craftbeerpi4-pione/cbpi/controller/log_file_controller.py

221 lines
8.5 KiB
Python

import datetime
import glob
import logging
import os
from logging.handlers import RotatingFileHandler
from time import strftime, localtime
import pandas as pd
import zipfile
import base64
import urllib3
from cbpi.api import *
from cbpi.api.config import ConfigType
from cbpi.api.base import CBPiBase
import asyncio
class LogController:
def __init__(self, cbpi):
'''
:param cbpi: craftbeerpi object
'''
self.cbpi = cbpi
self.logger = logging.getLogger(__name__)
self.configuration = False
self.datalogger = {}
def log_data(self, name: str, value: str) -> None:
self.logfiles = self.cbpi.config.get("CSVLOGFILES", "Yes")
self.influxdb = self.cbpi.config.get("INFLUXDB", "No")
if self.logfiles == "Yes":
if name not in self.datalogger:
max_bytes = self.cbpi.config.get("SENSOR_LOG_MAX_BYTES", 1048576)
backup_count = self.cbpi.config.get("SENSOR_LOG_BACKUP_COUNT", 3)
data_logger = logging.getLogger('cbpi.sensor.%s' % name)
data_logger.propagate = False
data_logger.setLevel(logging.DEBUG)
handler = RotatingFileHandler('./logs/sensor_%s.log' % name, maxBytes=max_bytes, backupCount=backup_count)
data_logger.addHandler(handler)
self.datalogger[name] = data_logger
formatted_time = strftime("%Y-%m-%d %H:%M:%S", localtime())
self.datalogger[name].info("%s,%s" % (formatted_time, str(value)))
if self.influxdb == "Yes":
self.influxdbcloud = self.cbpi.config.get("INFLUXDBCLOUD", "No")
self.influxdbaddr = self.cbpi.config.get("INFLUXDBADDR", None)
self.influxdbport = self.cbpi.config.get("INFLUXDBPORT", None)
self.influxdbname = self.cbpi.config.get("INFLUXDBNAME", None)
self.influxdbuser = self.cbpi.config.get("INFLUXDBUSER", None)
self.influxdbpwd = self.cbpi.config.get("INFLUXDBPWD", None)
id = name
try:
chars = {'ö':'oe','ä':'ae','ü':'ue','Ö':'Oe','Ä':'Ae','Ü':'Ue'}
sensor=self.cbpi.sensor.find_by_id(name)
if sensor is not None:
itemname=sensor.name.replace(" ", "_")
for char in chars:
itemname = itemname.replace(char,chars[char])
out="measurement,source=" + itemname + ",itemID=" + str(id) + " value="+str(value)
except Exception as e:
logging.error("InfluxDB ID Error: {}".format(e))
if self.influxdbcloud == "Yes":
self.influxdburl="https://" + self.influxdbaddr + "/api/v2/write?org=" + self.influxdbuser + "&bucket=" + self.influxdbname + "&precision=s"
try:
header = {'User-Agent': name, 'Authorization': "Token {}".format(self.influxdbpwd)}
http = urllib3.PoolManager()
req = http.request('POST',self.influxdburl, body=out, headers = header)
except Exception as e:
logging.error("InfluxDB cloud write Error: {}".format(e))
else:
self.base64string = base64.b64encode(('%s:%s' % (self.influxdbuser,self.influxdbpwd)).encode())
self.influxdburl='http://' + self.influxdbaddr + ':' + str(self.influxdbport) + '/write?db=' + self.influxdbname
try:
header = {'User-Agent': name, 'Content-Type': 'application/x-www-form-urlencoded','Authorization': 'Basic %s' % self.base64string.decode('utf-8')}
http = urllib3.PoolManager()
req = http.request('POST',self.influxdburl, body=out, headers = header)
except Exception as e:
logging.error("InfluxDB write Error: {}".format(e))
async def get_data(self, names, sample_rate='60s'):
logging.info("Start Log for {}".format(names))
'''
:param names: name as string or list of names as string
:param sample_rate: rate for resampling the data
:return:
'''
# make string to array
if isinstance(names, list) is False:
names = [names]
# remove duplicates
names = set(names)
result = None
def dateparse(time_in_secs):
'''
Internal helper for date parsing
:param time_in_secs:
:return:
'''
return datetime.datetime.strptime(time_in_secs, '%Y-%m-%d %H:%M:%S')
def datetime_to_str(o):
if isinstance(o, datetime.datetime):
return o.__str__()
for name in names:
# get all log names
all_filenames = glob.glob('./logs/sensor_%s.log*' % name)
# concat all logs
df = pd.concat([pd.read_csv(f, parse_dates=True, date_parser=dateparse, index_col='DateTime', names=['DateTime', name], header=None) for f in all_filenames])
logging.info("Read all files for {}".format(names))
# resample if rate provided
# if sample_rate is not None:
# df = df[name].resample(sample_rate).max()
# logging.info("Sampled now for {}".format(names))
df = df[name].dropna()
# take every nth row so that total number of rows does not exceed max_rows * 2
max_rows = 500
total_rows = df.shape[0]
if (total_rows > 0) and (total_rows > max_rows):
nth = int(total_rows/max_rows)
if nth > 1:
df = df.iloc[::nth]
if result is None:
result = df
else:
result = pd.merge(result, df, how='outer', left_index=True, right_index=True)
data = {"time": df.index.tolist()}
if len(names) > 1:
for name in names:
data[name] = result[name].interpolate(limit_direction='both', limit=10).tolist()
else:
data[name] = result.interpolate().tolist()
logging.info("Send Log for {}".format(names))
return data
async def get_data2(self, ids) -> dict:
def dateparse(time_in_secs):
return datetime.datetime.strptime(time_in_secs, '%Y-%m-%d %H:%M:%S')
result = dict()
for id in ids:
# df = pd.read_csv("./logs/sensor_%s.log" % id, parse_dates=True, date_parser=dateparse, index_col='DateTime', names=['DateTime',"Values"], header=None)
# concat all logs
all_filenames = glob.glob('./logs/sensor_%s.log*' % id)
df = pd.concat([pd.read_csv(f, parse_dates=True, date_parser=dateparse, index_col='DateTime', names=['DateTime', 'Values'], header=None) for f in all_filenames])
result[id] = {"time": df.index.astype(str).tolist(), "value":df.Values.tolist()}
return result
def get_logfile_names(self, name:str ) -> list:
'''
Get all log file names
:param name: log name as string. pattern /logs/sensor_%s.log*
:return: list of log file names
'''
return [os.path.basename(x) for x in glob.glob('./logs/sensor_%s.log*' % name)]
def clear_log(self, name:str ) -> str:
all_filenames = glob.glob('./logs/sensor_%s.log*' % name)
for f in all_filenames:
os.remove(f)
if name in self.datalogger:
del self.datalogger[name]
def get_all_zip_file_names(self, name: str) -> list:
'''
Return a list of all zip file names
:param name:
:return:
'''
return [os.path.basename(x) for x in glob.glob('./logs/*-sensor-%s.zip' % name)]
def clear_zip(self, name:str ) -> None:
"""
clear all zip files for a sensor
:param name: sensor name
:return: None
"""
all_filenames = glob.glob('./logs/*-sensor-%s.zip' % name)
for f in all_filenames:
os.remove(f)
def zip_log_data(self, name: str) -> str:
"""
:param name: sensor name
:return: zip_file_name
"""
formatted_time = strftime("%Y-%m-%d-%H_%M_%S", localtime())
file_name = './logs/%s-sensor-%s.zip' % (formatted_time, name)
zip = zipfile.ZipFile(file_name, 'w', zipfile.ZIP_DEFLATED)
all_filenames = glob.glob('./logs/sensor_%s.log*' % name)
for f in all_filenames:
zip.write(os.path.join(f))
zip.close()
return os.path.basename(file_name)