import psutil
import panel as pn
import pandas as pd
import holoviews as hv
from holoviews import dim, opts

hv.extension("bokeh")

# Define functions to get memory and CPU usage
def get_mem_data():
    vmem = psutil.virtual_memory()
    df = pd.DataFrame(dict(free=vmem.free/vmem.total,
                           used=vmem.used/vmem.total),
                      index=[pd.Timestamp.now()])
    return df*100

def get_cpu_data():
    cpu_percent = psutil.cpu_percent(percpu=True)
    df = pd.DataFrame(list(enumerate(cpu_percent)),
                      columns=['CPU', 'Utilization'])
    df['time'] = pd.Timestamp.now()
    return df

# Define DynamicMap callbacks returning Elements
def mem_stack(data):
    data = pd.melt(data, 'index', var_name='Type', value_name='Usage')
    areas = hv.Dataset(data).to(hv.Area, 'index', 'Usage')
    return hv.Area.stack(areas.overlay()).relabel('Memory')


def cpu_box(data):
    return hv.BoxWhisker(data, 'CPU', 'Utilization', label='CPU Usage')

# Set up StreamingDataFrame and add async callback
cpu_stream = hv.streams.Buffer(get_cpu_data(), 800, index=False)
mem_stream = hv.streams.Buffer(get_mem_data())

# Define DynamicMaps and display plot
cpu_dmap = hv.DynamicMap(cpu_box, streams=[cpu_stream])
mem_dmap = hv.DynamicMap(mem_stack, streams=[mem_stream])

plot = (cpu_dmap + mem_dmap).opts(
    opts.Area(height=400, width=400, ylim=(0, 100), framewise=True),
    opts.BoxWhisker(box_fill_color=dim('CPU').str(), cmap='Category20',
                    width=500, height=400, ylim=(0, 100))
)

# Create PeriodicCallback which run every 500 milliseconds
def cb():
    cpu_stream.send(get_cpu_data())
    mem_stream.send(get_mem_data())

callback = pn.io.PeriodicCallback(callback=cb, period=500)
callback.start()

# Show plot inside notebook
plot