Pyplan, the easiest way to create and share a data analytic app.

Last year, I had read this article by Vincent Tatan, in which he proposed and taught how to build your own data dashboard using free tools.

As a data analytic consultant, I am very familiar with the problems we face every day when we expose our insights to people without technical knowledge in data science. A good dashboard is essential to communicate the processes, data and results to these users. Inspired by Tatan’s article, I decided to share the most ideal platform I found, designed to easily create data analytics apps: Pyplan.

“Can we build a dashboard Web Application with Python for free?” — Vincent Tatan

What is Pyplan?

Pyplan is a graphical Integrated Development Environment (IDE) for creating and sharing Data Analytics Apps. It is intended for easily creating and deploying data analytics apps by Data Analysts, highly integrated with Pandas, Numpy and Xarray libraries.

Different than a Jupyter Notebook, where code is structured as a list of sentences, in Pyplan code is contained in nodes that work as calculation steps, organized in a hierarchical influence diagram. Nodes can be evaluated, and their results inspected through a console output or displayed as a table or graph.

User interfaces are created by dragging nodes on a canvas and final apps can be shared with one click to the general audience.

Pyplan approach to code borrows the best of traditional workflows (ETL pipelines) and interactive programming (Jupyter Notebooks). The idea looks promising and it is worth to try it but I think it will take some time before we find it widespread as a programming alternative.

To make it easier for the reader understand what are the differences between Jupyter Notebooks and Pyplan, I made a table comparing main features:

Pyplan vs Jupyter notebooks for creating apps

There are two ways to use Pyplan:

One is through the web app, no installation is required, just create a free account in https://my.pyplan.org and you can start working with Pyplan.

The other one will be familiar to any python user and is to install it via PIP, Conda or Anaconda Navigator to run Pyplan local in your computer. Instructions explaining how to do it can be found here.

Update: After publishing this article, I was contacted by Pyplan to ask me if they could add this model to their online examples library, so now you can access it from the file manager of your free account in the following path: Pyplan> Public > Examples> 09 — Lazada Trends.

Now, to the project:

• The data used for this model is available in my Github here.
• You can see the final published app we will make here.

In his model, Vincent Tatan uses a dataset from Lazada (e-commerce platform), generating data visualizations to learn about price changes in Lazada products over date and time.

I will focus on creating the app so that the starting point will be reading the database with the prices.

Once logged in, we will create our new model just clicking in the orange button and giving it a name:

Creating a new model in Pyplan

Let’s create the first node (drag and drop) and define it as:

result = pd.read_csv(‘https://raw.githubusercontent.com/damiangino/Working-with-Pyplan/master/lazada.csv')

Code in a console window in Pyplan

Keep in mind that the id name of the node in the properties panel is how you will call it in the code:

Properties panel of a node

Remember that any node must end defining the variable result, and indicating which output will be displayed when the node is evaluated (by doing double click or ctrl+R). In this case, if we evaluate the created node, the result will be as follows:

Result view of a table in Pyplan

Now we will create a selector to choose the products about we want to inspect price evolution:

Let’s drag and drop a decision node “Decision” (rectangular shape) and define it as:

list_of_products = product_df["product_title"].unique().tolist()
result = selector(list_of_products, [0,5], True)

That node will help us select the items we want to inspect.

Then we will create the table of filtered items. Let’s drag e new node and define it as “product df filter” and define it as:

# Filter database by the elements contained in the selector
_filter = product_df['product_title'].isin( product_selector.value )
df = product_df[_filter]
df = df.sort_values('datetime', ascending=False)
result = df

And last, let’s create two outputs, dragging and dropping two new nodes: Product rating defined as:

result = product_df_filter[['product_title', 'rating_count']].set_index('product_title')

And Product Price Trends defined as:

result = product_df_filter[['product_title', 'datetime', 'product_price']].set_index(['product_title', 'datetime'])

We should end up with a diagram like this:

You can toggle the view of the diagram between titles and ids with Ctrl+Y:

With Ctrl+Y you can toggle between titles an ids views

Now we are ready to create our user interface.

In order to do that we will click on the “Interfaces” icon menu and then on “New Interface”. We will name our interface as Price Optimization Dashboard.

Creating a new interface in Pyplan

Now that we have an empty interface we will create the app and share it in just three steps. No coding required!!!.

Step 1 — Designing the UI layout

Start by making double click in the new interface to open it.

Once in the Interface canvas, let’s first add a new widget (click on +) and then choose the diagram icon. We will use the diagram later to drag nodes from it.

Now let’s split the upper widget vertically (right arrow on the top right menu) and then the split the left widget horizontally. The image below shows this process.

Designing the interface layout in Pyplan

Step 2 — Add I/O elements to the UI

Now Shift+Click on the selector (green node “Choose a product name”) and then click on the destination widget (upper left corner) and then click on “Format” and then select “Multi select”

Configuring the selector

The next step will be to drag the “product df filtered” node to the upper right corner to display the list of attributes filtered by the selector. We will keep that widget as a table.

Next let’s drag the “Product Rating” to the empty widget and configure it as horizontal bar. Let’s also drag the “product_title” dimension as the header of the bars.

Configuring a widget on the interface layout in Pyplan

Finally, let’s display in the dashboard the Price Trend. In order to do that, start clearing the content of the diagram widget in the dashboard, next in the options menu, choose the graph and you will be prompted to select a node, search for product_price_trends and then change graph type from bars to linear and accept it to save the new UI configuration.

Changing the type of graphic element in a Pyplan Interface

Step 3 — Share the App

Now that we have the final UI we are ready to share our app. Just click on the share icon and select “Share external” so the user not need to be logged into Pyplan to use the app. Copy and share the link with anyone to allow for instant access to a full operational app.

Sharing a dashboard in Pyplan

Conclusion:

From this example, you can figure out, the universe of possibilities offered by this type of solution to those doing data analytics. The intuitive interface, the transparency in the calculation logic and the ease to create and share interactive apps, make Pyplan the ideal platform for those of us who must conduct data analysis in any organization.

I prefer Pyplan instead of other alternatives because it requires to write less code and also because users with minimal programming skills can start producing data analytic very rapidly

I hope that discovering this tool will be as useful to you as it was for me.