Test Coverage

Test Statistics

Current Test Status (November 2025):

• Total Tests: 607 tests
• Passing: 100%
• Execution Time: ~55 seconds

Test Organization

Test Classification

PETsARD test suite is organized by functional modules, using pytest markers for classification:

• unit: Unit tests for isolated component functionality
• integration: Integration tests for module interactions
• stress: Stress tests for large datasets and extreme scenarios
• excel: Excel file tests requiring openpyxl package

Test Directory Structure

tests/
├── test_petsard.py          # End-to-end workflow tests
├── test_executor.py          # Executor configuration & logging
├── test_adapter.py           # Configuration adapter tests
├── loader/                   # Data loading module
├── processor/                # Data processing module
├── metadater/               # Metadata management module
├── synthesizer/             # Synthesizer module
├── constrainer/             # Constrainer module
├── evaluator/               # Evaluator module
├── describer/               # Describer module
└── reporter/                # Reporter module

Core Functionality Tests

End-to-End Tests (test_petsard.py)

Tests complete PETsARD workflows:

• Basic Workflow: Load → Preprocess → Synthesize → Postprocess
• Data Preprocessing: Missing value handling, encoding, standardization
• Constraint Application: Numeric ranges, field proportion control
• Evaluation Metrics: Quality reports, statistical comparison
• Minimal Pipeline: Load and report only
• Error Handling: Invalid configuration detection

Executor (test_executor.py)

Tests Executor configuration and execution management:

• Configuration validation (log level, output type)
• Logger system initialization and reconfiguration
• YAML configuration loading
• Default value handling

Data Loading Tests

Loader (loader/test_loader.py)

Core Functionality:

• CSV/Excel file loading
• Custom NA values and header handling
• Data-Schema auto-reconciliation (field alignment, type inference)

Logical Type System:

• Text types: email, url, uuid, categorical, ip_address
• Numeric types: percentage, currency, latitude, longitude
• Identifier types: primary_key
• Type compatibility validation and conflict resolution

Schema Parameter System:

• Global parameters: compute_stats, optimize_dtypes, sample_size
• Field parameters: logical_type, leading_zeros
• Parameter conflict detection

Stress Tests (marked stress):

• 100MB - 5GB file processing
• Memory usage monitoring
• Type inference edge cases

Benchmarker (loader/test_benchmarker.py)

Tests benchmark dataset management:

• Dataset download and caching
• SHA-256 integrity verification (warns on mismatch)
• benchmark:// protocol handling
• LoaderAdapter integration

Splitter (loader/test_splitter.py)

Tests data splitting functionality:

• Train/test splitting
• Multiple sample generation
• Stratified sampling
• Random seed control

Data Processing Tests

Processor (processor/)

Missing Value Handlers (test_missing.py):

• MissingMean, MissingMedian, MissingSimple, MissingDrop
• Pandas nullable integer type support
• Banker’s rounding for integer types

Outlier Handlers:

• IQR, Z-score, Isolation Forest methods
• Outlier clipping or removal
• Pandas array compatibility

Encoders:

• Label, OneHot, Target encoding
• Automatic type inference

Scalers:

• Standard, MinMax, Robust scaling
• Numeric feature processing

Metadater (metadater/)

Tests three-layer metadata architecture:

Three-Layer Architecture:

• Metadata layer: Multi-table management
• Schema layer: Field definitions
• Attribute layer: Field properties

Core Functionality:

• Create metadata from data/configuration
• Difference comparison (diff)
• Data alignment (align)
• Statistics calculation
• YAML roundtrip compatibility

Evaluation & Reporting Tests

Describer (describer/)

DescriberDescribe (test_describer_describe.py):

• Statistical methods: basic, percentile, na, cardinality
• Granularity: global, columnwise
• Data types: numeric, categorical, mixed
• Edge cases: empty data, extreme values, high cardinality

DescriberCompare (test_describer_compare.py):

• JS Divergence calculation
• Code reuse architecture (reuses DescriberDescribe)
• NA value handling
• base/target parameter naming

Evaluator (evaluator/)

SDMetrics Integration (test_sdmetrics.py):

• Quality report evaluation
• Diagnostic report generation
• Single/multi-table scenarios
• Granularity control (global/columnwise/pairwise)

MPUCCS Privacy (test_mpuccs.py):

• Membership inference attack evaluation
• Privacy risk indicators
• Multi-class classification support

Custom Evaluator (test_custom_evaluator.py):

• Evaluator registration mechanism
• Custom evaluation logic
• Factory pattern integration

Constrainer (constrainer/)

Constraint Types:

• NaN Group Constraints (test_nan_group_constrainer.py):
  • Delete, Erase, Copy actions
  • Multi-field processing
• Field Constraints (test_field_constrainer.py):
  • Numeric ranges, conditional expressions
  • Complex logic combinations
  • String literal handling (with operators)
• Field Proportions (test_constrainer.py):
  • Category distribution maintenance
  • Missing value proportion control
  • Resampling mechanism

Reporter (reporter/)

ReporterSaveData:

• CSV/Excel/Pickle output
• Multi-format simultaneous saving
• Custom output paths

ReporterSaveSchema (test_reporter_save_schema.py):

• Schema CSV summary generation
• YAML detailed output
• Multi-source module support
• Schema flattening and structuring

Configuration Adapter Tests

Adapter (test_adapter.py)

Tests YAML configuration processing:

Three-Layer YAML Architecture:

• Module layer: Module type and name
• Experiment layer: Experiment configuration
• Parameters layer: Parameter details

Functionality Tests:

• benchmark:// protocol handling
• Postprocessor precision and rounding
• Multi-postprocessor sequences
• Pipeline data flow validation