Streamlit.io

a First Look

Was ist das Schwierigste Problem in der IT?

Latest Version: 0.71 

$ pipenv install streamlit
$ pipenv shell
$ streamlit hello

Welcome to Streamlit. Check out our demo in your browser.

  Local URL: http://localhost:8502
  Network URL: http://5.179.250.48:8502

 

 

Hello World

​import streamlit as st
import numpy as np
import pandas as pd

st.title('My first app')
st.write("Using data to create a table:")
st.write(pd.DataFrame({
    'first column': [1, 2, 3, 4],
    'second column': [10, 20, 30, 40]
}))
df = pd.DataFrame({
    'first column': [1, 2, 3, 4],
    'second column': [10, 20, 30, 40]
})

df

chart_data = pd.DataFrame(
    np.random.randn(20, 3),
    columns=['a', 'b', 'c'])

st.line_chart(chart_data)

test_app.py

$ streamlit run test_app.py
​# showing geographical map
st.map()

# Markdown text if checkbox activated
if st.checkbox("Click me"):
    st.markdown("# Header")

# a unique select box in the sidebar
option = st.sidebar.selectbox("Select number", [1,2,3], key="unique")
"You selected", option

# using columns and buttons
left_column, right_column = st.beta_columns(2)
pressed = left_column.button('Press me?')
if pressed:
    right_column.write("Woohoo!")​

# changing order of appearance
slot_1 = st.empty()
st.text("This will appear last")
slot_1.text("This will appear first")

Mini Tutorial

​df = pd.DataFrame()
# IDE says: "Statement seems to have no effect"
# but generates a table view of the dataframe
df

# it is not possible to center the content, yet
st.write(df.describe())

# By default info() does only return None
# So we need some workaround to see the output  
buf = io.StringIO()
df.info(buf=buf, verbose=True)
st.text(buf.getvalue())

# we can use matplotlib and even seaborn plots
# set_theme will globally change the layout of matplotlib plots
import seaborne as sns
sns.set_theme()
st.pyplot(sns.relplot(data=dax, kind="line"))

# Select the property ".figure" from the pandas inbuilt plotting
st.pyplot(df.plot().figure)

# hist returns a numpy array of subplots
# we need to select the first one
hist = df.hist(bins=100)
st.pyplot(hist[0, 0].figure)

TIPS & Quirks

  1. ​Within your company​
    • "​​Streamlit for teams" not yet available
  2. ​In the Cloud (needs github repo)
    • ​​request an invite
    • select a repo, branch and file
    • click deploy
    • any time you push will update your app immediatelly
    • Continuous Deployment! Very Nice!
  3. Examples
    1. https://share.streamlit.io/asmaier/streamlit-demo/main/stocky.py
    2. https://share.streamlit.io/asmaier/streamlit-demo/main/errorbars.py

Deployment

STREAMLIT VS. DASH

Plotly Dash Streamlit
focus production prototyping
code verbose minimal
aesthetics ugly, but fully customisable beautiful, but fixed
interactivity callbacks linear
performance high low
caching serious caching with redis, etc.. only @st.cache
plotting  plotly dash streamlit, matplotlib, bokeh, altair, vega lite, pydeck, graphviz, plotly dash, ...
deployment Heroku (free), Dash Enterprise Github (free), Streamlit for Teams
authentication/security yes what is that?
testing yes, pytest fixtures lalalalala

see also medium.com: Plotly Dash vs Streamlit

Big Picture

jupyter notebooks < streamlit < plotly dash

 

Competitors

  • Voila - Converts Jupyter notebooks to webapps
  • Panel - webapps build on top of bokeh
  • Shiny - R powered webapps (especially https://www.shinyapps.io/)

 

OPINION

medium.com: A roundup of Data Visualization Frameworks

panel.holoviz.org - Comparisons

Maybe we will finally see

something similar to

 https://demonstrations.wolfram.com ,

but this time open source

and based on

Streamlit and Python

???

FUTURE

THE END

Made with Slides.com