Add doc example to lib.rs

Doc fixes

crates/liblinear/src/ffi.rs

@@ -1,4 +1,4 @@
-//! FFI bindings for liblinear.
+//! FFI bindings for LIBLINEAR.
 
 use std::os::raw::c_char;
 

crates/liblinear/src/lib.rs

@@ -2,6 +2,40 @@
 //!
 //! `liblinear` is a Rust wrapper for the [LIBLINEAR](https://www.csie.ntu.edu.tw/~cjlin/liblinear/)
 //! C/C++ machine learning library.
+//!
+//! ## Usage
+//! ```
+//! use liblinear::{
+//! Model,
+//! Parameters,
+//! TrainingInput,
+//! PredictionInput,
+//! solver::L2R_LR,
+//! model::traits::*,
+//! parameter::traits::*,
+//! solver::traits::*,
+//! };
+//!
+//! let x: Vec<Vec<(u32, f64)>> = vec![
+//! vec![(1, 0.1), (3, 0.2)],
+//! vec![(3, 9.9)],
+//! vec![(1, 0.2), (2, 3.2)],
+//! ];
+//! let y = vec![0.0, 1.0, 0.0];
+//!
+//! let mut params = Parameters::<L2R_LR>::default();
+//! params
+//! .bias(0f64)
+//! .stopping_tolerance(0.1f64)
+//! .constraints_violation_cost(0.1f64);
+//!
+//! let model = Model::train(&TrainingInput::from_sparse_features(y, x).unwrap(), &params).unwrap();
+//!
+//! let predicted_class = model
+//! .predict(&PredictionInput::from_sparse_features(vec![(3u32, 9.9f64)]).unwrap())
+//! .unwrap();
+//! println!("{}",predicted_class);
+//! ```
 #[macro_use]
 extern crate num_derive;
 
@@ -12,12 +46,16 @@ pub mod parameter;
 pub mod solver;
 pub mod util;
 
-/// The version of the bundled liblinear C-library.
+pub use model::Model;
+pub use parameter::Parameters;
+pub use util::{PredictionInput, TrainingInput, TrainingInstance};
+
+/// The version of the bundled LIBLINEAR C/C++ library.
 pub fn liblinear_version() -> i32 {
     unsafe { ffi::liblinear_version }
 }
 
-/// Toggles the log output liblinear prints to the program's `stdout`.
+/// Toggles the log output LIBLINEAR prints to the program's `stdout`.
 pub fn toggle_liblinear_stdout_output(state: bool) {
     unsafe {
         match state {

crates/liblinear/src/model.rs

@@ -1,4 +1,4 @@
-//! Types and traits that wrap models.
+//! Types and traits that wrap LIBLINEAR models.
 
 use std::{ffi::CStr, marker::PhantomData};
 
@@ -130,7 +130,6 @@ pub mod traits {
     pub trait NonSingleClassModel {
         /// Returns the bias term corresponding to the class with the given index.
         ///
-        /// For classification models, if label index is not in a valid range, an !!!ERROR value will be returned.
         /// For regression models, the label index is ignored.
         fn label_bias(&self, label_index: u32) -> Result<f64, ModelError>;
     }

crates/liblinear/src/parameter.rs

@@ -1,4 +1,4 @@
-//! Types and traits that wrap hyper parameters.
+//! Types and traits that wrap LIBLINEAR hyper parameters.
 
 use std::marker::PhantomData;
 
@@ -23,6 +23,8 @@ pub mod traits {
     /// [`SupportsInitialSolutions`](crate::solver::traits::SupportsInitialSolutions) trait.
     pub trait SetInitialSolutions {
         /// Set the initial solution specification.
+        ///
+        /// Default: `None`
         fn initial_solutions(&mut self, init_solutions: Vec<f64>) -> &mut Self;
     }
 
@@ -51,7 +53,7 @@ pub mod traits {
     pub trait SetRegressionLossSensitivity {
         /// Set the tolerance margin/loss sensitivity of support vector regression (parameter `p`).
         ///
-        /// Default: `0.1
+        /// Default: `0.1`
         fn regression_loss_sensitivity(&mut self, p: f64) -> &mut Self;
     }
 }