For many Python data analyses, you only want to share a specific user-facing part of it rather than the whole code or notebook. This often is of the form of a standalone product that non-technical people can view directly from their existing tools - e.g. browsers, email, Slack, etc., and without the overhead and requirements of Python and Jupyter.
Datapane allows you to programmatically create reports from the objects in your Python analyses, such as pandas DataFrames, plots from visualization libraries, and Markdown text.
Datapane provides a Python API that allows you to create, save, and upload reports comprised of a collection of data-centric blocks.
For instance, Datapane provides a
Table block that takes a pandas
DataFrame. We can create a
Table block by passing a
DataFrame into it, and create a
Report with that single block in it as follows:
simple_report.pyimport pandas as pdimport datapane as dpdf = pd.read_csv('https://covid.ourworldindata.org/data/vaccinations/vaccinations-by-manufacturer.csv', parse_dates=['date'])df = df.groupby(['vaccine', 'date'])['total_vaccinations'].sum().tail(1000).reset_index()report = dp.Report(dp.Table(df))report.save(path='report.html', open=True)
As seen above,
Reports can be saved to local
HTML files. Copying this code into a new script and running it will generate the report.
$ python3 simple_report.py
If you send this HTML file to somebody (or publish it on Datapane Public), they will be able to view your dataset, sort and filter it, and download it as a CSV.
That report was pretty basic, but we can jazz it up by adding some plots and Markdown text. Unlike a traditional BI tool, Datapane does not rely on a proprietary visualization engine; instead, it natively supports Python visualization libraries such as Altair, Plotly, Bokeh, and Folium.
We also support an advanced Table component, called
DataTable, which allows sorting, filtering, and interactive analysis - however it requires uploading your report to a Datapane server to function.
Let's take the example above, and plot some data using the Python library Altair and add some text.
richer_report.pyimport pandas as pdimport altair as altimport datapane as dp# download data & group by manufacturerdf = pd.read_csv('https://covid.ourworldindata.org/data/vaccinations/vaccinations-by-manufacturer.csv', parse_dates=['date'])df = df.groupby(['vaccine', 'date'])['total_vaccinations'].sum().tail(1000).reset_index()# plot vaccinations over time using Altairplot = alt.Chart(df).mark_area(opacity=0.4, stroke='black').encode(x='date:T',y=alt.Y('total_vaccinations:Q'),color=alt.Color('vaccine:N', scale=alt.Scale(scheme='set1')),tooltip='vaccine:N').interactive().properties(width='container')# tablulate total vaccinations by manufacturertotal_df = df[df["date"] == df["date"].max()].sort_values("total_vaccinations", ascending=False).reset_index(drop=True)total_styled = total_df.style.bar(subset=["total_vaccinations"], color='#5fba7d', vmax=total_df["total_vaccinations"].sum())# embed into a Datapane Reportreport = dp.Report("## Vaccination Report",dp.Plot(plot, caption="Vaccinations by manufacturer over time"),dp.Table(total_styled, caption="Current vaccination totals by manufacturer"),dp.Table(df, caption="Initial Dataset"))report.save("report.html", open=True)
When this python script is run, we get the following report.
As described above, we can easily view our report in a browser. However, there are other ways to view and share our report whilst developing it.
Datapane has special integration into Jupyter Notebooks: if you're iterating a report, instead of having to open a new window to view it, you can preview a report directly from inside your notebook by calling
report.preview(), embedding it live into your notebook.
Next, we will explore how to upload and optionally share reports online, either on the free default Datapane server, or on your own Datapane Teams instance.