Reporter

Reporter(method, **kwargs)

Generates output files for experiment results and evaluation reports.

Parameters

method (str): Report generation method
- ‘save_data’: Save dataset to CSV
  - Additional parameter required:
    - source (str | List[str]): Target module or experiment name
- ‘save_report’: Generate evaluation report
  - Additional parameters required:
    - granularity (str | List[str]): Report detail level
      - Single granularity: ‘global’, ‘columnwise’, ‘pairwise’, ‘details’, ’tree’
      - Multiple granularities: [‘global’, ‘columnwise’] or [‘details’, ’tree’]
    - eval (str | List[str], optional): Target evaluation experiment name
    - naming_strategy (str, optional): Output filename naming strategy
      - ’traditional’: Traditional format (default) - petsard[Report]_eval_[granularity].csv
      - ‘compact’: Compact format - petsard.report.Rp.eval.G.csv
- ‘save_timing’: Save timing information
  - Additional parameters optional:
    - time_unit (str): Time unit (‘seconds’, ‘minutes’, ‘hours’, ‘days’)
    - module (str | List[str]): Filter by specific modules
output (str, optional): Output filename prefix
- Default: ‘petsard’

Examples

from petsard.reporter import Reporter


# Save synthetic data
reporter = Reporter('save_data', source='Synthesizer')
reporter.create({('Synthesizer', 'exp1'): synthetic_df})
reporter.report()  # Creates: petsard_Synthesizer[exp1].csv

# Generate evaluation report (traditional naming)
reporter = Reporter('save_report', granularity='global', naming_strategy='traditional')
reporter.create({('Evaluator', 'eval1_[global]'): results})
reporter.report()  # Creates: petsard[Report]_[global].csv

# Generate evaluation report (compact naming)
reporter = Reporter('save_report', granularity='global', naming_strategy='compact')
reporter.create({('Evaluator', 'eval1_[global]'): results})
reporter.report()  # Creates: petsard.report.Rp.eval1.G.csv

# Generate evaluation report (multiple granularities)
reporter = Reporter('save_report', granularity=['global', 'columnwise'])
reporter.create({
    ('Evaluator', 'eval1_[global]'): global_results,
    ('Evaluator', 'eval1_[columnwise]'): columnwise_results
})
reporter.report()  # Creates multiple reports for each granularity

# Generate evaluation report with new granularity types
reporter = Reporter('save_report', granularity=['details', 'tree'])
reporter.create({
    ('Evaluator', 'eval1_[details]'): details_results,
    ('Evaluator', 'eval1_[tree]'): tree_results
})
reporter.report()  # Creates detailed and tree-structured reports

# Save timing information
reporter = Reporter('save_timing', time_unit='minutes', module=['Loader', 'Synthesizer'])
reporter.create({'timing_data': timing_df})
reporter.report()  # Creates: petsard_timing_report.csv

Methods

`create(data)`

Initialize reporter with data using functional design pattern.

Parameters

data (dict): Report data where:
- Keys: Experiment tuples (module_name, experiment_name, …)
- Values: Data to be reported (pd.DataFrame)
- Optional ’exist_report’ key for merging with previous results
- For save_timing: ’timing_data’ key with timing DataFrame

Returns

dict | pd.DataFrame | None: Processed data ready for reporting
- For save_data: Dictionary of processed DataFrames
- For save_report: Dictionary with granularity-specific results
- For save_timing: DataFrame with timing information
- Returns None if no data to process

`report(processed_data)`

Generate and save report as CSV using functional design pattern.

Parameters

processed_data: Output from create() method

Output filename formats:

For save_data: {output}_{module-expt_name-pairs}.csv
For save_report (traditional): {output}[Report]_{eval}_[{granularity}].csv
For save_report (compact): {output}.report.{module_abbrev}.{eval}.{granularity_abbrev}.csv
For save_timing: {output}_timing_report.csv

Granularity Types

Traditional Granularities

global: Overall summary statistics
columnwise: Column-by-column analysis
pairwise: Pairwise column relationships

New Granularity Types (v2.0+)

details: Detailed breakdown with additional metrics
tree: Hierarchical tree-structured analysis

Multi-Granularity Support

The Reporter now supports processing multiple granularities in a single operation:

# Process multiple granularities simultaneously
reporter = Reporter('save_report', granularity=['global', 'columnwise', 'details'])
result = reporter.create(evaluation_data)
reporter.report(result)  # Generates separate reports for each granularity

Naming Strategies

Traditional Naming (Default)

Format: petsard[Report]_{eval}_[{granularity}].csv
Example: petsard[Report]_quality_eval_[global].csv
Maintains backward compatibility with existing workflows

Compact Naming

Format: petsard.report.{module}.{eval}.{granularity}.csv
Module abbreviations: Synthesizer→Sy, Evaluator→Ev, Reporter→Rp, etc.
Granularity abbreviations: global→G, columnwise→C, pairwise→P, details→D, tree→T
Example: petsard.report.Ev.quality_eval.G.csv
Provides cleaner, more readable filenames

# Traditional naming (default)
reporter = Reporter('save_report', granularity='global')
# Output: petsard[Report]_[global].csv

# Compact naming
reporter = Reporter('save_report', granularity='global', naming_strategy='compact')
# Output: petsard.report.Rp.eval.G.csv

Functional Design

The Reporter uses a functional “throw out and throw back in” design pattern:

create() processes data without storing it in instance variables
report() takes the processed data and generates output files
No internal state is maintained, reducing memory usage
Supports flexible naming strategies for different use cases

Describer Adapter