use num_complex::Complex64;
use runmat_accelerate_api::GpuTensorHandle;
use runmat_builtins::{CharArray, ComplexTensor, Tensor, Value};
use runmat_macros::runtime_builtin;
use crate::builtins::common::random_args::{complex_tensor_into_value, 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 PI: f64 = std::f64::consts::PI;
const SQRT_TWO_PI: f64 = 2.506_628_274_631_000_5;
const LANCZOS_G: f64 = 7.0;
const EPSILON: f64 = 1e-12;
const LANCZOS_COEFFS: [f64; 8] = [
676.5203681218851,
-1259.1392167224028,
771.3234287776531,
-176.6150291621406,
12.507343278686905,
-0.13857109526572012,
9.984_369_578_019_572e-6,
1.5056327351493116e-7,
];
#[runmat_macros::register_gpu_spec(builtin_path = "crate::builtins::math::elementwise::gamma")]
pub const GPU_SPEC: BuiltinGpuSpec = BuiltinGpuSpec {
name: "gamma",
op_kind: GpuOpKind::Elementwise,
supported_precisions: &[ScalarType::F32, ScalarType::F64],
broadcast: BroadcastSemantics::Matlab,
provider_hooks: &[ProviderHook::Unary { name: "unary_gamma" }],
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 execute gamma directly on device buffers via unary_gamma; runtimes gather to the host when the hook is unavailable.",
};
#[runmat_macros::register_fusion_spec(builtin_path = "crate::builtins::math::elementwise::gamma")]
pub const FUSION_SPEC: BuiltinFusionSpec = BuiltinFusionSpec {
name: "gamma",
shape: ShapeRequirements::BroadcastCompatible,
constant_strategy: ConstantStrategy::InlineLiteral,
elementwise: None,
reduction: None,
emits_nan: false,
notes: "Fusion planner currently falls back to host evaluation; providers may supply specialised kernels in the future.",
};
const BUILTIN_NAME: &str = "gamma";
fn builtin_error(message: impl Into<String>) -> RuntimeError {
build_runtime_error(message)
.with_builtin(BUILTIN_NAME)
.build()
}
#[runtime_builtin(
name = "gamma",
category = "math/elementwise",
summary = "Element-wise gamma function for scalars, vectors, matrices, or N-D tensors.",
keywords = "gamma,factorial,special,gpu",
accel = "unary",
type_resolver(numeric_unary_type),
builtin_path = "crate::builtins::math::elementwise::gamma"
)]
async fn gamma_builtin(value: Value, rest: Vec<Value>) -> BuiltinResult<Value> {
let output = parse_output_template(&rest)?;
let base = match value {
Value::GpuTensor(handle) => gamma_gpu(handle).await?,
Value::Complex(re, im) => gamma_complex_scalar_value(Complex64::new(re, im)),
Value::ComplexTensor(ct) => gamma_complex_tensor(ct)?,
Value::CharArray(ca) => gamma_char_array(ca)?,
Value::LogicalArray(logical) => {
let tensor = tensor::logical_to_tensor(&logical)
.map_err(|e| builtin_error(format!("gamma: {e}")))?;
gamma_tensor(tensor).map(tensor::tensor_into_value)?
}
Value::String(_) | Value::StringArray(_) => {
return Err(builtin_error("gamma: expected numeric input"))
}
Value::Tensor(tensor) => gamma_tensor(tensor).map(tensor::tensor_into_value)?,
Value::Num(n) => Value::Num(gamma_real_scalar(n)),
Value::Int(i) => Value::Num(gamma_real_scalar(i.to_f64())),
Value::Bool(b) => Value::Num(gamma_real_scalar(if b { 1.0 } else { 0.0 })),
other => {
return Err(builtin_error(format!(
"gamma: unsupported input type {:?}; expected numeric or gpuArray input",
other
)))
}
};
apply_output_template(base, &output).await
}
async fn gamma_gpu(handle: GpuTensorHandle) -> BuiltinResult<Value> {
if let Some(provider) = runmat_accelerate_api::provider_for_handle(&handle) {
if let Ok(out) = provider.unary_gamma(&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(gamma_tensor(tensor)?))
}
fn gamma_tensor(tensor: Tensor) -> BuiltinResult<Tensor> {
let mut data = Vec::with_capacity(tensor.data.len());
for &v in &tensor.data {
data.push(gamma_real_scalar(v));
}
Tensor::new(data, tensor.shape.clone()).map_err(|e| builtin_error(format!("gamma: {e}")))
}
fn gamma_complex_tensor(ct: ComplexTensor) -> BuiltinResult<Value> {
let mut out = Vec::with_capacity(ct.data.len());
for &(re, im) in &ct.data {
let res = gamma_complex_scalar(Complex64::new(re, im));
out.push((res.re, res.im));
}
let tensor = ComplexTensor::new(out, ct.shape.clone())
.map_err(|e| builtin_error(format!("gamma: {e}")))?;
Ok(complex_tensor_into_value(tensor))
}
fn gamma_complex_scalar_value(z: Complex64) -> Value {
let res = gamma_complex_scalar(z);
if res.im.abs() <= EPSILON * res.re.abs().max(1.0) {
Value::Num(res.re)
} else {
Value::Complex(res.re, res.im)
}
}
fn gamma_char_array(ca: CharArray) -> BuiltinResult<Value> {
let data = ca
.data
.iter()
.map(|&ch| gamma_real_scalar(ch as u32 as f64))
.collect::<Vec<_>>();
let tensor = Tensor::new(data, vec![ca.rows, ca.cols])
.map_err(|e| builtin_error(format!("gamma: {e}")))?;
Ok(tensor::tensor_into_value(tensor))
}
fn gamma_real_scalar(x: f64) -> f64 {
if x.is_nan() {
return f64::NAN;
}
if x.is_infinite() {
return if x.is_sign_positive() {
f64::INFINITY
} else {
f64::NAN
};
}
if is_non_positive_integer(x) {
return f64::INFINITY;
}
let result = gamma_complex_scalar(Complex64::new(x, 0.0));
if result.im.abs() <= EPSILON * result.re.abs().max(1.0) {
result.re
} else {
f64::NAN
}
}
fn gamma_complex_scalar(z: Complex64) -> Complex64 {
if z.re.is_nan() || z.im.is_nan() {
return Complex64::new(f64::NAN, f64::NAN);
}
if z.im.abs() <= EPSILON && z.re.is_infinite() {
return Complex64::new(f64::INFINITY, 0.0);
}
if is_complex_pole(z) {
return Complex64::new(f64::INFINITY, 0.0);
}
if z.re < 0.5 {
let sin_term = (Complex64::new(PI, 0.0) * z).sin();
if sin_term.norm_sqr() <= EPSILON * EPSILON {
return Complex64::new(f64::INFINITY, 0.0);
}
let gamma_one_minus_z = gamma_complex_scalar(Complex64::new(1.0, 0.0) - z);
return Complex64::new(PI, 0.0) / (sin_term * gamma_one_minus_z);
}
lanczos_gamma(z)
}
fn lanczos_gamma(z: Complex64) -> Complex64 {
let z_minus_one = z - Complex64::new(1.0, 0.0);
let mut sum = Complex64::new(0.999_999_999_999_809_9, 0.0);
for (idx, coeff) in LANCZOS_COEFFS.iter().enumerate() {
let denom = z_minus_one + Complex64::new((idx + 1) as f64, 0.0);
sum += Complex64::new(*coeff, 0.0) / denom;
}
let t = z_minus_one + Complex64::new(LANCZOS_G + 0.5, 0.0);
let power = t.powc(z_minus_one + Complex64::new(0.5, 0.0));
let exponential = (-t).exp();
Complex64::new(SQRT_TWO_PI, 0.0) * power * exponential * sum
}
fn is_non_positive_integer(x: f64) -> bool {
x <= 0.0 && is_close_to_integer(x)
}
fn is_complex_pole(z: Complex64) -> bool {
z.im.abs() <= EPSILON && is_non_positive_integer(z.re)
}
fn is_close_to_integer(x: f64) -> bool {
if !x.is_finite() {
return false;
}
let nearest = x.round();
let diff = (x - nearest).abs();
if nearest == 0.0 {
diff <= EPSILON * EPSILON
} else {
diff <= EPSILON * nearest.abs().max(1.0)
}
}
#[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(builtin_error("gamma: expected prototype after 'like'"))
} else {
Err(builtin_error("gamma: unrecognised argument for gamma"))
}
}
2 => {
if matches!(keyword_of(&args[0]).as_deref(), Some("like")) {
Ok(OutputTemplate::Like(args[1].clone()))
} else {
Err(builtin_error(
"gamma: unsupported option; only 'like' is accepted",
))
}
}
_ => Err(builtin_error("gamma: too many input arguments")),
}
}
async fn apply_output_template(value: Value, template: &OutputTemplate) -> BuiltinResult<Value> {
match template {
OutputTemplate::Default => Ok(value),
OutputTemplate::Like(proto) => apply_like_template(value, proto).await,
}
}
async fn apply_like_template(value: Value, prototype: &Value) -> BuiltinResult<Value> {
let analysis = analyse_like_prototype(prototype).await?;
match analysis.class {
PrototypeClass::Real => match analysis.device {
DevicePreference::Host => convert_to_host_real(value).await,
DevicePreference::Gpu => convert_to_gpu_real(value),
},
PrototypeClass::Complex => match analysis.device {
DevicePreference::Host => convert_to_host_complex(value).await,
DevicePreference::Gpu => Err(builtin_error(
"gamma: complex GPU prototypes are not supported yet",
)),
},
}
}
fn convert_to_gpu_real(value: Value) -> BuiltinResult<Value> {
let provider = runmat_accelerate_api::provider().ok_or_else(|| {
builtin_error(
"gamma: GPU output requested via 'like' but no acceleration provider is active",
)
})?;
match value {
Value::GpuTensor(handle) => Ok(Value::GpuTensor(handle)),
Value::Tensor(tensor) => {
let view = runmat_accelerate_api::HostTensorView {
data: &tensor.data,
shape: &tensor.shape,
};
let handle = provider
.upload(&view)
.map_err(|e| builtin_error(format!("gamma: {e}")))?;
Ok(Value::GpuTensor(handle))
}
Value::Num(n) => {
let tensor = Tensor::new(vec![n], vec![1, 1])
.map_err(|e| builtin_error(format!("gamma: {e}")))?;
convert_to_gpu_real(Value::Tensor(tensor))
}
Value::Int(i) => convert_to_gpu_real(Value::Num(i.to_f64())),
Value::Bool(b) => convert_to_gpu_real(Value::Num(if b { 1.0 } else { 0.0 })),
Value::LogicalArray(logical) => {
let tensor = tensor::logical_to_tensor(&logical)
.map_err(|e| builtin_error(format!("gamma: {e}")))?;
convert_to_gpu_real(Value::Tensor(tensor))
}
Value::Complex(_, _) | Value::ComplexTensor(_) => Err(builtin_error(
"gamma: GPU prototypes for 'like' only support real numeric outputs",
)),
other => Err(builtin_error(format!(
"gamma: unsupported result type for GPU output via 'like' ({other:?})"
))),
}
}
async fn convert_to_host_real(value: Value) -> BuiltinResult<Value> {
match value {
Value::GpuTensor(handle) => {
let proxy = Value::GpuTensor(handle);
gpu_helpers::gather_value_async(&proxy)
.await
.map_err(|flow| map_control_flow_with_builtin(flow, BUILTIN_NAME))
}
other => Ok(other),
}
}
#[async_recursion::async_recursion(?Send)]
async fn convert_to_host_complex(value: Value) -> BuiltinResult<Value> {
match value {
Value::Complex(_, _) | Value::ComplexTensor(_) => Ok(value),
Value::Num(n) => Ok(Value::Complex(n, 0.0)),
Value::Tensor(tensor) => {
let data = tensor.data.iter().map(|&re| (re, 0.0)).collect::<Vec<_>>();
let complex = ComplexTensor::new(data, tensor.shape.clone())
.map_err(|e| builtin_error(format!("gamma: {e}")))?;
Ok(complex_tensor_into_value(complex))
}
Value::GpuTensor(handle) => {
let gathered = gpu_helpers::gather_tensor_async(&handle)
.await
.map_err(|flow| map_control_flow_with_builtin(flow, BUILTIN_NAME))?;
convert_to_host_complex(Value::Tensor(gathered)).await
}
Value::LogicalArray(logical) => {
let tensor = tensor::logical_to_tensor(&logical)
.map_err(|e| builtin_error(format!("gamma: {e}")))?;
convert_to_host_complex(Value::Tensor(tensor)).await
}
Value::Bool(b) => convert_to_host_complex(Value::Num(if b { 1.0 } else { 0.0 })).await,
Value::Int(i) => convert_to_host_complex(Value::Num(i.to_f64())).await,
other => Err(builtin_error(format!(
"gamma: cannot convert {other:?} to complex output via 'like'"
))),
}
}
#[derive(Clone, Copy)]
enum DevicePreference {
Host,
Gpu,
}
#[derive(Clone, Copy)]
enum PrototypeClass {
Real,
Complex,
}
struct LikeAnalysis {
class: PrototypeClass,
device: DevicePreference,
}
#[async_recursion::async_recursion(?Send)]
async fn analyse_like_prototype(proto: &Value) -> BuiltinResult<LikeAnalysis> {
match proto {
Value::GpuTensor(_) => Ok(LikeAnalysis {
class: PrototypeClass::Real,
device: DevicePreference::Gpu,
}),
Value::Tensor(_)
| Value::Num(_)
| Value::Int(_)
| Value::Bool(_)
| Value::LogicalArray(_)
| Value::CharArray(_) => Ok(LikeAnalysis {
class: PrototypeClass::Real,
device: DevicePreference::Host,
}),
Value::Complex(_, _) | Value::ComplexTensor(_) => Ok(LikeAnalysis {
class: PrototypeClass::Complex,
device: DevicePreference::Host,
}),
other => {
let gathered = crate::dispatcher::gather_if_needed_async(other)
.await
.map_err(|flow| map_control_flow_with_builtin(flow, BUILTIN_NAME))?;
analyse_like_prototype(&gathered).await
}
}
}
#[cfg(test)]
pub(crate) mod tests {
use super::*;
use crate::builtins::common::test_support;
use futures::executor::block_on;
use runmat_accelerate_api::HostTensorView;
use runmat_builtins::{IntValue, ResolveContext, Tensor, Type};
fn gamma_builtin(value: Value, rest: Vec<Value>) -> BuiltinResult<Value> {
block_on(super::gamma_builtin(value, rest))
}
#[test]
fn gamma_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 gamma_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);
}
fn approx_eq(a: f64, b: f64, tol: f64) {
assert!((a - b).abs() <= tol, "expected {b}, got {a} (tol {tol})");
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_positive_integer() {
match gamma_builtin(Value::Num(5.0), Vec::new()).expect("gamma") {
Value::Num(v) => approx_eq(v, 24.0, 1e-12),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_half_integer() {
match gamma_builtin(Value::Num(0.5), Vec::new()).expect("gamma") {
Value::Num(v) => approx_eq(v, PI.sqrt(), 1e-12),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_negative_non_integer() {
match gamma_builtin(Value::Num(-0.5), Vec::new()).expect("gamma") {
Value::Num(v) => approx_eq(v, -2.0 * PI.sqrt(), 1e-12),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_matrix() {
let tensor = Tensor::new(vec![1.0, 3.0, 2.0, 4.0], vec![2, 2]).unwrap();
match gamma_builtin(Value::Tensor(tensor), Vec::new()).expect("gamma") {
Value::Tensor(t) => {
assert_eq!(t.shape, vec![2, 2]);
approx_eq(t.data[0], 1.0, 1e-12);
approx_eq(t.data[1], 2.0, 1e-12);
approx_eq(t.data[2], 1.0, 1e-12);
approx_eq(t.data[3], 6.0, 1e-12);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_pole_returns_inf() {
match gamma_builtin(Value::Num(0.0), Vec::new()).expect("gamma") {
Value::Num(v) => assert!(v.is_infinite() && v.is_sign_positive()),
other => panic!("expected scalar result, got {other:?}"),
}
match gamma_builtin(Value::Num(-3.0), Vec::new()).expect("gamma") {
Value::Num(v) => assert!(v.is_infinite()),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_small_negative_not_infinite() {
match gamma_builtin(Value::Num(-1.0e-10), Vec::new()).expect("gamma") {
Value::Num(v) => {
assert!(v.is_finite(), "expected finite value, got {v}");
assert!(v.is_sign_negative(), "expected negative value, got {v}");
assert!(v.abs() > 1.0e9, "expected large magnitude, got {v}");
}
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_complex_identity() {
let z = Complex64::new(0.5, 1.0);
let gamma_z = gamma_complex_scalar(z);
let gamma_z_plus_one = gamma_complex_scalar(z + Complex64::new(1.0, 0.0));
let lhs = gamma_z_plus_one;
let rhs = z * gamma_z;
approx_eq(lhs.re, rhs.re, 1e-10);
approx_eq(lhs.im, rhs.im, 1e-10);
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_char_array() {
let chars = CharArray::new("ab".chars().collect(), 1, 2).unwrap();
match gamma_builtin(Value::CharArray(chars), Vec::new()).expect("gamma") {
Value::Tensor(t) => {
assert_eq!(t.shape, vec![1, 2]);
approx_eq(t.data[0], gamma_real_scalar(97.0), 1e-12);
approx_eq(t.data[1], gamma_real_scalar(98.0), 1e-12);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_string_input_errors() {
let err = gamma_builtin(Value::from("hello"), Vec::new()).expect_err("expected error");
assert!(err.message().contains("expected numeric input"));
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_gpu_provider_fallback() {
test_support::with_test_provider(|provider| {
let tensor = Tensor::new(vec![0.5, 1.5, 2.5], vec![1, 3]).unwrap();
let view = HostTensorView {
data: &tensor.data,
shape: &tensor.shape,
};
let handle = provider.upload(&view).expect("upload");
let result = gamma_builtin(Value::GpuTensor(handle), Vec::new()).expect("gamma");
let gathered = test_support::gather(result).expect("gather");
approx_eq(gathered.data[0], PI.sqrt(), 1e-12);
approx_eq(gathered.data[1], 0.8862269254527579, 1e-12);
approx_eq(gathered.data[2], 1.329340388179137, 1e-12);
});
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
#[cfg(feature = "wgpu")]
fn gamma_wgpu_matches_cpu() {
let _ = runmat_accelerate::backend::wgpu::provider::register_wgpu_provider(
runmat_accelerate::backend::wgpu::provider::WgpuProviderOptions::default(),
);
let tensor = Tensor::new(vec![0.5, 1.5, 2.5, -0.5, 4.2, -1.3], vec![3, 2]).expect("tensor");
let cpu = gamma_tensor(tensor.clone()).expect("cpu gamma");
let provider = runmat_accelerate_api::provider().expect("provider");
let view = HostTensorView {
data: &tensor.data,
shape: &tensor.shape,
};
let handle = provider.upload(&view).expect("upload");
let gpu_value = block_on(gamma_gpu(handle)).expect("gamma gpu");
let gathered = test_support::gather(gpu_value).expect("gather");
assert_eq!(gathered.shape, cpu.shape);
let tol = match provider.precision() {
runmat_accelerate_api::ProviderPrecision::F64 => 1e-10,
runmat_accelerate_api::ProviderPrecision::F32 => 1e-4,
};
for (got, expected) in gathered.data.iter().zip(cpu.data.iter()) {
approx_eq(*got, *expected, tol);
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_like_gpu_prototype() {
test_support::with_test_provider(|provider| {
let tensor = Tensor::new(vec![1.0, 2.0], vec![2, 1]).unwrap();
let proto_view = HostTensorView {
data: &[0.0],
shape: &[1, 1],
};
let proto = provider.upload(&proto_view).expect("upload");
let result = gamma_builtin(
Value::Tensor(tensor.clone()),
vec![Value::from("like"), Value::GpuTensor(proto.clone())],
)
.expect("gamma");
match result {
Value::GpuTensor(handle) => {
let gathered = test_support::gather(Value::GpuTensor(handle)).expect("gather");
approx_eq(gathered.data[0], 1.0, 1e-12);
approx_eq(gathered.data[1], 1.0, 1e-12);
}
other => panic!("expected GPU tensor, got {other:?}"),
}
});
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_like_gpu_complex_result_errors() {
test_support::with_test_provider(|provider| {
let proto_view = HostTensorView {
data: &[0.0],
shape: &[1, 1],
};
let proto = provider.upload(&proto_view).expect("upload");
let err = gamma_builtin(
Value::Complex(0.5, 0.75),
vec![Value::from("like"), Value::GpuTensor(proto)],
)
.expect_err("expected error");
assert!(err.message().contains("only support real numeric outputs"));
});
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_like_complex_requires_host() {
let result = gamma_builtin(
Value::Num(2.0),
vec![Value::from("like"), Value::Complex(1.0, 0.0)],
)
.expect("gamma");
match result {
Value::Complex(re, im) => {
approx_eq(re, 1.0, 1e-12);
approx_eq(im, 0.0, 1e-12);
}
other => panic!("expected complex result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_like_rejects_extra_args() {
let err = gamma_builtin(
Value::Num(1.0),
vec![Value::from("like"), Value::Num(0.0), Value::Num(1.0)],
)
.expect_err("expected error");
assert!(err.message().contains("too many input arguments"));
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn gamma_int_promotes() {
let value = Value::Int(IntValue::I32(4));
match gamma_builtin(value, Vec::new()).expect("gamma") {
Value::Num(v) => approx_eq(v, 6.0, 1e-12),
other => panic!("expected scalar result, got {other:?}"),
}
}
}