]> git.proxmox.com Git - rustc.git/blobdiff - vendor/rand-0.7.3/src/distributions/normal.rs
Merge tag 'debian/1.52.1+dfsg1-1_exp2' into proxmox/buster
[rustc.git] / vendor / rand-0.7.3 / src / distributions / normal.rs
diff --git a/vendor/rand-0.7.3/src/distributions/normal.rs b/vendor/rand-0.7.3/src/distributions/normal.rs
new file mode 100644 (file)
index 0000000..ec62fa9
--- /dev/null
@@ -0,0 +1,177 @@
+// Copyright 2018 Developers of the Rand project.
+// Copyright 2013 The Rust Project Developers.
+//
+// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
+// https://www.apache.org/licenses/LICENSE-2.0> or the MIT license
+// <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your
+// option. This file may not be copied, modified, or distributed
+// except according to those terms.
+
+//! The normal and derived distributions.
+#![allow(deprecated)]
+
+use crate::distributions::utils::ziggurat;
+use crate::distributions::{ziggurat_tables, Distribution, Open01};
+use crate::Rng;
+
+/// Samples floating-point numbers according to the normal distribution
+/// `N(0, 1)` (a.k.a. a standard normal, or Gaussian). This is equivalent to
+/// `Normal::new(0.0, 1.0)` but faster.
+///
+/// See `Normal` for the general normal distribution.
+///
+/// Implemented via the ZIGNOR variant[^1] of the Ziggurat method.
+///
+/// [^1]: Jurgen A. Doornik (2005). [*An Improved Ziggurat Method to
+///       Generate Normal Random Samples*](
+///       https://www.doornik.com/research/ziggurat.pdf).
+///       Nuffield College, Oxford
+#[deprecated(since = "0.7.0", note = "moved to rand_distr crate")]
+#[derive(Clone, Copy, Debug)]
+pub struct StandardNormal;
+
+impl Distribution<f64> for StandardNormal {
+    fn sample<R: Rng + ?Sized>(&self, rng: &mut R) -> f64 {
+        #[inline]
+        fn pdf(x: f64) -> f64 {
+            (-x * x / 2.0).exp()
+        }
+        #[inline]
+        fn zero_case<R: Rng + ?Sized>(rng: &mut R, u: f64) -> f64 {
+            // compute a random number in the tail by hand
+
+            // strange initial conditions, because the loop is not
+            // do-while, so the condition should be true on the first
+            // run, they get overwritten anyway (0 < 1, so these are
+            // good).
+            let mut x = 1.0f64;
+            let mut y = 0.0f64;
+
+            while -2.0 * y < x * x {
+                let x_: f64 = rng.sample(Open01);
+                let y_: f64 = rng.sample(Open01);
+
+                x = x_.ln() / ziggurat_tables::ZIG_NORM_R;
+                y = y_.ln();
+            }
+
+            if u < 0.0 {
+                x - ziggurat_tables::ZIG_NORM_R
+            } else {
+                ziggurat_tables::ZIG_NORM_R - x
+            }
+        }
+
+        ziggurat(
+            rng,
+            true, // this is symmetric
+            &ziggurat_tables::ZIG_NORM_X,
+            &ziggurat_tables::ZIG_NORM_F,
+            pdf,
+            zero_case,
+        )
+    }
+}
+
+/// The normal distribution `N(mean, std_dev**2)`.
+///
+/// This uses the ZIGNOR variant of the Ziggurat method, see [`StandardNormal`]
+/// for more details.
+///
+/// Note that [`StandardNormal`] is an optimised implementation for mean 0, and
+/// standard deviation 1.
+///
+/// [`StandardNormal`]: crate::distributions::StandardNormal
+#[deprecated(since = "0.7.0", note = "moved to rand_distr crate")]
+#[derive(Clone, Copy, Debug)]
+pub struct Normal {
+    mean: f64,
+    std_dev: f64,
+}
+
+impl Normal {
+    /// Construct a new `Normal` distribution with the given mean and
+    /// standard deviation.
+    ///
+    /// # Panics
+    ///
+    /// Panics if `std_dev < 0`.
+    #[inline]
+    pub fn new(mean: f64, std_dev: f64) -> Normal {
+        assert!(std_dev >= 0.0, "Normal::new called with `std_dev` < 0");
+        Normal { mean, std_dev }
+    }
+}
+impl Distribution<f64> for Normal {
+    fn sample<R: Rng + ?Sized>(&self, rng: &mut R) -> f64 {
+        let n = rng.sample(StandardNormal);
+        self.mean + self.std_dev * n
+    }
+}
+
+
+/// The log-normal distribution `ln N(mean, std_dev**2)`.
+///
+/// If `X` is log-normal distributed, then `ln(X)` is `N(mean, std_dev**2)`
+/// distributed.
+#[deprecated(since = "0.7.0", note = "moved to rand_distr crate")]
+#[derive(Clone, Copy, Debug)]
+pub struct LogNormal {
+    norm: Normal,
+}
+
+impl LogNormal {
+    /// Construct a new `LogNormal` distribution with the given mean
+    /// and standard deviation.
+    ///
+    /// # Panics
+    ///
+    /// Panics if `std_dev < 0`.
+    #[inline]
+    pub fn new(mean: f64, std_dev: f64) -> LogNormal {
+        assert!(std_dev >= 0.0, "LogNormal::new called with `std_dev` < 0");
+        LogNormal {
+            norm: Normal::new(mean, std_dev),
+        }
+    }
+}
+impl Distribution<f64> for LogNormal {
+    fn sample<R: Rng + ?Sized>(&self, rng: &mut R) -> f64 {
+        self.norm.sample(rng).exp()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::{LogNormal, Normal};
+    use crate::distributions::Distribution;
+
+    #[test]
+    fn test_normal() {
+        let norm = Normal::new(10.0, 10.0);
+        let mut rng = crate::test::rng(210);
+        for _ in 0..1000 {
+            norm.sample(&mut rng);
+        }
+    }
+    #[test]
+    #[should_panic]
+    fn test_normal_invalid_sd() {
+        Normal::new(10.0, -1.0);
+    }
+
+
+    #[test]
+    fn test_log_normal() {
+        let lnorm = LogNormal::new(10.0, 10.0);
+        let mut rng = crate::test::rng(211);
+        for _ in 0..1000 {
+            lnorm.sample(&mut rng);
+        }
+    }
+    #[test]
+    #[should_panic]
+    fn test_log_normal_invalid_sd() {
+        LogNormal::new(10.0, -1.0);
+    }
+}