optionstratlib_simulation/simulation/

simulator.rs

use crate::simulation::WalkParams;
use crate::simulation::randomwalk::RandomWalk;
use crate::simulation::steps::Step;
use optionstratlib_core::model::Options;
use optionstratlib_core::model::Positive;
use optionstratlib_core::utils::{Len, deterministic_rng};
use optionstratlib_pricing::error::PricingError;
use optionstratlib_pricing::pricing::Profit;
use optionstratlib_pricing::pricing::monte_carlo::price_option_monte_carlo;
use optionstratlib_pricing::pricing::unified::MonteCarloPricer;
use rand::RngExt;
use rayon::prelude::*;
use rust_decimal::Decimal;
use std::fmt::Display;
use std::ops::{AddAssign, Index, IndexMut};
use tracing::warn;

/// Number of walks from which [`Simulator::new`] builds them on the rayon
/// pool rather than serially (#860). Measured on an Apple M5 Max (18 cores)
/// with 30-step geometric Brownian walks: one or two walks build no faster on
/// the pool than serially, three already build faster. Longer walks only
/// move the crossover lower.
pub const PARALLEL_MIN_WALKS: usize = 3;

/// Walks [`Simulator::new`] builds per round on the rayon pool. Each round
/// draws its walks' seeds in order and returns at its first failing walk, so
/// a huge `size` never holds more than one round of seeds and results, and a
/// failing build stops within a round of its first error.
const PARALLEL_ROUND_WALKS: usize = 1024;

/// Represents a generic simulator for managing and simulating random walks.
///
/// # Type Parameters
/// * `X`: A type that represents the state or value within the random walk. It must adhere to the following bounds:
///    - `Copy`: Allows for efficient copying of values.
///    - `TryInto<Positive>`: Ensures values can be converted into a `Positive` type (potentially for validation or numerical operations).
///    - `AddAssign`: Allows addition and assignment (`+=`) operations.
///    - `Display`: Enables the formatting of values as strings for user-facing output.
///
/// * `Y`: A type that represents the step or transition within the random walk. It must adhere to the following bounds:
///    - `TryInto<Positive>`: Ensures values can be converted into a `Positive` type.
///    - `Display`: Enables the formatting of values as strings for user-facing output.
///    - `Clone`: Allows for creating deep copies of the values.
///
/// # Fields
/// * `title` (`String`): The name or description of the simulator, primarily used for identification or display purposes.
/// * `random_walks` (`Vec<RandomWalk<X, Y>>`): A collection of `RandomWalk` instances, where each random walk adheres to the defined types `X` and `Y`.
///
/// # Usage
/// This struct is used as a high-level container to manage multiple random walks and perform simulations. Adding specific
/// functionality such as initializing, running simulations, or generating statistical data depends on additional methods provided
/// separately.
///
/// Note: This struct is generic and requires types provided for both state (`X`) and step/transition (`Y`) that meet the respective
/// trait bounds.
#[derive(Debug, Clone)]
pub struct Simulator<X, Y>
where
    X: Copy + TryInto<Positive> + AddAssign + Display,
    Y: TryInto<Positive> + Display + Clone,
{
    title: String,
    random_walks: Vec<RandomWalk<X, Y>>,
}

impl<X, Y> Simulator<X, Y>
where
    X: Copy + TryInto<Positive> + AddAssign + Display,
    Y: TryInto<Positive> + Display + Clone,
{
    /// Creates a new simulator that builds `size` random walks from the
    /// supplied fallible generator.
    ///
    /// The generator is invoked once per random walk; if any walk fails, no
    /// partial `Simulator` is returned and the error is the one of the
    /// lowest-indexed failing walk.
    ///
    /// From [`PARALLEL_MIN_WALKS`] walks up, the walks are built on the rayon
    /// pool and collected in index order (#860). The seed of every walk is
    /// drawn up front, in order, from the master seed, so a seeded simulator
    /// is bit-identical to the serial build whatever the thread count. Below
    /// the threshold the walks are built serially, which is faster there.
    /// The serial build stops at the first failing walk; the parallel build
    /// works in rounds of 1024 walks and stops at the end of the round that
    /// holds the first failing walk, whose error it returns.
    ///
    /// # Parameters
    ///
    /// * `title` - A descriptive title; individual walks are titled
    ///   `"{title}_{i}"`.
    /// * `size` - Number of random walks to generate.
    /// * `params` - Walk parameters shared across all generated walks.
    ///   When `params.seed` is `Some(seed)`, walk `i` receives the `i`-th
    ///   `u64` drawn from
    ///   [`optionstratlib_core::utils::deterministic_rng`]`(seed)` as its own
    ///   seed: the walks differ from one another, and the whole simulator is
    ///   reproducible bit for bit. With `None` every walk draws from the
    ///   thread RNG.
    /// * `generator` - A fallible step generator. Cloned per walk; pass a
    ///   function pointer or a stateless closure for best ergonomics.
    ///
    /// # Returns
    ///
    /// `Ok(Simulator)` on success, or `Err(E)` from the lowest-indexed walk
    /// whose generator invocation fails.
    ///
    /// # Errors
    ///
    /// Returns the error type produced by the supplied generator. For
    /// chain-backed generators (e.g. [`crate::simulation::generator_positive`])
    /// this is [`crate::error::SimulationError`].
    pub fn new<F, E>(
        title: String,
        size: usize,
        params: &WalkParams<X, Y>,
        generator: F,
    ) -> Result<Self, E>
    where
        F: Fn(&WalkParams<X, Y>) -> Result<Vec<Step<X, Y>>, E> + Clone + Send + Sync,
        E: Send,
        X: Copy + TryInto<Positive> + AddAssign + Display + Send + Sync,
        Y: TryInto<Positive> + Display + Clone + Send + Sync,
    {
        // The reservation is only a growth hint, and `E` is the generator's
        // error type, which cannot carry an allocation failure. A `size` whose
        // walks cannot be reserved up front used to abort in
        // `Vec::with_capacity`; it now grows walk by walk, and each walk's own
        // allocation reports through the generator (#788).
        let mut random_walks = Vec::new();
        if let Err(e) = random_walks.try_reserve_exact(size) {
            warn!(
                size,
                error = %e,
                "Simulator::new: cannot reserve the walks up front; growing on demand"
            );
        }
        // Walk `i` is seeded with the `i`-th `u64` of
        // `deterministic_rng(seed)`: distinct paths, reproducible run to run,
        // and independent of how many walks follow. The seeds are drawn in
        // order before any walk is built, so the parallel build below gives
        // every walk the seed the serial build gives it.
        let mut seeds = params.seed.map(deterministic_rng);
        if size < PARALLEL_MIN_WALKS {
            let mut walk_params = params.clone();
            for i in 0..size {
                if let Some(seeds) = seeds.as_mut() {
                    walk_params.seed = Some(seeds.random::<u64>());
                }
                let walk_title = format!("{title}_{i}");
                random_walks.push(RandomWalk::new(
                    walk_title,
                    &walk_params,
                    generator.clone(),
                )?);
            }
        } else {
            // One round at a time: its seeds are drawn in order, its walks
            // are built on the pool, and they are appended in index order up
            // to the first error. Memory beyond the walks themselves stays
            // bounded by the round, and a failing build stops after the round
            // that holds its first failure.
            let mut indices = 0..size;
            loop {
                let jobs: Vec<(usize, Option<u64>)> = indices
                    .by_ref()
                    .take(PARALLEL_ROUND_WALKS)
                    .map(|i| (i, seeds.as_mut().map(|seeds| seeds.random::<u64>())))
                    .collect();
                if jobs.is_empty() {
                    break;
                }
                let built: Vec<Result<RandomWalk<X, Y>, E>> = jobs
                    .into_par_iter()
                    .map_init(
                        || params.clone(),
                        |walk_params, (i, seed)| {
                            if seed.is_some() {
                                walk_params.seed = seed;
                            }
                            RandomWalk::new(format!("{title}_{i}"), walk_params, generator.clone())
                        },
                    )
                    .collect();
                for walk in built {
                    random_walks.push(walk?);
                }
            }
        }
        Ok(Self {
            title,
            random_walks,
        })
    }

    /// Returns the title of the random walk.
    ///
    /// # Returns
    ///
    /// A string slice containing the title of the random walk.
    #[must_use]
    pub fn get_title(&self) -> &str {
        &self.title
    }

    /// Updates the title of the random walk.
    ///
    /// # Parameters
    ///
    /// * `title` - The new title to set
    pub fn set_title(&mut self, title: String) {
        self.title = title;
    }

    /// Retrieves the steps of the random walks contained within the current object.
    ///
    /// This method returns a vector of references to `RandomWalk` instances stored
    /// in the `random_walks` collection member of the struct. Each `RandomWalk`
    /// instance represents a step in the random walk process.
    ///
    /// # Returns
    ///
    /// A `Vec` containing references to `RandomWalk<X, Y>` values, where
    /// `X` and `Y` are the types used within the random walk structure.
    ///
    /// # Note
    ///
    /// The returned vector contains borrowed references to the `RandomWalk`
    /// elements within the struct, and the lifetime of these references
    /// is tied to the lifetime of the parent object.
    #[must_use]
    pub fn get_random_walks(&self) -> Vec<&RandomWalk<X, Y>> {
        self.random_walks.iter().collect::<Vec<&RandomWalk<X, Y>>>()
    }

    /// Retrieves a reference to the `RandomWalk` at the specified index.
    ///
    /// # Arguments
    ///
    /// * `index` - The index of the desired `RandomWalk` within the `random_walks` collection.
    ///
    /// # Returns
    ///
    /// * `Some(&RandomWalk<X, Y>)` - The walk at `index`.
    /// * `None` - If `index` is out of bounds.
    ///
    /// Named `get_*`, so it follows [`slice::get`] rather than
    /// [`std::ops::Index`]: an out-of-bounds index is reported, not aborted
    /// on. Use the `simulator[index]` operator when the index is an invariant
    /// of the caller and a panic is the wanted contract.
    #[must_use]
    pub fn get_random_walk(&self, index: usize) -> Option<&RandomWalk<X, Y>> {
        self.random_walks.get(index)
    }

    /// Retrieves a mutable reference to a `RandomWalk` at the specified index.
    ///
    /// # Arguments
    ///
    /// * `index` - The zero-based index of the `RandomWalk` to access within the `random_walks` collection.
    ///
    /// # Returns
    ///
    /// * `Some(&mut RandomWalk<X, Y>)` - The walk at `index`.
    /// * `None` - If `index` is out of bounds.
    ///
    /// Mutable counterpart of [`Simulator::get_random_walk`], with the same
    /// out-of-bounds contract.
    pub fn get_random_walk_mut(&mut self, index: usize) -> Option<&mut RandomWalk<X, Y>> {
        self.random_walks.get_mut(index)
    }

    /// Returns a reference to the first `RandomWalk` element in the `random_walks` collection, if it exists.
    ///
    /// # Returns
    /// - `Some(&RandomWalk<X, Y>)` if the `random_walks` collection is not empty.
    /// - `None` if the `random_walks` collection is empty.
    ///
    #[must_use]
    pub fn first(&self) -> Option<&RandomWalk<X, Y>> {
        self.random_walks.first()
    }

    /// Returns the last random walk in the collection, if it exists.
    ///
    /// # Returns
    /// - `Some(&RandomWalk<X, Y>)`: A reference to the last `RandomWalk` in the collection.
    /// - `None`: If the collection is empty.
    ///
    /// # Note
    /// The `last` method does not consume the collection; it returns a read-only reference to the last element.
    #[must_use]
    pub fn last(&self) -> Option<&RandomWalk<X, Y>> {
        self.random_walks.last()
    }

    /// Retrieves a nested vector of references to `Step<X, Y>` objects.
    ///
    /// This function iterates over the elements of the current container (`self`)
    /// assuming it implements `IntoIterator`, and for each element,
    /// calls its `get_steps` method. The results are then collected into a
    /// two-dimensional `Vec` structure.
    ///
    /// # Returns
    /// A `Vec` where each inner vector contains references to `Step<X, Y>` objects.
    ///
    /// # Type Parameters
    /// - `X`: The type of the first generic parameter in `Step`.
    /// - `Y`: The type of the second generic parameter in `Step`.
    ///
    #[must_use]
    pub fn get_steps(&self) -> Vec<Vec<&Step<X, Y>>> {
        self.into_iter().map(|step| step.get_steps()).collect()
    }

    /// Returns a `Vec` containing the last `Step` of each item in the iterator.
    ///
    /// This method assumes that each item in the iterator is an iterable itself.
    /// It retrieves the last element of each iterable and collects them into a new `Vec`.
    ///
    /// # Returns
    /// A vector of references to the last `Step<X, Y>` of each non-empty
    /// random walk. Empty walks are silently skipped, so the returned
    /// vector may be shorter than `self.len()`.
    #[must_use]
    pub fn get_last_steps(&self) -> Vec<&Step<X, Y>> {
        self.into_iter().filter_map(|step| step.last()).collect()
    }

    /// Retrieves the last value of each non-empty random walk.
    ///
    /// # Returns
    /// A `Vec` of references to the last `Step<X, Y>` of each non-empty
    /// random walk. Empty walks are silently skipped, so the returned
    /// vector may be shorter than `self.len()`.
    #[must_use]
    pub fn get_last_values(&self) -> Vec<&Step<X, Y>> {
        self.into_iter().filter_map(|step| step.last()).collect()
    }

    /// Retrieves the last set of positive values from the internal state.
    ///
    /// This method extracts the positive values from the most recent set of steps retrieved
    /// by the `last_values` method and returns them as a vector of `Positive` items.
    ///
    /// # Returns
    /// * `Vec<Positive>` - A vector containing the last positive values derived from the steps.
    ///
    /// # Notes
    /// * The `last_values` method is called internally to obtain the most recent set of steps.
    /// * The positive value for each step is retrieved via the `get_positive_value` method;
    ///   a step whose value is not a valid `Positive` is left out of the result.
    #[must_use]
    pub fn get_last_positive_values(&self) -> Vec<Positive> {
        let last_values = self.get_last_values();
        last_values
            .iter()
            .filter_map(|step| step.get_positive_value().ok())
            .collect::<Vec<Positive>>()
    }

    /// Calculates the price of a financial option using Monte Carlo simulation.
    ///
    /// This method computes the price of the provided `option` based on a Monte Carlo
    /// simulation approach. It retrieves the most recent positive values of the
    /// underlying asset, which are then used in the simulation to estimate the option's price.
    ///
    /// # Arguments
    /// * `option` - A reference to an `Options` object, representing the financial option
    ///   to be priced.
    ///
    /// # Returns
    /// * If successful, it returns a `Positive` value wrapped in `Ok`, which represents
    ///   the computed price of the option.
    /// * If an error occurs, it returns a `Box<dyn Error>` wrapped in `Err`, indicating
    ///   the failure during the pricing process.
    ///
    /// # Errors
    /// This function will return an error if:
    /// * Retrieving the last positive values fails.
    /// * The Monte Carlo pricing function (`price_option_monte_carlo`) encounters an issue
    ///   during execution.
    ///
    /// # Note
    /// The implementation assumes that the underlying asset's most recent positive values
    /// are available and meaningful for Monte Carlo simulation. Ensure that the input data
    /// and the `option` are valid before invoking this method.
    pub fn get_mc_option_price(&self, option: &Options) -> Result<Positive, PricingError> {
        let last_values = self.get_last_positive_values();
        price_option_monte_carlo(option, &last_values)
    }
}

impl MonteCarloPricer for Simulator<Positive, Positive> {
    /// Delegates to [`Simulator::get_mc_option_price`]: the option is
    /// priced from the terminal value of every walk the simulator holds.
    /// Any failure is reported as [`PricingError::SimulationError`], as
    /// the `MonteCarlo` arm of `price_option` has always done.
    #[inline]
    fn price_monte_carlo(&self, option: &Options) -> Result<Positive, PricingError> {
        self.get_mc_option_price(option)
            .map_err(|e| PricingError::simulation_error(&e.to_string()))
    }
}

impl<X, Y> Len for Simulator<X, Y>
where
    X: Copy + TryInto<Positive> + AddAssign + Display,
    Y: TryInto<Positive> + Display + Clone,
{
    /// Returns the number of elements in the `random_walks` collection.
    ///
    /// # Returns
    /// - `usize`: The total count of elements in the `random_walks` collection.
    ///
    /// This method is typically used when you need to determine the size
    /// of the internal `random_walks` data structure.
    fn len(&self) -> usize {
        self.random_walks.len()
    }

    /// Checks if the `random_walks` collection is empty.
    ///
    /// # Returns
    /// * `true` - If the `random_walks` collection contains no elements.
    /// * `false` - If the `random_walks` collection contains one or more elements.
    ///
    fn is_empty(&self) -> bool {
        self.random_walks.is_empty()
    }
}

impl<X, Y> Index<usize> for Simulator<X, Y>
where
    X: Copy + TryInto<Positive> + AddAssign + Display,
    Y: TryInto<Positive> + Display + Clone,
{
    /// Defines an alias `Output` for the type `RandomWalk<X, Y>`.
    ///
    /// # Type Parameters
    /// - `X`: Represents the type of the first parameter used in the `RandomWalk`.
    /// - `Y`: Represents the type of the second parameter used in the `RandomWalk`.
    ///
    /// `Output` can be used as a shorthand to refer to a `RandomWalk` instance
    /// with specific `X` and `Y` types, improving code readability and reducing
    /// verbosity in the type definitions or method signatures.
    type Output = RandomWalk<X, Y>;

    /// Retrieves a reference to the element at the specified index in the `random_walks` vector.
    ///
    /// # Parameters
    /// - `index`: The zero-based index of the element to retrieve from the `random_walks` vector.
    ///
    /// # Returns
    /// A reference to the element at the specified `index` in the `random_walks` vector.
    ///
    /// # Panics
    /// This function will panic if the given `index` is out of bounds, i.e., greater than or equal to
    /// the length of the `random_walks` vector, as indexing a `Vec` does. Use
    /// [`Simulator::get_random_walk`] for the checked form, which returns `None` instead.
    ///
    /// Note: This implementation assumes that `Self` implements the `Index` trait and
    /// that `random_walks` is a field in the implementing struct.
    #[allow(clippy::indexing_slicing)]
    fn index(&self, index: usize) -> &Self::Output {
        // `Index` / `IndexMut` follow the std contract: an out-of-bounds index
        // is a caller bug and panics, exactly like `Vec`. The checked
        // accessor is `get_random_walk`, which returns `None` instead (#341, #788).
        &self.random_walks[index]
    }
}

impl<X, Y> IndexMut<usize> for Simulator<X, Y>
where
    X: Copy + TryInto<Positive> + AddAssign + Display,
    Y: TryInto<Positive> + Display + Clone,
{
    /// Retrieves a mutable reference to the walk at `index`.
    ///
    /// # Panics
    ///
    /// Panics if `index` is out of bounds, as indexing a `Vec` does. Use
    /// [`Simulator::get_random_walk_mut`] for the checked form, which returns
    /// `None` instead.
    #[allow(clippy::indexing_slicing)]
    fn index_mut(&mut self, index: usize) -> &mut Self::Output {
        // `Index` / `IndexMut` follow the std contract: an out-of-bounds index
        // is a caller bug and panics, exactly like `Vec`. The checked
        // accessor is `get_random_walk_mut`, which returns `None` instead (#341, #788).
        &mut self.random_walks[index]
    }
}

impl<X, Y> Display for Simulator<X, Y>
where
    X: Copy + TryInto<Positive> + AddAssign + Display,
    Y: TryInto<Positive> + Display + Clone,
{
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        writeln!(f, "{}", self.title)?;
        for random_walk in &self.random_walks {
            writeln!(f, "\t{random_walk}")?;
        }
        Ok(())
    }
}

impl<X, Y> Profit for Simulator<X, Y>
where
    X: AddAssign + Copy + Display + TryInto<Positive>,
    Y: Clone + Display + TryInto<Positive>,
{
    fn calculate_profit_at(&self, _price: &Positive) -> Result<Decimal, PricingError> {
        Err(PricingError::other(
            "Profit calculation not implemented for Simulator",
        ))
    }
}

impl<'a, X, Y> IntoIterator for &'a Simulator<X, Y>
where
    X: Copy + TryInto<Positive> + AddAssign + Display,
    Y: TryInto<Positive> + Display + Clone,
{
    type Item = &'a RandomWalk<X, Y>;
    type IntoIter = std::slice::Iter<'a, RandomWalk<X, Y>>;

    fn into_iter(self) -> Self::IntoIter {
        self.random_walks.iter()
    }
}

#[cfg(test)]
#[allow(irrefutable_let_patterns)]
mod tests {
    use super::*;
    use crate::error::SimulationError;
    use crate::simulation::generator_positive;
    use crate::simulation::{
        WalkParams, WalkType, WalkTypeAble,
        steps::{Step, Xstep, Ystep},
    };
    use optionstratlib_core::model::ExpirationDate;
    use optionstratlib_core::pos_or_panic;
    use optionstratlib_core::utils::{TimeFrame, time::convert_time_frame};
    use optionstratlib_pricing::pricing::OptionPricing;
    use rust_decimal_macros::dec;
    use std::convert::Infallible;
    use tracing::{debug, info};

    /// Seed of the simulated walks, so every run draws the same stream.
    const SEED: u64 = 685;

    // Helper structs and functions for testing
    #[derive(Clone)]
    struct TestWalker;

    impl TestWalker {
        fn new() -> Self {
            TestWalker {}
        }
    }
    impl WalkTypeAble<Positive, Positive> for TestWalker {}

    fn test_generator(
        params: &WalkParams<Positive, Positive>,
    ) -> Result<Vec<Step<Positive, Positive>>, Infallible> {
        Ok(vec![params.init_step.clone()])
    }

    /// The engine prices through `MonteCarloPricer` exactly what the
    /// simulator produces from the same paths, whether the engine owns the
    /// simulator or borrows it.
    #[test]
    fn test_monte_carlo_pricer_matches_the_simulator_on_the_same_paths() {
        use optionstratlib_core::model::types::{OptionStyle, OptionType, Side};
        use optionstratlib_pricing::pricing::{GenericPricingEngine, price_option_with};

        let prices: Vec<Positive> = (0..12)
            .map(|i| pos_or_panic!(100.0 + f64::from(i) * 1.5))
            .collect();
        let params = WalkParams {
            size: prices.len(),
            init_step: Step::new(
                Positive::ONE,
                TimeFrame::Day,
                ExpirationDate::Days(pos_or_panic!(30.0)),
                Positive::HUNDRED,
            ),
            walker: Box::new(TestWalker),
            walk_type: WalkType::Historical {
                timeframe: TimeFrame::Day,
                prices,
                symbol: None,
            },
            seed: None,
        };
        let simulator =
            Simulator::new("MC parity".to_string(), 8, &params, generator_positive).unwrap();
        let option = Options {
            option_type: OptionType::European,
            side: Side::Long,
            underlying_symbol: "TEST".to_string(),
            strike_price: Positive::HUNDRED,
            expiration_date: ExpirationDate::Days(pos_or_panic!(30.0)),
            implied_volatility: pos_or_panic!(0.2),
            quantity: Positive::ONE,
            contract_size: Positive::ONE,
            underlying_price: Positive::HUNDRED,
            risk_free_rate: dec!(0.05),
            option_style: OptionStyle::Call,
            dividend_yield: Positive::ZERO,
            exotic_params: None,
        };

        let expected = simulator.get_mc_option_price(&option).unwrap();
        assert!(expected > Positive::ZERO);
        assert_eq!(expected, simulator.price_monte_carlo(&option).unwrap());

        let generic = GenericPricingEngine::MonteCarlo {
            simulator: simulator.clone(),
        };
        assert_eq!(price_option_with(&option, &generic).unwrap(), expected);

        let borrowed = GenericPricingEngine::MonteCarlo {
            simulator: &simulator,
        };
        assert_eq!(price_option_with(&option, &borrowed).unwrap(), expected);
    }

    /// Moved from `pricing::monte_carlo` (#508): the Monte Carlo price
    /// from the terminal values of a simulated year of paths lands near
    /// the Black-Scholes price.
    #[test]
    fn test_monte_carlo_price_from_simulator_matches_black_scholes()
    -> Result<(), Box<dyn std::error::Error>> {
        use optionstratlib_core::assert_pos_relative_eq;
        use optionstratlib_core::model::utils::create_sample_option;
        use optionstratlib_core::model::{OptionStyle, Side};
        use optionstratlib_pricing::pricing::monte_carlo::price_option_monte_carlo;

        let walker = Box::new(TestWalker);
        let initial_price = pos_or_panic!(1000.0);
        let days = pos_or_panic!(365.0);
        let volatility = pos_or_panic!(0.2);
        let mut option = create_sample_option(
            OptionStyle::Call,
            Side::Long,
            initial_price,
            Positive::ONE,
            initial_price,
            volatility,
        );
        option.risk_free_rate = dec!(0.05);
        option.dividend_yield = pos_or_panic!(0.02);
        option.expiration_date = ExpirationDate::Days(days);

        let init_step = Step {
            x: Xstep::new(Positive::ONE, TimeFrame::Day, ExpirationDate::Days(days)),
            y: Ystep::new(0, initial_price),
        };

        let dt = convert_time_frame(Positive::ONE, &TimeFrame::Day, &TimeFrame::Year)?;
        let walk_params = WalkParams {
            size: 365,
            init_step,
            walk_type: WalkType::Custom {
                dt,
                drift: dec!(0.02),
                volatility,
                vov: pos_or_panic!(0.01),
                vol_speed: Default::default(),
                vol_mean: pos_or_panic!(0.2),
            },
            walker,
            seed: Some(SEED),
        };

        let Ok(simulator) = Simulator::new(
            "Test Simulator".to_string(),
            100,
            &walk_params,
            generator_positive,
        ) else {
            panic!("simulator setup failed");
        };

        let get_last_positive_values = simulator.get_last_positive_values();

        let result = price_option_monte_carlo(&option, &get_last_positive_values);
        assert!(result.is_ok());

        let bs = option.calculate_price_black_scholes().unwrap();
        assert_pos_relative_eq!(
            result.unwrap(),
            Positive::new_decimal(bs).unwrap(),
            pos_or_panic!(10.0)
        );
        Ok(())
    }

    // Test Simulator creation
    #[test]
    fn test_simulator_creation() -> Result<(), Box<dyn std::error::Error>> {
        let walker = Box::new(TestWalker);
        let initial_price = Positive::HUNDRED;
        let init_step = Step {
            x: Xstep::new(
                Positive::ONE,
                TimeFrame::Minute,
                ExpirationDate::Days(pos_or_panic!(30.0)),
            ),
            y: Ystep::new(0, initial_price),
        };

        let walk_params = WalkParams {
            size: 5,
            init_step,
            walk_type: WalkType::GeometricBrownian {
                dt: convert_time_frame(
                    Positive::ONE / pos_or_panic!(30.0),
                    &TimeFrame::Minute,
                    &TimeFrame::Day,
                )?,
                drift: dec!(0.0),
                volatility: pos_or_panic!(0.2),
            },
            walker,
            seed: Some(SEED),
        };

        let Ok(simulator) = Simulator::new(
            "Test Simulator".to_string(),
            5,
            &walk_params,
            test_generator,
        ) else {
            unreachable!()
        };

        assert_eq!(simulator.get_title(), "Test Simulator");
        assert_eq!(simulator.len(), 5);
        assert!(!simulator.is_empty());
        Ok(())
    }

    // Test title methods
    #[test]
    fn test_simulator_title_methods() -> Result<(), Box<dyn std::error::Error>> {
        let walker = Box::new(TestWalker);
        let initial_price = Positive::HUNDRED;
        let init_step = Step {
            x: Xstep::new(
                Positive::ONE,
                TimeFrame::Minute,
                ExpirationDate::Days(pos_or_panic!(30.0)),
            ),
            y: Ystep::new(0, initial_price),
        };

        let walk_params = WalkParams {
            size: 3,
            init_step,
            walk_type: WalkType::GeometricBrownian {
                dt: convert_time_frame(
                    Positive::ONE / pos_or_panic!(30.0),
                    &TimeFrame::Minute,
                    &TimeFrame::Day,
                )?,
                drift: dec!(0.0),
                volatility: pos_or_panic!(0.2),
            },
            walker,
            seed: Some(SEED),
        };

        let Ok(mut simulator) = Simulator::new(
            "Original Title".to_string(),
            3,
            &walk_params,
            test_generator,
        ) else {
            unreachable!()
        };

        assert_eq!(simulator.get_title(), "Original Title");

        simulator.set_title("New Title".to_string());
        assert_eq!(simulator.get_title(), "New Title");
        Ok(())
    }

    // Test step access methods
    #[test]
    fn test_simulator_step_access() -> Result<(), Box<dyn std::error::Error>> {
        let walker = Box::new(TestWalker);
        let initial_price = Positive::HUNDRED;
        let init_step = Step {
            x: Xstep::new(
                Positive::ONE,
                TimeFrame::Minute,
                ExpirationDate::Days(pos_or_panic!(30.0)),
            ),
            y: Ystep::new(0, initial_price),
        };

        let walk_params = WalkParams {
            size: 3,
            init_step,
            walk_type: WalkType::GeometricBrownian {
                dt: convert_time_frame(
                    Positive::ONE / pos_or_panic!(30.0),
                    &TimeFrame::Minute,
                    &TimeFrame::Day,
                )?,
                drift: dec!(0.0),
                volatility: pos_or_panic!(0.2),
            },
            walker,
            seed: Some(SEED),
        };

        let Ok(simulator) = Simulator::new(
            "Test Simulator".to_string(),
            3,
            &walk_params,
            test_generator,
        ) else {
            unreachable!()
        };

        // Test get_steps
        let steps = simulator.get_random_walks();
        assert_eq!(steps.len(), 3);

        // Test get_step
        let step = simulator.get_random_walk(1).unwrap();
        assert_eq!(step.get_title(), "Test Simulator_1");
        // Out of bounds is reported, not aborted on; the panicking contract
        // lives on the `Index` operator, covered by
        // `test_simulator_index_out_of_bounds`.
        assert!(simulator.get_random_walk(99).is_none());

        // Test first and last
        assert!(simulator.first().is_some());
        assert!(simulator.last().is_some());
        assert_eq!(
            simulator.first().expect("should be Ok").get_title(),
            "Test Simulator_0"
        );
        assert_eq!(
            simulator.last().expect("should be Ok").get_title(),
            "Test Simulator_2"
        );
        Ok(())
    }

    // Test Index and IndexMut traits
    #[test]
    fn test_simulator_indexing() -> Result<(), Box<dyn std::error::Error>> {
        let walker = Box::new(TestWalker);
        let initial_price = Positive::HUNDRED;
        let init_step = Step {
            x: Xstep::new(
                Positive::ONE,
                TimeFrame::Minute,
                ExpirationDate::Days(pos_or_panic!(30.0)),
            ),
            y: Ystep::new(0, initial_price),
        };

        let walk_params = WalkParams {
            size: 3,
            init_step,
            walk_type: WalkType::GeometricBrownian {
                dt: convert_time_frame(
                    Positive::ONE / pos_or_panic!(30.0),
                    &TimeFrame::Minute,
                    &TimeFrame::Day,
                )?,
                drift: dec!(0.0),
                volatility: pos_or_panic!(0.2),
            },
            walker,
            seed: Some(SEED),
        };

        let Ok(mut simulator) = Simulator::new(
            "Test Simulator".to_string(),
            3,
            &walk_params,
            test_generator,
        ) else {
            unreachable!()
        };

        // Test immutable indexing
        assert_eq!(simulator[0].get_title(), "Test Simulator_0");
        assert_eq!(simulator[1].get_title(), "Test Simulator_1");
        assert_eq!(simulator[2].get_title(), "Test Simulator_2");

        // Test mutable indexing
        simulator[1].set_title("Modified Title".to_string());
        assert_eq!(simulator[1].get_title(), "Modified Title");
        Ok(())
    }

    // Test display formatting
    #[test]
    fn test_simulator_display() -> Result<(), Box<dyn std::error::Error>> {
        let walker = Box::new(TestWalker);
        let initial_price = Positive::HUNDRED;
        let init_step = Step {
            x: Xstep::new(
                Positive::ONE,
                TimeFrame::Minute,
                ExpirationDate::Days(pos_or_panic!(30.0)),
            ),
            y: Ystep::new(0, initial_price),
        };

        let walk_params = WalkParams {
            size: 2,
            init_step,
            walk_type: WalkType::GeometricBrownian {
                dt: convert_time_frame(
                    Positive::ONE / pos_or_panic!(30.0),
                    &TimeFrame::Minute,
                    &TimeFrame::Day,
                )?,
                drift: dec!(0.0),
                volatility: pos_or_panic!(0.2),
            },
            walker,
            seed: Some(SEED),
        };

        let Ok(simulator) =
            Simulator::new("Display Test".to_string(), 2, &walk_params, test_generator)
        else {
            unreachable!()
        };

        let display_output = format!("{simulator}");
        assert!(display_output.starts_with("Display Test"));
        assert!(display_output.contains("Display Test_0"));
        assert!(display_output.contains("Display Test_1"));
        Ok(())
    }

    // Test simulator with empty collection
    #[test]
    fn test_simulator_empty() {
        let simulator: Simulator<Positive, Positive> = Simulator {
            title: "Empty Simulator".to_string(),
            random_walks: Vec::new(),
        };

        assert_eq!(simulator.get_title(), "Empty Simulator");
        assert_eq!(simulator.len(), 0);
        assert!(simulator.is_empty());
        assert!(simulator.first().is_none());
        assert!(simulator.last().is_none());
    }

    #[test]
    fn test_simulator_new_propagates_generator_error_short_circuits()
    -> Result<(), Box<dyn std::error::Error>> {
        // Regression for #349: ensure the simulator constructor returns
        // the first generator error and does not silently build a
        // partial simulator.
        use std::sync::atomic::{AtomicU32, Ordering};
        let walker = Box::new(TestWalker);
        let initial_price = Positive::HUNDRED;
        let init_step = Step {
            x: Xstep::new(
                Positive::ONE,
                TimeFrame::Minute,
                ExpirationDate::Days(pos_or_panic!(30.0)),
            ),
            y: Ystep::new(0, initial_price),
        };

        let walk_params = WalkParams {
            size: 1,
            init_step,
            walk_type: WalkType::GeometricBrownian {
                dt: convert_time_frame(
                    Positive::ONE / pos_or_panic!(30.0),
                    &TimeFrame::Minute,
                    &TimeFrame::Day,
                )?,
                drift: dec!(0.0),
                volatility: pos_or_panic!(0.2),
            },
            walker,
            seed: Some(SEED),
        };

        // The generator is shared across the rayon pool from
        // `PARALLEL_MIN_WALKS` walks up, so it must be `Sync`: the call count
        // is an atomic. One walk short of the threshold the walks are built
        // serially, which stops at the first error.
        let calls = AtomicU32::new(0);
        let size = PARALLEL_MIN_WALKS - 1;
        let result: Result<Simulator<Positive, Positive>, &'static str> =
            Simulator::new("Err Sim".to_string(), size, &walk_params, |p| {
                let n = calls.fetch_add(1, Ordering::SeqCst);
                if n == 0 {
                    Err("boom")
                } else {
                    Ok(vec![p.init_step.clone()])
                }
            });

        match result {
            Err(msg) => assert_eq!(msg, "boom"),
            Ok(_) => panic!("expected generator error to propagate"),
        }
        // The first walk failed, so no later walk was built.
        assert_eq!(calls.load(Ordering::SeqCst), 1);
        Ok(())
    }

    // Test panic scenarios (these would typically be in separate test functions)
    #[test]
    #[should_panic(expected = "index out of bounds")]
    fn test_simulator_index_out_of_bounds() {
        let walker = Box::new(TestWalker);
        let initial_price = Positive::HUNDRED;
        let init_step = Step {
            x: Xstep::new(
                Positive::ONE,
                TimeFrame::Minute,
                ExpirationDate::Days(pos_or_panic!(30.0)),
            ),
            y: Ystep::new(0, initial_price),
        };

        let walk_params = WalkParams {
            size: 3,
            init_step,
            walk_type: WalkType::GeometricBrownian {
                dt: match convert_time_frame(
                    Positive::ONE / pos_or_panic!(30.0),
                    &TimeFrame::Minute,
                    &TimeFrame::Day,
                ) {
                    Ok(value) => value,
                    Err(error) => panic!("convert_time_frame failed: {error}"),
                },
                drift: dec!(0.0),
                volatility: pos_or_panic!(0.2),
            },
            walker,
            seed: Some(SEED),
        };

        let Ok(simulator) =
            Simulator::new("Panic Test".to_string(), 3, &walk_params, test_generator)
        else {
            unreachable!()
        };

        // This should panic
        let _ = simulator[3];
    }

    #[test]
    fn test_full_simulation() -> Result<(), SimulationError> {
        let simulator_size: usize = 5;
        let n_steps = 10;
        let initial_price = Positive::HUNDRED;
        let std_dev = pos_or_panic!(20.0);
        let walker = Box::new(TestWalker::new());
        let days = Positive::TWO;

        let walk_params = WalkParams {
            size: n_steps,
            init_step: Step {
                x: Xstep::new(Positive::ONE, TimeFrame::Hour, ExpirationDate::Days(days)),
                y: Ystep::new(0, initial_price),
            },
            walk_type: WalkType::GeometricBrownian {
                dt: convert_time_frame(Positive::ONE / days, &TimeFrame::Hour, &TimeFrame::Day)?,
                drift: dec!(0.0),
                volatility: std_dev,
            },
            walker,
            seed: Some(SEED),
        };

        assert_eq!(walk_params.size, n_steps);
        assert_eq!(walk_params.init_step.get_value(), &Positive::HUNDRED);
        assert_eq!(walk_params.y(), &Positive::HUNDRED);

        let simulator = Simulator::new(
            "Simulator".to_string(),
            simulator_size,
            &walk_params,
            generator_positive,
        )?;
        debug!("Simulator: {}", simulator);
        assert_eq!(simulator.get_title(), "Simulator");
        assert_eq!(simulator.len(), simulator_size);

        let random_walk = simulator[0].clone();
        assert_eq!(random_walk.get_title(), "Simulator_0");
        assert_eq!(random_walk.len(), n_steps);

        let step = random_walk[0].clone();
        assert_eq!(*step.get_index(), Positive::ONE);
        let step_string = format!("{step}");
        assert_eq!(step.to_string(), step_string);

        let y_step = step.get_y_step();
        assert_eq!(*y_step.index(), 0);
        assert_eq!(*y_step.value(), Positive::HUNDRED);

        let x_step = step.get_x_step();
        assert_eq!(*x_step.index(), 0);
        assert_eq!(*x_step.step_size_in_time(), Positive::ONE);
        assert_eq!(x_step.time_unit(), &TimeFrame::Hour);
        assert_eq!(x_step.days_left()?, Positive::TWO);

        let next_step = step.next(pos_or_panic!(200.0)).expect("should be Ok");
        assert_eq!(next_step.get_value(), &pos_or_panic!(200.0));
        let next_step_string = format!("{next_step}");
        assert_eq!(next_step.to_string(), next_step_string);

        let previous_step = step.previous(pos_or_panic!(50.0))?;
        assert_eq!(previous_step.get_value(), &pos_or_panic!(50.0));
        let previous_step_string = format!("{previous_step}");
        assert_eq!(previous_step.to_string(), previous_step_string);

        let x_step = step.get_x_step();
        let next_x_step = x_step.next().expect("should be Ok");
        assert_eq!(*next_x_step.index(), 1);
        assert_eq!(*next_x_step.step_size_in_time(), Positive::ONE);
        let next_x_step_string = format!("{next_x_step}");
        assert_eq!(next_x_step.to_string(), next_x_step_string);

        let y_step = step.get_y_step();
        assert_eq!(*y_step.index(), 0);
        assert_eq!(*y_step.value(), Positive::HUNDRED);
        assert_eq!(y_step.positive().unwrap(), Positive::HUNDRED);

        let last_steps: Vec<&Step<Positive, Positive>> = simulator
            .into_iter()
            .map(|step| step.last().expect("should be Ok"))
            .collect();
        info!("Last Steps: {:?}", last_steps);
        assert_eq!(last_steps.len(), simulator_size);

        let last_values: Vec<&Positive> = simulator
            .into_iter()
            .map(|step| step.last().expect("should be Ok").get_value())
            .collect();
        info!("Last Values: {:?}", last_values);
        assert_eq!(last_values.len(), simulator_size);

        Ok(())
    }
}

/// `Simulator::new` reserved `Vec::with_capacity(size)` up front, which hit
/// `capacity overflow` for a `size` whose walks cannot fit in memory before
/// the generator ever ran (#788).
#[cfg(test)]
mod tests_new_panic_freedom {
    use super::*;
    use crate::simulation::{WalkType, WalkTypeAble};
    use optionstratlib_core::model::ExpirationDate;
    use optionstratlib_core::utils::TimeFrame;
    use rust_decimal::Decimal;

    #[derive(Debug, Clone)]
    struct BareWalker;

    impl WalkTypeAble<Positive, Positive> for BareWalker {}

    #[test]
    fn test_new_unreservable_size_reaches_the_generator() {
        let params = WalkParams {
            size: 3,
            init_step: Step::new(
                Positive::ONE,
                TimeFrame::Day,
                ExpirationDate::Days(Positive::HUNDRED),
                Positive::HUNDRED,
            ),
            walk_type: WalkType::Brownian {
                dt: Positive::ONE,
                drift: Decimal::ZERO,
                volatility: Positive::ONE,
            },
            walker: Box::new(BareWalker),
            seed: Some(788),
        };
        let result: Result<Simulator<Positive, Positive>, &'static str> =
            Simulator::new("Huge".to_string(), usize::MAX, &params, |_| Err("boom"));
        assert!(matches!(result, Err("boom")));
    }
}