Coming from Orange

If you're an Orange user, you're familiar with widget-based visual programming for data science. Sigilweaver shares the visual workflow approach but focuses on data transformation rather than machine learning.

Terminology Mapping

Orange	Sigilweaver	Notes
Workflow	Workflow	Saved pipeline file
Widget	Tool	Dragged from palette onto canvas
Channel	Wire	Connections between widgets/tools
File	Input tool	Load CSV, Parquet, Excel
Save Data	Output tool	Save transformed data
Select Columns	Select tool	Choose, rename, reorder columns
Select Rows	Filter tool	Filter rows by condition
Group	Summarize tool	Aggregate by groups
Merge Data	Join tool	Combine datasets
Concatenate	Union tool	Stack datasets vertically
Feature Constructor	Formula tool	Create calculated columns

Common Tasks

Loading Data

Orange:
Drag File widget → Browse for file → Data automatically loads

Sigilweaver:
Drag Input tool → Pick your file → Preview after execution

Filtering Rows

Orange:
Drag Select Rows widget → Choose conditions visually

Sigilweaver:
Drag Filter tool → Write expression like pl.col("class") == "A" → True/False outputs

Key difference: Orange uses GUI builders. Sigilweaver uses Polars expressions - more powerful but requires learning syntax.

Selecting Features (Columns)

Orange:
Drag Select Columns widget → Drag columns between Available/Features/Meta

Sigilweaver:
Drag Select tool → Check columns to keep → Configure rename, reorder, or type casting

Creating New Features

Orange:
Drag Feature Constructor widget → Define new feature with expression

Sigilweaver:
Drag Formula tool → Expression: pl.col("height") / pl.col("weight") → Name output column

Grouping and Aggregating

Orange:
Drag Aggregate widget (if available) or use Group widget

Sigilweaver:
Drag Summarize tool → Set columns to "Group By" → Choose aggregation (Sum, Mean, Count, etc.) for metrics

Merging Datasets

Orange:
Drag Merge Data widget → Connect two data sources → Configure merge type

Sigilweaver:
Drag Join tool → Wire left (L) and right (R) inputs → Configure join keys → Three outputs: matched (J), unmatched left (L), unmatched right (R)

Key Differences

1. Scope: Data Prep vs. Machine Learning

Orange is a visual data mining and machine learning platform:

• Classification, regression, clustering widgets
• Model evaluation and visualization
• Interactive data exploration
• Educational focus for teaching ML concepts

Sigilweaver focuses on data transformation and ETL:

• Loading, cleaning, filtering, joining data
• Preparing datasets for analysis
• No ML models (yet)

Think of Sigilweaver as handling the data wrangling phase before you move to Orange for machine learning.

2. Expression Language

Orange uses visual builders and simple Python-like expressions.

Sigilweaver uses Polars expressions (Python-based):

# Filter rows
pl.col("petal_length") > 5.0

# Create feature
pl.col("sepal_length") / pl.col("sepal_width")

# Conditional feature
pl.when(pl.col("species") == "setosa")
  .then(1)
  .otherwise(0)

The Expressions Guide covers the syntax in detail.

3. Target Audience

Orange is designed for:

• Students learning data science
• Researchers exploring data visually
• Users who want to try ML without coding

Sigilweaver is designed for:

• Data analysts who need repeatable pipelines
• Anyone preparing data for downstream analysis
• Users comfortable with expressions but who prefer visual workflow structure

4. Performance and Scale

Orange loads entire datasets into memory. Sigilweaver uses lazy evaluation:

• Workflow is optimized before execution
• Processes data in streaming fashion when possible
• Handles datasets larger than available memory

Your First Workflow

Let's build a simple Orange-style workflow: load data, filter, create a calculated column, and aggregate.

Steps:

1. Load data:

   • Drag Input tool to canvas
   • Pick your CSV file (e.g., iris.csv)
   • Click to preview

2. Filter for specific species:

   • Drag Filter tool, wire from Input
   • Expression: pl.col("species") == "versicolor"
   • Use the T (true) output

3. Calculate sepal ratio:

   • Drag Formula tool, wire from Filter's T output
   • Expression: pl.col("sepal_length") / pl.col("sepal_width")
   • Output column name: sepal_ratio

4. Aggregate statistics:

   • Drag Summarize tool, wire from Formula
   • Set species to Group By (if you want per-species stats, or leave blank for global)
   • Set sepal_ratio to Mean
   • Set petal_length to Mean
   • Set species to Count (row count)

5. Save results:

   • Drag Output tool, wire from Summarize
   • Configure output format (CSV, Parquet, Excel)
   • Execute workflow

What's Missing (Compared to Orange)?

Sigilweaver is focused on data transformation, so it doesn't have:

• Machine learning models (classification, regression, clustering)
• Interactive visualizations (scatter plots, box plots, etc.)
• Model evaluation widgets (confusion matrix, ROC curves)
• Educational add-ons for teaching

If you need ML and visualization, Orange is excellent. But for data cleaning, joining, and preparation at scale, Sigilweaver is faster and more efficient.

Combining Orange and Sigilweaver

You can use both tools together:

1. Sigilweaver: Clean, filter, join, and prepare large datasets → Save to CSV/Parquet
2. Orange: Load the prepared data → Build ML models → Visualize results

This separation keeps each tool doing what it does best.

Next Steps

• Interface Guide - Learn canvas navigation and workflow building
• Expressions Basics - Master Polars expression syntax
• Tools Reference - Detailed documentation for every tool

If you like Orange's visual approach, you'll feel comfortable with Sigilweaver's canvas. The main learning curve is expressions - they're more powerful than Orange's GUI builders once you get the hang of them.