add benchmarks ; add explicit #[inline] annotations

This diff adds benchmarks to get more info regarding Issue #1.

It appears that the remaining difference between the "simple"
case and the "cargo" case is the result of a difference in
performance between using `match` and `if` for tight loops.

I suspect it's because of the way that match arms get reordered:
if I manually reorder the "if" statement, I can reproduce the
match performance.

Also added a couple #[inline] annotations in tables.rs, though
the difference in performance in my measurements is negligible.

Bumped version number to 0.1.1.

Author: kwantam

.gitignore

@@ -1,2 +1,3 @@
 target
 Cargo.lock
+scripts/tmp

.travis.yml

@@ -6,6 +6,7 @@ script:
   - cargo clean
   - cargo build --verbose --features default
   - cargo test --verbose --features default
+  - cargo bench --verbose --features default
   - rustdoc --test README.md -L target/debug -L target/debug/deps
   - cargo doc
 after_success: |

Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 
 name = "unicode-width"
-version = "0.1.0"
+version = "0.1.1"
 authors = ["kwantam <[email protected]>"]
 
 homepage = "https://github.com/unicode-rs/unicode-width"

README.md

@@ -33,5 +33,5 @@ to your `Cargo.toml`:
 
 ```toml
 [dependencies]
-unicode-width = "0.1.0"
+unicode-width = "0.1.1"
 ```

scripts/unicode.py

@@ -206,6 +206,7 @@ def emit_charwidth_module(f, width_table):
     #[cfg(feature = "no_std")]
     use core::result::Result::{Ok, Err};
 
+    #[inline]
     fn bsearch_range_value_table(c: char, is_cjk: bool, r: &'static [(char, char, u8, u8)]) -> u8 {
         #[cfg(feature = "no_std")]
         use core::cmp::Ordering::{Equal, Less, Greater};
@@ -226,6 +227,7 @@ def emit_charwidth_module(f, width_table):
 """)
 
     f.write("""
+    #[inline]
     pub fn width(c: char, is_cjk: bool) -> Option<usize> {
         match c as usize {
             _c @ 0 => Some(0),          // null is zero width

src/lib.rs

@@ -39,7 +39,7 @@
 //!
 //! ```toml
 //! [dependencies]
-//! unicode-width = "0.1.0"
+//! unicode-width = "0.1.1"
 //! ```
 
 #![deny(missing_docs, unsafe_code)]
@@ -49,6 +49,8 @@
 #![cfg_attr(feature = "no_std", no_std)]
 #![cfg_attr(feature = "no_std", feature(no_std, core))]
 
+#![cfg_attr(test, feature(test, unicode))]
+
 #[cfg(feature = "no_std")]
 #[macro_use]
 extern crate core;
@@ -57,6 +59,9 @@ extern crate core;
 #[macro_use]
 extern crate std;
 
+#[cfg(test)]
+extern crate test;
+
 #[cfg(feature = "no_std")]
 use core::prelude::*;
 
@@ -70,6 +75,9 @@ use std::ops::Add;
 
 mod tables;
 
+#[cfg(test)]
+mod tests;
+
 /// Methods for determining displayed width of Unicode characters.
 pub trait UnicodeWidthChar {
     /// Returns the character's displayed width in columns, or `None` if the
@@ -92,8 +100,10 @@ pub trait UnicodeWidthChar {
 }
 
 impl UnicodeWidthChar for char {
+    #[inline]
     fn width(self) -> Option<usize> { cw::width(self, false) }
 
+    #[inline]
     fn width_cjk(self) -> Option<usize> { cw::width(self, true) }
 }
 
@@ -121,75 +131,13 @@ pub trait UnicodeWidthStr {
 }
 
 impl UnicodeWidthStr for str {
+    #[inline]
     fn width(&self) -> usize {
         self.chars().map(|c| cw::width(c, false).unwrap_or(0)).fold(0, Add::add)
     }
 
+    #[inline]
     fn width_cjk(&self) -> usize {
         self.chars().map(|c| cw::width(c, true).unwrap_or(0)).fold(0, Add::add)
     }
 }
-
-#[cfg(test)]
-mod tests {
-    #[test]
-    fn test_str() {
-        use super::UnicodeWidthStr;
-
-        assert_eq!(UnicodeWidthStr::width("ｈｅｌｌｏ"), 10);
-        assert_eq!("ｈｅｌｌｏ".width_cjk(), 10);
-        assert_eq!(UnicodeWidthStr::width("\0\0\0\x01\x01"), 0);
-        assert_eq!("\0\0\0\x01\x01".width_cjk(), 0);
-        assert_eq!(UnicodeWidthStr::width(""), 0);
-        assert_eq!("".width_cjk(), 0);
-        assert_eq!(UnicodeWidthStr::width("\u{2081}\u{2082}\u{2083}\u{2084}"), 4);
-        assert_eq!("\u{2081}\u{2082}\u{2083}\u{2084}".width_cjk(), 8);
-    }
-
-    #[test]
-    fn test_char() {
-        use super::UnicodeWidthChar;
-        #[cfg(feature = "no_std")]
-        use core::option::Option::{Some, None};
-
-        assert_eq!(UnicodeWidthChar::width('ｈ'), Some(2));
-        assert_eq!('ｈ'.width_cjk(), Some(2));
-        assert_eq!(UnicodeWidthChar::width('\x00'), Some(0));
-        assert_eq!('\x00'.width_cjk(), Some(0));
-        assert_eq!(UnicodeWidthChar::width('\x01'), None);
-        assert_eq!('\x01'.width_cjk(), None);
-        assert_eq!(UnicodeWidthChar::width('\u{2081}'), Some(1));
-        assert_eq!('\u{2081}'.width_cjk(), Some(2));
-    }
-
-    #[test]
-    fn test_char2() {
-        use super::UnicodeWidthChar;
-        #[cfg(feature = "no_std")]
-        use core::option::Option::{Some, None};
-
-        assert_eq!(UnicodeWidthChar::width('\x00'),Some(0));
-        assert_eq!('\x00'.width_cjk(),Some(0));
-
-        assert_eq!(UnicodeWidthChar::width('\x0A'),None);
-        assert_eq!('\x0A'.width_cjk(),None);
-
-        assert_eq!(UnicodeWidthChar::width('w'),Some(1));
-        assert_eq!('w'.width_cjk(),Some(1));
-
-        assert_eq!(UnicodeWidthChar::width('ｈ'),Some(2));
-        assert_eq!('ｈ'.width_cjk(),Some(2));
-
-        assert_eq!(UnicodeWidthChar::width('\u{AD}'),Some(1));
-        assert_eq!('\u{AD}'.width_cjk(),Some(1));
-
-        assert_eq!(UnicodeWidthChar::width('\u{1160}'),Some(0));
-        assert_eq!('\u{1160}'.width_cjk(),Some(0));
-
-        assert_eq!(UnicodeWidthChar::width('\u{a1}'),Some(1));
-        assert_eq!('\u{a1}'.width_cjk(),Some(2));
-
-        assert_eq!(UnicodeWidthChar::width('\u{300}'),Some(0));
-        assert_eq!('\u{300}'.width_cjk(),Some(0));
-    }
-}

src/tables.rs

@@ -24,6 +24,7 @@ pub mod charwidth {
     #[cfg(feature = "no_std")]
     use core::result::Result::{Ok, Err};
 
+    #[inline]
     fn bsearch_range_value_table(c: char, is_cjk: bool, r: &'static [(char, char, u8, u8)]) -> u8 {
         #[cfg(feature = "no_std")]
         use core::cmp::Ordering::{Equal, Less, Greater};
@@ -42,6 +43,7 @@ pub mod charwidth {
         }
     }
 
+    #[inline]
     pub fn width(c: char, is_cjk: bool) -> Option<usize> {
         match c as usize {
             _c @ 0 => Some(0),          // null is zero width

src/tests.rs

@@ -0,0 +1,148 @@
+// Copyright 2012-2015 The Rust Project Developers. See the COPYRIGHT
+// file at the top-level directory of this distribution and at
+// http://rust-lang.org/COPYRIGHT.
+//
+// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
+// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
+// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
+// option. This file may not be copied, modified, or distributed
+// except according to those terms.
+
+use std::iter;
+use test::{self, Bencher};
+
+use super::UnicodeWidthChar;
+
+#[cfg(feature = "no_std")]
+use std::prelude::v1::*;
+
+#[bench]
+fn cargo(b: &mut Bencher) {
+    let string = iter::repeat('a').take(4096).collect::<String>();
+
+    b.iter(|| {
+        for c in string.chars() {
+            test::black_box(UnicodeWidthChar::width(c));
+        }
+    });
+}
+
+#[bench]
+fn stdlib(b: &mut Bencher) {
+    let string = iter::repeat('a').take(4096).collect::<String>();
+
+    b.iter(|| {
+        for c in string.chars() {
+            test::black_box(c.width(false));
+        }
+    });
+}
+
+#[bench]
+fn simple_if(b: &mut Bencher) {
+    let string = iter::repeat('a').take(4096).collect::<String>();
+
+    b.iter(|| {
+        for c in string.chars() {
+            test::black_box(simple_width_if(c));
+        }
+    });
+}
+
+#[bench]
+fn simple_match(b: &mut Bencher) {
+    let string = iter::repeat('a').take(4096).collect::<String>();
+
+    b.iter(|| {
+        for c in string.chars() {
+            test::black_box(simple_width_match(c));
+        }
+    });
+}
+
+#[inline]
+fn simple_width_if(c: char) -> Option<usize> {
+    let cu = c as u32;
+    if cu < 127 {
+        if cu > 31 {
+            Some(1)
+        } else if cu == 0 {
+            Some(0)
+        } else {
+            None
+        }
+    } else {
+        UnicodeWidthChar::width(c)
+    }
+}
+
+#[inline]
+fn simple_width_match(c: char) -> Option<usize> {
+    match c as u32 {
+        cu if cu == 0 => Some(0),
+        cu if cu < 0x20 => None,
+        cu if cu < 0x7f => Some(1),
+        _ => UnicodeWidthChar::width(c)
+    }
+}
+
+#[test]
+fn test_str() {
+    use super::UnicodeWidthStr;
+
+    assert_eq!(UnicodeWidthStr::width("ｈｅｌｌｏ"), 10);
+    assert_eq!("ｈｅｌｌｏ".width_cjk(), 10);
+    assert_eq!(UnicodeWidthStr::width("\0\0\0\x01\x01"), 0);
+    assert_eq!("\0\0\0\x01\x01".width_cjk(), 0);
+    assert_eq!(UnicodeWidthStr::width(""), 0);
+    assert_eq!("".width_cjk(), 0);
+    assert_eq!(UnicodeWidthStr::width("\u{2081}\u{2082}\u{2083}\u{2084}"), 4);
+    assert_eq!("\u{2081}\u{2082}\u{2083}\u{2084}".width_cjk(), 8);
+}
+
+#[test]
+fn test_char() {
+    use super::UnicodeWidthChar;
+    #[cfg(feature = "no_std")]
+    use core::option::Option::{Some, None};
+
+    assert_eq!(UnicodeWidthChar::width('ｈ'), Some(2));
+    assert_eq!('ｈ'.width_cjk(), Some(2));
+    assert_eq!(UnicodeWidthChar::width('\x00'), Some(0));
+    assert_eq!('\x00'.width_cjk(), Some(0));
+    assert_eq!(UnicodeWidthChar::width('\x01'), None);
+    assert_eq!('\x01'.width_cjk(), None);
+    assert_eq!(UnicodeWidthChar::width('\u{2081}'), Some(1));
+    assert_eq!('\u{2081}'.width_cjk(), Some(2));
+}
+
+#[test]
+fn test_char2() {
+    use super::UnicodeWidthChar;
+    #[cfg(feature = "no_std")]
+    use core::option::Option::{Some, None};
+
+    assert_eq!(UnicodeWidthChar::width('\x00'),Some(0));
+    assert_eq!('\x00'.width_cjk(),Some(0));
+
+    assert_eq!(UnicodeWidthChar::width('\x0A'),None);
+    assert_eq!('\x0A'.width_cjk(),None);
+
+    assert_eq!(UnicodeWidthChar::width('w'),Some(1));
+    assert_eq!('w'.width_cjk(),Some(1));
+
+    assert_eq!(UnicodeWidthChar::width('ｈ'),Some(2));
+    assert_eq!('ｈ'.width_cjk(),Some(2));
+
+    assert_eq!(UnicodeWidthChar::width('\u{AD}'),Some(1));
+    assert_eq!('\u{AD}'.width_cjk(),Some(1));
+
+    assert_eq!(UnicodeWidthChar::width('\u{1160}'),Some(0));
+    assert_eq!('\u{1160}'.width_cjk(),Some(0));
+
+    assert_eq!(UnicodeWidthChar::width('\u{a1}'),Some(1));
+    assert_eq!('\u{a1}'.width_cjk(),Some(2));
+
+    assert_eq!(UnicodeWidthChar::width('\u{300}'),Some(0));
+    assert_eq!('\u{300}'.width_cjk(),Some(0));
+}