#!./venv/bin/python3
import pandas as pd
import numpy as np
import json
import matplotlib.pyplot as plt
import os
from datetime import timedelta

## Read the measurements data file ##
DATA_MEAS_DIR = 'data/measurements'
# Always plot latest datafile - replace [-1] with another index if you want to plot a specific file.
MEAS_LOG_FILE = sorted(os.listdir(DATA_MEAS_DIR))[-1]

# Store each dictionary of the measurements json in a list
with open(os.path.join(DATA_MEAS_DIR, MEAS_LOG_FILE)) as f:
    meas_data = [json.loads(line) for line in f]

# Use setpoint logger (only necessary for part two of the exercise "collecting fresh data")
use_setpoint_log = False


## Read the setpoints data file ##
if use_setpoint_log:
    DATA_SP_DIR = 'data/setpoints'
    # Always plot latest datafile
    SP_LOG_FILE = sorted(os.listdir(DATA_SP_DIR))[-1]

    # Store each dictionary of the setpoints json in a list
    with open(os.path.join(DATA_SP_DIR, SP_LOG_FILE)) as f:
        sp_data = [json.loads(line) for line in f]

    # Merge measurements and setpoints in one list
    data = meas_data + sp_data

else:
    data = meas_data

################################################################################
################## Part 2 ######################################################
################################################################################

def overshoot(df, T1, T2):
    yT1, yT2 = df[T1], df[T2]
    over = 1 / (yT1 - yT2) * np.max(yT2 - df)
    return over

SETPOINT_UNIX = 1718013959.6977577
SETPOINT_TS = pd.to_datetime(SETPOINT_UNIX, unit='s')
WINDOW = pd.to_datetime(SETPOINT_UNIX+25, unit='s')

## The controller is reasonably fast at reacting to changes; the sum of in and
## out is at zero roughly 5-10 seconds after a change.

# Construct a dataframe and pivot it to obtain a dataframe with a column per unit, and a row per timestamp.
df = pd.DataFrame.from_records(data)
df['time'] = pd.to_datetime(df['time'], unit='s')
df_pivot = df.pivot_table(values='value', columns='unit', index='time')
df_resampled = df_pivot.resample('0.1s').mean()
df_resampled.interpolate(method='linear', inplace=True)
df_resampled = pd.DataFrame(df_resampled)

# Plot the data. Note, that the data will mostly not be plotted with lines.
plt.ion()  # Turn interactive mode on
plt.figure()
ax1, ax2 = plt.subplot(211), plt.subplot(212)
df_resampled[[c for c in df_resampled.columns if '_p' in c]].plot(marker='.', ax=ax1, linewidth=3)
ax2.plot(df_resampled['pcc_p'][SETPOINT_TS:WINDOW], marker='.', linewidth=3, label='pcc_p')
ax2.plot(df_resampled['dumpload_p'][SETPOINT_TS:WINDOW], marker='.', linewidth=3, label='dumpload')
plt.legend()

# print(overshoot(df_resampled['pcc_p'][SETPOINT_TS:WINDOW], SETPOINT_TS, WINDOW))

plt.show(block=True)