Docs
Python API
Processor API (WIP)
get_changes()

get_changes()

Compare differences between current configuration and default configuration.

Syntax 

def get_changes() -> pd.DataFrame

Parameters

None

Returns

  • pd.DataFrame
    • DataFrame recording configuration differences
    • Columns: processor (processor type), col (column name), current (current config), default (default config)

Description

The get_changes() method is used to view which configurations have been customized. This method will:

  1. Compare current configuration with default configuration
  2. List all differences
  3. Return results in DataFrame format

This is very useful for debugging and documentation.

Basic Example 

from petsard import Loader, Processor

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

# Create processor and update configuration
processor = Processor(metadata=schema)
processor.update_config({
    'missing': {
        'age': 'missing_median',
        'income': 'missing_mean'
    },
    'encoder': {
        'gender': 'encoder_onehot'
    },
    'scaler': {
        'age': 'scaler_minmax'
    }
})

# View which configurations were modified
changes = processor.get_changes()
print(changes)

Output Example 

  processor      col           current         default
0   missing      age    MissingMedian     MissingMean
1   encoder   gender   EncoderOneHot  EncoderUniform
2    scaler      age    ScalerMinMax  ScalerStandard

Check Changes for Specific Processor Type 

processor = Processor(metadata=schema, config=custom_config)
changes = processor.get_changes()

# View only encoder changes
encoder_changes = changes[changes['processor'] == 'encoder']
print("Encoder changes:")
print(encoder_changes)

# View changes for specific column
age_changes = changes[changes['col'] == 'age']
print("\nChanges for age column:")
print(age_changes)

Count Number of Changes 

changes = processor.get_changes()

print(f"Total {len(changes)} configurations modified")
print(f"Modified processor types: {changes['processor'].unique()}")
print(f"Affected columns: {changes['col'].unique()}")

Export Change Records 

changes = processor.get_changes()

# Save as CSV
changes.to_csv('processor_config_changes.csv', index=False)

# Or save as Markdown table
print(changes.to_markdown(index=False))

Use Cases

1. Validate Custom Configuration 

# Ensure important columns use custom configuration
processor = Processor(metadata=schema, config=custom_config)
changes = processor.get_changes()

# Check if age uses median imputation
age_missing = changes[
    (changes['col'] == 'age') & 
    (changes['processor'] == 'missing')
]
assert 'Median' in age_missing['current'].values[0]
print("Age column configuration validation passed!")

2. Documentation 

# Record configuration used in experiment
processor = Processor(metadata=schema, config=experiment_config)
changes = processor.get_changes()

print("Experiment configuration record:")
print(f"Date: {datetime.now()}")
print(f"Number of modifications: {len(changes)}")
print("\nDetailed changes:")
print(changes.to_string(index=False))

3. Compare Different Configurations 

# Create two processors with different configurations
processor1 = Processor(metadata=schema, config=config1)
processor2 = Processor(metadata=schema, config=config2)

changes1 = processor1.get_changes()
changes2 = processor2.get_changes()

print("Configuration 1 changes:")
print(changes1)
print("\nConfiguration 2 changes:")
print(changes2)

4. Detect Unexpected Configuration Changes 

processor = Processor(metadata=schema)

# Perform some operations...
processor.update_config(some_config)

# Check for unexpected changes
changes = processor.get_changes()
unexpected_cols = ['critical_field1', 'critical_field2']

unexpected_changes = changes[changes['col'].isin(unexpected_cols)]
if not unexpected_changes.empty:
    print("Warning: Critical column configurations were modified!")
    print(unexpected_changes)

Output Format

DataFrame contains the following columns:

ColumnDescriptionExample
processorProcessor type'missing', 'encoder', 'scaler'
colColumn name'age', 'gender', 'income'
currentCurrently used processor'MissingMedian', 'EncoderOneHot'
defaultDefault processor'MissingMean', 'EncoderUniform'

Notes

  • Only displays items different from default configuration
  • Returns empty DataFrame if there are no changes
  • Can be called at any time, does not require calling fit() first
  • Both current and default configurations show class names (not instances)
  • If a column’s processor is set to None, it may also appear in the change list
  • Global transformation processors (e.g., outlier_isolationforest) affect display for all columns
update_config()