The Plugin Orchestrator

DATADOC operates as an intelligent orchestrator. It doesn't modify your data haphazardly. Instead, it passes your dataset through an isolated chain of plugins in a strict priority order. Each plugin analyzes the data, computes optimal transformations, and builds a reproducible execution graph.

flowchart TD
    A["Raw Dataset CSV"] --> B("DATADOC Engine")
    
    subgraph PluginPipeline ["Priority-Ordered Plugin Pipeline"]
        direction TB
        P1["10: MissingValuePlugin
Imputes nulls"]
        P2["20: OutlierPlugin
Clips via IQR"]
        P3["30: DatetimePlugin
Extracts components"]
        P4["40: CategoricalEncoder
One-Hot Encodings"]
        P5["45: ScalingPlugin
Standard Scaling"]
        
        P1 --> P2 --> P3 --> P4 --> P5
    end
    
    B --> PluginPipeline
    
    PluginPipeline --> C{"Output Formats"}
    C -->|"datadoc engineer"| D["Cleaned CSV"]
    C -->|"datadoc pipeline"| E["Deterministic Python Script"]
    C -->|"datadoc analyze"| F["Terminal Report"]
                

How Plugins Work

Every plugin implements a strict BasePlugin interface ensuring absolute safety and determinism:

• analyze(): Evaluates the column to check if the plugin's logic applies.
• apply(): Applies the transformation mathematically and efficiently using Polars.
• rollback(): Safely undoes the transformation if needed.
• generate_code(): Generates equivalent Python source code for pipeline exportation.