runmat-runtime 0.5.0

//! MATLAB-compatible `factorial` builtin with GPU-aware semantics for RunMat.
//!
//! Implements element-wise factorial for numerical inputs, mirroring MATLAB’s
//! restrictions to non-negative integers while providing documented fallbacks
//! when GPU providers lack a dedicated kernel.

use once_cell::sync::Lazy;
use runmat_accelerate_api::{GpuTensorHandle, HostTensorView};
use runmat_builtins::{
    BuiltinCompletionPolicy, BuiltinDescriptor, BuiltinErrorDescriptor, BuiltinOutputMode,
    BuiltinParamArity, BuiltinParamDescriptor, BuiltinParamType, BuiltinSignatureDescriptor,
    Tensor, Value,
};
use runmat_macros::runtime_builtin;

use crate::builtins::common::random_args::keyword_of;
use crate::builtins::common::spec::{
    BroadcastSemantics, BuiltinFusionSpec, BuiltinGpuSpec, ConstantStrategy, GpuOpKind,
    ProviderHook, ReductionNaN, ResidencyPolicy, ScalarType, ShapeRequirements,
};
use crate::builtins::common::{gpu_helpers, map_control_flow_with_builtin, tensor};
use crate::builtins::math::type_resolvers::numeric_unary_type;
use crate::{build_runtime_error, BuiltinResult, RuntimeError};

const MAX_FACTORIAL_N: usize = 170;

static FACT_TABLE: Lazy<[f64; MAX_FACTORIAL_N + 1]> = Lazy::new(|| {
    let mut table = [1.0f64; MAX_FACTORIAL_N + 1];
    let mut acc = 1.0;
    for (n, slot) in table.iter_mut().enumerate().skip(1) {
        acc *= n as f64;
        *slot = acc;
    }
    table
});

#[runmat_macros::register_gpu_spec(builtin_path = "crate::builtins::math::elementwise::factorial")]
pub const GPU_SPEC: BuiltinGpuSpec = BuiltinGpuSpec {
    name: "factorial",
    op_kind: GpuOpKind::Elementwise,
    supported_precisions: &[ScalarType::F32, ScalarType::F64],
    broadcast: BroadcastSemantics::Matlab,
    provider_hooks: &[ProviderHook::Unary {
        name: "unary_factorial",
    }],
    constant_strategy: ConstantStrategy::InlineLiteral,
    residency: ResidencyPolicy::NewHandle,
    nan_mode: ReductionNaN::Include,
    two_pass_threshold: None,
    workgroup_size: None,
    accepts_nan_mode: false,
    notes: "Providers may implement unary_factorial; otherwise the runtime gathers to host and mirrors MATLAB overflow/NaN behaviour.",
};

#[runmat_macros::register_fusion_spec(
    builtin_path = "crate::builtins::math::elementwise::factorial"
)]
pub const FUSION_SPEC: BuiltinFusionSpec = BuiltinFusionSpec {
    name: "factorial",
    shape: ShapeRequirements::BroadcastCompatible,
    constant_strategy: ConstantStrategy::InlineLiteral,
    elementwise: None,
    reduction: None,
    emits_nan: true,
    notes: "Factorial is evaluated as a scalar helper; fusion currently bypasses it and executes the standalone host or provider kernel.",
};

const BUILTIN_NAME: &str = "factorial";

const FACTORIAL_OUTPUT: [BuiltinParamDescriptor; 1] = [BuiltinParamDescriptor {
    name: "Y",
    ty: BuiltinParamType::NumericArray,
    arity: BuiltinParamArity::Required,
    default: None,
    description: "Elementwise factorial result.",
}];

const FACTORIAL_INPUTS_X: [BuiltinParamDescriptor; 1] = [BuiltinParamDescriptor {
    name: "X",
    ty: BuiltinParamType::Any,
    arity: BuiltinParamArity::Required,
    default: None,
    description: "Real numeric input.",
}];

const FACTORIAL_INPUTS_X_LIKE: [BuiltinParamDescriptor; 3] = [
    BuiltinParamDescriptor {
        name: "X",
        ty: BuiltinParamType::Any,
        arity: BuiltinParamArity::Required,
        default: None,
        description: "Real numeric input.",
    },
    BuiltinParamDescriptor {
        name: "like",
        ty: BuiltinParamType::StringScalar,
        arity: BuiltinParamArity::Required,
        default: None,
        description: "Literal string \"like\".",
    },
    BuiltinParamDescriptor {
        name: "prototype",
        ty: BuiltinParamType::LikePrototype,
        arity: BuiltinParamArity::Required,
        default: None,
        description: "Output class/device prototype.",
    },
];

const FACTORIAL_SIGNATURES: [BuiltinSignatureDescriptor; 2] = [
    BuiltinSignatureDescriptor {
        label: "Y = factorial(X)",
        inputs: &FACTORIAL_INPUTS_X,
        outputs: &FACTORIAL_OUTPUT,
    },
    BuiltinSignatureDescriptor {
        label: "Y = factorial(X, \"like\", prototype)",
        inputs: &FACTORIAL_INPUTS_X_LIKE,
        outputs: &FACTORIAL_OUTPUT,
    },
];

const FACTORIAL_ERROR_INVALID_ARGUMENT: BuiltinErrorDescriptor = BuiltinErrorDescriptor {
    code: "RM.FACTORIAL.INVALID_ARGUMENT",
    identifier: Some("RunMat:factorial:InvalidArgument"),
    when: "Optional arguments are malformed or unsupported.",
    message: "factorial: invalid argument",
};

const FACTORIAL_ERROR_INVALID_INPUT: BuiltinErrorDescriptor = BuiltinErrorDescriptor {
    code: "RM.FACTORIAL.INVALID_INPUT",
    identifier: Some("RunMat:factorial:InvalidInput"),
    when: "Input value or prototype cannot be converted to supported real numeric form.",
    message: "factorial: invalid input",
};

const FACTORIAL_ERROR_GPU_UNSUPPORTED: BuiltinErrorDescriptor = BuiltinErrorDescriptor {
    code: "RM.FACTORIAL.GPU_UNSUPPORTED",
    identifier: Some("RunMat:factorial:GpuUnsupported"),
    when: "GPU output via \"like\" is requested but no compatible provider is active.",
    message: "factorial: gpu output not supported",
};

const FACTORIAL_ERROR_INTERNAL: BuiltinErrorDescriptor = BuiltinErrorDescriptor {
    code: "RM.FACTORIAL.INTERNAL",
    identifier: Some("RunMat:factorial:Internal"),
    when: "Internal gather/provider/tensor construction failed.",
    message: "factorial: internal error",
};

const FACTORIAL_ERRORS: [BuiltinErrorDescriptor; 4] = [
    FACTORIAL_ERROR_INVALID_ARGUMENT,
    FACTORIAL_ERROR_INVALID_INPUT,
    FACTORIAL_ERROR_GPU_UNSUPPORTED,
    FACTORIAL_ERROR_INTERNAL,
];

pub const FACTORIAL_DESCRIPTOR: BuiltinDescriptor = BuiltinDescriptor {
    signatures: &FACTORIAL_SIGNATURES,
    output_mode: BuiltinOutputMode::Fixed,
    completion_policy: BuiltinCompletionPolicy::Public,
    errors: &FACTORIAL_ERRORS,
};

fn factorial_error_with_detail(
    error: &'static BuiltinErrorDescriptor,
    detail: impl std::fmt::Display,
) -> RuntimeError {
    let mut builder =
        build_runtime_error(format!("{}: {}", error.message, detail)).with_builtin(BUILTIN_NAME);
    if let Some(identifier) = error.identifier {
        builder = builder.with_identifier(identifier);
    }
    builder.build()
}

#[runtime_builtin(
    name = "factorial",
    category = "math/elementwise",
    summary = "Compute element-wise factorial values.",
    keywords = "factorial,n!,permutation,gpu",
    accel = "unary",
    type_resolver(numeric_unary_type),
    descriptor(crate::builtins::math::elementwise::factorial::FACTORIAL_DESCRIPTOR),
    builtin_path = "crate::builtins::math::elementwise::factorial"
)]
async fn factorial_builtin(value: Value, rest: Vec<Value>) -> BuiltinResult<Value> {
    let output = parse_output_template(&rest)?;
    let base = match value {
        Value::GpuTensor(handle) => factorial_gpu(handle).await?,
        Value::Complex(_, _) | Value::ComplexTensor(_) => {
            return Err(factorial_error_with_detail(
                &FACTORIAL_ERROR_INVALID_INPUT,
                "complex inputs are not supported; use gamma(z + 1) instead",
            ))
        }
        Value::String(_) | Value::StringArray(_) | Value::CharArray(_) => {
            return Err(factorial_error_with_detail(
                &FACTORIAL_ERROR_INVALID_INPUT,
                "expected numeric or logical input",
            ))
        }
        other => {
            let tensor = tensor::value_into_tensor_for("factorial", other)
                .map_err(|e| factorial_error_with_detail(&FACTORIAL_ERROR_INVALID_INPUT, e))?;
            factorial_tensor(tensor).map(tensor::tensor_into_value)?
        }
    };
    apply_output_template(base, &output).await
}

#[derive(Clone)]
enum OutputTemplate {
    Default,
    Like(Value),
}

fn parse_output_template(args: &[Value]) -> BuiltinResult<OutputTemplate> {
    match args.len() {
        0 => Ok(OutputTemplate::Default),
        1 => {
            if matches!(keyword_of(&args[0]).as_deref(), Some("like")) {
                Err(factorial_error_with_detail(
                    &FACTORIAL_ERROR_INVALID_ARGUMENT,
                    "expected prototype after 'like'",
                ))
            } else {
                Err(factorial_error_with_detail(
                    &FACTORIAL_ERROR_INVALID_ARGUMENT,
                    "unrecognised option; only 'like' is supported",
                ))
            }
        }
        2 => {
            if matches!(keyword_of(&args[0]).as_deref(), Some("like")) {
                Ok(OutputTemplate::Like(args[1].clone()))
            } else {
                Err(factorial_error_with_detail(
                    &FACTORIAL_ERROR_INVALID_ARGUMENT,
                    "unrecognised option; only 'like' is supported",
                ))
            }
        }
        _ => Err(factorial_error_with_detail(
            &FACTORIAL_ERROR_INVALID_ARGUMENT,
            "too many input arguments",
        )),
    }
}

async fn factorial_gpu(handle: GpuTensorHandle) -> BuiltinResult<Value> {
    if let Some(provider) = runmat_accelerate_api::provider_for_handle(&handle) {
        if let Ok(out) = provider.unary_factorial(&handle).await {
            return Ok(Value::GpuTensor(out));
        }
    }
    let tensor = gpu_helpers::gather_tensor_async(&handle)
        .await
        .map_err(|flow| map_control_flow_with_builtin(flow, BUILTIN_NAME))?;
    Ok(tensor::tensor_into_value(factorial_tensor(tensor)?))
}

fn factorial_tensor(tensor: Tensor) -> BuiltinResult<Tensor> {
    let mut data = Vec::with_capacity(tensor.data.len());
    for &value in &tensor.data {
        data.push(factorial_scalar(value));
    }
    Tensor::new(data, tensor.shape.clone())
        .map_err(|e| factorial_error_with_detail(&FACTORIAL_ERROR_INTERNAL, e))
}

fn factorial_scalar(value: f64) -> f64 {
    if value.is_nan() {
        return f64::NAN;
    }
    if value == 0.0 {
        return 1.0;
    }
    if value.is_infinite() {
        return if value.is_sign_positive() {
            f64::INFINITY
        } else {
            f64::NAN
        };
    }
    if value < 0.0 {
        return f64::NAN;
    }
    let Some(n) = classify_nonnegative_integer(value) else {
        return f64::NAN;
    };
    if n > MAX_FACTORIAL_N {
        return f64::INFINITY;
    }
    FACT_TABLE[n]
}

fn classify_nonnegative_integer(value: f64) -> Option<usize> {
    if !value.is_finite() {
        return None;
    }
    if value < 0.0 {
        return None;
    }
    let rounded = value.round();
    let tol = f64::EPSILON * value.abs().max(1.0);
    if (value - rounded).abs() > tol {
        return None;
    }
    if rounded < 0.0 {
        return None;
    }
    Some(rounded as usize)
}

async fn apply_output_template(value: Value, template: &OutputTemplate) -> BuiltinResult<Value> {
    match template {
        OutputTemplate::Default => Ok(value),
        OutputTemplate::Like(proto) => {
            let analysis = analyse_like_prototype(proto).await?;
            match analysis.device {
                DevicePreference::Host => convert_to_host_like(value).await,
                DevicePreference::Gpu => convert_to_gpu_like(value),
            }
        }
    }
}

#[derive(Clone, Copy)]
enum DevicePreference {
    Host,
    Gpu,
}

struct LikeAnalysis {
    device: DevicePreference,
}

#[async_recursion::async_recursion(?Send)]
async fn analyse_like_prototype(proto: &Value) -> BuiltinResult<LikeAnalysis> {
    match proto {
        Value::GpuTensor(_) => Ok(LikeAnalysis {
            device: DevicePreference::Gpu,
        }),
        Value::Tensor(_) | Value::Num(_) | Value::Int(_) | Value::Bool(_) => Ok(LikeAnalysis {
            device: DevicePreference::Host,
        }),
        Value::LogicalArray(_) => Ok(LikeAnalysis {
            device: DevicePreference::Host,
        }),
        Value::Complex(_, _) | Value::ComplexTensor(_) => Err(factorial_error_with_detail(
            &FACTORIAL_ERROR_INVALID_INPUT,
            "complex prototypes for 'like' are not supported; results are always real",
        )),
        Value::String(_) | Value::StringArray(_) | Value::CharArray(_) => {
            Err(factorial_error_with_detail(
                &FACTORIAL_ERROR_INVALID_INPUT,
                "prototype must be numeric or a gpuArray",
            ))
        }
        other => {
            let gathered = gpu_helpers::gather_value_async(other)
                .await
                .map_err(|flow| map_control_flow_with_builtin(flow, BUILTIN_NAME))?;
            analyse_like_prototype(&gathered).await
        }
    }
}

async fn convert_to_host_like(value: Value) -> BuiltinResult<Value> {
    match value {
        Value::GpuTensor(handle) => gpu_helpers::gather_value_async(&Value::GpuTensor(handle))
            .await
            .map_err(|flow| map_control_flow_with_builtin(flow, BUILTIN_NAME)),
        other => Ok(other),
    }
}

fn convert_to_gpu_like(value: Value) -> BuiltinResult<Value> {
    let provider = runmat_accelerate_api::provider().ok_or_else(|| {
        factorial_error_with_detail(
            &FACTORIAL_ERROR_GPU_UNSUPPORTED,
            "GPU output requested via 'like' but no acceleration provider is active",
        )
    })?;
    match value {
        Value::GpuTensor(handle) => Ok(Value::GpuTensor(handle)),
        Value::Tensor(tensor) => upload_tensor(provider, tensor),
        Value::Num(n) => {
            let tensor = Tensor::new(vec![n], vec![1, 1])
                .map_err(|e| factorial_error_with_detail(&FACTORIAL_ERROR_INTERNAL, e))?;
            upload_tensor(provider, tensor)
        }
        Value::Int(i) => convert_to_gpu_like(Value::Num(i.to_f64())),
        Value::Bool(b) => convert_to_gpu_like(Value::Num(if b { 1.0 } else { 0.0 })),
        Value::LogicalArray(logical) => {
            let tensor = tensor::logical_to_tensor(&logical)
                .map_err(|e| factorial_error_with_detail(&FACTORIAL_ERROR_INTERNAL, e))?;
            upload_tensor(provider, tensor)
        }
        other => Err(factorial_error_with_detail(
            &FACTORIAL_ERROR_INVALID_INPUT,
            format!("cannot place value {other:?} on the GPU via 'like'"),
        )),
    }
}

fn upload_tensor(
    provider: &'static dyn runmat_accelerate_api::AccelProvider,
    tensor: Tensor,
) -> BuiltinResult<Value> {
    let view = HostTensorView {
        data: &tensor.data,
        shape: &tensor.shape,
    };
    let handle = provider.upload(&view).map_err(|e| {
        factorial_error_with_detail(
            &FACTORIAL_ERROR_INTERNAL,
            format!("failed to upload GPU result: {e}"),
        )
    })?;
    Ok(Value::GpuTensor(handle))
}

#[cfg(test)]
pub(crate) mod tests {
    use super::*;
    use crate::builtins::common::test_support;
    use futures::executor::block_on;
    use runmat_builtins::{IntValue, LogicalArray, ResolveContext, Tensor, Type};

    fn factorial_builtin(value: Value, rest: Vec<Value>) -> BuiltinResult<Value> {
        block_on(super::factorial_builtin(value, rest))
    }

    #[test]
    fn factorial_descriptor_signatures_cover_core_forms() {
        let labels: Vec<&str> = FACTORIAL_DESCRIPTOR
            .signatures
            .iter()
            .map(|sig| sig.label)
            .collect();
        assert!(labels.contains(&"Y = factorial(X)"));
        assert!(labels.contains(&"Y = factorial(X, \"like\", prototype)"));
    }

    #[test]
    fn factorial_type_preserves_tensor_shape() {
        let out = numeric_unary_type(
            &[Type::Tensor {
                shape: Some(vec![Some(2), Some(3)]),
            }],
            &ResolveContext::new(Vec::new()),
        );
        assert_eq!(
            out,
            Type::Tensor {
                shape: Some(vec![Some(2), Some(3)])
            }
        );
    }

    #[test]
    fn factorial_type_scalar_tensor_returns_num() {
        let out = numeric_unary_type(
            &[Type::Tensor {
                shape: Some(vec![Some(1), Some(1)]),
            }],
            &ResolveContext::new(Vec::new()),
        );
        assert_eq!(out, Type::Num);
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_scalar_positive() {
        let result = factorial_builtin(Value::Num(5.0), Vec::new()).expect("factorial");
        assert_eq!(result, Value::Num(120.0));
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_zero_is_one() {
        let result = factorial_builtin(Value::Num(0.0), Vec::new()).expect("factorial");
        assert_eq!(result, Value::Num(1.0));
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_vector_inputs() {
        let tensor = Tensor::new(vec![0.0, 1.0, 3.0, 5.0], vec![4, 1]).unwrap();
        let result = factorial_builtin(Value::Tensor(tensor), Vec::new()).expect("factorial");
        match result {
            Value::Tensor(out) => {
                assert_eq!(out.shape, vec![4, 1]);
                assert_eq!(out.data, vec![1.0, 1.0, 6.0, 120.0]);
            }
            other => panic!("expected tensor result, got {other:?}"),
        }
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_non_integer_produces_nan() {
        let result = factorial_builtin(Value::Num(2.5), Vec::new()).expect("factorial");
        match result {
            Value::Num(v) => assert!(v.is_nan()),
            other => panic!("expected scalar NaN, got {other:?}"),
        }
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_negative_produces_nan() {
        let tensor = Tensor::new(vec![-1.0, 3.0], vec![2, 1]).unwrap();
        let result = factorial_builtin(Value::Tensor(tensor), Vec::new()).expect("factorial");
        match result {
            Value::Tensor(out) => {
                assert!(out.data[0].is_nan());
                assert_eq!(out.data[1], 6.0);
            }
            other => panic!("expected tensor result, got {other:?}"),
        }
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_small_positive_non_integer_nan() {
        let result = factorial_builtin(Value::Num(1e-12), Vec::new()).expect("factorial");
        match result {
            Value::Num(v) => assert!(v.is_nan()),
            other => panic!("expected scalar NaN, got {other:?}"),
        }
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_overflow_returns_inf() {
        let result = factorial_builtin(Value::Num(171.0), Vec::new()).expect("factorial");
        match result {
            Value::Num(v) => assert!(v.is_infinite()),
            other => panic!("expected scalar Inf, got {other:?}"),
        }
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_like_missing_prototype_errors() {
        let err = factorial_builtin(Value::Num(3.0), vec![Value::from("like")])
            .expect_err("expected error");
        assert!(err.message().contains("prototype"));
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_like_gpu_prototype_uploads() {
        test_support::with_test_provider(|provider| {
            let tensor = Tensor::new(vec![3.0, 4.0], vec![2, 1]).unwrap();
            let proto_view = HostTensorView {
                data: &[0.0],
                shape: &[1, 1],
            };
            let proto = provider.upload(&proto_view).expect("upload");
            let result = factorial_builtin(
                Value::Tensor(tensor.clone()),
                vec![Value::from("like"), Value::GpuTensor(proto)],
            )
            .expect("factorial");
            match result {
                Value::GpuTensor(handle) => {
                    let gathered = test_support::gather(Value::GpuTensor(handle)).expect("gather");
                    assert_eq!(gathered.shape, vec![2, 1]);
                    assert_eq!(gathered.data, vec![6.0, 24.0]);
                }
                other => panic!("expected GPU tensor, got {other:?}"),
            }
        });
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_gpu_provider_roundtrip() {
        test_support::with_test_provider(|provider| {
            let tensor = Tensor::new(vec![0.0, 1.0, 3.0, 5.0], vec![4, 1]).unwrap();
            let view = HostTensorView {
                data: &tensor.data,
                shape: &tensor.shape,
            };
            let handle = provider.upload(&view).expect("upload");
            let result = factorial_builtin(Value::GpuTensor(handle), Vec::new()).expect("fact");
            let gathered = test_support::gather(result).expect("gather");
            assert_eq!(gathered.shape, vec![4, 1]);
            assert_eq!(gathered.data, vec![1.0, 1.0, 6.0, 120.0]);
        });
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_like_host_with_gpu_input_gathers() {
        test_support::with_test_provider(|provider| {
            let tensor = Tensor::new(vec![3.0, 4.0], vec![2, 1]).unwrap();
            let view = HostTensorView {
                data: &tensor.data,
                shape: &tensor.shape,
            };
            let handle = provider.upload(&view).expect("upload");
            let result = factorial_builtin(
                Value::GpuTensor(handle),
                vec![Value::from("like"), Value::Num(0.0)],
            )
            .expect("factorial");
            match result {
                Value::Tensor(t) => {
                    assert_eq!(t.data, vec![6.0, 24.0]);
                }
                other => panic!("expected host tensor, got {other:?}"),
            }
        });
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_logical_input_promotes() {
        let logical = LogicalArray::new(vec![1, 0, 1], vec![3, 1]).unwrap();
        let result = factorial_builtin(Value::LogicalArray(logical), Vec::new()).expect("fact");
        match result {
            Value::Tensor(t) => assert_eq!(t.data, vec![1.0, 1.0, 1.0]),
            other => panic!("expected tensor result, got {other:?}"),
        }
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_int_input_promotes_to_double() {
        let value = Value::Int(IntValue::U16(5));
        let result = factorial_builtin(value, Vec::new()).expect("factorial");
        assert_eq!(result, Value::Num(120.0));
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_nan_propagates() {
        let result = factorial_builtin(Value::Num(f64::NAN), Vec::new()).expect("factorial");
        match result {
            Value::Num(v) => assert!(v.is_nan()),
            other => panic!("expected scalar NaN, got {other:?}"),
        }
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_complex_input_errors() {
        let err = factorial_builtin(Value::Complex(1.0, 0.5), Vec::new())
            .expect_err("expected complex rejection");
        assert_eq!(err.identifier(), FACTORIAL_ERROR_INVALID_INPUT.identifier);
        assert!(err.message().contains("complex"));
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_string_input_errors() {
        let err = factorial_builtin(Value::from("hello"), Vec::new())
            .expect_err("expected string rejection");
        assert_eq!(err.identifier(), FACTORIAL_ERROR_INVALID_INPUT.identifier);
        assert!(err.message().contains("numeric"));
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    fn factorial_like_complex_prototype_rejected() {
        let err = factorial_builtin(
            Value::Num(3.0),
            vec![Value::from("like"), Value::Complex(0.0, 1.0)],
        )
        .expect_err("expected complex prototype rejection");
        assert!(err.message().contains("complex"));
    }

    #[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
    #[test]
    #[cfg(feature = "wgpu")]
    fn factorial_wgpu_matches_cpu_after_gather() {
        let _ = runmat_accelerate::backend::wgpu::provider::register_wgpu_provider(
            runmat_accelerate::backend::wgpu::provider::WgpuProviderOptions::default(),
        );
        let tensor = Tensor::new(vec![0.0, 1.0, 4.0], vec![3, 1]).unwrap();
        let cpu = factorial_builtin(Value::Tensor(tensor.clone()), Vec::new()).expect("cpu");
        let view = HostTensorView {
            data: &tensor.data,
            shape: &tensor.shape,
        };
        let handle = runmat_accelerate_api::provider()
            .unwrap()
            .upload(&view)
            .unwrap();
        let gpu = block_on(factorial_gpu(handle)).expect("gpu");
        let gathered = test_support::gather(gpu).expect("gather");
        let cpu_tensor = match cpu {
            Value::Tensor(t) => t,
            Value::Num(n) => Tensor::new(vec![n], vec![1, 1]).unwrap(),
            other => panic!("unexpected cpu result {other:?}"),
        };
        assert_eq!(gathered.shape, cpu_tensor.shape);
        assert_eq!(gathered.data, cpu_tensor.data);
    }
}