DescriberAdapter

DescriberAdapter handles data description and comparison, supporting single dataset description and multi-dataset comparative analysis.

Class Architecture

classDiagram

    class DescriberAdapter {
        +config: dict
        +source: dict
        +describer: Describer
        +__init__(config)
        +run() dict~str, DataFrame~
    }

    class Describer {
        +config: dict
        +mode: str
        +method: str
        +create()
        +eval(data) dict~str, DataFrame~
    }

    class BaseEvaluator {
        <<abstract>>
        +evaluate() dict
    }

    class DescriberDescribe {
        +evaluate() dict
    }

    class DescriberCompare {
        +evaluate() dict
    }

    DescriberAdapter ..> Describer : uses for description
    Describer --> BaseEvaluator : creates
    BaseEvaluator <|-- DescriberDescribe
    BaseEvaluator <|-- DescriberCompare

    %% Style definitions
    class DescriberAdapter {
        <<Main Class>>
    }
    style DescriberAdapter fill:#E6E6FA

    class Describer {
        <<Core Module>>
    }
    style Describer fill:#4169E1,color:#fff

    style BaseEvaluator fill:#9370DB,color:#fff
    style DescriberDescribe fill:#FFE4E1
    style DescriberCompare fill:#FFE4E1

    note for DescriberAdapter "1. Describe mode: Single dataset description\n2. Compare mode: Two datasets comparison\n3. Flexible source specification\n4. Auto-aligns data types before description"

Legend:

• Light purple box: DescriberAdapter main class
• Blue box: Core description modules
• Purple box: Data alignment modules
• Light pink box: Configuration classes
• ..>: Dependency relationship
• -->: Ownership relationship

Main Features

• Unified data description interface
• Flexible data source selection (via source parameter)
• Two modes support: describe (single dataset), compare (dataset comparison)
• Automatic data type alignment (using Schema)
• Support for various statistical methods and JS Divergence calculation

Method Reference

__init__(config: dict)

Initialize a DescriberAdapter instance.

Parameters:

• config: dict, required
  • Configuration parameters dictionary
  • Must include source key (data source)
  • Optional method key: default, describe, compare
  • Optional mode key: automatically determined by source count

run(input: dict)

Execute data description or comparison, including automatic data type alignment.

Parameters:

• input: dict, required
  • Input parameters dictionary
  • Contains data dictionary (datasets)
  • Optional metadata for data type alignment

Returns: No direct return value. Use get_result() to retrieve results.

set_input(status)

Set input data for the describer.

Parameters:

• status: Status, required
  • System status object
  • Extracts data based on source configuration

Returns:

• dict: Dictionary containing data required for description

get_result()

Retrieve description results.

Returns:

• dict[str, pd.DataFrame]: Dictionary of description results

Usage Examples

Single Dataset Description

from petsard.adapter import DescriberAdapter

# Describe single dataset
adapter = DescriberAdapter({
    "source": "Loader",  # or ["Loader"]
    "method": "describe",
    "describe_method": ["mean", "median", "std", "corr"]
})

# Execute description
adapter.run({})

# Get results
results = adapter.get_result()

Dataset Comparison

# Compare two datasets
adapter = DescriberAdapter({
    "source": {
        "base": "Splitter.train",
        "target": "Synthesizer"
    },
    "method": "compare",
    "stats_method": ["mean", "std", "jsdivergence"],
    "compare_method": "pct_change"
})

# Execute comparison
adapter.run({})

# Get results
comparison_results = adapter.get_result()

Workflow

1. Source Parsing: Parse source parameter to determine data sources
2. Mode Determination:
   • 1 source: describe mode
   • 2 sources: compare mode
3. Data Collection: Collect specified data from Status
4. Schema Retrieval: Attempt to get metadata for data alignment
5. Data Type Alignment (when Schema is available)
6. Execute Description or Comparison

Source Parameter Format

Describe Mode (Single Data Source)

# String format
source: "Loader"

# List format
source: ["Synthesizer"]

Compare Mode (Two Data Sources)

# Dictionary format (recommended)
source:
  base: "Splitter.train"
  target: "Synthesizer"

# Backward compatibility format
source:
  ori: "Splitter.train"
  syn: "Synthesizer"

Data Source Syntax

• Simple format: "ModuleName" - Takes first available data from module
• Precise format: "ModuleName.key" - Takes specific keyed data from module
  • Examples: "Splitter.train", "Splitter.validation"

Notes

• This is an internal API, direct usage is not recommended
• Use YAML configuration files and Executor instead
• Compare mode reuses DescriberDescribe’s statistical functionality
• Parameter naming recommends using base/target instead of legacy ori/syn
• Results are cached until next run() call