: # Number 4 dataset # load the data fields = [ 'n_tokens_title', 'n_tokens_content', 'n_unique_tokens', 'n_non_stop_unique_tokens' ] y_column = 2 # fields = ['n_tokens_content', 'n_unique_tokens'] # y_column = 1 # should be the order in the input file, not in the "fields" order. # fields = ['n_tokens_title','n_tokens_content','n_unique_tokens'] # fields = ['n_tokens_content', 'n_unique_tokens','n_non_stop_unique_tokens'] # #fields = [ 'n_tokens_content', 'n_unique_tokens', 'n_non_stop_unique_tokens'] # y_column = 1 # should be the order in the input file, not in the "fields" order. data = tools.load_csv(file4, fields, y_column) if dataID == 5: # Number 5 dataset # ''' # load the data fields = [ 'Temperature', 'Exhaust_Vacuum', 'Ambient_Pressure', 'Relative_Humidity', 'energy_output' ] y_column = 4 # should be the order in the input file, not in the "fields" order. data = tools.load_csv(file5, fields, y_column) if dataID == 6: # Number 6 dataset fields = [ 'year', 'c1', 'c2', 'c3', 'c4', 'c5', 'c6', 'c7', 'c8', 'c9', 'c10', 'c11', 'c12', 'c13', 'c14', 'c15', 'c16', 'c17', 'c18', 'c19', 'c20', 'c21', 'c22', 'c23', 'c24', 'c25', 'c26', 'c27', 'c28', 'c29', 'c30', 'c31', 'c32', 'c33', 'c34', 'c35', 'c36', 'c37', 'c38', 'c39', 'c40', 'c41', 'c42', 'c43', 'c44', 'c45', 'c46', 'c47', 'c48', 'c49', 'c50', 'c51', 'c52', 'c53', 'c54', 'c55', 'c56', 'c57', 'c58', 'c59', 'c60', 'c61', 'c62', 'c63', 'c64', 'c65', 'c66', 'c67', 'c68', 'c69', 'c70', 'c71', 'c72', 'c73', 'c74', 'c75', 'c76', 'c77', 'c78', 'c79', 'c80', 'c81', 'c82', 'c83', 'c84', 'c85', 'c86', 'c87', 'c88', 'c89', 'c90' ] fields = ['year', 'c1', 'c2', 'c3'] y_column = 0 # should be the order in the input file, not in the "fields" order. data = tools.load_csv(file6, fields, y_column) if dataID == 7: # Number 7 dataset # load the data # fields = ['duration','width','height','bitrate','framerate','i','p','b','frames','i_size','p_size','b_size','size','o_bitrate','o_framerate','o_width','o_height','umem','utime'] fields = [ 'Time_(seconds)', 'Methane_conc_(ppm)', 'Ethylene_conc_(ppm)', 'c1', 'c2', 'c3', 'c4', 'c5', 'c6', 'c7', 'c8', 'c9', 'c10', 'c11', 'c12', 'c13', 'c14', 'c15', 'c16' ] fields = ['c1', 'c2'] y_column = 1 # should be the order in the input file, not in the "fields" order. fields = ['c1', 'c2', 'c4'] fields = [ 'Methane_conc_(ppm)', 'Ethylene_conc_(ppm)', 'c1', 'c2', 'c3', 'c4' ] y_column = 3 # should be the order in the input file, not in the "fields" order. # data = tools.load_csv("datasets/1online_video_dataset/1transcoding_mesurment.csv",fields,y_column) data = tools.load_csv(file7, fields, y_column, sep=' ') if dataID == 8: # Number 8 dataset # load the data fields = [ 'timestamp', 'Global_active_power', 'Global_reactive_power', 'Voltage', 'Global_intensity', 'Sub_metering_1', 'Sub_metering_2', 'Sub_metering_3', 'energy' ] # fields = ['Date', 'Global_active_power', 'Global_reactive_power', 'Voltage', 'Global_intensity', # 'Sub_metering_1', 'Sub_metering_2', 'Sub_metering_3'] # fields = ['Global_active_power', 'Global_reactive_power', 'energy'] y_column = 8 # should be the order in the input file, not in the "fields" order. data = tools.load_csv(file8, fields, y_column, sep=',') return data