Redesing `evaluate` interface for dispatch, modularity and extensibility

[mvanzulli]

Motivation

The current package supports a variety of financial models (e.g., Black-Scholes, Heston, Vasicek) and instruments (e.g., European Options, American Options, FX Options), but tightly couples them via a shared evaluate function. This design limits flexibility and maintainability, making it difficult to:

1. Add new pricing models without changing existing instrument code
2. Add new financial instruments without altering model implementations
3. Handle diverse market data inputs in a consistent way
4. Introduce model calibration or validation workflows
5. Test components independently

Proposal: Abstract Interfaces

To decouple components and improve extensibility, we propose defining a set of abstract interfaces:

abstract type FinancialInstrument end
abstract type PricingModel end
abstract type PricingEngine end
abstract type MarketData end

Required Methods

All models and instruments should implement:

required_parameters(::Type{T}) where T <: Union{FinancialInstrument, PricingModel}
validate_parameters(x::Union{FinancialInstrument, PricingModel})
price(engine::PricingEngine, instrument::FinancialInstrument, model::PricingModel, market_data::MarketData)

Type Hierarchy

Instruments

• Options: EuropeanOption, AmericanOption, BarrierOption, BinaryOption, FXOption, etc.
• Forwards/Futures: EquityForward, FXForward, InterestRateFuture, etc.
• Swaps: InterestRateSwap, CrossCurrencySwap, EquitySwap
• Bonds: FixedRateBond, FloatingRateBond, ZeroCouponBond
• Credit Derivatives: CreditDefaultSwap, CreditSpreadOption

Market Data

• EquityMarketData
• InterestRateMarketData
• FXMarketData

Pricing Engines

• BlackScholesEngine
• HestonEngine
• BinomialEngine
• MonteCarloEngine
• FiniteDifferenceEngine

Models

• BlackScholesModel
• HestonModel
• BinomialModel
• CoxRossRubinsteinModel
• JarrowRuddModel
• LeisenReimerModel

Benefits

• Modularity: Models and instruments are decoupled, enabling easier maintenance and extension.

• Testability: Components can be tested independently.

• Consistency: Unified interface for handling market data and validation.

• Extensibility: New capabilities can be added easily, such as:

  • Model calibration
  • Sensitivity and risk analysis
  • Integration with external market data feeds
  • Support for custom pricing engines