GitHub

MERA.jl Testing Framework

codecov (current development version coverage)

MERA.jl uses a sophisticated testing approach designed for scientific computing packages that handle large-scale astrophysical simulation data. We test locally on Julia 1.10 and 1.11 with heavy data due to the computational and memory requirements of realistic testing scenarios.

Why Local Testing with Heavy Data

Scientific computing packages require testing with production-scale data because:

  • Data Scale: Realistic RAMSES simulations contain millions of AMR cells across 10+ refinement levels
  • Physical Validity: Results must obey conservation laws (mass, momentum, energy)
  • Memory Requirements: Testing requires 16-32GB RAM for realistic datasets
  • Extended Runtime: Comprehensive testing takes 2-8 hours vs. CI limits of 1-2 hours

Testing Areas and Types

Test Categories

Unit Tests

  • Individual function validation
  • API compatibility across Julia versions
  • Data structure correctness
  • Parameter validation

Integration Tests

  • Data loading workflows (getinfo() → gethydro() → analysis)
  • Multi-component analysis (hydro/gravity/particle combinations)
  • I/O workflows (savedata/loaddata cycles)

End-to-End Tests

  • Complete scientific analysis pipelines
  • Projection operations with conservation law validation
  • Performance and memory management under load

Physical Validation Tests

  • Conservation laws (mass, momentum, energy preservation)
  • AMR boundary effects and grid hierarchy handling
  • Coordinate transformations and numerical precision

Performance Tests

  • Memory management with large datasets
  • Multi-threading consistency and efficiency
  • I/O optimization and cache performance

Quality Assurance Tests

  • Package health checks with Aqua.jl
  • Dependency compatibility verification
  • Code quality and consistency validation

Tutorial-Based Testing

Our documentation tutorials serve as comprehensive end-to-end tests:

  • Tutorial Source: Mera-Docs Jupyter Notebooks
  • Validation Method: Manual review of error-free execution with correct results and plots
  • Coverage: All major MERA.jl workflows and features
  • Stability Indicator: If tutorials run successfully, MERA.jl is stable for production use

Test Execution

CI Testing (Lightweight)

# Automated testing on GitHub Actions
MERA_SKIP_HEAVY=true julia --project=. -e "using Pkg; Pkg.test()"
  • Duration: 5-15 minutes
  • Purpose: Rapid development feedback
  • Coverage: Basic functionality, API compatibility

Local Comprehensive Testing

# Full validation with heavy data
MERA_LOCAL_COVERAGE=true julia --project=. -e "using Pkg; Pkg.test()"
  • Duration: 2-8 hours
  • Purpose: Complete validation before release
  • Data: Production RAMSES simulation outputs (2-50GB)
  • Julia Versions: Both 1.10 (LTS) and 1.11 (current)

Testing Philosophy

MERA.jl's testing ensures both software reliability and scientific validity through a combination of traditional software testing practices and scientific computing validation requirements. The multi-environment approach balances development efficiency with the rigorous validation needs of astrophysical simulation analysis software.