update_config()

Update the processor’s configuration settings.

Syntax

def update_config(
    config: dict
) -> None

Parameters

config : dict, required
- New processor configuration
- Required parameter
- Structure: {processor_type: {field_name: processing_method}}

Returns

None (method modifies instance state)

Description

The update_config() method updates the processor configuration. It can:

Override default processing methods
Set custom processors for specific columns
Disable processing for specific columns (set to None or "none")

Basic Example

from petsard import Loader, Processor

# Load data
loader = Loader('data.csv', schema='schema.yaml')
data, schema = loader.load()

# Create processor
processor = Processor(metadata=schema)

# Update configuration
new_config = {
    'missing': {
        'age': 'missing_median',
        'income': 'missing_mean'
    },
    'encoder': {
        'gender': 'encoder_onehot',
        'education': 'encoder_label'
    }
}
processor.update_config(new_config)

# Use updated configuration
processor.fit(data)
processed_data = processor.transform(data)

Configuration Formats

1. Using Processor Name (String)

config = {
    'missing': {
        'age': 'missing_mean',
        'income': 'missing_median'
    },
    'outlier': {
        'age': 'outlier_zscore',
        'income': 'outlier_iqr'
    },
    'encoder': {
        'gender': 'encoder_onehot'
    },
    'scaler': {
        'age': 'scaler_minmax',
        'income': 'scaler_standard'
    }
}
processor.update_config(config)

2. Using Processor with Parameters (Dictionary)

config = {
    'missing': {
        'age': {
            'method': 'missing_simple',
            'value': 0.0
        }
    },
    'scaler': {
        'created_at': {
            'method': 'scaler_timeanchor',
            'reference': 'event_time',
            'unit': 'D'
        }
    },
    'encoder': {
        'doc_date': {
            'method': 'encoder_date',
            'input_format': '%MinguoY-%m-%d',
            'date_type': 'date'
        }
    }
}
processor.update_config(config)

3. Disable Specific Processing

config = {
    'outlier': {
        'age': None,  # Don't process age outliers
        'income': 'outlier_iqr'
    },
    'scaler': {
        'gender': 'none'  # String "none" also disables
    }
}
processor.update_config(config)

Partial Update

# Update only some columns, others keep default
processor = Processor(metadata=schema)

# Update only missing value handling for age
processor.update_config({
    'missing': {
        'age': 'missing_median'
    }
})

# Other columns still use default configuration

Multiple Updates

processor = Processor(metadata=schema)

# First update
processor.update_config({
    'missing': {'age': 'missing_median'}
})

# Second update (will override or add)
processor.update_config({
    'missing': {'income': 'missing_mean'},
    'encoder': {'gender': 'encoder_onehot'}
})

# Final configuration includes all updates

Setting Configuration at Initialization

# Can also provide configuration when creating processor
custom_config = {
    'missing': {
        'age': 'missing_median',
        'income': 'missing_mean'
    },
    'encoder': {
        'gender': 'encoder_onehot'
    }
}

processor = Processor(metadata=schema, config=custom_config)
# No need to call update_config()

Verify Configuration Update

processor = Processor(metadata=schema)

# Update configuration
new_config = {
    'missing': {'age': 'missing_median'},
    'encoder': {'gender': 'encoder_onehot'}
}
processor.update_config(new_config)

# Verify update
config = processor.get_config(col=['age', 'gender'])
print("age missing:", type(config['missing']['age']).__name__)
print("gender encoder:", type(config['encoder']['gender']).__name__)

Available Processor Names

Missing Value Processors

missing_mean: Fill with mean
missing_median: Fill with median
missing_mode: Fill with mode
missing_simple: Fill with custom value (requires value parameter)
missing_drop: Drop rows with missing values

Outlier Processors

outlier_zscore: Z-Score method
outlier_iqr: Interquartile Range method
outlier_isolationforest: Isolation Forest (global)
outlier_lof: Local Outlier Factor (global)

Encoders

encoder_uniform: Uniform encoding
encoder_label: Label encoding
encoder_onehot: One-Hot encoding
encoder_date: Date format conversion (requires parameters)

Scalers

scaler_standard: Standardization
scaler_minmax: Min-Max scaling
scaler_zerocenter: Zero centering
scaler_log: Logarithmic transformation
scaler_log1p: log(1+x) transformation
scaler_timeanchor: Time anchor scaling (requires parameters)

Discretization

discretizing_kbins: K-bins discretization (requires parameters)

Notes

Updating configuration overwrites default settings for that column
Must call this method before fit()
If updating after fit(), need to retrain
Invalid processor names will raise ConfigError
Setting to None or "none" disables that processing
Processors with parameters must use dictionary format
Updates are cumulative, won’t clear other columns’ configuration

get_config()get_changes()