Data Wrangler in Fabric Notebook-More than the Power Query In Python

📌 Overview

Data Wrangler in Microsoft Fabric provides an interactive, no-code interface for cleaning, transforming, and exploring your data directly inside a Fabric Notebook.It bridges the gap between visual data prep and reproducible code — letting you explore data interactively and automatically generate Python (Pandas or PySpark) scripts for production.

💡 Why Use Data Wrangler?

Capability	Description
🧩 Visual Interface	Clean spreadsheet-like grid for exploring and transforming data
🧠 Smart Profiling	Auto-detects data types, unique counts, nulls, and distributions
⚙️ Code Generation	Exports equivalent Python/PySpark code for reproducibility
🔗 Native Fabric Integration	Works directly with OneLake, Lakehouses, and DataFrames
🚀 Accelerated Prep	Perfect for early-stage data cleaning, profiling, and transformation

Launching Data Wrangler interface in a Fabric Notebook

🧩 When to Use Data Wrangler

✅ Best for:

• Rapid data exploration and cleaning

• Preparing datasets before modeling or visualization

• Low-code environments and collaborative data analysis

🚫 Avoid for:

• Large-scale ETL or multi-table joins

• Automated, scheduled data pipelines

• Transformations requiring advanced Spark operations

🧪 Step-by-Step Tutorial

Step 1: Load Your Dataset

Start by loading a table or file into a Spark DataFrame.

DataFrame displayed in notebook ready for Data Wrangler

Step 2: Launch Data Wrangler

From the DataFrame output cell:

• Click the “⋮ (More options)” icon.

• Select “Data Wrangler”.

Menu showing option to open Data Wrangler

Step 3: Explore and Profile Your Data

Once open:

• View data summaries (nulls, types, unique counts).

• Sort or filter directly.

• Identify anomalies or missing values visually.

Data profiling view showing column summaries and stats

Step 4: Apply Transformations

Perform transformations directly from the UI:

• 🧹 Remove duplicates

• 🔄 Change column data types

• ✂️ Split or merge columns

• 🧮 Create calculated fields

• 🧍 Filter out invalid rows

Each transformation is automatically tracked in the “Applied Steps” pane.

Applied Steps panel showing a list of transformations

Step 5: Export Generated Code

Once you’re satisfied:

• Click “Add code to Notebook”

• Review the auto-generated Python/PySpark code.

Example output:

# Code generated by Data Wrangler for pandas DataFrame

def clean_data(wrangler_sample_df):

    # Sort by column: 'Ticket' (ascending)

    wrangler_sample_df = wrangler_sample_df.sort_values(['Ticket'])

    # Drop duplicate rows across all columns

    wrangler_sample_df = wrangler_sample_df.drop_duplicates()

    return wrangler_sample_df

wrangler_sample_df_clean = clean_data(wrangler_sample_df.copy())

wrangler_sample_df_clean.head()

Generated PySpark code ready to insert into Notebook

Step 6: Save or Continue Your Workflow

Save your transformed dataset back to your Lakehouse

You can now use this data for:

• Power BI Semantic Models

• Dataflows

• Further Fabric Notebooks or Pipelines

  
  

⚙️ Best Practices

✅ Use sampled datasets for quicker performance✅ Review and optimize generated code before production✅ Use Wrangler for interactive prep, and Dataflows or Pipelines for automation

⚠️ Known Limitations

Limitation	Workaround
⚡ May slow down on very large datasets	Use data sampling or filter before launching
🔗 Single-table operations only	Use PySpark or Dataflows for joins
⏱ No scheduling or automation	Integrate with Fabric Pipelines or Notebook jobs

⚡ Final Thoughts

Data Wrangler bridges the gap between visual data preparation and code-based transformation.It empowers analysts and data enthusiasts to:

• Clean and prepare data efficiently

• Learn PySpark/Python naturally

• Transition smoothly into advanced notebook-based workflows

If you’ve ever wanted to learn PySpark but didn’t know where to begin — start with Data Wrangler.You’ll be writing transformations confidently before you realize it.

End-to-end flow: from raw data → Wrangler → cleaned dataset → Lakehouse

Related contents:

Live demo on Data Wrangler in Fabric : https://www.youtube.com/watch?v=Ge0VWZMa50I

Microsoft Learn : https://learn.microsoft.com/en-us/fabric/data-science/data-wrangler