More idiomatic serde support

* Follow the suggestions from the Serde docs.
* Rename the feature from `serde` to `serde1`.
* Fix a doctest.
* Mention the feature in the README.

Cargo.toml

@@ -10,6 +10,9 @@ readme = "README.md"
 repository = "https://github.com/vks/average"
 version = "0.8.0"
 
+[features]
+serde1 = ["serde", "serde_derive"]
+
 [[bench]]
 harness = false
 name = "mean"
@@ -26,16 +29,13 @@ name = "kurtosis"
 num-traits = "0.2"
 num-integer = "0.1"
 float-ord = "0.2"
+serde = { version = "1", optional = true }
+serde_derive = { version = "1", optional = true }
 
 [dependencies.conv]
 default-features = false
 version = "0.3"
 
-[dependencies.serde]
-features = ["derive"]
-optional = true
-version = "1"
-
 [dev-dependencies]
 bencher = "0.1"
 rand = "0.5"

README.md

@@ -15,6 +15,7 @@ easily parallelized by using `merge`.
 [Latest Version]: https://img.shields.io/crates/v/average.svg
 [crates.io]: https://crates.io/crates/average
 
+
 ## Implemented statistics
 
 * Mean and its error.
@@ -24,6 +25,14 @@ easily parallelized by using `merge`.
 * Quantile.
 * Histogram.
 
+
+## Crate features
+
+The following optional feature is available:
+
+* `serde1` enables serialization, via Serde version 1.
+
+
 ## Related Projects
 
 * [`quantiles`](https://crates.io/crates/quantiles):

src/lib.rs

@@ -89,9 +89,10 @@
 
 extern crate conv;
 extern crate float_ord;
-#[cfg(feature = "serde")]
+#[cfg(feature = "serde1")]
-#[macro_use]
 extern crate serde;
+#[cfg(feature = "serde1")]
+#[macro_use] extern crate serde_derive;
 extern crate num_traits;
 extern crate num_integer;
 

src/minmax.rs

@@ -24,7 +24,7 @@ fn max(a: f64, b: f64) -> f64 {
 /// println!("The minimum is {}.", a.min());
 /// ```
 #[derive(Debug, Clone)]
-#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "serde1", derive(Serialize, Deserialize))]
 pub struct Min {
     x: f64,
 }
@@ -104,7 +104,7 @@ impl Merge for Min {
 /// assert_eq!(a.max(), 5.);
 /// ```
 #[derive(Debug, Clone)]
-#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "serde1", derive(Serialize, Deserialize))]
 pub struct Max {
     x: f64,
 }

src/moments/kurtosis.rs

@@ -3,7 +3,7 @@
 ///
 /// This can be used to estimate the standard error of the mean.
 #[derive(Debug, Clone)]
-#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "serde1", derive(Serialize, Deserialize))]
 pub struct Kurtosis {
     /// Estimator of mean, variance and skewness.
     avg: Skewness,

src/moments/mean.rs

@@ -10,7 +10,7 @@
 /// println!("The mean is {}.", a.mean());
 /// ```
 #[derive(Debug, Clone)]
-#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "serde1", derive(Serialize, Deserialize))]
 pub struct Mean {
     /// Mean value.
     avg: f64,

src/moments/mod.rs

@@ -32,6 +32,10 @@ pub type MeanWithError = Variance;
 /// # extern crate conv;
 /// # extern crate num_integer;
 /// # extern crate num_traits;
+/// #[cfg(feature = "serde1")]
+/// extern crate serde;
+/// #[cfg(feature = "serde1")]
+/// #[macro_use] extern crate serde_derive;
 /// # #[macro_use] extern crate average;
 /// # fn main() {
 /// define_moments!(Moments4, 4);
@@ -65,7 +69,7 @@ macro_rules! define_moments {
         /// Estimate the first N moments of a sequence of numbers a sequence of numbers
         /// ("population").
         #[derive(Debug, Clone)]
-        #[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+        #[cfg_attr(feature = "serde1", derive(Serialize, Deserialize))]
         pub struct $name {
             /// Number of samples.
             ///
@@ -126,13 +130,12 @@ macro_rules! define_moments {
             #[inline]
             pub fn standardized_moment(&self, p: usize) -> f64 {
                 match p {
-                    0 => n,
+                    0 => f64::approx_from(self.n).unwrap(),
                     1 => 0.,
                     2 => 1.,
                     _ => {
                         let variance = self.central_moment(2);
                         assert_ne!(variance, 0.);
-                        let n = f64::approx_from(self.n).unwrap();
                         self.central_moment(p) / pow(variance.sqrt(), p)
                     },
                 }

src/moments/skewness.rs

@@ -3,7 +3,7 @@
 ///
 /// This can be used to estimate the standard error of the mean.
 #[derive(Debug, Clone)]
-#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "serde1", derive(Serialize, Deserialize))]
 pub struct Skewness {
     /// Estimator of mean and variance.
     avg: MeanWithError,

src/moments/variance.rs

@@ -13,7 +13,7 @@
 /// println!("The mean is {} ± {}.", a.mean(), a.error());
 /// ```
 #[derive(Debug, Clone)]
-#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "serde1", derive(Serialize, Deserialize))]
 pub struct Variance {
     /// Estimator of average.
     avg: Mean,

src/quantile.rs

@@ -15,7 +15,7 @@ use super::Estimate;
 // This uses the P² algorithm introduced here:
 // http://www.cs.wustl.edu/~jain/papers/ftp/psqr.pdf
 #[derive(Debug, Clone)]
-#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "serde1", derive(Serialize, Deserialize))]
 pub struct Quantile {
     /// Marker heights.
     q: [f64; 5],

src/weighted_mean.rs

@@ -17,7 +17,7 @@ use super::{MeanWithError, Estimate, Merge};
 /// println!("The weighted mean is {}.", a.mean());
 /// ```
 #[derive(Debug, Clone)]
-#[cfg_attr(feature = "serde", derive(Serialize, Deserialize))]
+#[cfg_attr(feature = "serde1", derive(Serialize, Deserialize))]
 pub struct WeightedMean  {
     /// Sum of the weights.
     weight_sum: f64,

tests/moments.rs

@@ -59,7 +59,7 @@ fn simple_serde() {
     let a: Moments4 = (1..6).map(f64::from).collect();
     let b = serde_json::to_string(&a).unwrap();
     assert_eq!(&b, "{\"n\":5,\"avg\":3.0,\"m\":[10.0,1.7763568394002506e-15,34.00000000000001]}");
-    let mut c: Moments = serde_json::from_str(&b).unwrap();
+    let mut c: Moments4 = serde_json::from_str(&b).unwrap();
     assert_eq!(c.len(), 5);
     assert_eq!(c.mean(), 3.0);
     assert_eq!(c.central_moment(0), 1.0);