How to make an interactive time serie plot using plotly?

Question

I am trying to make an interactive time serie visualization using plotly and jupyter notebook. I want to have a simple plot where I can filter the index of a dataframe using plotly and ipywidget and store the new index I have. But, I have no idea how to do so. I am investigating the documentation without any success. What I am doing so far:

import pandas as pd
import numpy as np
import plotly.graph_objs as go
from ipywidgets import interactive

index = pd.date_range(start='2020-01-01', end='2020-01-15', freq='D')
timeserie = pd.DataFrame(np.random.normal(0,1,size=index.size), index=index, columns=['sensor'])
fig = go.FigureWidget([
    go.Scatter(
        x=timeserie.index.values,
        y=timeserie.values,
        mode='markers'
    )
])
 
def update_training_dataset(index_min, index_max, sensor):
    scatter = fig.data[0]
    index = timeserie.loc[(timeserie.index >= index_min) & (timeserie.index <= index_max)].index
    sensor_value = timeserie.loc[scatter.x, sensor].values
    with fig.batch_update():
        fig.layout.yaxis.title = sensor
        scatter.x = index
        scatter.y = sensor_value
 
interactive(update_training_dataset, index_min=index, index_max=index, sensor=timeserie.columns)

But, it leads to a strange error.. KeyError: "None of [Int64Index([15778368000000000000, ... are in the [index]" This is weird as the index of my timeserie has datetimeindex as type. This code would lead to updating the dataframe according to the values of sensor, index_min, index_max that the user set. Also, I note that the date are provided in a select widget... I would love to have a date picker here. Can someone help me? Provide any code that I can get some insights from? Thank you:)

EDIT

The solution is provided below thanks to Serge:)

fig = go.FigureWidget([
    go.Scatter(
        x=timeserie.index,
        y=timeserie.values,
        mode='markers'
    )
])
 
def update_training_dataset(index_min, index_max, Sensor):
    scatter = fig.data[0]
    index = timeserie.loc[(timeserie.index >= index_min) & (timeserie.index <= index_max)].index
    sensor_value = timeserie.loc[scatter.x, Sensor].values
    with fig.batch_update():
        fig.layout.yaxis.title = Sensor
        scatter.x = index
        scatter.y = sensor_value
 
        
date_picker_max = DatePicker(
    description='End date',
    disabled=False,
    value = index.max()
)
 
date_picker_min = DatePicker(
    description='Start date',
    disabled=False,
    value = index.min()
)
 
interact(
    update_training_dataset, 
    index_min=date_picker_min, 
    index_max=date_picker_max, 
    Sensor=timeserie.columns
)

I am still working on a way to have hours:minutes:seconds in the date picker.

EDIT 2 By the way, no need to use interact instead of interactive: they seem to support widgets as parameters. Also, you need to import ipydatetime as below to get datetime picker.

# usual imports
from ipydatetime import DatetimePicker

fig = go.FigureWidget([
    go.Scatter(
        x=timeserie.index,
        y=timeserie.values,
        mode='markers'
    )
])
 
def update_training_dataset(index_min, index_max, Sensor):
    scatter = fig.data[0]
    index = timeserie.loc[(timeserie.index >= index_min) & (timeserie.index <= index_max)].index
    sensor_value = timeserie.loc[scatter.x, Sensor].values
    with fig.batch_update():
        fig.layout.yaxis.title = Sensor
        scatter.x = index
        scatter.y = sensor_value
 
        
date_picker_max = DatetimePicker(
    description='End date',
    disabled=False,
    value = index.max()
)
 
date_picker_min = DatetimePicker(
    description='Start date',
    disabled=False,
    value = index.min()
)
 
interact(
    update_training_dataset, 
    index_min=date_picker_min, 
    index_max=date_picker_max, 
    Sensor=timeserie.columns
)

Answer 1

Actually, your code is all good. You did a simple mistake in the definition of fig . Try the following

fig = go.FigureWidget([
    go.Scatter(
        x=timeserie.index,
        y=timeserie.values,
        mode='markers'
    )
])
 
def update_training_dataset(index_min, index_max, sensor):
    scatter = fig.data[0]
    index = timeserie.loc[(timeserie.index >= index_min) & (timeserie.index <= index_max)].index
    sensor_value = timeserie.loc[scatter.x, sensor].values
    with fig.batch_update():
        fig.layout.yaxis.title = sensor
        scatter.x = index
        scatter.y = sensor_value
 
interactive(update_training_dataset, index_min=index, index_max=index, sensor=timeserie.columns)

You simly made the error of defining x=timeserie.index.values when it actually should be x=timeserie.index .

The result is fine when this is changed.