Plotting histogram instead of bar chart

I am trying to make a stacked bar graph but when I give a list to the parameter “y” to the function “plotly.express.bar”, plotly creates a histogram instead of a bar chart.

import pandas as pd
import plotly.express as px

tweets = pd.read_csv("Tweets.csv")
tweets.head()

bar = tweets[['airline','airline_sentiment']]
bar.reset_index()

new_df = pd.DataFrame(columns = ["airline", "positive", "negative", "neutral"])

airlines = bar['airline'].unique()

for i in range(len(airlines)):
    new_df.loc[len(new_df),'airline']=airlines[i]
    x = bar[bar['airline'] == airlines[i]]
    new_df.loc[len(new_df)-1,'positive'] = (x['airline_sentiment'] == 'positive').sum()
    new_df.loc[len(new_df)-1,'negative'] = len(x[x['airline_sentiment'] == 'negative'])
    new_df.loc[len(new_df)-1,'neutral'] = len(x[x['airline_sentiment'] == 'neutral'])
             
new_df

fig = px.bar(new_df, x = "airline", y = ["positive","negative", "neutral"], title = "Sentiments")
fig.show()

The CSV file and Jupyter notebook are included in the zipped file. StackedBar.zip

[nicolaskruchten]

It would be nice for Plotly to do this, yes, but it’s more of a data-later responsibility here IMO than a vis-layer concern. If Pandas is unable to get the type right in this case, it’s unlikely that Plotly will be able to do a better job 😃

[SinghShreyas]

Yeah, you are right Nicolas.

Anyways, as a programmer, we should be able to debug the problem and your comment helped me in understanding the problem which was great learning for me.