update distributable to include readme and license

The code has suffered and needs refactoring/commenting. BUT... IT WORKS!

Cargo.toml

@@ -1,14 +1,14 @@
 [package]
 publish = false
-name = "xrep"
+name = "ripgrep"
 version = "0.1.0"  #:version
 authors = ["Andrew Gallant <jamslam@gmail.com>"]
 description = """
 Line oriented search tool using Rust's regex library.
 """
-documentation = "https://github.com/BurntSushi/xrep"
-homepage = "https://github.com/BurntSushi/xrep"
-repository = "https://github.com/BurntSushi/xrep"
+documentation = "https://github.com/BurntSushi/ripgrep"
+homepage = "https://github.com/BurntSushi/ripgrep"
+repository = "https://github.com/BurntSushi/ripgrep"
 readme = "README.md"
 keywords = ["regex", "grep", "egrep", "search", "pattern"]
 license = "Unlicense/MIT"
@@ -16,7 +16,7 @@ license = "Unlicense/MIT"
 [[bin]]
 bench = false
 path = "src/main.rs"
-name = "xrep"
+name = "rg"
 
 [dependencies]
 crossbeam = "0.2"

appveyor.yml

@@ -1,6 +1,6 @@
 environment:
   global:
-    PROJECT_NAME: xrep
+    PROJECT_NAME: rg
   matrix:
     # Nightly channel
     - TARGET: i686-pc-windows-gnu
@@ -32,16 +32,14 @@ build: false
 # Equivalent to Travis' `script` phase
 # TODO modify this phase as you see fit
 test_script:
-  # - cargo build --verbose
-  - cargo test matchslash2 -- --nocapture
+  - cargo test --verbose
 
 before_deploy:
   # Generate artifacts for release
-  - SET RUSTFLAGS="-C target-feature=+ssse3"
-  - cargo build --release --features simd-accel
+  # TODO(burntsushi): How can we enable SSSE3 on Windows?
+  - cargo build --release
   - mkdir staging
-  # TODO update this part to copy the artifacts that make sense for your project
-  - copy target\release\xrep.exe staging
+  - copy target\release\rg.exe staging
   - cd staging
     # release zipfile will look like 'rust-everywhere-v1.2.3-x86_64-pc-windows-msvc'
   - 7z a ../%PROJECT_NAME%-%APPVEYOR_REPO_TAG_NAME%-%TARGET%.zip *

ci/before_deploy.sh

@@ -6,23 +6,22 @@ set -ex
 
 # Generate artifacts for release
 mk_artifacts() {
-    RUSTFLAGS="-C target-feature=+ssse3" cargo build --target $TARGET --release --features simd-accel
+    RUSTFLAGS="-C target-feature=+ssse3" \
+      cargo build --target $TARGET --release --features simd-accel
 }
 
 mk_tarball() {
     # create a "staging" directory
     local td=$(mktempd)
     local out_dir=$(pwd)
+    local name="${PROJECT_NAME}-${TRAVIS_TAG}-${TARGET}"
+    mkdir "$td/$name"
 
-    # TODO update this part to copy the artifacts that make sense for your project
-    # NOTE All Cargo build artifacts will be under the 'target/$TARGET/{debug,release}'
-    cp target/$TARGET/release/xrep $td
+    cp target/$TARGET/release/rg "$td/$name/"
+    cp {README,UNLICENSE,COPYING,LICENSE_MIT} "$td/$name/"
 
     pushd $td
-
-    # release tarball will look like 'rust-everywhere-v1.2.3-x86_64-unknown-linux-gnu.tar.gz'
-    tar czf $out_dir/${PROJECT_NAME}-${TRAVIS_TAG}-${TARGET}.tar.gz *
-
+    tar czf "$out_dir/$name.tar.gz" *
     popd
     rm -r $td
 }

ci/script.sh

@@ -42,7 +42,7 @@ run_test_suite() {
     cargo test --target $TARGET
 
     # sanity check the file type
-    file target/$TARGET/debug/xrep
+    file target/$TARGET/debug/rg
 }
 
 main() {

grep/Cargo.toml

@@ -6,14 +6,15 @@ authors = ["Andrew Gallant <jamslam@gmail.com>"]
 description = """
 Fast line oriented regex searching as a library.
 """
-documentation = "https://github.com/BurntSushi/xrep"
-homepage = "https://github.com/BurntSushi/xrep"
-repository = "https://github.com/BurntSushi/xrep"
+documentation = "https://github.com/BurntSushi/ripgrep"
+homepage = "https://github.com/BurntSushi/ripgrep"
+repository = "https://github.com/BurntSushi/ripgrep"
 readme = "README.md"
 keywords = ["regex", "grep", "egrep", "search", "pattern"]
 license = "Unlicense/MIT"
 
 [dependencies]
+log = "0.3"
 memchr = "0.1"
 memmap = "0.2"
 regex = "0.1.75"

grep/src/lib.rs

@@ -4,6 +4,8 @@
 A fast line oriented regex searcher.
 */
 
+#[macro_use]
+extern crate log;
 extern crate memchr;
 extern crate regex;
 extern crate regex_syntax as syntax;

grep/src/literals.rs

@@ -1,13 +1,22 @@
+/*!
+The literals module is responsible for extracting *inner* literals out of the
+AST of a regular expression. Normally this is the job of the regex engine
+itself, but the regex engine doesn't look for inner literals. Since we're doing
+line based searching, we can use them, so we need to do it ourselves.
+
+Note that this implementation is incredibly suspicious. We need something more
+principled.
+*/
 use std::cmp;
 use std::iter;
 
 use regex::bytes::Regex;
 use syntax::{
     Expr, Literals, Lit,
-    Repeater,
+    ByteClass, ByteRange, CharClass, ClassRange, Repeater,
 };
 
-#[derive(Debug)]
+#[derive(Clone, Debug)]
 pub struct LiteralSets {
     prefixes: Literals,
     suffixes: Literals,
@@ -27,6 +36,7 @@ impl LiteralSets {
 
     pub fn to_regex(&self) -> Option<Regex> {
         if self.prefixes.all_complete() && !self.prefixes.is_empty() {
+            debug!("literal prefixes detected: {:?}", self.prefixes);
             // When this is true, the regex engine will do a literal scan.
             return None;
         }
@@ -56,13 +66,27 @@ impl LiteralSets {
         if suf_lcs.len() > lit.len() {
             lit = suf_lcs;
         }
-        if req.len() > lit.len() {
+        if req_lits.len() == 1 && req.len() > lit.len() {
             lit = req;
         }
-        if lit.is_empty() {
+
+        // Special case: if we detected an alternation of inner required
+        // literals and its longest literal is bigger than the longest
+        // prefix/suffix, then choose the alternation. In practice, this
+        // helps with case insensitive matching, which can generate lots of
+        // inner required literals.
+        let any_empty = req_lits.iter().any(|lit| lit.is_empty());
+        if req.len() > lit.len() && req_lits.len() > 1 && !any_empty {
+            debug!("required literals found: {:?}", req_lits);
+            let alts: Vec<String> =
+                req_lits.into_iter().map(|x| bytes_to_regex(x)).collect();
+            // Literals always compile.
+            Some(Regex::new(&alts.join("|")).unwrap())
+        } else if lit.is_empty() {
             None
         } else {
             // Literals always compile.
+            debug!("required literal found: {:?}", show(lit));
             Some(Regex::new(&bytes_to_regex(lit)).unwrap())
         }
     }
@@ -75,14 +99,30 @@ fn union_required(expr: &Expr, lits: &mut Literals) {
             let s: String = chars.iter().cloned().collect();
             lits.cross_add(s.as_bytes());
         }
-        Literal { casei: true, .. } => {
-            lits.cut();
+        Literal { ref chars, casei: true } => {
+            for &c in chars {
+                let cls = CharClass::new(vec![
+                    ClassRange { start: c, end: c },
+                ]).case_fold();
+                if !lits.add_char_class(&cls) {
+                    lits.cut();
+                    return;
+                }
+            }
         }
         LiteralBytes { ref bytes, casei: false } => {
             lits.cross_add(bytes);
         }
-        LiteralBytes { casei: true, .. } => {
-            lits.cut();
+        LiteralBytes { ref bytes, casei: true } => {
+            for &b in bytes {
+                let cls = ByteClass::new(vec![
+                    ByteRange { start: b, end: b },
+                ]).case_fold();
+                if !lits.add_byte_class(&cls) {
+                    lits.cut();
+                    return;
+                }
+            }
         }
         Class(_) => {
             lits.cut();
@@ -205,3 +245,18 @@ fn bytes_to_regex(bs: &[u8]) -> String {
     }
     s
 }
+
+/// Converts arbitrary bytes to a nice string.
+fn show(bs: &[u8]) -> String {
+    // Why aren't we using this to feed to the regex? Doesn't really matter
+    // I guess. ---AG
+    use std::ascii::escape_default;
+    use std::str;
+
+    let mut nice = String::new();
+    for &b in bs {
+        let part: Vec<u8> = escape_default(b).collect();
+        nice.push_str(str::from_utf8(&part).unwrap());
+    }
+    nice
+}

grep/src/search.rs

@@ -152,6 +152,7 @@ impl GrepBuilder {
                  .unicode(true)
                  .case_insensitive(self.opts.case_insensitive)
                  .parse(&self.pattern));
+        debug!("regex ast:\n{:#?}", expr);
         Ok(try!(nonl::remove(expr, self.opts.line_terminator)))
     }
 }
@@ -253,7 +254,7 @@ impl<'b, 's> Iterator for Iter<'b, 's> {
             self.start = self.buf.len();
             return None;
         }
-        self.start = mat.end + 1;
+        self.start = mat.end;
         Some(mat)
     }
 }

src/args.rs

@@ -16,6 +16,7 @@ use ignore::Ignore;
 use out::Out;
 use printer::Printer;
 use search::{InputBuffer, Searcher};
+use search_buffer::BufferSearcher;
 use sys;
 use types::{FileTypeDef, Types, TypesBuilder};
 use walk;
@@ -27,13 +28,13 @@ use Result;
 /// If you've never heard of Docopt before, see: http://docopt.org
 /// (TL;DR: The CLI parser is generated from the usage string below.)
 const USAGE: &'static str = "
-Usage: xrep [options] <pattern> [<path> ...]
-       xrep [options] --files [<path> ...]
-       xrep [options] --type-list
-       xrep --help
-       xrep --version
+Usage: rg [options] <pattern> [<path> ...]
+       rg [options] --files [<path> ...]
+       rg [options] --type-list
+       rg --help
+       rg --version
 
-xrep is like the silver searcher and grep, but faster than both.
+rg combines the usability of the silver search with the raw speed of grep.
 
 Common options:
     -a, --text                 Search binary files as if they were text.
@@ -75,6 +76,11 @@ Less common options:
     -C, --context NUM
         Show NUM lines before and after each match.
 
+    --column
+        Show column numbers (1 based) in output. This only shows the column
+        numbers for the first match on each line. Note that this doesn't try
+        to account for Unicode. One byte is equal to one column.
+
     --context-separator ARG
         The string to use when separating non-continuous context lines. Escape
         sequences may be used. [default: --]
@@ -106,8 +112,16 @@ Less common options:
         The byte to use for a line terminator. Escape sequences may be used.
         [default: \\n]
 
+    --mmap
+        Search using memory maps when possible. This is enabled by default
+        when ripgrep thinks it will be faster. (Note that mmap searching
+        doesn't current support the various context related options.)
+
+    --no-mmap
+        Never use memory maps, even when they might be faster.
+
     --no-ignore
-        Don't respect ignore files (.gitignore, .xrepignore, etc.)
+        Don't respect ignore files (.gitignore, .rgignore, etc.)
 
     --no-ignore-parent
         Don't respect ignore files in parent directories.
@@ -123,7 +137,7 @@ Less common options:
         (capped at 6). [default: 0]
 
     --version
-        Show the version number of xrep and exit.
+        Show the version number of ripgrep and exit.
 
 File type management options:
     --type-list
@@ -138,7 +152,7 @@ File type management options:
 ";
 
 /// RawArgs are the args as they are parsed from Docopt. They aren't used
-/// directly by the rest of xrep.
+/// directly by the rest of ripgrep.
 #[derive(Debug, RustcDecodable)]
 pub struct RawArgs {
     arg_pattern: String,
@@ -146,6 +160,7 @@ pub struct RawArgs {
     flag_after_context: usize,
     flag_before_context: usize,
     flag_color: String,
+    flag_column: bool,
     flag_context: usize,
     flag_context_separator: String,
     flag_count: bool,
@@ -160,10 +175,12 @@ pub struct RawArgs {
     flag_line_number: bool,
     flag_line_terminator: String,
     flag_literal: bool,
+    flag_mmap: bool,
     flag_no_heading: bool,
     flag_no_ignore: bool,
     flag_no_ignore_parent: bool,
     flag_no_line_number: bool,
+    flag_no_mmap: bool,
     flag_pretty: bool,
     flag_quiet: bool,
     flag_replace: Option<String>,
@@ -186,17 +203,20 @@ pub struct Args {
     after_context: usize,
     before_context: usize,
     color: bool,
+    column: bool,
     context_separator: Vec<u8>,
     count: bool,
     eol: u8,
     files: bool,
     follow: bool,
     glob_overrides: Option<Gitignore>,
+    grep: Grep,
     heading: bool,
     hidden: bool,
     ignore_case: bool,
     invert_match: bool,
     line_number: bool,
+    mmap: bool,
     no_ignore: bool,
     no_ignore_parent: bool,
     quiet: bool,
@@ -210,7 +230,7 @@ pub struct Args {
 }
 
 impl RawArgs {
-    /// Convert arguments parsed into a configuration used by xrep.
+    /// Convert arguments parsed into a configuration used by ripgrep.
     fn to_args(&self) -> Result<Args> {
         let pattern = {
             let pattern =
@@ -243,6 +263,19 @@ impl RawArgs {
             } else {
                 (self.flag_after_context, self.flag_before_context)
             };
+        let mmap =
+            if before_context > 0 || after_context > 0 || self.flag_no_mmap {
+                false
+            } else if self.flag_mmap {
+                true
+            } else {
+                // If we're only searching a few paths and all of them are
+                // files, then memory maps are probably faster.
+                paths.len() <= 10 && paths.iter().all(|p| p.is_file())
+            };
+        if mmap {
+            debug!("will try to use memory maps");
+        }
         let eol = {
             let eol = unescape(&self.flag_line_terminator);
             if eol.is_empty() {
@@ -283,23 +316,32 @@ impl RawArgs {
         btypes.add_defaults();
         try!(self.add_types(&mut btypes));
         let types = try!(btypes.build());
+        let grep = try!(
+            GrepBuilder::new(&pattern)
+                .case_insensitive(self.flag_ignore_case)
+                .line_terminator(eol)
+                .build()
+        );
         let mut args = Args {
             pattern: pattern,
             paths: paths,
             after_context: after_context,
             before_context: before_context,
             color: color,
+            column: self.flag_column,
             context_separator: unescape(&self.flag_context_separator),
             count: self.flag_count,
             eol: eol,
             files: self.flag_files,
             follow: self.flag_follow,
             glob_overrides: glob_overrides,
+            grep: grep,
             heading: !self.flag_no_heading && self.flag_heading,
             hidden: self.flag_hidden,
             ignore_case: self.flag_ignore_case,
             invert_match: self.flag_invert_match,
             line_number: !self.flag_no_line_number && self.flag_line_number,
+            mmap: mmap,
             no_ignore: self.flag_no_ignore,
             no_ignore_parent: self.flag_no_ignore_parent,
             quiet: self.flag_quiet,
@@ -345,7 +387,7 @@ impl Args {
     ///
     /// If a CLI usage error occurred, then exit the process and print a usage
     /// or error message. Similarly, if the user requested the version of
-    /// xrep, then print the version and exit.
+    /// ripgrep, then print the version and exit.
     ///
     /// Also, initialize a global logger.
     pub fn parse() -> Result<Args> {
@@ -367,7 +409,7 @@ impl Args {
         raw.to_args().map_err(From::from)
     }
 
-    /// Returns true if xrep should print the files it will search and exit
+    /// Returns true if ripgrep should print the files it will search and exit
     /// (but not do any actual searching).
     pub fn files(&self) -> bool {
         self.files
@@ -378,12 +420,8 @@ impl Args {
     /// basic searching of regular expressions in a single buffer.
     ///
     /// The pattern and other flags are taken from the command line.
-    pub fn grep(&self) -> Result<Grep> {
-        GrepBuilder::new(&self.pattern)
-            .case_insensitive(self.ignore_case)
-            .line_terminator(self.eol)
-            .build()
-            .map_err(From::from)
+    pub fn grep(&self) -> Grep {
+        self.grep.clone()
     }
 
     /// Creates a new input buffer that is used in searching.
@@ -393,10 +431,16 @@ impl Args {
         inp
     }
 
+    /// Whether we should prefer memory maps for searching or not.
+    pub fn mmap(&self) -> bool {
+        self.mmap
+    }
+
     /// Create a new printer of individual search results that writes to the
     /// writer given.
     pub fn printer<W: Send + io::Write>(&self, wtr: W) -> Printer<W> {
         let mut p = Printer::new(wtr, self.color)
+            .column(self.column)
             .context_separator(self.context_separator.clone())
             .eol(self.eol)
             .heading(self.heading)
@@ -446,6 +490,24 @@ impl Args {
             .text(self.text)
     }
 
+    /// Create a new line based searcher whose configuration is taken from the
+    /// command line. This search operates on an entire file all once (which
+    /// may have been memory mapped).
+    pub fn searcher_buffer<'a, W: Send + io::Write>(
+        &self,
+        printer: &'a mut Printer<W>,
+        grep: &'a Grep,
+        path: &'a Path,
+        buf: &'a [u8],
+    ) -> BufferSearcher<'a, W> {
+        BufferSearcher::new(printer, grep, path, buf)
+            .count(self.count)
+            .eol(self.eol)
+            .line_number(self.line_number)
+            .invert_match(self.invert_match)
+            .text(self.text)
+    }
+
     /// Returns the number of worker search threads that should be used.
     pub fn threads(&self) -> usize {
         self.threads
@@ -456,8 +518,8 @@ impl Args {
         &self.type_defs
     }
 
-    /// Returns true if xrep should print the type definitions currently loaded
-    /// and then exit.
+    /// Returns true if ripgrep should print the type definitions currently
+    /// loaded and then exit.
     pub fn type_list(&self) -> bool {
         self.type_list
     }

src/gitignore.rs

@@ -9,7 +9,7 @@ The motivation for this submodule is performance and portability:
 2. We could shell out to a `git` sub-command like ls-files or status, but it
    seems better to not rely on the existence of external programs for a search
    tool. Besides, we need to implement this logic anyway to support things like
-   an .xrepignore file.
+   an .rgignore file.
 
 The key implementation detail here is that a single gitignore file is compiled
 into a single RegexSet, which can be used to report which globs match a
@@ -379,7 +379,7 @@ mod tests {
         };
     }
 
-    const ROOT: &'static str = "/home/foobar/rust/xrep";
+    const ROOT: &'static str = "/home/foobar/rust/rg";
 
     ignored!(ig1, ROOT, "months", "months");
     ignored!(ig2, ROOT, "*.lock", "Cargo.lock");

src/glob.rs

@@ -29,7 +29,6 @@ to make its way into `glob` proper.
 use std::error::Error as StdError;
 use std::fmt;
 use std::iter;
-use std::path;
 use std::str;
 
 use regex;
@@ -214,7 +213,7 @@ impl Pattern {
     /// regular expression and will represent the matching semantics of this
     /// glob pattern and the options given.
     pub fn to_regex_with(&self, options: &MatchOptions) -> String {
-        let sep = regex::quote(&path::MAIN_SEPARATOR.to_string());
+        let seps = regex::quote(r"/\");
         let mut re = String::new();
         re.push_str("(?-u)");
         if options.case_insensitive {
@@ -235,26 +234,27 @@ impl Pattern {
                 }
                 Token::Any => {
                     if options.require_literal_separator {
-                        re.push_str(&format!("[^{}]", sep));
+                        re.push_str(&format!("[^{}]", seps));
                     } else {
                         re.push_str(".");
                     }
                 }
                 Token::ZeroOrMore => {
                     if options.require_literal_separator {
-                        re.push_str(&format!("[^{}]*", sep));
+                        re.push_str(&format!("[^{}]*", seps));
                     } else {
                         re.push_str(".*");
                     }
                 }
                 Token::RecursivePrefix => {
-                    re.push_str(&format!("(?:{sep}?|.*{sep})", sep=sep));
+                    re.push_str(&format!("(?:[{sep}]?|.*[{sep}])", sep=seps));
                 }
                 Token::RecursiveSuffix => {
-                    re.push_str(&format!("(?:{sep}?|{sep}.*)", sep=sep));
+                    re.push_str(&format!("(?:[{sep}]?|[{sep}].*)", sep=seps));
                 }
                 Token::RecursiveZeroOrMore => {
-                    re.push_str(&format!("(?:{sep}|{sep}.*{sep})", sep=sep));
+                    re.push_str(&format!("(?:[{sep}]|[{sep}].*[{sep}])",
+                                         sep=seps));
                 }
                 Token::Class { negated, ref ranges } => {
                     re.push('[');
@@ -480,9 +480,9 @@ mod tests {
                 let pat = Pattern::new($pat).unwrap();
                 let path = &Path::new($path).to_str().unwrap();
                 let re = Regex::new(&pat.to_regex_with(&$options)).unwrap();
-                println!("PATTERN: {}", $pat);
-                println!("REGEX: {:?}", re);
-                println!("PATH: {}", path);
+                // println!("PATTERN: {}", $pat);
+                // println!("REGEX: {:?}", re);
+                // println!("PATH: {}", path);
                 assert!(!re.is_match(path.as_bytes()));
             }
         };
@@ -564,12 +564,11 @@ mod tests {
         case_insensitive: true,
         require_literal_separator: false,
     };
-    const SEP: char = ::std::path::MAIN_SEPARATOR;
 
     toregex!(re_casei, "a", "(?i)^a$", &CASEI);
 
-    toregex!(re_slash1, "?", format!("^[^{}]$", SEP), SLASHLIT);
-    toregex!(re_slash2, "*", format!("^[^{}]*$", SEP), SLASHLIT);
+    toregex!(re_slash1, "?", r"^[^/\\]$", SLASHLIT);
+    toregex!(re_slash2, "*", r"^[^/\\]*$", SLASHLIT);
 
     toregex!(re1, "a", "^a$");
     toregex!(re2, "?", "^.$");

src/ignore.rs

@@ -5,7 +5,7 @@ whether a *single* file path should be searched or not.
 In general, there are two ways to ignore a particular file:
 
 1. Specify an ignore rule in some "global" configuration, such as a
-   $HOME/.xrepignore or on the command line.
+   $HOME/.rgignore or on the command line.
 2. A specific ignore file (like .gitignore) found during directory traversal.
 
 The `IgnoreDir` type handles ignore patterns for any one particular directory
@@ -24,7 +24,7 @@ use types::Types;
 const IGNORE_NAMES: &'static [&'static str] = &[
     ".gitignore",
     ".agignore",
-    ".xrepignore",
+    ".rgignore",
 ];
 
 /// Represents an error that can occur when parsing a gitignore file.
@@ -257,8 +257,8 @@ pub struct IgnoreDir {
     /// A single accumulation of glob patterns for this directory, matched
     /// using gitignore semantics.
     ///
-    /// This will include patterns from xrepignore as well. The patterns are
-    /// ordered so that precedence applies automatically (e.g., xrepignore
+    /// This will include patterns from rgignore as well. The patterns are
+    /// ordered so that precedence applies automatically (e.g., rgignore
     /// patterns procede gitignore patterns).
     gi: Option<Gitignore>,
     // TODO(burntsushi): Matching other types of glob patterns that don't
@@ -422,7 +422,7 @@ mod tests {
         };
     }
 
-    const ROOT: &'static str = "/home/foobar/rust/xrep";
+    const ROOT: &'static str = "/home/foobar/rust/rg";
 
     ignored_dir!(id1, ROOT, "src/main.rs", "", "src/main.rs");
     ignored_dir!(id2, ROOT, "", "src/main.rs", "src/main.rs");

src/main.rs

@@ -34,6 +34,8 @@ use std::thread;
 
 use crossbeam::sync::chase_lev::{self, Steal, Stealer};
 use grep::Grep;
+use memmap::{Mmap, Protection};
+use term::Terminal;
 use walkdir::DirEntry;
 
 use args::Args;
@@ -61,6 +63,7 @@ mod ignore;
 mod out;
 mod printer;
 mod search;
+mod search_buffer;
 mod sys;
 mod terminal;
 mod types;
@@ -99,7 +102,7 @@ fn run(args: Args) -> Result<u64> {
                 chan_work: stealer.clone(),
                 inpbuf: args.input_buffer(),
                 outbuf: Some(vec![]),
-                grep: try!(args.grep()),
+                grep: args.grep(),
                 match_count: 0,
             };
             workers.push(thread::spawn(move || worker.run()));
@@ -196,11 +199,11 @@ impl Worker {
             let mut printer = self.args.printer(outbuf);
             self.do_work(&mut printer, work);
             let outbuf = printer.into_inner();
-            if !outbuf.is_empty() {
+            if !outbuf.get_ref().is_empty() {
                 let mut out = self.out.lock().unwrap();
                 out.write(&outbuf);
             }
-            self.outbuf = Some(outbuf);
+            self.outbuf = Some(outbuf.into_inner());
         }
         self.match_count
     }
@@ -221,7 +224,11 @@ impl Worker {
                 if let Ok(p) = path.strip_prefix("./") {
                     path = p;
                 }
-                self.search(printer, path, file)
+                if self.args.mmap() {
+                    self.search_mmap(printer, path, &file)
+                } else {
+                    self.search(printer, path, file)
+                }
             }
         };
         match result {
@@ -248,4 +255,23 @@ impl Worker {
             rdr,
         ).run().map_err(From::from)
     }
+
+    fn search_mmap<W: Send + io::Write>(
+        &mut self,
+        printer: &mut Printer<W>,
+        path: &Path,
+        file: &File,
+    ) -> Result<u64> {
+        if try!(file.metadata()).len() == 0 {
+            // Opening a memory map with an empty file results in an error.
+            return Ok(0);
+        }
+        let mmap = try!(Mmap::open(file, Protection::Read));
+        Ok(self.args.searcher_buffer(
+            printer,
+            &self.grep,
+            path,
+            unsafe { mmap.as_slice() },
+        ).run())
+    }
 }

src/out.rs

@@ -1,5 +1,11 @@
 use std::io::{self, Write};
 
+use term::{StdoutTerminal, Terminal};
+#[cfg(windows)]
+use term::WinConsole;
+
+use printer::Writer;
+
 /// Out controls the actual output of all search results for a particular file
 /// to the end user.
 ///
@@ -8,15 +14,32 @@ use std::io::{self, Write};
 /// file as a whole. For example, it knows when to print a file separator.)
 pub struct Out<W: io::Write> {
     wtr: io::BufWriter<W>,
+    term: Option<Box<StdoutTerminal>>,
     printed: bool,
     file_separator: Option<Vec<u8>>,
 }
 
+/// This is like term::stdout, but on Windows always uses WinConsole instead
+/// of trying for a TerminfoTerminal. This may be a mistake.
+#[cfg(windows)]
+fn term_stdout() -> Option<Box<StdoutTerminal>> {
+    WinConsole::new(io::stdout())
+        .ok()
+        .map(|t| Box::new(t) as Box<StdoutTerminal>)
+}
+
+#[cfg(not(windows))]
+fn term_stdout() -> Option<Box<StdoutTerminal>> {
+    // We never use this crap on *nix.
+    None
+}
+
 impl<W: io::Write> Out<W> {
     /// Create a new Out that writes to the wtr given.
     pub fn new(wtr: W) -> Out<W> {
         Out {
             wtr: io::BufWriter::new(wtr),
+            term: term_stdout(),
             printed: false,
             file_separator: None,
         }
@@ -33,14 +56,31 @@ impl<W: io::Write> Out<W> {
 
     /// Write the search results of a single file to the underlying wtr and
     /// flush wtr.
-    pub fn write(&mut self, buf: &[u8]) {
+    pub fn write(&mut self, buf: &Writer<Vec<u8>>) {
         if let Some(ref sep) = self.file_separator {
             if self.printed {
                 let _ = self.wtr.write_all(sep);
                 let _ = self.wtr.write_all(b"\n");
             }
         }
-        let _ = self.wtr.write_all(buf);
+        match *buf {
+            Writer::Colored(ref tt) => {
+                let _ = self.wtr.write_all(tt.get_ref());
+            }
+            Writer::Windows(ref w) => {
+                match self.term {
+                    None => {
+                        let _ = self.wtr.write_all(w.get_ref());
+                    }
+                    Some(ref mut stdout) => {
+                        w.print_stdout(stdout);
+                    }
+                }
+            }
+            Writer::NoColor(ref buf) => {
+                let _ = self.wtr.write_all(buf);
+            }
+        }
         let _ = self.wtr.flush();
         self.printed = true;
     }

src/printer.rs

@@ -3,7 +3,7 @@ use std::path::Path;
 use std::sync::Arc;
 
 use regex::bytes::Regex;
-use term::{self, Terminal};
+use term::{self, StdoutTerminal, Terminal};
 use term::color::*;
 use term::terminfo::TermInfo;
 
@@ -22,6 +22,8 @@ pub struct Printer<W> {
     wtr: Writer<W>,
     /// Whether anything has been printed to wtr yet.
     has_printed: bool,
+    /// Whether to show column numbers for the first match or not.
+    column: bool,
     /// The string to use to separate non-contiguous runs of context lines.
     context_separator: Vec<u8>,
     /// The end-of-line terminator used by the printer. In general, eols are
@@ -48,6 +50,7 @@ impl<W: Send + io::Write> Printer<W> {
         Printer {
             wtr: Writer::new(wtr, color),
             has_printed: false,
+            column: false,
             context_separator: "--".to_string().into_bytes(),
             eol: b'\n',
             heading: false,
@@ -57,6 +60,13 @@ impl<W: Send + io::Write> Printer<W> {
         }
     }
 
+    /// When set, column numbers will be printed for the first match on each
+    /// line.
+    pub fn column(mut self, yes: bool) -> Printer<W> {
+        self.column = yes;
+        self
+    }
+
     /// Set the context separator. The default is `--`.
     pub fn context_separator(mut self, sep: Vec<u8>) -> Printer<W> {
         self.context_separator = sep;
@@ -105,9 +115,9 @@ impl<W: Send + io::Write> Printer<W> {
     }
 
     /// Flushes the underlying writer and returns it.
-    pub fn into_inner(mut self) -> W {
+    pub fn into_inner(mut self) -> Writer<W> {
         let _ = self.wtr.flush();
-        self.wtr.into_inner()
+        self.wtr
     }
 
     /// Prints a type definition.
@@ -173,6 +183,11 @@ impl<W: Send + io::Write> Printer<W> {
         if let Some(line_number) = line_number {
             self.line_number(line_number, b':');
         }
+        if self.column {
+            let c = re.find(&buf[start..end]).map(|(s, _)| s + 1).unwrap_or(0);
+            self.write(c.to_string().as_bytes());
+            self.write(b":");
+        }
         if self.replace.is_some() {
             let line = re.replace_all(
                 &buf[start..end], &**self.replace.as_ref().unwrap());
@@ -227,7 +242,8 @@ impl<W: Send + io::Write> Printer<W> {
 
     fn write_heading<P: AsRef<Path>>(&mut self, path: P) {
         if self.wtr.is_color() {
-            let _ = self.wtr.fg(GREEN);
+            let _ = self.wtr.fg(BRIGHT_GREEN);
+            let _ = self.wtr.attr(term::Attr::Bold);
         }
         self.write(path.as_ref().to_string_lossy().as_bytes());
         self.write_eol();
@@ -238,7 +254,7 @@ impl<W: Send + io::Write> Printer<W> {
 
     fn line_number(&mut self, n: u64, sep: u8) {
         if self.wtr.is_color() {
-            let _ = self.wtr.fg(YELLOW);
+            let _ = self.wtr.fg(BRIGHT_BLUE);
             let _ = self.wtr.attr(term::Attr::Bold);
         }
         self.write(n.to_string().as_bytes());
@@ -262,11 +278,32 @@ impl<W: Send + io::Write> Printer<W> {
     }
 }
 
-enum Writer<W> {
+/// Writer corresponds to the final output buffer for search results. All
+/// search results are written to a Writer and then a Writer is flushed to
+/// stdout only after the full search has completed.
+pub enum Writer<W> {
     Colored(TerminfoTerminal<W>),
+    Windows(WindowsWriter<W>),
     NoColor(W),
 }
 
+pub struct WindowsWriter<W> {
+    wtr: W,
+    pos: usize,
+    colors: Vec<WindowsColor>,
+}
+
+pub struct WindowsColor {
+    pos: usize,
+    opt: WindowsOption,
+}
+
+pub enum WindowsOption {
+    Foreground(Color),
+    Background(Color),
+    Reset,
+}
+
 lazy_static! {
     static ref TERMINFO: Option<Arc<TermInfo>> = {
         match term::terminfo::TermInfo::from_env() {
@@ -284,7 +321,13 @@ impl<W: Send + io::Write> Writer<W> {
         // If we want color, build a TerminfoTerminal and see if the current
         // environment supports coloring. If not, bail with NoColor. To avoid
         // losing our writer (ownership), do this the long way.
-        if !color || TERMINFO.is_none() {
+        if !color {
+            return NoColor(wtr);
+        }
+        if cfg!(windows) {
+            return Windows(WindowsWriter { wtr: wtr, pos: 0, colors: vec![] });
+        }
+        if TERMINFO.is_none() {
             return NoColor(wtr);
         }
         let info = TERMINFO.clone().unwrap();
@@ -299,28 +342,35 @@ impl<W: Send + io::Write> Writer<W> {
     fn is_color(&self) -> bool {
         match *self {
             Colored(_) => true,
+            Windows(_) => true,
             NoColor(_) => false,
         }
     }
 
-    fn map_result<F>(
+    fn map_result<F, G>(
         &mut self,
         mut f: F,
+        mut g: G,
     ) -> term::Result<()>
-    where F: FnMut(&mut TerminfoTerminal<W>) -> term::Result<()> {
+    where F: FnMut(&mut TerminfoTerminal<W>) -> term::Result<()>,
+          G: FnMut(&mut WindowsWriter<W>) -> term::Result<()> {
         match *self {
             Colored(ref mut w) => f(w),
+            Windows(ref mut w) => g(w),
             NoColor(_) => Err(term::Error::NotSupported),
         }
     }
 
-    fn map_bool<F>(
+    fn map_bool<F, G>(
         &self,
         mut f: F,
+        mut g: G,
     ) -> bool
-    where F: FnMut(&TerminfoTerminal<W>) -> bool {
+    where F: FnMut(&TerminfoTerminal<W>) -> bool,
+          G: FnMut(&WindowsWriter<W>) -> bool {
         match *self {
             Colored(ref w) => f(w),
+            Windows(ref w) => g(w),
             NoColor(_) => false,
         }
     }
@@ -330,6 +380,7 @@ impl<W: Send + io::Write> io::Write for Writer<W> {
     fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
         match *self {
             Colored(ref mut w) => w.write(buf),
+            Windows(ref mut w) => w.write(buf),
             NoColor(ref mut w) => w.write(buf),
         }
     }
@@ -337,6 +388,7 @@ impl<W: Send + io::Write> io::Write for Writer<W> {
     fn flush(&mut self) -> io::Result<()> {
         match *self {
             Colored(ref mut w) => w.flush(),
+            Windows(ref mut w) => w.flush(),
             NoColor(ref mut w) => w.flush(),
         }
     }
@@ -346,48 +398,49 @@ impl<W: Send + io::Write> term::Terminal for Writer<W> {
     type Output = W;
 
     fn fg(&mut self, fg: term::color::Color) -> term::Result<()> {
-        self.map_result(|w| w.fg(fg))
+        self.map_result(|w| w.fg(fg), |w| w.fg(fg))
     }
 
     fn bg(&mut self, bg: term::color::Color) -> term::Result<()> {
-        self.map_result(|w| w.bg(bg))
+        self.map_result(|w| w.bg(bg), |w| w.bg(bg))
     }
 
     fn attr(&mut self, attr: term::Attr) -> term::Result<()> {
-        self.map_result(|w| w.attr(attr))
+        self.map_result(|w| w.attr(attr), |w| w.attr(attr))
     }
 
     fn supports_attr(&self, attr: term::Attr) -> bool {
-        self.map_bool(|w| w.supports_attr(attr))
+        self.map_bool(|w| w.supports_attr(attr), |w| w.supports_attr(attr))
     }
 
     fn reset(&mut self) -> term::Result<()> {
-        self.map_result(|w| w.reset())
+        self.map_result(|w| w.reset(), |w| w.reset())
     }
 
     fn supports_reset(&self) -> bool {
-        self.map_bool(|w| w.supports_reset())
+        self.map_bool(|w| w.supports_reset(), |w| w.supports_reset())
     }
 
     fn supports_color(&self) -> bool {
-        self.map_bool(|w| w.supports_color())
+        self.map_bool(|w| w.supports_color(), |w| w.supports_color())
     }
 
     fn cursor_up(&mut self) -> term::Result<()> {
-        self.map_result(|w| w.cursor_up())
+        self.map_result(|w| w.cursor_up(), |w| w.cursor_up())
     }
 
     fn delete_line(&mut self) -> term::Result<()> {
-        self.map_result(|w| w.delete_line())
+        self.map_result(|w| w.delete_line(), |w| w.delete_line())
     }
 
     fn carriage_return(&mut self) -> term::Result<()> {
-        self.map_result(|w| w.carriage_return())
+        self.map_result(|w| w.carriage_return(), |w| w.carriage_return())
     }
 
     fn get_ref(&self) -> &W {
         match *self {
             Colored(ref w) => w.get_ref(),
+            Windows(ref w) => w.get_ref(),
             NoColor(ref w) => w,
         }
     }
@@ -395,6 +448,7 @@ impl<W: Send + io::Write> term::Terminal for Writer<W> {
     fn get_mut(&mut self) -> &mut W {
         match *self {
             Colored(ref mut w) => w.get_mut(),
+            Windows(ref mut w) => w.get_mut(),
             NoColor(ref mut w) => w,
         }
     }
@@ -402,7 +456,110 @@ impl<W: Send + io::Write> term::Terminal for Writer<W> {
     fn into_inner(self) -> W {
         match self {
             Colored(w) => w.into_inner(),
+            Windows(w) => w.into_inner(),
             NoColor(w) => w,
         }
     }
 }
+
+impl<W: Send + io::Write> WindowsWriter<W> {
+    fn push(&mut self, opt: WindowsOption) {
+        let pos = self.pos;
+        self.colors.push(WindowsColor { pos: pos, opt: opt });
+    }
+}
+
+impl WindowsWriter<Vec<u8>> {
+    /// Print the contents to the given terminal.
+    pub fn print_stdout(&self, tt: &mut Box<StdoutTerminal>) {
+        let mut last = 0;
+        for col in &self.colors {
+            let _ = tt.write_all(&self.wtr[last..col.pos]);
+            match col.opt {
+                WindowsOption::Foreground(c) => {
+                    let _ = tt.fg(c);
+                }
+                WindowsOption::Background(c) => {
+                    let _ = tt.bg(c);
+                }
+                WindowsOption::Reset => {
+                    let _ = tt.reset();
+                }
+            }
+            last = col.pos;
+        }
+        let _ = tt.write_all(&self.wtr[last..]);
+        let _ = tt.flush();
+    }
+}
+
+impl<W: Send + io::Write> io::Write for WindowsWriter<W> {
+    fn write(&mut self, buf: &[u8]) -> io::Result<usize> {
+        let n = try!(self.wtr.write(buf));
+        self.pos += n;
+        Ok(n)
+    }
+
+    fn flush(&mut self) -> io::Result<()> {
+        self.wtr.flush()
+    }
+}
+
+impl<W: Send + io::Write> term::Terminal for WindowsWriter<W> {
+    type Output = W;
+
+    fn fg(&mut self, fg: term::color::Color) -> term::Result<()> {
+        self.push(WindowsOption::Foreground(fg));
+        Ok(())
+    }
+
+    fn bg(&mut self, bg: term::color::Color) -> term::Result<()> {
+        self.push(WindowsOption::Background(bg));
+        Ok(())
+    }
+
+    fn attr(&mut self, attr: term::Attr) -> term::Result<()> {
+        Err(term::Error::NotSupported)
+    }
+
+    fn supports_attr(&self, attr: term::Attr) -> bool {
+        false
+    }
+
+    fn reset(&mut self) -> term::Result<()> {
+        self.push(WindowsOption::Reset);
+        Ok(())
+    }
+
+    fn supports_reset(&self) -> bool {
+        true
+    }
+
+    fn supports_color(&self) -> bool {
+        true
+    }
+
+    fn cursor_up(&mut self) -> term::Result<()> {
+        Err(term::Error::NotSupported)
+    }
+
+    fn delete_line(&mut self) -> term::Result<()> {
+        Err(term::Error::NotSupported)
+    }
+
+    fn carriage_return(&mut self) -> term::Result<()> {
+        Err(term::Error::NotSupported)
+    }
+
+    fn get_ref(&self) -> &W {
+        &self.wtr
+    }
+
+    fn get_mut(&mut self) -> &mut W {
+        &mut self.wtr
+    }
+
+    fn into_inner(self) -> W {
+        self.wtr
+    }
+}

src/search.rs

@@ -74,14 +74,14 @@ pub struct Searcher<'a, R, W: 'a> {
 
 /// Options for configuring search.
 #[derive(Clone)]
-struct Options {
-    after_context: usize,
-    before_context: usize,
-    count: bool,
-    eol: u8,
-    invert_match: bool,
-    line_number: bool,
-    text: bool,
+pub struct Options {
+    pub after_context: usize,
+    pub before_context: usize,
+    pub count: bool,
+    pub eol: u8,
+    pub invert_match: bool,
+    pub line_number: bool,
+    pub text: bool,
 }
 
 impl Default for Options {
@@ -219,14 +219,11 @@ impl<'a, R: io::Read, W: Send + io::Write> Searcher<'a, R, W> {
                         self.print_inverted_matches(upto);
                     }
                 } else if matched {
-                    self.match_count += 1;
-                    if !self.opts.count {
-                        let start = self.last_match.start();
-                        let end = self.last_match.end();
-                        self.print_after_context(start);
-                        self.print_before_context(start);
-                        self.print_match(start, end);
-                    }
+                    let start = self.last_match.start();
+                    let end = self.last_match.end();
+                    self.print_after_context(start);
+                    self.print_before_context(start);
+                    self.print_match(start, end);
                 }
                 if matched {
                     self.inp.pos = self.last_match.end();
@@ -275,11 +272,8 @@ impl<'a, R: io::Read, W: Send + io::Write> Searcher<'a, R, W> {
         debug_assert!(self.opts.invert_match);
         let mut it = IterLines::new(self.opts.eol, self.inp.pos);
         while let Some((start, end)) = it.next(&self.inp.buf[..upto]) {
-            if !self.opts.count {
-                self.print_match(start, end);
-            }
+            self.print_match(start, end);
             self.inp.pos = end;
-            self.match_count += 1;
         }
     }
 
@@ -325,11 +319,15 @@ impl<'a, R: io::Read, W: Send + io::Write> Searcher<'a, R, W> {
 
     #[inline(always)]
     fn print_match(&mut self, start: usize, end: usize) {
+        self.match_count += 1;
+        if self.opts.count {
+            return;
+        }
         self.print_separator(start);
         self.count_lines(start);
         self.add_line(end);
         self.printer.matched(
-            self.grep.regex(), &self.path,
+            self.grep.regex(), self.path,
             &self.inp.buf, start, end, self.line_count);
         self.last_printed = end;
         self.after_context_remaining = self.opts.after_context;
@@ -535,7 +533,7 @@ impl InputBuffer {
 ///
 /// Note that this may return both false positives and false negatives.
 #[inline(always)]
-fn is_binary(buf: &[u8]) -> bool {
+pub fn is_binary(buf: &[u8]) -> bool {
     if buf.len() >= 4 && &buf[0..4] == b"%PDF" {
         return true;
     }
@@ -544,7 +542,7 @@ fn is_binary(buf: &[u8]) -> bool {
 
 /// Count the number of lines in the given buffer.
 #[inline(always)]
-fn count_lines(mut buf: &[u8], eol: u8) -> u64 {
+pub fn count_lines(mut buf: &[u8], eol: u8) -> u64 {
     let mut count = 0;
     while let Some(pos) = memchr(eol, buf) {
         count += 1;
@@ -575,7 +573,7 @@ fn replace_buf(buf: &mut [u8], a: u8, b: u8) {
 /// advance over the positions of each line. We neglect that approach to avoid
 /// the borrow in the search code. (Because the borrow prevents composition
 /// through other mutable methods.)
-struct IterLines {
+pub struct IterLines {
     eol: u8,
     pos: usize,
 }
@@ -585,7 +583,7 @@ impl IterLines {
     ///
     /// The buffer is passed to the `next` method.
     #[inline(always)]
-    fn new(eol: u8, start: usize) -> IterLines {
+    pub fn new(eol: u8, start: usize) -> IterLines {
         IterLines {
             eol: eol,
             pos: start,
@@ -597,7 +595,7 @@ impl IterLines {
     ///
     /// The range returned includes the new line.
     #[inline(always)]
-    fn next(&mut self, buf: &[u8]) -> Option<(usize, usize)> {
+    pub fn next(&mut self, buf: &[u8]) -> Option<(usize, usize)> {
         match memchr(self.eol, &buf[self.pos..]) {
             None => {
                 if self.pos < buf.len() {
@@ -689,6 +687,7 @@ mod tests {
     use std::path::Path;
 
     use grep::{Grep, GrepBuilder};
+    use term::Terminal;
 
     use printer::Printer;
 
@@ -747,7 +746,7 @@ fn main() {
                 &mut inp, &mut pp, &grep, test_path(), hay(haystack));
             map(searcher).run().unwrap()
         };
-        (count, String::from_utf8(pp.into_inner()).unwrap())
+        (count, String::from_utf8(pp.into_inner().into_inner()).unwrap())
     }
 
     fn search<F: FnMut(TestSearcher) -> TestSearcher>(
@@ -763,7 +762,7 @@ fn main() {
                 &mut inp, &mut pp, &grep, test_path(), hay(haystack));
             map(searcher).run().unwrap()
         };
-        (count, String::from_utf8(pp.into_inner()).unwrap())
+        (count, String::from_utf8(pp.into_inner().into_inner()).unwrap())
     }
 
     #[test]
@@ -870,7 +869,7 @@ fn main() {
     }
 
     #[test]
-    fn basic_search() {
+    fn basic_search1() {
         let (count, out) = search_smallcap("Sherlock", &*SHERLOCK, |s|s);
         assert_eq!(2, count);
         assert_eq!(out, "\
@@ -887,7 +886,6 @@ fn main() {
         assert_eq!(out, "");
     }
 
-
     #[test]
     fn binary_text() {
         let text = "Sherlock\n\x00Holmes\n";

src/search_buffer.rs

@@ -0,0 +1,285 @@
+use std::cmp;
+use std::io;
+use std::path::Path;
+
+use grep::Grep;
+
+use printer::Printer;
+use search::{IterLines, Options, count_lines, is_binary};
+
+pub struct BufferSearcher<'a, W: 'a> {
+    opts: Options,
+    printer: &'a mut Printer<W>,
+    grep: &'a Grep,
+    path: &'a Path,
+    buf: &'a [u8],
+    match_count: u64,
+    line_count: Option<u64>,
+    last_line: usize,
+}
+
+impl<'a, W: Send + io::Write> BufferSearcher<'a, W> {
+    pub fn new(
+        printer: &'a mut Printer<W>,
+        grep: &'a Grep,
+        path: &'a Path,
+        buf: &'a [u8],
+    ) -> BufferSearcher<'a, W> {
+        BufferSearcher {
+            opts: Options::default(),
+            printer: printer,
+            grep: grep,
+            path: path,
+            buf: buf,
+            match_count: 0,
+            line_count: None,
+            last_line: 0,
+        }
+    }
+
+    /// If enabled, searching will print a count instead of each match.
+    ///
+    /// Disabled by default.
+    pub fn count(mut self, yes: bool) -> Self {
+        self.opts.count = yes;
+        self
+    }
+
+    /// Set the end-of-line byte used by this searcher.
+    pub fn eol(mut self, eol: u8) -> Self {
+        self.opts.eol = eol;
+        self
+    }
+
+    /// If enabled, matching is inverted so that lines that *don't* match the
+    /// given pattern are treated as matches.
+    pub fn invert_match(mut self, yes: bool) -> Self {
+        self.opts.invert_match = yes;
+        self
+    }
+
+    /// If enabled, compute line numbers and prefix each line of output with
+    /// them.
+    pub fn line_number(mut self, yes: bool) -> Self {
+        self.opts.line_number = yes;
+        self
+    }
+
+    /// If enabled, search binary files as if they were text.
+    pub fn text(mut self, yes: bool) -> Self {
+        self.opts.text = yes;
+        self
+    }
+
+    #[inline(never)]
+    pub fn run(mut self) -> u64 {
+        let binary_upto = cmp::min(4096, self.buf.len());
+        if !self.opts.text && is_binary(&self.buf[..binary_upto]) {
+            return 0;
+        }
+
+        self.match_count = 0;
+        self.line_count = if self.opts.line_number { Some(0) } else { None };
+        let mut last_end = 0;
+        for m in self.grep.iter(self.buf) {
+            if self.opts.invert_match {
+                self.print_inverted_matches(last_end, m.start());
+            } else {
+                self.print_match(m.start(), m.end());
+            }
+            last_end = m.end();
+        }
+        if self.opts.invert_match {
+            let upto = self.buf.len();
+            self.print_inverted_matches(last_end, upto);
+        }
+        if self.opts.count && self.match_count > 0 {
+            self.printer.path_count(self.path, self.match_count);
+        }
+        self.match_count
+    }
+
+    #[inline(always)]
+    pub fn print_match(&mut self, start: usize, end: usize) {
+        self.match_count += 1;
+        if self.opts.count {
+            return;
+        }
+        self.count_lines(start);
+        self.add_line(end);
+        self.printer.matched(
+            self.grep.regex(), self.path, self.buf,
+            start, end, self.line_count);
+    }
+
+    #[inline(always)]
+    fn print_inverted_matches(&mut self, start: usize, end: usize) {
+        debug_assert!(self.opts.invert_match);
+        let mut it = IterLines::new(self.opts.eol, start);
+        while let Some((s, e)) = it.next(&self.buf[..end]) {
+            self.print_match(s, e);
+        }
+    }
+
+    #[inline(always)]
+    fn count_lines(&mut self, upto: usize) {
+        if let Some(ref mut line_count) = self.line_count {
+            *line_count += count_lines(
+                &self.buf[self.last_line..upto], self.opts.eol);
+            self.last_line = upto;
+        }
+    }
+
+    #[inline(always)]
+    fn add_line(&mut self, line_end: usize) {
+        if let Some(ref mut line_count) = self.line_count {
+            *line_count += 1;
+            self.last_line = line_end;
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::path::Path;
+
+    use grep::{Grep, GrepBuilder};
+    use term::Terminal;
+
+    use printer::Printer;
+
+    use super::BufferSearcher;
+
+    lazy_static! {
+        static ref SHERLOCK: &'static str = "\
+For the Doctor Watsons of this world, as opposed to the Sherlock
+Holmeses, success in the province of detective work must always
+be, to a very large extent, the result of luck. Sherlock Holmes
+can extract a clew from a wisp of straw or a flake of cigar ash;
+but Doctor Watson has to have it taken out for him and dusted,
+and exhibited clearly, with a label attached.\
+";
+        static ref CODE: &'static str = "\
+extern crate snap;
+
+use std::io;
+
+fn main() {
+    let stdin = io::stdin();
+    let stdout = io::stdout();
+
+    // Wrap the stdin reader in a Snappy reader.
+    let mut rdr = snap::Reader::new(stdin.lock());
+    let mut wtr = stdout.lock();
+    io::copy(&mut rdr, &mut wtr).expect(\"I/O operation failed\");
+}
+";
+    }
+
+    fn matcher(pat: &str) -> Grep {
+        GrepBuilder::new(pat).build().unwrap()
+    }
+
+    fn test_path() -> &'static Path {
+        &Path::new("/baz.rs")
+    }
+
+    type TestSearcher<'a> = BufferSearcher<'a, Vec<u8>>;
+
+    fn search<F: FnMut(TestSearcher) -> TestSearcher>(
+        pat: &str,
+        haystack: &str,
+        mut map: F,
+    ) -> (u64, String) {
+        let mut pp = Printer::new(vec![], false).with_filename(true);
+        let grep = GrepBuilder::new(pat).build().unwrap();
+        let count = {
+            let searcher = BufferSearcher::new(
+                &mut pp, &grep, test_path(), haystack.as_bytes());
+            map(searcher).run()
+        };
+        (count, String::from_utf8(pp.into_inner().into_inner()).unwrap())
+    }
+
+    #[test]
+    fn basic_search() {
+        let (count, out) = search("Sherlock", &*SHERLOCK, |s|s);
+        assert_eq!(2, count);
+        assert_eq!(out, "\
+/baz.rs:For the Doctor Watsons of this world, as opposed to the Sherlock
+/baz.rs:be, to a very large extent, the result of luck. Sherlock Holmes
+");
+    }
+
+    #[test]
+    fn binary() {
+        let text = "Sherlock\n\x00Holmes\n";
+        let (count, out) = search("Sherlock|Holmes", text, |s|s);
+        assert_eq!(0, count);
+        assert_eq!(out, "");
+    }
+
+
+    #[test]
+    fn binary_text() {
+        let text = "Sherlock\n\x00Holmes\n";
+        let (count, out) = search("Sherlock|Holmes", text, |s| s.text(true));
+        assert_eq!(2, count);
+        assert_eq!(out, "/baz.rs:Sherlock\n/baz.rs:\x00Holmes\n");
+    }
+
+    #[test]
+    fn line_numbers() {
+        let (count, out) = search(
+            "Sherlock", &*SHERLOCK, |s| s.line_number(true));
+        assert_eq!(2, count);
+        assert_eq!(out, "\
+/baz.rs:1:For the Doctor Watsons of this world, as opposed to the Sherlock
+/baz.rs:3:be, to a very large extent, the result of luck. Sherlock Holmes
+");
+    }
+
+    #[test]
+    fn count() {
+        let (count, out) = search(
+            "Sherlock", &*SHERLOCK, |s| s.count(true));
+        assert_eq!(2, count);
+        assert_eq!(out, "/baz.rs:2\n");
+    }
+
+    #[test]
+    fn invert_match() {
+        let (count, out) = search(
+            "Sherlock", &*SHERLOCK, |s| s.invert_match(true));
+        assert_eq!(4, count);
+        assert_eq!(out, "\
+/baz.rs:Holmeses, success in the province of detective work must always
+/baz.rs:can extract a clew from a wisp of straw or a flake of cigar ash;
+/baz.rs:but Doctor Watson has to have it taken out for him and dusted,
+/baz.rs:and exhibited clearly, with a label attached.
+");
+    }
+
+    #[test]
+    fn invert_match_line_numbers() {
+        let (count, out) = search("Sherlock", &*SHERLOCK, |s| {
+            s.invert_match(true).line_number(true)
+        });
+        assert_eq!(4, count);
+        assert_eq!(out, "\
+/baz.rs:2:Holmeses, success in the province of detective work must always
+/baz.rs:4:can extract a clew from a wisp of straw or a flake of cigar ash;
+/baz.rs:5:but Doctor Watson has to have it taken out for him and dusted,
+/baz.rs:6:and exhibited clearly, with a label attached.
+");
+    }
+
+    #[test]
+    fn invert_match_count() {
+        let (count, out) = search("Sherlock", &*SHERLOCK, |s| {
+            s.invert_match(true).count(true)
+        });
+        assert_eq!(4, count);
+        assert_eq!(out, "/baz.rs:4\n");
+    }
+}

src/sys.rs

@@ -1,19 +1,19 @@
 /*!
 This io module contains various platform specific functions for detecting
-how xrep is being used. e.g., Is stdin being piped into it? Is stdout being
+how ripgrep is being used. e.g., Is stdin being piped into it? Is stdout being
 redirected to a file? etc... We use this information to tweak various default
 configuration parameters such as colors and match formatting.
 */
 
-use libc;
-
 #[cfg(unix)]
 pub fn stdin_is_atty() -> bool {
+    use libc;
     0 < unsafe { libc::isatty(libc::STDIN_FILENO) }
 }
 
 #[cfg(unix)]
 pub fn stdout_is_atty() -> bool {
+    use libc;
     0 < unsafe { libc::isatty(libc::STDOUT_FILENO) }
 }