Rewrite fzf in Go

32 files changed, 3464 insertions, 47 deletions

diff --git a/.gitignore b/.gitignore
index 1627430d..09154676 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,3 +1,5 @@
+bin
+src/fzf/fzf_*
 pkg
 Gemfile.lock
 .DS_Store
diff --git a/install b/install
index 3176b27d..6b64a3f2 100755
--- a/install
+++ b/install
@@ -3,60 +3,81 @@
 cd `dirname $BASH_SOURCE`
 fzf_base=`pwd`
 
-# ruby executable
-echo -n "Checking Ruby executable ... "
-ruby=`which ruby`
-if [ $? -ne 0 ]; then
-  echo "ruby executable not found!"
-  exit 1
-fi
-
-# System ruby is preferred
-system_ruby=/usr/bin/ruby
-if [ -x $system_ruby -a $system_ruby != "$ruby" ]; then
-  $system_ruby --disable-gems -rcurses -e0 2> /dev/null
-  [ $? -eq 0 ] && ruby=$system_ruby
-fi
-
-echo "OK ($ruby)"
+ARCHI=$(uname -sm)
 
-# Curses-support
-echo -n "Checking Curses support ... "
-"$ruby" -rcurses -e0 2> /dev/null
-if [ $? -eq 0 ]; then
-  echo "OK"
-else
-  echo "Not found"
-  echo "Installing 'curses' gem ... "
-  if (( EUID )); then
-    /usr/bin/env gem install curses --user-install
+download() {
+  echo "Downloading fzf executable ($1) ..."
+  if curl -fLo "$fzf_base"/bin/fzf https://github.com/junegunn/fzf-bin/releases/download/snapshot/$1; then
+    chmod +x "$fzf_base"/bin/fzf
   else
-    /usr/bin/env gem install curses
+    echo "Failed to download $1"
+    exit 1
   fi
+}
+
+mkdir -p "$fzf_base"/bin
+if [ "$ARCHI" = "Darwin x86_64" ]; then
+  download fzf_darwin_amd64
+elif [ "$ARCHI" = "Linux x86_64" ]; then
+  download fzf_linux_amd64
+else # No prebuilt executable
+  echo "No prebuilt binary for $ARCHI ... Installing legacy Ruby version ..."
+
+  # ruby executable
+  echo -n "Checking Ruby executable ... "
+  ruby=`which ruby`
   if [ $? -ne 0 ]; then
-    echo
-    echo "Failed to install 'curses' gem."
-    if [[ $(uname -r) =~ 'ARCH' ]]; then
-      echo "Make sure that base-devel package group is installed."
-    fi
+    echo "ruby executable not found!"
     exit 1
   fi
-fi
 
-# Ruby version
-echo -n "Checking Ruby version ... "
-"$ruby" -e 'exit RUBY_VERSION >= "1.9"'
-if [ $? -eq 0 ]; then
-  echo ">= 1.9"
-  "$ruby" --disable-gems -rcurses -e0 2> /dev/null
+  # System ruby is preferred
+  system_ruby=/usr/bin/ruby
+  if [ -x $system_ruby -a $system_ruby != "$ruby" ]; then
+    $system_ruby --disable-gems -rcurses -e0 2> /dev/null
+    [ $? -eq 0 ] && ruby=$system_ruby
+  fi
+
+  echo "OK ($ruby)"
+
+  # Curses-support
+  echo -n "Checking Curses support ... "
+  "$ruby" -rcurses -e0 2> /dev/null
+  if [ $? -eq 0 ]; then
+    echo "OK"
+  else
+    echo "Not found"
+    echo "Installing 'curses' gem ... "
+    if (( EUID )); then
+      /usr/bin/env gem install curses --user-install
+    else
+      /usr/bin/env gem install curses
+    fi
+    if [ $? -ne 0 ]; then
+      echo
+      echo "Failed to install 'curses' gem."
+      if [[ $(uname -r) =~ 'ARCH' ]]; then
+        echo "Make sure that base-devel package group is installed."
+      fi
+      exit 1
+    fi
+  fi
+
+  # Ruby version
+  echo -n "Checking Ruby version ... "
+  "$ruby" -e 'exit RUBY_VERSION >= "1.9"'
   if [ $? -eq 0 ]; then
-    fzf_cmd="$ruby --disable-gems $fzf_base/fzf"
+    echo ">= 1.9"
+    "$ruby" --disable-gems -rcurses -e0 2> /dev/null
+    if [ $? -eq 0 ]; then
+      fzf_cmd="$ruby --disable-gems $fzf_base/fzf"
+    else
+      fzf_cmd="$ruby $fzf_base/fzf"
+    fi
   else
+    echo "< 1.9"
     fzf_cmd="$ruby $fzf_base/fzf"
   fi
-else
-  echo "< 1.9"
-  fzf_cmd="$ruby $fzf_base/fzf"
 fi
 
 # Auto-completion
@@ -85,10 +106,17 @@ for shell in bash zsh; do
 # Setup fzf function
 # ------------------
 unalias fzf 2> /dev/null
-fzf() {
-  $fzf_cmd "\$@"
-}
-export -f fzf > /dev/null
+unset fzf 2> /dev/null
+if [ -x "$fzf_base/bin/fzf" ]; then
+  if [[ ! "\$PATH" =~ "$fzf_base/bin" ]]; then
+    export PATH="$fzf_base/bin:\$PATH"
+  fi
+else
+  fzf() {
+    $fzf_cmd "\$@"
+  }
+  export -f fzf > /dev/null
+fi
 
 # Auto-completion
 # ---------------
diff --git a/src/Dockerfile b/src/Dockerfile
new file mode 100644
index 00000000..3c062eea
--- /dev/null
+++ b/src/Dockerfile
@@ -0,0 +1,33 @@
+FROM ubuntu:14.04
+MAINTAINER Junegunn Choi <junegunn.c@gmail.com>
+
+# apt-get
+RUN apt-get update && apt-get -y upgrade
+RUN apt-get install -y --force-yes git vim-nox curl procps sudo \
+                                   build-essential libncurses-dev
+
+# Setup jg user with sudo privilege
+RUN useradd -s /bin/bash -m jg && echo 'jg:jg' | chpasswd && \
+    echo 'jg ALL=(ALL) NOPASSWD: ALL' > /etc/sudoers.d/jg
+
+# Setup dotfiles
+USER jg
+RUN cd ~ && git clone https://github.com/junegunn/dotfiles.git && \
+    dotfiles/install > /dev/null
+
+# Install Go 1.4
+RUN cd ~ && curl https://storage.googleapis.com/golang/go1.4.linux-amd64.tar.gz | tar -xz && \
+    mv go go1.4 && \
+    echo 'export GOROOT=~/go1.4' >> ~/dotfiles/bashrc-extra && \
+    echo 'export PATH=~/go1.4/bin:$PATH' >> ~/dotfiles/bashrc-extra
+
+# Symlink fzf directory
+RUN mkdir -p ~jg/go/src/github.com/junegunn && \
+    ln -s /fzf ~jg/go/src/github.com/junegunn/fzf
+
+# Volume
+VOLUME /fzf
+
+# Default CMD
+CMD cd ~jg/go/src/github.com/junegunn/fzf/src && /bin/bash -l
+
diff --git a/src/LICENSE b/src/LICENSE
new file mode 100644
index 00000000..fe4c31ae
--- /dev/null
+++ b/src/LICENSE
@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2015 Junegunn Choi
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
diff --git a/src/Makefile b/src/Makefile
new file mode 100644
index 00000000..bae4c906
--- /dev/null
+++ b/src/Makefile
@@ -0,0 +1,49 @@
+BINARY := fzf/fzf
+
+UNAME_S := $(shell uname -s)
+ifeq ($(UNAME_S),Darwin)
+	BINARY := $(BINARY)_darwin
+else ifeq ($(UNAME_S),Linux)
+	BINARY := $(BINARY)_linux
+endif
+
+UNAME_M := $(shell uname -m)
+ifneq ($(filter i386 i686,$(UNAME_M)),)
+$(error "filtered is not supported, yet.")
+endif
+
+ifeq ($(UNAME_M),x86_64)
+	BINARY := $(BINARY)_amd64
+else ifneq ($(filter i386 i686,$(UNAME_M)),)
+	BINARY := $(BINARY)_386
+else # TODO
+$(error "$(UNAME_M) is not supported, yet.")
+endif
+
+BINDIR  = ../bin
+SOURCES = $(wildcard *.go fzf/*.go)
+
+all: build
+
+build: $(BINARY)
+
+$(BINARY): $(SOURCES)
+	go get
+	go test -v
+	cd fzf && go build -o $(notdir $(BINARY))
+
+install: $(BINARY)
+	mkdir -p $(BINDIR)
+	cp -f $(BINARY) $(BINDIR)/fzf
+
+clean:
+	rm -f $(BINARY)
+
+docker:
+	docker build -t junegunn/ubuntu-sandbox .
+
+linux64:
+	docker run -i -t -u jg -v $(shell cd ..; pwd):/fzf junegunn/ubuntu-sandbox \
+		/bin/bash -ci 'cd ~jg/go/src/github.com/junegunn/fzf/src; make build'
+
+.PHONY: build install linux64 clean docker run
diff --git a/src/README.md b/src/README.md
new file mode 100644
index 00000000..2f3ca3bb
--- /dev/null
+++ b/src/README.md
@@ -0,0 +1,59 @@
+fzf in Go
+=========
+
+This directory contains the source code for the new fzf implementation in Go.
+This new version has the following benefits over the previous Ruby version.
+
+- Immensely faster
+    - No GIL. Performance is linearly proportional to the number of cores.
+    - It's so fast that I even decided to remove the sort limit (`--sort=N`)
+- Does not require Ruby and distributed as an executable binary
+    - Ruby dependency is especially painful on Ruby 2.1 or above which
+      ships without curses gem
+
+Build
+-----
+
+```sh
+# Build fzf executable
+make
+
+# Install the executable to ../bin directory
+make install
+
+# Build executable for Linux x86_64 using Docker
+make linux64
+```
+
+
+Prebuilt binaries
+-----------------
+
+- Darwin x86_64
+- Linux x86_64
+
+Third-party libraries used
+--------------------------
+
+- [ncurses](https://www.gnu.org/software/ncurses/)
+- [mattn/go-runewidth](https://github.com/mattn/go-runewidth)
+    - Licensed under [MIT](http://mattn.mit-license.org/2013)
+- [mattn/go-shellwords](https://github.com/mattn/go-shellwords)
+    - Licensed under [MIT](http://mattn.mit-license.org/2014)
+
+Contribution
+------------
+
+For the moment, I will not add or accept any new features until we can be sure
+that the implementation is stable and we have a sufficient number of test
+cases. However, fixes for obvious bugs and new test cases are welcome.
+
+I also care much about the performance of the implementation (that's the
+reason I rewrote the whole thing in Go, right?), so please make sure that your
+change does not result in performance regression. Please be minded that we
+still don't have a quantitative measure of the performance.
+
+License
+-------
+
+- [MIT](LICENSE)
diff --git a/src/algo.go b/src/algo.go
new file mode 100644
index 00000000..16790ba9
--- /dev/null
+++ b/src/algo.go
@@ -0,0 +1,152 @@
+package fzf
+
+import "strings"
+
+/*
+ * String matching algorithms here do not use strings.ToLower to avoid
+ * performance penalty. And they assume pattern runes are given in lowercase
+ * letters when caseSensitive is false.
+ *
+ * In short: They try to do as little work as possible.
+ */
+
+func FuzzyMatch(caseSensitive bool, input *string, pattern []rune) (int, int) {
+	runes := []rune(*input)
+
+	// 0. (FIXME) How to find the shortest match?
+	//    a_____b__c__abc
+	//    ^^^^^^^^^^  ^^^
+	// 1. forward scan (abc)
+	//   *-----*-----*>
+	//   a_____b___abc__
+	// 2. reverse scan (cba)
+	//   a_____b___abc__
+	//            <***
+	pidx := 0
+	sidx := -1
+	eidx := -1
+
+	for index, char := range runes {
+		// This is considerably faster than blindly applying strings.ToLower to the
+		// whole string
+		if !caseSensitive && char >= 65 && char <= 90 {
+			char += 32
+		}
+		if char == pattern[pidx] {
+			if sidx < 0 {
+				sidx = index
+			}
+			if pidx += 1; pidx == len(pattern) {
+				eidx = index + 1
+				break
+			}
+		}
+	}
+
+	if sidx >= 0 && eidx >= 0 {
+		pidx -= 1
+		for index := eidx - 1; index >= sidx; index-- {
+			char := runes[index]
+			if !caseSensitive && char >= 65 && char <= 90 {
+				char += 32
+			}
+			if char == pattern[pidx] {
+				if pidx -= 1; pidx < 0 {
+					sidx = index
+					break
+				}
+			}
+		}
+		return sidx, eidx
+	}
+	return -1, -1
+}
+
+func ExactMatchStrings(caseSensitive bool, input *string, pattern []rune) (int, int) {
+	var str string
+	if caseSensitive {
+		str = *input
+	} else {
+		str = strings.ToLower(*input)
+	}
+
+	if idx := strings.Index(str, string(pattern)); idx >= 0 {
+		prefixRuneLen := len([]rune((*input)[:idx]))
+		return prefixRuneLen, prefixRuneLen + len(pattern)
+	}
+	return -1, -1
+}
+
+/*
+ * This is a basic string searching algorithm that handles case sensitivity.
+ * Although naive, it still performs better than the combination of
+ * strings.ToLower + strings.Index for typical fzf use cases where input
+ * strings and patterns are not very long.
+ *
+ * We might try to implement better algorithms in the future:
+ * http://en.wikipedia.org/wiki/String_searching_algorithm
+ */
+func ExactMatchNaive(caseSensitive bool, input *string, pattern []rune) (int, int) {
+	runes := []rune(*input)
+	numRunes := len(runes)
+	plen := len(pattern)
+	if len(runes) < plen {
+		return -1, -1
+	}
+
+	pidx := 0
+	for index := 0; index < numRunes; index++ {
+		char := runes[index]
+		if !caseSensitive && char >= 65 && char <= 90 {
+			char += 32
+		}
+		if pattern[pidx] == char {
+			pidx += 1
+			if pidx == plen {
+				return index - plen + 1, index + 1
+			}
+		} else {
+			index -= pidx
+			pidx = 0
+		}
+	}
+	return -1, -1
+}
+
+func PrefixMatch(caseSensitive bool, input *string, pattern []rune) (int, int) {
+	runes := []rune(*input)
+	if len(runes) < len(pattern) {
+		return -1, -1
+	}
+
+	for index, r := range pattern {
+		char := runes[index]
+		if !caseSensitive && char >= 65 && char <= 90 {
+			char += 32
+		}
+		if char != r {
+			return -1, -1
+		}
+	}
+	return 0, len(pattern)
+}
+
+func SuffixMatch(caseSensitive bool, input *string, pattern []rune) (int, int) {
+	runes := []rune(strings.TrimRight(*input, " "))
+	trimmedLen := len(runes)
+	diff := trimmedLen - len(pattern)
+	if diff < 0 {
+		return -1, -1
+	}
+
+	for index, r := range pattern {
+		char := runes[index+diff]
+		if !caseSensitive && char >= 65 && char <= 90 {
+			char += 32
+		}
+		if char != r {
+			return -1, -1
+		}
+	}
+	return trimmedLen - len(pattern), trimmedLen
+}
diff --git a/src/algo_test.go b/src/algo_test.go
new file mode 100644
index 00000000..5da01a6d
--- /dev/null
+++ b/src/algo_test.go
@@ -0,0 +1,44 @@
+package fzf
+
+import (
+	"strings"
+	"testing"
+)
+
+func assertMatch(t *testing.T, fun func(bool, *string, []rune) (int, int), caseSensitive bool, input string, pattern string, sidx int, eidx int) {
+	if !caseSensitive {
+		pattern = strings.ToLower(pattern)
+	}
+	s, e := fun(caseSensitive, &input, []rune(pattern))
+	if s != sidx {
+		t.Errorf("Invalid start index: %d (expected: %d, %s / %s)", s, sidx, input, pattern)
+	}
+	if e != eidx {
+		t.Errorf("Invalid end index: %d (expected: %d, %s / %s)", e, eidx, input, pattern)
+	}
+}
+
+func TestFuzzyMatch(t *testing.T) {
+	assertMatch(t, FuzzyMatch, false, "fooBarbaz", "oBZ", 2, 9)
+	assertMatch(t, FuzzyMatch, true, "fooBarbaz", "oBZ", -1, -1)
+	assertMatch(t, FuzzyMatch, true, "fooBarbaz", "oBz", 2, 9)
+	assertMatch(t, FuzzyMatch, true, "fooBarbaz", "fooBarbazz", -1, -1)
+}
+
+func TestExactMatchNaive(t *testing.T) {
+	assertMatch(t, ExactMatchNaive, false, "fooBarbaz", "oBA", 2, 5)
+	assertMatch(t, ExactMatchNaive, true, "fooBarbaz", "oBA", -1, -1)
+	assertMatch(t, ExactMatchNaive, true, "fooBarbaz", "fooBarbazz", -1, -1)
+}
+
+func TestPrefixMatch(t *testing.T) {
+	assertMatch(t, PrefixMatch, false, "fooBarbaz", "Foo", 0, 3)
+	assertMatch(t, PrefixMatch, true, "fooBarbaz", "Foo", -1, -1)
+	assertMatch(t, PrefixMatch, false, "fooBarbaz", "baz", -1, -1)
+}
+
+func TestSuffixMatch(t *testing.T) {
+	assertMatch(t, SuffixMatch, false, "fooBarbaz", "Foo", -1, -1)
+	assertMatch(t, SuffixMatch, false, "fooBarbaz", "baz", 6, 9)
+	assertMatch(t, SuffixMatch, true, "fooBarbaz", "Baz", -1, -1)
+}
diff --git a/src/atomicbool.go b/src/atomicbool.go
new file mode 100644
index 00000000..f2f4894f
--- /dev/null
+++ b/src/atomicbool.go
@@ -0,0 +1,27 @@
+package fzf
+
+import "sync"
+
+type AtomicBool struct {
+	mutex sync.Mutex
+	state bool
+}
+
+func NewAtomicBool(initialState bool) *AtomicBool {
+	return &AtomicBool{
+		mutex: sync.Mutex{},
+		state: initialState}
+}
+
+func (a *AtomicBool) Get() bool {
+	a.mutex.Lock()
+	defer a.mutex.Unlock()
+	return a.state
+}
+
+func (a *AtomicBool) Set(newState bool) bool {
+	a.mutex.Lock()
+	defer a.mutex.Unlock()
+	a.state = newState
+	return a.state
+}
diff --git a/src/atomicbool_test.go b/src/atomicbool_test.go
new file mode 100644
index 00000000..0af45701
--- /dev/null
+++ b/src/atomicbool_test.go
@@ -0,0 +1,17 @@
+package fzf
+
+import "testing"
+
+func TestAtomicBool(t *testing.T) {
+	if !NewAtomicBool(true).Get() || NewAtomicBool(false).Get() {
+		t.Error("Invalid initial value")
+	}
+
+	ab := NewAtomicBool(true)
+	if ab.Set(false) {
+		t.Error("Invalid return value")
+	}
+	if ab.Get() {
+		t.Error("Invalid state")
+	}
+}
diff --git a/src/cache.go b/src/cache.go
new file mode 100644
index 00000000..340f3258
--- /dev/null
+++ b/src/cache.go
@@ -0,0 +1,47 @@
+package fzf
+
+import "sync"
+
+type QueryCache map[string][]*Item
+type ChunkCache struct {
+	mutex sync.Mutex
+	cache map[*Chunk]*QueryCache
+}
+
+func NewChunkCache() ChunkCache {
+	return ChunkCache{sync.Mutex{}, make(map[*Chunk]*QueryCache)}
+}
+
+func (cc *ChunkCache) Add(chunk *Chunk, key string, list []*Item) {
+	if len(key) == 0 || !chunk.IsFull() {
+		return
+	}
+
+	cc.mutex.Lock()
+	defer cc.mutex.Unlock()
+
+	qc, ok := cc.cache[chunk]
+	if !ok {
+		cc.cache[chunk] = &QueryCache{}
+		qc = cc.cache[chunk]
+	}
+	(*qc)[key] = list
+}
+
+func (cc *ChunkCache) Find(chunk *Chunk, key string) ([]*Item, bool) {
+	if len(key) == 0 || !chunk.IsFull() {
+		return nil, false
+	}
+
+	cc.mutex.Lock()
+	defer cc.mutex.Unlock()
+
+	qc, ok := cc.cache[chunk]
+	if ok {
+		list, ok := (*qc)[key]
+		if ok {
+			return list, true
+		}
+	}
+	return nil, false
+}
diff --git a/src/chunklist.go b/src/chunklist.go
new file mode 100644
index 00000000..b1f9638d
--- /dev/null
+++ b/src/chunklist.go
@@ -0,0 +1,73 @@
+package fzf
+
+import "sync"
+
+const CHUNK_SIZE int = 100
+
+type Chunk []*Item // >>> []Item
+
+type Transformer func(*string, int) *Item
+
+type ChunkList struct {
+	chunks []*Chunk
+	count  int
+	mutex  sync.Mutex
+	trans  Transformer
+}
+
+func NewChunkList(trans Transformer) *ChunkList {
+	return &ChunkList{
+		chunks: []*Chunk{},
+		count:  0,
+		mutex:  sync.Mutex{},
+		trans:  trans}
+}
+
+func (c *Chunk) push(trans Transformer, data *string, index int) {
+	*c = append(*c, trans(data, index))
+}
+
+func (c *Chunk) IsFull() bool {
+	return len(*c) == CHUNK_SIZE
+}
+
+func (cl *ChunkList) lastChunk() *Chunk {
+	return cl.chunks[len(cl.chunks)-1]
+}
+
+func CountItems(cs []*Chunk) int {
+	if len(cs) == 0 {
+		return 0
+	}
+	return CHUNK_SIZE*(len(cs)-1) + len(*(cs[len(cs)-1]))
+}
+
+func (cl *ChunkList) Count() int {
+	return cl.count
+}
+
+func (cl *ChunkList) Chunks() []*Chunk {
+	return cl.chunks
+}
+
+func (cl *ChunkList) Push(data string) {
+	cl.mutex.Lock()
+	defer cl.mutex.Unlock()
+
+	if len(cl.chunks) == 0 || cl.lastChunk().IsFull() {
+		newChunk := Chunk(make([]*Item, 0, CHUNK_SIZE))
+		cl.chunks = append(cl.chunks, &newChunk)
+	}
+
+	cl.lastChunk().push(cl.trans, &data, cl.count)
+	cl.count += 1
+}
+
+func (cl *ChunkList) Snapshot() []*Chunk {
+	cl.mutex.Lock()
+	defer cl.mutex.Unlock()
+
+	ret := make([]*Chunk, len(cl.chunks))
+	copy(ret, cl.chunks)
+	return ret
+}
diff --git a/src/chunklist_test.go b/src/chunklist_test.go
new file mode 100644
index 00000000..a7daa47e
--- /dev/null
+++ b/src/chunklist_test.go
@@ -0,0 +1,66 @@
+package fzf
+
+import (
+	"fmt"
+	"testing"
+)
+
+func TestChunkList(t *testing.T) {
+	cl := NewChunkList(func(s *string, i int) *Item {
+		return &Item{text: s, index: i * 2}
+	})
+
+	// Snapshot
+	snapshot := cl.Snapshot()
+	if len(snapshot) > 0 {
+		t.Error("Snapshot should be empty now")
+	}
+
+	// Add some data
+	cl.Push("hello")
+	cl.Push("world")
+
+	// Previously created snapshot should remain the same
+	if len(snapshot) > 0 {
+		t.Error("Snapshot should not have changed")
+	}
+
+	// But the new snapshot should contain the added items
+	snapshot = cl.Snapshot()
+	if len(snapshot) != 1 {
+		t.Error("Snapshot should not be empty now")
+	}
+
+	// Check the content of the ChunkList
+	chunk1 := snapshot[0]
+	if len(*chunk1) != 2 {
+		t.Error("Snapshot should contain only two items")
+	}
+	if *(*chunk1)[0].text != "hello" || (*chunk1)[0].index != 0 ||
+		*(*chunk1)[1].text != "world" || (*chunk1)[1].index != 2 {
+		t.Error("Invalid data")
+	}
+	if chunk1.IsFull() {
+		t.Error("Chunk should not have been marked full yet")
+	}
+
+	// Add more data
+	for i := 0; i < CHUNK_SIZE*2; i++ {
+		cl.Push(fmt.Sprintf("item %d", i))
+	}
+
+	// Previous snapshot should remain the same
+	if len(snapshot) != 1 {
+		t.Error("Snapshot should stay the same")
+	}
+
+	// New snapshot
+	snapshot = cl.Snapshot()
+	if len(snapshot) != 3 || !snapshot[0].IsFull() ||
+		!snapshot[1].IsFull() || snapshot[2].IsFull() {
+		t.Error("Expected two full chunks and one more chunk")
+	}
+	if len(*snapshot[2]) != 2 {
+		t.Error("Unexpected number of items")
+	}
+}
diff --git a/src/constants.go b/src/constants.go
new file mode 100644
index 00000000..b0b64dbb
--- /dev/null
+++ b/src/constants.go
@@ -0,0 +1,12 @@
+package fzf
+
+const VERSION = "0.9.0"
+
+const (
+	EVT_READ_NEW EventType = iota
+	EVT_READ_FIN
+	EVT_SEARCH_NEW
+	EVT_SEARCH_PROGRESS
+	EVT_SEARCH_FIN
+	EVT_CLOSE
+)
diff --git a/src/core.go b/src/core.go
new file mode 100644
index 00000000..2601397a
--- /dev/null
+++ b/src/core.go
@@ -0,0 +1,153 @@
+package fzf
+
+import (
+	"fmt"
+	"os"
+	"runtime"
+	"time"
+)
+
+const COORDINATOR_DELAY time.Duration = 100 * time.Millisecond
+
+func initProcs() {
+	runtime.GOMAXPROCS(runtime.NumCPU())
+}
+
+/*
+Reader   -> EVT_READ_FIN
+Reader   -> EVT_READ_NEW        -> Matcher  (restart)
+Terminal -> EVT_SEARCH_NEW      -> Matcher  (restart)
+Matcher  -> EVT_SEARCH_PROGRESS -> Terminal (update info)
+Matcher  -> EVT_SEARCH_FIN      -> Terminal (update list)
+*/
+
+func Run(options *Options) {
+	initProcs()
+
+	opts := ParseOptions()
+
+	if opts.Version {
+		fmt.Println(VERSION)
+		os.Exit(0)
+	}
+
+	// Event channel
+	eventBox := NewEventBox()
+
+	// Chunk list
+	var chunkList *ChunkList
+	if len(opts.WithNth) == 0 {
+		chunkList = NewChunkList(func(data *string, index int) *Item {
+			return &Item{text: data, index: index}
+		})
+	} else {
+		chunkList = NewChunkList(func(data *string, index int) *Item {
+			item := Item{text: data, index: index}
+			tokens := Tokenize(item.text, opts.Delimiter)
+			item.origText = item.text
+			item.text = Transform(tokens, opts.WithNth).whole
+			return &item
+		})
+	}
+
+	// Reader
+	reader := Reader{func(str string) { chunkList.Push(str) }, eventBox}
+	go reader.ReadSource()
+
+	// Matcher
+	patternBuilder := func(runes []rune) *Pattern {
+		return BuildPattern(
+			opts.Mode, opts.Case, opts.Nth, opts.Delimiter, runes)
+	}
+	matcher := NewMatcher(patternBuilder, opts.Sort > 0, eventBox)
+
+	// Defered-interactive / Non-interactive
+	//   --select-1 | --exit-0 | --filter
+	if filtering := opts.Filter != nil; filtering || opts.Select1 || opts.Exit0 {
+		limit := 0
+		var patternString string
+		if filtering {
+			patternString = *opts.Filter
+		} else {
+			if opts.Select1 || opts.Exit0 {
+				limit = 1
+			}
+			patternString = opts.Query
+		}
+		pattern := patternBuilder([]rune(patternString))
+
+		looping := true
+		for looping {
+			eventBox.Wait(func(events *Events) {
+				for evt, _ := range *events {
+					switch evt {
+					case EVT_READ_FIN:
+						looping = false
+						return
+					}
+				}
+			})
+			time.Sleep(COORDINATOR_DELAY)
+		}
+
+		matches, cancelled := matcher.scan(MatchRequest{
+			chunks:  chunkList.Snapshot(),
+			pattern: pattern}, limit)
+
+		if !cancelled && (filtering || opts.Exit0) {
+			if opts.PrintQuery {
+				fmt.Println(patternString)
+			}
+			for _, item := range matches {
+				item.Print()
+			}
+			os.Exit(0)
+		}
+	}
+
+	// Go interactive
+	go matcher.Loop()
+
+	// Terminal I/O
+	terminal := NewTerminal(opts, eventBox)
+	go terminal.Loop()
+
+	// Event coordination
+	reading := true
+	ticks := 0
+	for {
+		delay := true
+		ticks += 1
+		eventBox.Wait(func(events *Events) {
+			defer events.Clear()
+			for evt, value := range *events {
+				switch evt {
+
+				case EVT_READ_NEW, EVT_READ_FIN:
+					reading = reading && evt == EVT_READ_NEW
+					terminal.UpdateCount(chunkList.Count(), !reading)
+					matcher.Reset(chunkList.Snapshot(), terminal.Input(), false)
+
+				case EVT_SEARCH_NEW:
+					matcher.Reset(chunkList.Snapshot(), terminal.Input(), true)
+					delay = false
+
+				case EVT_SEARCH_PROGRESS:
+					switch val := value.(type) {
+					case float32:
+						terminal.UpdateProgress(val)
+					}
+
+				case EVT_SEARCH_FIN:
+					switch val := value.(type) {
+					case []*Item:
+						terminal.UpdateList(val)
+					}
+				}
+			}
+		})
+		if ticks > 3 && delay && reading {
+			time.Sleep(COORDINATOR_DELAY)
+		}
+	}
+}
diff --git a/src/curses/curses.go b/src/curses/curses.go
new file mode 100644
index 00000000..945a3ce4
--- /dev/null
+++ b/src/curses/curses.go
@@ -0,0 +1,424 @@
+package curses
+
+// #include <ncurses.h>
+// #include <locale.h>
+// #cgo LDFLAGS: -lncurses
+import "C"
+
+import (
+	"os"
+	"os/signal"
+	"syscall"
+	"time"
+	"unicode/utf8"
+)
+
+const (
+	RUNE = iota
+
+	CTRL_A
+	CTRL_B
+	CTRL_C
+	CTRL_D
+	CTRL_E
+	CTRL_F
+	CTRL_G
+	CTRL_H
+	TAB
+	CTRL_J
+	CTRL_K
+	CTRL_L
+	CTRL_M
+	CTRL_N
+	CTRL_O
+	CTRL_P
+	CTRL_Q
+	CTRL_R
+	CTRL_S
+	CTRL_T
+	CTRL_U
+	CTRL_V
+	CTRL_W
+	CTRL_X
+	CTRL_Y
+	CTRL_Z
+	ESC
+
+	INVALID
+	MOUSE
+
+	BTAB
+
+	DEL
+	PGUP
+	PGDN
+
+	ALT_B
+	ALT_F
+	ALT_D
+	ALT_BS
+)
+
+const (
+	COL_NORMAL = iota
+	COL_PROMPT
+	COL_MATCH
+	COL_CURRENT
+	COL_CURRENT_MATCH
+	COL_SPINNER
+	COL_INFO
+	COL_CURSOR
+	COL_SELECTED
+)
+
+const (
+	DOUBLE_CLICK_DURATION = 500 * time.Millisecond
+)
+
+type Event struct {
+	Type       int
+	Char       rune
+	MouseEvent *MouseEvent
+}
+
+type MouseEvent struct {
+	Y      int
+	X      int
+	S      int
+	Down   bool
+	Double bool
+	Mod    bool
+}
+
+var (
+	_buf          []byte
+	_in           *os.File
+	_color        func(int, bool) C.int
+	_prevDownTime time.Time
+	_prevDownY    int
+	_clickY       []int
+)
+
+func init() {
+	_prevDownTime = time.Unix(0, 0)
+	_clickY = []int{}
+}
+
+func attrColored(pair int, bold bool) C.int {
+	var attr C.int = 0
+	if pair > COL_NORMAL {
+		attr = C.COLOR_PAIR(C.int(pair))
+	}
+	if bold {
+		attr = attr | C.A_BOLD
+	}
+	return attr
+}
+
+func attrMono(pair int, bold bool) C.int {
+	var attr C.int = 0
+	switch pair {
+	case COL_CURRENT:
+		if bold {
+			attr = C.A_REVERSE
+		}
+	case COL_MATCH:
+		attr = C.A_UNDERLINE
+	case COL_CURRENT_MATCH:
+		attr = C.A_UNDERLINE | C.A_REVERSE
+	}
+	if bold {
+		attr = attr | C.A_BOLD
+	}
+	return attr
+}
+
+func MaxX() int {
+	return int(C.COLS)
+}
+
+func MaxY() int {
+	return int(C.LINES)
+}
+
+func getch(nonblock bool) int {
+	b := make([]byte, 1)
+	syscall.SetNonblock(int(_in.Fd()), nonblock)
+	_, err := _in.Read(b)
+	if err != nil {
+		return -1
+	}
+	return int(b[0])
+}
+
+func Init(color bool, color256 bool, black bool, mouse bool) {
+	{
+		in, err := os.OpenFile("/dev/tty", syscall.O_RDONLY, 0)
+		if err != nil {
+			panic("Failed to open /dev/tty")
+		}
+		_in = in
+		// Break STDIN
+		// syscall.Dup2(int(in.Fd()), int(os.Stdin.Fd()))
+	}
+
+	swapOutput()
+
+	C.setlocale(C.LC_ALL, C.CString(""))
+	C.initscr()
+	if mouse {
+		C.mousemask(C.ALL_MOUSE_EVENTS, nil)
+	}
+	C.cbreak()
+	C.noecho()
+	C.raw()          // stty dsusp undef
+	C.set_tabsize(4) // FIXME
+
+	intChan := make(chan os.Signal, 1)
+	signal.Notify(intChan, os.Interrupt, os.Kill)
+	go func() {
+		<-intChan
+		Close()
+		os.Exit(1)
+	}()
+
+	if color {
+		C.start_color()
+		var bg C.short
+		if black {
+			bg = C.COLOR_BLACK
+		} else {
+			C.use_default_colors()
+			bg = -1
+		}
+		if color256 {
+			C.init_pair(COL_PROMPT, 110, bg)
+			C.init_pair(COL_MATCH, 108, bg)
+			C.init_pair(COL_CURRENT, 254, 236)
+			C.init_pair(COL_CURRENT_MATCH, 151, 236)
+			C.init_pair(COL_SPINNER, 148, bg)
+			C.init_pair(COL_INFO, 144, bg)
+			C.init_pair(COL_CURSOR, 161, 236)
+			C.init_pair(COL_SELECTED, 168, 236)
+		} else {
+			C.init_pair(COL_PROMPT, C.COLOR_BLUE, bg)
+			C.init_pair(COL_MATCH, C.COLOR_GREEN, bg)
+			C.init_pair(COL_CURRENT, C.COLOR_YELLOW, C.COLOR_BLACK)
+			C.init_pair(COL_CURRENT_MATCH, C.COLOR_GREEN, C.COLOR_BLACK)
+			C.init_pair(COL_SPINNER, C.COLOR_GREEN, bg)
+			C.init_pair(COL_INFO, C.COLOR_WHITE, bg)
+			C.init_pair(COL_CURSOR, C.COLOR_RED, C.COLOR_BLACK)
+			C.init_pair(COL_SELECTED, C.COLOR_MAGENTA, C.COLOR_BLACK)
+		}
+		_color = attrColored
+	} else {
+		_color = attrMono
+	}
+}
+
+func Close() {
+	C.endwin()
+	swapOutput()
+}
+
+func swapOutput() {
+	syscall.Dup2(2, 3)
+	syscall.Dup2(1, 2)
+	syscall.Dup2(3, 1)
+}
+
+func GetBytes() []byte {
+	c := getch(false)
+	_buf = append(_buf, byte(c))
+
+	for {
+		c = getch(true)
+		if c == -1 {
+			break
+		}
+		_buf = append(_buf, byte(c))
+	}
+
+	return _buf
+}
+
+// 27 (91 79) 77 type x y
+func mouseSequence(sz *int) Event {
+	if len(_buf) < 6 {
+		return Event{INVALID, 0, nil}
+	}
+	*sz = 6
+	switch _buf[3] {
+	case 32, 36, 40, 48, // mouse-down / shift / cmd / ctrl
+		35, 39, 43, 51: // mouse-up / shift / cmd / ctrl
+		mod := _buf[3] >= 36
+		down := _buf[3]%2 == 0
+		x := int(_buf[4] - 33)
+		y := int(_buf[5] - 33)
+		double := false
+		if down {
+			now := time.Now()
+			if now.Sub(_prevDownTime) < DOUBLE_CLICK_DURATION {
+				_clickY = append(_clickY, y)
+			} else {
+				_clickY = []int{y}
+			}
+			_prevDownTime = now
+		} else {
+			if len(_clickY) > 1 && _clickY[0] == _clickY[1] &&
+				time.Now().Sub(_prevDownTime) < DOUBLE_CLICK_DURATION {
+				double = true
+			}
+		}
+		return Event{MOUSE, 0, &MouseEvent{y, x, 0, down, double, mod}}
+	case 96, 100, 104, 112, // scroll-up / shift / cmd / ctrl
+		97, 101, 105, 113: // scroll-down / shift / cmd / ctrl
+		mod := _buf[3] >= 100
+		s := 1 - int(_buf[3]%2)*2
+		return Event{MOUSE, 0, &MouseEvent{0, 0, s, false, false, mod}}
+	}
+	return Event{INVALID, 0, nil}
+}
+
+func escSequence(sz *int) Event {
+	if len(_buf) < 2 {
+		return Event{ESC, 0, nil}
+	}
+	*sz = 2
+	switch _buf[1] {
+	case 98:
+		return Event{ALT_B, 0, nil}
+	case 100:
+		return Event{ALT_D, 0, nil}
+	case 102:
+		return Event{ALT_F, 0, nil}
+	case 127:
+		return Event{ALT_BS, 0, nil}
+	case 91, 79:
+		if len(_buf) < 3 {
+			return Event{INVALID, 0, nil}
+		}
+		*sz = 3
+		switch _buf[2] {
+		case 68:
+			return Event{CTRL_B, 0, nil}
+		case 67:
+			return Event{CTRL_F, 0, nil}
+		case 66:
+			return Event{CTRL_J, 0, nil}
+		case 65:
+			return Event{CTRL_K, 0, nil}
+		case 90:
+			return Event{BTAB, 0, nil}
+		case 72:
+			return Event{CTRL_A, 0, nil}
+		case 70:
+			return Event{CTRL_E, 0, nil}
+		case 77:
+			return mouseSequence(sz)
+		case 49, 50, 51, 52, 53, 54:
+			if len(_buf) < 4 {
+				return Event{INVALID, 0, nil}
+			}
+			*sz = 4
+			switch _buf[2] {
+			case 50:
+				return Event{INVALID, 0, nil} // INS
+			case 51:
+				return Event{DEL, 0, nil}
+			case 52:
+				return Event{CTRL_E, 0, nil}
+			case 53:
+				return Event{PGUP, 0, nil}
+			case 54:
+				return Event{PGDN, 0, nil}
+			case 49:
+				switch _buf[3] {
+				case 126:
+					return Event{CTRL_A, 0, nil}
+				case 59:
+					if len(_buf) != 6 {
+						return Event{INVALID, 0, nil}
+					}
+					*sz = 6
+					switch _buf[4] {
+					case 50:
+						switch _buf[5] {
+						case 68:
+							return Event{CTRL_A, 0, nil}
+						case 67:
+							return Event{CTRL_E, 0, nil}
+						}
+					case 53:
+						switch _buf[5] {
+						case 68:
+							return Event{ALT_B, 0, nil}
+						case 67:
+							return Event{ALT_F, 0, nil}
+						}
+					} // _buf[4]
+				} // _buf[3]
+			} // _buf[2]
+		} // _buf[2]
+	} // _buf[1]
+	return Event{INVALID, 0, nil}
+}
+
+func GetChar() Event {
+	if len(_buf) == 0 {
+		_buf = GetBytes()
+	}
+	if len(_buf) == 0 {
+		panic("Empty _buffer")
+	}
+
+	sz := 1
+	defer func() {
+		_buf = _buf[sz:]
+	}()
+
+	switch _buf[0] {
+	case CTRL_C, CTRL_G, CTRL_Q:
+		return Event{CTRL_C, 0, nil}
+	case 127:
+		return Event{CTRL_H, 0, nil}
+	case ESC:
+		return escSequence(&sz)
+	}
+
+	// CTRL-A ~ CTRL-Z
+	if _buf[0] <= CTRL_Z {
+		return Event{int(_buf[0]), 0, nil}
+	}
+	r, rsz := utf8.DecodeRune(_buf)
+	sz = rsz
+	return Event{RUNE, r, nil}
+}
+
+func Move(y int, x int) {
+	C.move(C.int(y), C.int(x))
+}
+
+func MoveAndClear(y int, x int) {
+	Move(y, x)
+	C.clrtoeol()
+}
+
+func Print(text string) {
+	C.addstr(C.CString(text))
+}
+
+func CPrint(pair int, bold bool, text string) {
+	attr := _color(pair, bold)
+	C.attron(attr)
+	Print(text)
+	C.attroff(attr)
+}
+
+func Clear() {
+	C.clear()
+}
+
+func Refresh() {
+	C.refresh()
+}
diff --git a/src/eventbox.go b/src/eventbox.go
new file mode 100644
index 00000000..6685e7c8
--- /dev/null
+++ b/src/eventbox.go
@@ -0,0 +1,48 @@
+package fzf
+
+import "sync"
+
+type EventType int
+
+type Events map[EventType]interface{}
+
+type EventBox struct {
+	events Events
+	cond   *sync.Cond
+}
+
+func NewEventBox() *EventBox {
+	return &EventBox{make(Events), sync.NewCond(&sync.Mutex{})}
+}
+
+func (b *EventBox) Wait(callback func(*Events)) {
+	b.cond.L.Lock()
+	defer b.cond.L.Unlock()
+
+	if len(b.events) == 0 {
+		b.cond.Wait()
+	}
+
+	callback(&b.events)
+}
+
+func (b *EventBox) Set(event EventType, value interface{}) {
+	b.cond.L.Lock()
+	defer b.cond.L.Unlock()
+	b.events[event] = value
+	b.cond.Broadcast()
+}
+
+// Unsynchronized; should be called within Wait routine
+func (events *Events) Clear() {
+	for event := range *events {
+		delete(*events, event)
+	}
+}
+
+func (b *EventBox) Peak(event EventType) bool {
+	b.cond.L.Lock()
+	defer b.cond.L.Unlock()
+	_, ok := b.events[event]
+	return ok
+}
diff --git a/src/fzf/main.go b/src/fzf/main.go
new file mode 100644
index 00000000..29d4767c
--- /dev/null
+++ b/src/fzf/main.go
@@ -0,0 +1,7 @@
+package main
+
+import "github.com/junegunn/fzf/src"
+
+func main() {
+	fzf.Run(fzf.ParseOptions())
+}
diff --git a/src/item.go b/src/item.go
new file mode 100644
index 00000000..b70da93f
--- /dev/null
+++ b/src/item.go
@@ -0,0 +1,135 @@
+package fzf
+
+import (
+	"fmt"
+	"sort"
+)
+
+type Offset [2]int
+
+type Item struct {
+	text        *string
+	origText    *string
+	offsets     []Offset
+	index       int
+	rank        Rank
+	transformed *Transformed
+}
+
+type Rank [3]int
+
+var NilRank = Rank{-1, 0, 0}
+
+func (i *Item) Rank() Rank {
+	if i.rank[0] > 0 {
+		return i.rank
+	}
+	sort.Sort(ByOrder(i.offsets))
+	matchlen := 0
+	prevEnd := 0
+	for _, offset := range i.offsets {
+		begin := offset[0]
+		end := offset[1]
+		if prevEnd > begin {
+			begin = prevEnd
+		}
+		if end > prevEnd {
+			prevEnd = end
+		}
+		if end > begin {
+			matchlen += end - begin
+		}
+	}
+	i.rank = Rank{matchlen, len(*i.text), i.index}
+	return i.rank
+}
+
+func (i *Item) Print() {
+	if i.origText != nil {
+		fmt.Println(*i.origText)
+	} else {
+		fmt.Println(*i.text)
+	}
+}
+
+type ByOrder []Offset
+
+func (a ByOrder) Len() int {
+	return len(a)
+}
+
+func (a ByOrder) Swap(i, j int) {
+	a[i], a[j] = a[j], a[i]
+}
+
+func (a ByOrder) Less(i, j int) bool {
+	ioff := a[i]
+	joff := a[j]
+	return (ioff[0] < joff[0]) || (ioff[0] == joff[0]) && (ioff[1] <= joff[1])
+}
+
+type ByRelevance []*Item
+
+func (a ByRelevance) Len() int {
+	return len(a)
+}
+
+func (a ByRelevance) Swap(i, j int) {
+	a[i], a[j] = a[j], a[i]
+}
+
+func (a ByRelevance) Less(i, j int) bool {
+	irank := a[i].Rank()
+	jrank := a[j].Rank()
+
+	return compareRanks(irank, jrank)
+}
+
+func compareRanks(irank Rank, jrank Rank) bool {
+	for idx := range irank {
+		if irank[idx] < jrank[idx] {
+			return true
+		} else if irank[idx] > jrank[idx] {
+			return false
+		}
+	}
+	return true
+}
+
+func SortMerge(partialResults [][]*Item) []*Item {
+	if len(partialResults) == 1 {
+		return partialResults[0]
+	}
+
+	merged := []*Item{}
+
+	for len(partialResults) > 0 {
+		minRank := Rank{0, 0, 0}
+		minIdx := -1
+
+		for idx, partialResult := range partialResults {
+			if len(partialResult) > 0 {
+				rank := partialResult[0].Rank()
+				if minIdx < 0 || compareRanks(rank, minRank) {
+					minRank = rank
+					minIdx = idx
+				}
+			}
+		}
+
+		if minIdx >= 0 {
+			merged = append(merged, partialResults[minIdx][0])
+			partialResults[minIdx] = partialResults[minIdx][1:]
+		}
+
+		nonEmptyPartialResults := make([][]*Item, 0, len(partialResults))
+		for _, partialResult := range partialResults {
+			if len(partialResult) > 0 {
+				nonEmptyPartialResults = append(nonEmptyPartialResults, partialResult)
+			}
+		}
+		partialResults = nonEmptyPartialResults
+	}
+
+	return merged
+}
diff --git a/src/item_test.go b/src/item_test.go
new file mode 100644
index 00000000..1e316291
--- /dev/null
+++ b/src/item_test.go
@@ -0,0 +1,78 @@
+package fzf
+
+import (
+	"sort"
+	"testing"
+)
+
+func TestOffsetSort(t *testing.T) {
+	offsets := []Offset{
+		Offset{3, 5}, Offset{2, 7},
+		Offset{1, 3}, Offset{2, 9}}
+	sort.Sort(ByOrder(offsets))
+
+	if offsets[0][0] != 1 || offsets[0][1] != 3 ||
+		offsets[1][0] != 2 || offsets[1][1] != 7 ||
+		offsets[2][0] != 2 || offsets[2][1] != 9 ||
+		offsets[3][0] != 3 || offsets[3][1] != 5 {
+		t.Error("Invalid order:", offsets)
+	}
+}
+
+func TestRankComparison(t *testing.T) {
+	if compareRanks(Rank{3, 0, 5}, Rank{2, 0, 7}) ||
+		!compareRanks(Rank{3, 0, 5}, Rank{3, 0, 6}) ||
+		!compareRanks(Rank{1, 2, 3}, Rank{1, 3, 2}) ||
+		!compareRanks(NilRank, Rank{0, 0, 0}) ||
+		compareRanks(Rank{0, 0, 0}, NilRank) {
+		t.Error("Invalid order")
+	}
+}
+
+// Match length, string length, index
+func TestItemRank(t *testing.T) {
+	strs := []string{"foo", "foobar", "bar", "baz"}
+	item1 := Item{text: &strs[0], index: 1, offsets: []Offset{}}
+	rank1 := item1.Rank()
+	if rank1[0] != 0 || rank1[1] != 3 || rank1[2] != 1 {
+		t.Error(item1.Rank())
+	}
+	// Only differ in index
+	item2 := Item{text: &strs[0], index: 0, offsets: []Offset{}}
+
+	items := []*Item{&item1, &item2}
+	sort.Sort(ByRelevance(items))
+	if items[0] != &item2 || items[1] != &item1 {
+		t.Error(items)
+	}
+
+	items = []*Item{&item2, &item1, &item1, &item2}
+	sort.Sort(ByRelevance(items))
+	if items[0] != &item2 || items[1] != &item2 ||
+		items[2] != &item1 || items[3] != &item1 {
+		t.Error(items)
+	}
+
+	// Sort by relevance
+	item3 := Item{text: &strs[1], index: 2, offsets: []Offset{Offset{1, 3}, Offset{5, 7}}}
+	item4 := Item{text: &strs[1], index: 2, offsets: []Offset{Offset{1, 2}, Offset{6, 7}}}
+	item5 := Item{text: &strs[2], index: 2, offsets: []Offset{Offset{1, 3}, Offset{5, 7}}}
+	item6 := Item{text: &strs[2], index: 2, offsets: []Offset{Offset{1, 2}, Offset{6, 7}}}
+	items = []*Item{&item1, &item2, &item3, &item4, &item5, &item6}
+	sort.Sort(ByRelevance(items))
+	if items[0] != &item2 || items[1] != &item1 ||
+		items[2] != &item6 || items[3] != &item4 ||
+		items[4] != &item5 || items[5] != &item3 {
+		t.Error(items)
+	}
+
+	// Sort merged lists
+	lists := [][]*Item{
+		[]*Item{&item2, &item4, &item5}, []*Item{&item1, &item6}, []*Item{&item3}}
+	items = SortMerge(lists)
+	if items[0] != &item2 || items[1] != &item1 ||
+		items[2] != &item6 || items[3] != &item4 ||
+		items[4] != &item5 || items[5] != &item3 {
+		t.Error(items)
+	}
+}
diff --git a/src/matcher.go b/src/matcher.go
new file mode 100644
index 00000000..363b07fd
--- /dev/null
+++ b/src/matcher.go
@@ -0,0 +1,215 @@
+package fzf
+
+import (
+	"fmt"
+	"runtime"
+	"sort"
+	"time"
+)
+
+type MatchRequest struct {
+	chunks  []*Chunk
+	pattern *Pattern
+}
+
+type Matcher struct {
+	patternBuilder func([]rune) *Pattern
+	sort           bool
+	eventBox       *EventBox
+	reqBox         *EventBox
+	partitions     int
+	queryCache     QueryCache
+}
+
+const (
+	REQ_RETRY EventType = iota
+	REQ_RESET
+)
+
+const (
+	STAT_CANCELLED int = iota
+	STAT_QCH
+	STAT_CHUNKS
+)
+
+const (
+	PROGRESS_MIN_DURATION = 200 * time.Millisecond
+)
+
+func NewMatcher(patternBuilder func([]rune) *Pattern,
+	sort bool, eventBox *EventBox) *Matcher {
+	return &Matcher{
+		patternBuilder: patternBuilder,
+		sort:           sort,
+		eventBox:       eventBox,
+		reqBox:         NewEventBox(),
+		partitions:     runtime.NumCPU(),
+		queryCache:     make(QueryCache)}
+}
+
+func (m *Matcher) Loop() {
+	prevCount := 0
+
+	for {
+		var request MatchRequest
+
+		m.reqBox.Wait(func(events *Events) {
+			for _, val := range *events {
+				switch val := val.(type) {
+				case MatchRequest:
+					request = val
+				default:
+					panic(fmt.Sprintf("Unexpected type: %T", val))
+				}
+			}
+			events.Clear()
+		})
+
+		// Restart search
+		patternString := request.pattern.AsString()
+		allMatches := []*Item{}
+		cancelled := false
+		count := CountItems(request.chunks)
+
+		foundCache := false
+		if count == prevCount {
+			// Look up queryCache
+			if cached, found := m.queryCache[patternString]; found {
+				foundCache = true
+				allMatches = cached
+			}
+		} else {
+			// Invalidate queryCache
+			prevCount = count
+			m.queryCache = make(QueryCache)
+		}
+
+		if !foundCache {
+			allMatches, cancelled = m.scan(request, 0)
+		}
+
+		if !cancelled {
+			m.queryCache[patternString] = allMatches
+			m.eventBox.Set(EVT_SEARCH_FIN, allMatches)
+		}
+	}
+}
+
+func (m *Matcher) sliceChunks(chunks []*Chunk) [][]*Chunk {
+	perSlice := len(chunks) / m.partitions
+
+	// No need to parallelize
+	if perSlice == 0 {
+		return [][]*Chunk{chunks}
+	}
+
+	slices := make([][]*Chunk, m.partitions)
+	for i := 0; i < m.partitions; i++ {
+		start := i * perSlice
+		end := start + perSlice
+		if i == m.partitions-1 {
+			end = len(chunks)
+		}
+		slices[i] = chunks[start:end]
+	}
+	return slices
+}
+
+type partialResult struct {
+	index   int
+	matches []*Item
+}
+
+func (m *Matcher) scan(request MatchRequest, limit int) ([]*Item, bool) {
+	startedAt := time.Now()
+
+	numChunks := len(request.chunks)
+	if numChunks == 0 {
+		return []*Item{}, false
+	}
+	pattern := request.pattern
+	empty := pattern.IsEmpty()
+	cancelled := NewAtomicBool(false)
+
+	slices := m.sliceChunks(request.chunks)
+	numSlices := len(slices)
+	resultChan := make(chan partialResult, numSlices)
+	countChan := make(chan int, numSlices)
+
+	for idx, chunks := range slices {
+		go func(idx int, chunks []*Chunk) {
+			sliceMatches := []*Item{}
+			for _, chunk := range chunks {
+				var matches []*Item
+				if empty {
+					matches = *chunk
+				} else {
+					matches = request.pattern.Match(chunk)
+				}
+				sliceMatches = append(sliceMatches, matches...)
+				if cancelled.Get() {
+					return
+				}
+				countChan <- len(sliceMatches)
+			}
+			if !empty && m.sort {
+				sort.Sort(ByRelevance(sliceMatches))
+			}
+			resultChan <- partialResult{idx, sliceMatches}
+		}(idx, chunks)
+	}
+
+	count := 0
+	matchCount := 0
+	for matchesInChunk := range countChan {
+		count += 1
+		matchCount += matchesInChunk
+
+		if limit > 0 && matchCount > limit {
+			return nil, true // For --select-1 and --exit-0
+		}
+
+		if count == numChunks {
+			break
+		}
+
+		if !empty && m.reqBox.Peak(REQ_RESET) {
+			cancelled.Set(true)
+			return nil, true
+		}
+
+		if time.Now().Sub(startedAt) > PROGRESS_MIN_DURATION {
+			m.eventBox.Set(EVT_SEARCH_PROGRESS, float32(count)/float32(numChunks))
+		}
+	}
+
+	partialResults := make([][]*Item, numSlices)
+	for range slices {
+		partialResult := <-resultChan
+		partialResults[partialResult.index] = partialResult.matches
+	}
+
+	var allMatches []*Item
+	if empty || !m.sort {
+		allMatches = []*Item{}
+		for _, matches := range partialResults {
+			allMatches = append(allMatches, matches...)
+		}
+	} else {
+		allMatches = SortMerge(partialResults)
+	}
+
+	return allMatches, false
+}
+
+func (m *Matcher) Reset(chunks []*Chunk, patternRunes []rune, cancel bool) {
+	pattern := m.patternBuilder(patternRunes)
+
+	var event EventType
+	if cancel {
+		event = REQ_RESET
+	} else {
+		event = REQ_RETRY
+	}
+	m.reqBox.Set(event, MatchRequest{chunks, pattern})
+}
diff --git a/src/options.go b/src/options.go
new file mode 100644
index 00000000..4929dfd7
--- /dev/null
+++ b/src/options.go
@@ -0,0 +1,276 @@
+package fzf
+
+import (
+	"fmt"
+	"github.com/junegunn/go-shellwords"
+	"os"
+	"regexp"
+	"strings"
+)
+
+const USAGE = `usage: fzf [options]
+
+  Search
+    -x, --extended        Extended-search mode
+    -e, --extended-exact  Extended-search mode (exact match)
+    -i                    Case-insensitive match (default: smart-case match)
+    +i                    Case-sensitive match
+    -n, --nth=N[,..]      Comma-separated list of field index expressions
+                          for limiting search scope. Each can be a non-zero
+                          integer or a range expression ([BEGIN]..[END])
+        --with-nth=N[,..] Transform the item using index expressions for search
+    -d, --delimiter=STR   Field delimiter regex for --nth (default: AWK-style)
+
+  Search result
+    -s, --sort            Sort the result
+    +s, --no-sort         Do not sort the result. Keep the sequence unchanged.
+
+  Interface
+    -m, --multi           Enable multi-select with tab/shift-tab
+        --no-mouse        Disable mouse
+    +c, --no-color        Disable colors
+    +2, --no-256          Disable 256-color
+        --black           Use black background
+        --reverse         Reverse orientation
+        --prompt=STR      Input prompt (default: '> ')
+
+  Scripting
+    -q, --query=STR       Start the finder with the given query
+    -1, --select-1        Automatically select the only match
+    -0, --exit-0          Exit immediately when there's no match
+    -f, --filter=STR      Filter mode. Do not start interactive finder.
+        --print-query     Print query as the first line
+
+  Environment variables
+    FZF_DEFAULT_COMMAND   Default command to use when input is tty
+    FZF_DEFAULT_OPTS      Defaults options. (e.g. "-x -m")
+
+`
+
+type Mode int
+
+const (
+	MODE_FUZZY Mode = iota
+	MODE_EXTENDED
+	MODE_EXTENDED_EXACT
+)
+
+type Case int
+
+const (
+	CASE_SMART Case = iota
+	CASE_IGNORE
+	CASE_RESPECT
+)
+
+type Options struct {
+	Mode       Mode
+	Case       Case
+	Nth        []Range
+	WithNth    []Range
+	Delimiter  *regexp.Regexp
+	Sort       int
+	Multi      bool
+	Mouse      bool
+	Color      bool
+	Color256   bool
+	Black      bool
+	Reverse    bool
+	Prompt     string
+	Query      string
+	Select1    bool
+	Exit0      bool
+	Filter     *string
+	PrintQuery bool
+	Version    bool
+}
+
+func DefaultOptions() *Options {
+	return &Options{
+		Mode:       MODE_FUZZY,
+		Case:       CASE_SMART,
+		Nth:        make([]Range, 0),
+		WithNth:    make([]Range, 0),
+		Delimiter:  nil,
+		Sort:       1000,
+		Multi:      false,
+		Mouse:      true,
+		Color:      true,
+		Color256:   strings.Contains(os.Getenv("TERM"), "256"),
+		Black:      false,
+		Reverse:    false,
+		Prompt:     "> ",
+		Query:      "",
+		Select1:    false,
+		Exit0:      false,
+		Filter:     nil,
+		PrintQuery: false,
+		Version:    false}
+}
+
+func help(ok int) {
+	os.Stderr.WriteString(USAGE)
+	os.Exit(ok)
+}
+
+func errorExit(msg string) {
+	os.Stderr.WriteString(msg + "\n")
+	help(1)
+}
+
+func optString(arg string, prefix string) (bool, string) {
+	rx, _ := regexp.Compile(fmt.Sprintf("^(?:%s)(.*)$", prefix))
+	matches := rx.FindStringSubmatch(arg)
+	if len(matches) > 1 {
+		return true, matches[1]
+	} else {
+		return false, ""
+	}
+}
+
+func nextString(args []string, i *int, message string) string {
+	if len(args) > *i+1 {
+		*i++
+	} else {
+		errorExit(message)
+	}
+	return args[*i]
+}
+
+func optionalNumeric(args []string, i *int) int {
+	if len(args) > *i+1 {
+		if strings.IndexAny(args[*i+1], "0123456789") == 0 {
+			*i++
+		}
+	}
+	return 1 // Don't care
+}
+
+func splitNth(str string) []Range {
+	if match, _ := regexp.MatchString("^[0-9,-.]+$", str); !match {
+		errorExit("invalid format: " + str)
+	}
+
+	tokens := strings.Split(str, ",")
+	ranges := make([]Range, len(tokens))
+	for idx, s := range tokens {
+		r, ok := ParseRange(&s)
+		if !ok {
+			errorExit("invalid format: " + str)
+		}
+		ranges[idx] = r
+	}
+	return ranges
+}
+
+func delimiterRegexp(str string) *regexp.Regexp {
+	rx, e := regexp.Compile(str)
+	if e != nil {
+		str = regexp.QuoteMeta(str)
+	}
+
+	rx, e = regexp.Compile(fmt.Sprintf("(?:.*?%s)|(?:.+?$)", str))
+	if e != nil {
+		errorExit("invalid regular expression: " + e.Error())
+	}
+	return rx
+}
+
+func parseOptions(opts *Options, allArgs []string) {
+	for i := 0; i < len(allArgs); i++ {
+		arg := allArgs[i]
+		switch arg {
+		case "-h", "--help":
+			help(0)
+		case "-x", "--extended":
+			opts.Mode = MODE_EXTENDED
+		case "-e", "--extended-exact":
+			opts.Mode = MODE_EXTENDED_EXACT
+		case "+x", "--no-extended", "+e", "--no-extended-exact":
+			opts.Mode = MODE_FUZZY
+		case "-q", "--query":
+			opts.Query = nextString(allArgs, &i, "query string required")
+		case "-f", "--filter":
+			filter := nextString(allArgs, &i, "query string required")
+			opts.Filter = &filter
+		case "-d", "--delimiter":
+			opts.Delimiter = delimiterRegexp(nextString(allArgs, &i, "delimiter required"))
+		case "-n", "--nth":
+			opts.Nth = splitNth(nextString(allArgs, &i, "nth expression required"))
+		case "--with-nth":
+			opts.WithNth = splitNth(nextString(allArgs, &i, "nth expression required"))
+		case "-s", "--sort":
+			opts.Sort = optionalNumeric(allArgs, &i)
+		case "+s", "--no-sort":
+			opts.Sort = 0
+		case "-i":
+			opts.Case = CASE_IGNORE
+		case "+i":
+			opts.Case = CASE_RESPECT
+		case "-m", "--multi":
+			opts.Multi = true
+		case "+m", "--no-multi":
+			opts.Multi = false
+		case "--no-mouse":
+			opts.Mouse = false
+		case "+c", "--no-color":
+			opts.Color = false
+		case "+2", "--no-256":
+			opts.Color256 = false
+		case "--black":
+			opts.Black = true
+		case "--no-black":
+			opts.Black = false
+		case "--reverse":
+			opts.Reverse = true
+		case "--no-reverse":
+			opts.Reverse = false
+		case "-1", "--select-1":
+			opts.Select1 = true
+		case "+1", "--no-select-1":
+			opts.Select1 = false
+		case "-0", "--exit-0":
+			opts.Exit0 = true
+		case "+0", "--no-exit-0":
+			opts.Exit0 = false
+		case "--print-query":
+			opts.PrintQuery = true
+		case "--no-print-query":
+			opts.PrintQuery = false
+		case "--prompt":
+			opts.Prompt = nextString(allArgs, &i, "prompt string required")
+		case "--version":
+			opts.Version = true
+		default:
+			if match, value := optString(arg, "-q|--query="); match {
+				opts.Query = value
+			} else if match, value := optString(arg, "-f|--filter="); match {
+				opts.Filter = &value
+			} else if match, value := optString(arg, "-d|--delimiter="); match {
+				opts.Delimiter = delimiterRegexp(value)
+			} else if match, value := optString(arg, "--prompt="); match {
+				opts.Prompt = value
+			} else if match, value := optString(arg, "-n|--nth="); match {
+				opts.Nth = splitNth(value)
+			} else if match, value := optString(arg, "--with-nth="); match {
+				opts.WithNth = splitNth(value)
+			} else if match, _ := optString(arg, "-s|--sort="); match {
+				opts.Sort = 1 // Don't care
+			} else {
+				errorExit("unknown option: " + arg)
+			}
+		}
+	}
+}
+
+func ParseOptions() *Options {
+	opts := DefaultOptions()
+
+	// Options from Env var
+	words, _ := shellwords.Parse(os.Getenv("FZF_DEFAULT_OPTS"))
+	parseOptions(opts, words)
+
+	// Options from command-line arguments
+	parseOptions(opts, os.Args[1:])
+	return opts
+}
diff --git a/src/options_test.go b/src/options_test.go
new file mode 100644
index 00000000..f0aa3a0d
--- /dev/null
+++ b/src/options_test.go
@@ -0,0 +1,37 @@
+package fzf
+
+import "testing"
+
+func TestDelimiterRegex(t *testing.T) {
+	rx := delimiterRegexp("*")
+	tokens := rx.FindAllString("-*--*---**---", -1)
+	if tokens[0] != "-*" || tokens[1] != "--*" || tokens[2] != "---*" ||
+		tokens[3] != "*" || tokens[4] != "---" {
+		t.Errorf("%s %s %d", rx, tokens, len(tokens))
+	}
+}
+
+func TestSplitNth(t *testing.T) {
+	{
+		ranges := splitNth("..")
+		if len(ranges) != 1 ||
+			ranges[0].begin != RANGE_ELLIPSIS ||
+			ranges[0].end != RANGE_ELLIPSIS {
+			t.Errorf("%s", ranges)
+		}
+	}
+	{
+		ranges := splitNth("..3,1..,2..3,4..-1,-3..-2,..,2,-2")
+		if len(ranges) != 8 ||
+			ranges[0].begin != RANGE_ELLIPSIS || ranges[0].end != 3 ||
+			ranges[1].begin != 1 || ranges[1].end != RANGE_ELLIPSIS ||
+			ranges[2].begin != 2 || ranges[2].end != 3 ||
+			ranges[3].begin != 4 || ranges[3].end != -1 ||
+			ranges[4].begin != -3 || ranges[4].end != -2 ||
+			ranges[5].begin != RANGE_ELLIPSIS || ranges[5].end != RANGE_ELLIPSIS ||
+			ranges[6].begin != 2 || ranges[6].end != 2 ||
+			ranges[7].begin != -2 || ranges[7].end != -2 {
+			t.Errorf("%s", ranges)
+		}
+	}
+}
diff --git a/src/pattern.go b/src/pattern.go
new file mode 100644
index 00000000..533aa59f
--- /dev/null
+++ b/src/pattern.go
@@ -0,0 +1,305 @@
+package fzf
+
+import (
+	"regexp"
+	"sort"
+	"strings"
+)
+
+const UPPERCASE = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
+
+// fuzzy
+// 'exact
+// ^exact-prefix
+// exact-suffix$
+// !not-fuzzy
+// !'not-exact
+// !^not-exact-prefix
+// !not-exact-suffix$
+
+type TermType int
+
+const (
+	TERM_FUZZY TermType = iota
+	TERM_EXACT
+	TERM_PREFIX
+	TERM_SUFFIX
+)
+
+type Term struct {
+	typ      TermType
+	inv      bool
+	text     []rune
+	origText []rune
+}
+
+type Pattern struct {
+	mode          Mode
+	caseSensitive bool
+	text          []rune
+	terms         []Term
+	hasInvTerm    bool
+	delimiter     *regexp.Regexp
+	nth           []Range
+	procFun       map[TermType]func(bool, *string, []rune) (int, int)
+}
+
+var (
+	_patternCache map[string]*Pattern
+	_splitRegex   *regexp.Regexp
+	_cache        ChunkCache
+)
+
+func init() {
+	// We can uniquely identify the pattern for a given string since
+	// mode and caseMode do not change while the program is running
+	_patternCache = make(map[string]*Pattern)
+	_splitRegex = regexp.MustCompile("\\s+")
+	_cache = NewChunkCache()
+}
+
+func clearPatternCache() {
+	_patternCache = make(map[string]*Pattern)
+}
+
+func BuildPattern(mode Mode, caseMode Case,
+	nth []Range, delimiter *regexp.Regexp, runes []rune) *Pattern {
+
+	var asString string
+	switch mode {
+	case MODE_EXTENDED, MODE_EXTENDED_EXACT:
+		asString = strings.Trim(string(runes), " ")
+	default:
+		asString = string(runes)
+	}
+
+	cached, found := _patternCache[asString]
+	if found {
+		return cached
+	}
+
+	caseSensitive, hasInvTerm := true, false
+	terms := []Term{}
+
+	switch caseMode {
+	case CASE_SMART:
+		if !strings.ContainsAny(asString, UPPERCASE) {
+			runes, caseSensitive = []rune(strings.ToLower(asString)), false
+		}
+	case CASE_IGNORE:
+		runes, caseSensitive = []rune(strings.ToLower(asString)), false
+	}
+
+	switch mode {
+	case MODE_EXTENDED, MODE_EXTENDED_EXACT:
+		terms = parseTerms(mode, string(runes))
+		for _, term := range terms {
+			if term.inv {
+				hasInvTerm = true
+			}
+		}
+	}
+
+	ptr := &Pattern{
+		mode:          mode,
+		caseSensitive: caseSensitive,
+		text:          runes,
+		terms:         terms,
+		hasInvTerm:    hasInvTerm,
+		nth:           nth,
+		delimiter:     delimiter,
+		procFun:       make(map[TermType]func(bool, *string, []rune) (int, int))}
+
+	ptr.procFun[TERM_FUZZY] = FuzzyMatch
+	ptr.procFun[TERM_EXACT] = ExactMatchNaive
+	ptr.procFun[TERM_PREFIX] = PrefixMatch
+	ptr.procFun[TERM_SUFFIX] = SuffixMatch
+
+	_patternCache[asString] = ptr
+	return ptr
+}
+
+func parseTerms(mode Mode, str string) []Term {
+	tokens := _splitRegex.Split(str, -1)
+	terms := []Term{}
+	for _, token := range tokens {
+		typ, inv, text := TERM_FUZZY, false, token
+		origText := []rune(text)
+		if mode == MODE_EXTENDED_EXACT {
+			typ = TERM_EXACT
+		}
+
+		if strings.HasPrefix(text, "!") {
+			inv = true
+			text = text[1:]
+		}
+
+		if strings.HasPrefix(text, "'") {
+			if mode == MODE_EXTENDED {
+				typ = TERM_EXACT
+				text = text[1:]
+			}
+		} else if strings.HasPrefix(text, "^") {
+			typ = TERM_PREFIX
+			text = text[1:]
+		} else if strings.HasSuffix(text, "$") {
+			typ = TERM_SUFFIX
+			text = text[:len(text)-1]
+		}
+
+		if len(text) > 0 {
+			terms = append(terms, Term{
+				typ:      typ,
+				inv:      inv,
+				text:     []rune(text),
+				origText: origText})
+		}
+	}
+	return terms
+}
+
+func (p *Pattern) IsEmpty() bool {
+	if p.mode == MODE_FUZZY {
+		return len(p.text) == 0
+	} else {
+		return len(p.terms) == 0
+	}
+}
+
+func (p *Pattern) AsString() string {
+	return string(p.text)
+}
+
+func (p *Pattern) CacheKey() string {
+	if p.mode == MODE_FUZZY {
+		return p.AsString()
+	}
+	cacheableTerms := []string{}
+	for _, term := range p.terms {
+		if term.inv {
+			continue
+		}
+		cacheableTerms = append(cacheableTerms, string(term.origText))
+	}
+	sort.Strings(cacheableTerms)
+	return strings.Join(cacheableTerms, " ")
+}
+
+func (p *Pattern) Match(chunk *Chunk) []*Item {
+	space := chunk
+
+	// ChunkCache: Exact match
+	cacheKey := p.CacheKey()
+	if !p.hasInvTerm { // Because we're excluding Inv-term from cache key
+		if cached, found := _cache.Find(chunk, cacheKey); found {
+			return cached
+		}
+	}
+
+	// ChunkCache: Prefix match
+	foundPrefixCache := false
+	for idx := len(cacheKey) - 1; idx > 0; idx-- {
+		if cached, found := _cache.Find(chunk, cacheKey[:idx]); found {
+			cachedChunk := Chunk(cached)
+			space = &cachedChunk
+			foundPrefixCache = true
+			break
+		}
+	}
+
+	// ChunkCache: Suffix match
+	if !foundPrefixCache {
+		for idx := 1; idx < len(cacheKey); idx++ {
+			if cached, found := _cache.Find(chunk, cacheKey[idx:]); found {
+				cachedChunk := Chunk(cached)
+				space = &cachedChunk
+				break
+			}
+		}
+	}
+
+	var matches []*Item
+	if p.mode == MODE_FUZZY {
+		matches = p.fuzzyMatch(space)
+	} else {
+		matches = p.extendedMatch(space)
+	}
+
+	if !p.hasInvTerm {
+		_cache.Add(chunk, cacheKey, matches)
+	}
+	return matches
+}
+
+func (p *Pattern) fuzzyMatch(chunk *Chunk) []*Item {
+	matches := []*Item{}
+	for _, item := range *chunk {
+		input := p.prepareInput(item)
+		if sidx, eidx := p.iter(FuzzyMatch, input, p.text); sidx >= 0 {
+			matches = append(matches, &Item{
+				text:    item.text,
+				index:   item.index,
+				offsets: []Offset{Offset{sidx, eidx}},
+				rank:    NilRank})
+		}
+	}
+	return matches
+}
+
+func (p *Pattern) extendedMatch(chunk *Chunk) []*Item {
+	matches := []*Item{}
+	for _, item := range *chunk {
+		input := p.prepareInput(item)
+		offsets := []Offset{}
+	Loop:
+		for _, term := range p.terms {
+			pfun := p.procFun[term.typ]
+			if sidx, eidx := p.iter(pfun, input, term.text); sidx >= 0 {
+				if term.inv {
+					break Loop
+				}
+				offsets = append(offsets, Offset{sidx, eidx})
+			} else if term.inv {
+				offsets = append(offsets, Offset{0, 0})
+			}
+		}
+		if len(offsets) == len(p.terms) {
+			matches = append(matches, &Item{
+				text:    item.text,
+				index:   item.index,
+				offsets: offsets,
+				rank:    NilRank})
+		}
+	}
+	return matches
+}
+
+func (p *Pattern) prepareInput(item *Item) *Transformed {
+	if item.transformed != nil {
+		return item.transformed
+	}
+
+	var ret *Transformed
+	if len(p.nth) > 0 {
+		tokens := Tokenize(item.text, p.delimiter)
+		ret = Transform(tokens, p.nth)
+	} else {
+		trans := Transformed{
+			whole: item.text,
+			parts: []Token{Token{text: item.text, prefixLength: 0}}}
+		ret = &trans
+	}
+	item.transformed = ret
+	return ret
+}
+
+func (p *Pattern) iter(pfun func(bool, *string, []rune) (int, int),
+	inputs *Transformed, pattern []rune) (int, int) {
+	for _, part := range inputs.parts {
+		prefixLength := part.prefixLength
+		if sidx, eidx := pfun(p.caseSensitive, part.text, pattern); sidx >= 0 {
+			return sidx + prefixLength, eidx + prefixLength
+		}
+	}
+	return -1, -1
+}
diff --git a/src/pattern_test.go b/src/pattern_test.go
new file mode 100644
index 00000000..a1ce6263
--- /dev/null
+++ b/src/pattern_test.go
@@ -0,0 +1,87 @@
+package fzf
+
+import "testing"
+
+func TestParseTermsExtended(t *testing.T) {
+	terms := parseTerms(MODE_EXTENDED,
+		"aaa 'bbb ^ccc ddd$ !eee !'fff !^ggg !hhh$")
+	if len(terms) != 8 ||
+		terms[0].typ != TERM_FUZZY || terms[0].inv ||
+		terms[1].typ != TERM_EXACT || terms[1].inv ||
+		terms[2].typ != TERM_PREFIX || terms[2].inv ||
+		terms[3].typ != TERM_SUFFIX || terms[3].inv ||
+		terms[4].typ != TERM_FUZZY || !terms[4].inv ||
+		terms[5].typ != TERM_EXACT || !terms[5].inv ||
+		terms[6].typ != TERM_PREFIX || !terms[6].inv ||
+		terms[7].typ != TERM_SUFFIX || !terms[7].inv {
+		t.Errorf("%s", terms)
+	}
+	for idx, term := range terms {
+		if len(term.text) != 3 {
+			t.Errorf("%s", term)
+		}
+		if idx > 0 && len(term.origText) != 4+idx/5 {
+			t.Errorf("%s", term)
+		}
+	}
+}
+
+func TestParseTermsExtendedExact(t *testing.T) {
+	terms := parseTerms(MODE_EXTENDED_EXACT,
+		"aaa 'bbb ^ccc ddd$ !eee !'fff !^ggg !hhh$")
+	if len(terms) != 8 ||
+		terms[0].typ != TERM_EXACT || terms[0].inv || len(terms[0].text) != 3 ||
+		terms[1].typ != TERM_EXACT || terms[1].inv || len(terms[1].text) != 4 ||
+		terms[2].typ != TERM_PREFIX || terms[2].inv || len(terms[2].text) != 3 ||
+		terms[3].typ != TERM_SUFFIX || terms[3].inv || len(terms[3].text) != 3 ||
+		terms[4].typ != TERM_EXACT || !terms[4].inv || len(terms[4].text) != 3 ||
+		terms[5].typ != TERM_EXACT || !terms[5].inv || len(terms[5].text) != 4 ||
+		terms[6].typ != TERM_PREFIX || !terms[6].inv || len(terms[6].text) != 3 ||
+		terms[7].typ != TERM_SUFFIX || !terms[7].inv || len(terms[7].text) != 3 {
+		t.Errorf("%s", terms)
+	}
+}
+
+func TestParseTermsEmpty(t *testing.T) {
+	terms := parseTerms(MODE_EXTENDED, "' $ ^ !' !^ !$")
+	if len(terms) != 0 {
+		t.Errorf("%s", terms)
+	}
+}
+
+func TestExact(t *testing.T) {
+	defer clearPatternCache()
+	clearPatternCache()
+	pattern := BuildPattern(MODE_EXTENDED, CASE_SMART,
+		[]Range{}, nil, []rune("'abc"))
+	str := "aabbcc abc"
+	sidx, eidx := ExactMatchNaive(pattern.caseSensitive, &str, pattern.terms[0].text)
+	if sidx != 7 || eidx != 10 {
+		t.Errorf("%s / %d / %d", pattern.terms, sidx, eidx)
+	}
+}
+
+func TestCaseSensitivity(t *testing.T) {
+	defer clearPatternCache()
+	clearPatternCache()
+	pat1 := BuildPattern(MODE_FUZZY, CASE_SMART, []Range{}, nil, []rune("abc"))
+	clearPatternCache()
+	pat2 := BuildPattern(MODE_FUZZY, CASE_SMART, []Range{}, nil, []rune("Abc"))
+	clearPatternCache()
+	pat3 := BuildPattern(MODE_FUZZY, CASE_IGNORE, []Range{}, nil, []rune("abc"))
+	clearPatternCache()
+	pat4 := BuildPattern(MODE_FUZZY, CASE_IGNORE, []Range{}, nil, []rune("Abc"))
+	clearPatternCache()
+	pat5 := BuildPattern(MODE_FUZZY, CASE_RESPECT, []Range{}, nil, []rune("abc"))
+	clearPatternCache()
+	pat6 := BuildPattern(MODE_FUZZY, CASE_RESPECT, []Range{}, nil, []rune("Abc"))
+
+	if string(pat1.text) != "abc" || pat1.caseSensitive != false ||
+		string(pat2.text) != "Abc" || pat2.caseSensitive != true ||
+		string(pat3.text) != "abc" || pat3.caseSensitive != false ||
+		string(pat4.text) != "abc" || pat4.caseSensitive != false ||
+		string(pat5.text) != "abc" || pat5.caseSensitive != true ||
+		string(pat6.text) != "Abc" || pat6.caseSensitive != true {
+		t.Error("Invalid case conversion")
+	}
+}
diff --git a/src/reader.go b/src/reader.go
new file mode 100644
index 00000000..0e1f0a9c
--- /dev/null
+++ b/src/reader.go
@@ -0,0 +1,60 @@
+package fzf
+
+// #include <unistd.h>
+import "C"
+
+import (
+	"bufio"
+	"fmt"
+	"io"
+	"os"
+	"os/exec"
+)
+
+const DEFAULT_COMMAND = "find * -path '*/\\.*' -prune -o -type f -print -o -type l -print 2> /dev/null"
+
+type Reader struct {
+	pusher   func(string)
+	eventBox *EventBox
+}
+
+func (r *Reader) ReadSource() {
+	if int(C.isatty(C.int(os.Stdin.Fd()))) != 0 {
+		cmd := os.Getenv("FZF_DEFAULT_COMMAND")
+		if len(cmd) == 0 {
+			cmd = DEFAULT_COMMAND
+		}
+		r.readFromCommand(cmd)
+	} else {
+		r.readFromStdin()
+	}
+	r.eventBox.Set(EVT_READ_FIN, nil)
+}
+
+func (r *Reader) feed(src io.Reader) {
+	if scanner := bufio.NewScanner(src); scanner != nil {
+		for scanner.Scan() {
+			r.pusher(scanner.Text())
+			r.eventBox.Set(EVT_READ_NEW, nil)
+		}
+	}
+}
+
+func (r *Reader) readFromStdin() {
+	r.feed(os.Stdin)
+}
+
+func (r *Reader) readFromCommand(cmd string) {
+	arg := fmt.Sprintf("%q", cmd)
+	listCommand := exec.Command("sh", "-c", arg[1:len(arg)-1])
+	out, err := listCommand.StdoutPipe()
+	if err != nil {
+		return
+	}
+	err = listCommand.Start()
+	if err != nil {
+		return
+	}
+	defer listCommand.Wait()
+	r.feed(out)
+}
diff --git a/src/reader_test.go b/src/reader_test.go
new file mode 100644
index 00000000..f51ccab0
--- /dev/null
+++ b/src/reader_test.go
@@ -0,0 +1,52 @@
+package fzf
+
+import "testing"
+
+func TestReadFromCommand(t *testing.T) {
+	strs := []string{}
+	eb := NewEventBox()
+	reader := Reader{
+		pusher:   func(s string) { strs = append(strs, s) },
+		eventBox: eb}
+
+	// Check EventBox
+	if eb.Peak(EVT_READ_NEW) {
+		t.Error("EVT_READ_NEW should not be set yet")
+	}
+
+	// Normal command
+	reader.readFromCommand(`echo abc && echo def`)
+	if len(strs) != 2 || strs[0] != "abc" || strs[1] != "def" {
+		t.Errorf("%s", strs)
+	}
+
+	// Check EventBox again
+	if !eb.Peak(EVT_READ_NEW) {
+		t.Error("EVT_READ_NEW should be set yet")
+	}
+
+	// Wait should return immediately
+	eb.Wait(func(events *Events) {
+		if _, found := (*events)[EVT_READ_NEW]; !found {
+			t.Errorf("%s", events)
+		}
+		events.Clear()
+	})
+
+	// EventBox is cleared
+	if eb.Peak(EVT_READ_NEW) {
+		t.Error("EVT_READ_NEW should not be set yet")
+	}
+
+	// Failing command
+	reader.readFromCommand(`no-such-command`)
+	strs = []string{}
+	if len(strs) > 0 {
+		t.Errorf("%s", strs)
+	}
+
+	// Check EventBox again
+	if eb.Peak(EVT_READ_NEW) {
+		t.Error("Command failed. EVT_READ_NEW should be set")
+	}
+}
diff --git a/src/terminal.go b/src/terminal.go
new file mode 100644
index 00000000..b6c7154f
--- /dev/null
+++ b/src/terminal.go
@@ -0,0 +1,580 @@
+package fzf
+
+import (
+	"fmt"
+	C "github.com/junegunn/fzf/src/curses"
+	"github.com/junegunn/go-runewidth"
+	"os"
+	"regexp"
+	"sort"
+	"sync"
+	"time"
+)
+
+type Terminal struct {
+	prompt     string
+	reverse    bool
+	tac        bool
+	cx         int
+	cy         int
+	offset     int
+	yanked     []rune
+	input      []rune
+	multi      bool
+	printQuery bool
+	count      int
+	progress   int
+	reading    bool
+	list       []*Item
+	selected   map[*string]*string
+	reqBox     *EventBox
+	eventBox   *EventBox
+	mutex      sync.Mutex
+	initFunc   func()
+}
+
+var _spinner []string = []string{`-`, `\`, `|`, `/`, `-`, `\`, `|`, `/`}
+
+const (
+	REQ_PROMPT EventType = iota
+	REQ_INFO
+	REQ_LIST
+	REQ_REDRAW
+	REQ_CLOSE
+	REQ_QUIT
+)
+
+func NewTerminal(opts *Options, eventBox *EventBox) *Terminal {
+	input := []rune(opts.Query)
+	return &Terminal{
+		prompt:     opts.Prompt,
+		tac:        opts.Sort == 0,
+		reverse:    opts.Reverse,
+		cx:         displayWidth(input),
+		cy:         0,
+		offset:     0,
+		yanked:     []rune{},
+		input:      input,
+		multi:      opts.Multi,
+		printQuery: opts.PrintQuery,
+		list:       []*Item{},
+		selected:   make(map[*string]*string),
+		reqBox:     NewEventBox(),
+		eventBox:   eventBox,
+		mutex:      sync.Mutex{},
+		initFunc: func() {
+			C.Init(opts.Color, opts.Color256, opts.Black, opts.Mouse)
+		}}
+}
+
+func (t *Terminal) Input() []rune {
+	t.mutex.Lock()
+	defer t.mutex.Unlock()
+	return copySlice(t.input)
+}
+
+func (t *Terminal) UpdateCount(cnt int, final bool) {
+	t.mutex.Lock()
+	t.count = cnt
+	t.reading = !final
+	t.mutex.Unlock()
+	t.reqBox.Set(REQ_INFO, nil)
+}
+
+func (t *Terminal) UpdateProgress(progress float32) {
+	t.mutex.Lock()
+	t.progress = int(progress * 100)
+	t.mutex.Unlock()
+	t.reqBox.Set(REQ_INFO, nil)
+}
+
+func (t *Terminal) UpdateList(list []*Item) {
+	t.mutex.Lock()
+	t.progress = 100
+	t.list = list
+	t.mutex.Unlock()
+	t.reqBox.Set(REQ_INFO, nil)
+	t.reqBox.Set(REQ_LIST, nil)
+}
+
+func (t *Terminal) listIndex(y int) int {
+	if t.tac {
+		return len(t.list) - y - 1
+	} else {
+		return y
+	}
+}
+
+func (t *Terminal) output() {
+	if t.printQuery {
+		fmt.Println(string(t.input))
+	}
+	if len(t.selected) == 0 {
+		if len(t.list) > t.cy {
+			t.list[t.listIndex(t.cy)].Print()
+		}
+	} else {
+		for ptr, orig := range t.selected {
+			if orig != nil {
+				fmt.Println(*orig)
+			} else {
+				fmt.Println(*ptr)
+			}
+		}
+	}
+}
+
+func displayWidth(runes []rune) int {
+	l := 0
+	for _, r := range runes {
+		l += runewidth.RuneWidth(r)
+	}
+	return l
+}
+
+func (t *Terminal) move(y int, x int, clear bool) {
+	maxy := C.MaxY()
+	if !t.reverse {
+		y = maxy - y - 1
+	}
+
+	if clear {
+		C.MoveAndClear(y, x)
+	} else {
+		C.Move(y, x)
+	}
+}
+
+func (t *Terminal) placeCursor() {
+	t.move(0, len(t.prompt)+displayWidth(t.input[:t.cx]), false)
+}
+
+func (t *Terminal) printPrompt() {
+	t.move(0, 0, true)
+	C.CPrint(C.COL_PROMPT, true, t.prompt)
+	C.CPrint(C.COL_NORMAL, true, string(t.input))
+}
+
+func (t *Terminal) printInfo() {
+	t.move(1, 0, true)
+	if t.reading {
+		duration := int64(200) * int64(time.Millisecond)
+		idx := (time.Now().UnixNano() % (duration * int64(len(_spinner)))) / duration
+		C.CPrint(C.COL_SPINNER, true, _spinner[idx])
+	}
+
+	t.move(1, 2, false)
+	output := fmt.Sprintf("%d/%d", len(t.list), t.count)
+	if t.multi && len(t.selected) > 0 {
+		output += fmt.Sprintf(" (%d)", len(t.selected))
+	}
+	if t.progress > 0 && t.progress < 100 {
+		output += fmt.Sprintf(" (%d%%)", t.progress)
+	}
+	C.CPrint(C.COL_INFO, false, output)
+}
+
+func (t *Terminal) printList() {
+	t.constrain()
+
+	maxy := maxItems()
+	count := len(t.list) - t.offset
+	for i := 0; i < maxy; i++ {
+		t.move(i+2, 0, true)
+		if i < count {
+			t.printItem(t.list[t.listIndex(i+t.offset)], i == t.cy-t.offset)
+		}
+	}
+}
+
+func (t *Terminal) printItem(item *Item, current bool) {
+	_, selected := t.selected[item.text]
+	if current {
+		C.CPrint(C.COL_CURSOR, true, ">")
+		if selected {
+			C.CPrint(C.COL_CURRENT, true, ">")
+		} else {
+			C.CPrint(C.COL_CURRENT, true, " ")
+		}
+		t.printHighlighted(item, true, C.COL_CURRENT, C.COL_CURRENT_MATCH)
+	} else {
+		C.CPrint(C.COL_CURSOR, true, " ")
+		if selected {
+			C.CPrint(C.COL_SELECTED, true, ">")
+		} else {
+			C.Print(" ")
+		}
+		t.printHighlighted(item, false, 0, C.COL_MATCH)
+	}
+}
+
+func trimRight(runes []rune, width int) ([]rune, int) {
+	currentWidth := displayWidth(runes)
+	trimmed := 0
+
+	for currentWidth > width && len(runes) > 0 {
+		sz := len(runes)
+		currentWidth -= runewidth.RuneWidth(runes[sz-1])
+		runes = runes[:sz-1]
+		trimmed += 1
+	}
+	return runes, trimmed
+}
+
+func trimLeft(runes []rune, width int) ([]rune, int) {
+	currentWidth := displayWidth(runes)
+	trimmed := 0
+
+	for currentWidth > width && len(runes) > 0 {
+		currentWidth -= runewidth.RuneWidth(runes[0])
+		runes = runes[1:]
+		trimmed += 1
+	}
+	return runes, trimmed
+}
+
+func (*Terminal) printHighlighted(item *Item, bold bool, col1 int, col2 int) {
+	maxe := 0
+	for _, offset := range item.offsets {
+		if offset[1] > maxe {
+			maxe = offset[1]
+		}
+	}
+
+	// Overflow
+	text := []rune(*item.text)
+	offsets := item.offsets
+	maxWidth := C.MaxX() - 3
+	fullWidth := displayWidth(text)
+	if fullWidth > maxWidth {
+		// Stri..
+		matchEndWidth := displayWidth(text[:maxe])
+		if matchEndWidth <= maxWidth-2 {
+			text, _ = trimRight(text, maxWidth-2)
+			text = append(text, []rune("..")...)
+		} else {
+			// Stri..
+			if matchEndWidth < fullWidth-2 {
+				text = append(text[:maxe], []rune("..")...)
+			}
+			// ..ri..
+			var diff int
+			text, diff = trimLeft(text, maxWidth-2)
+
+			// Transform offsets
+			offsets = make([]Offset, len(item.offsets))
+			for idx, offset := range item.offsets {
+				b, e := offset[0], offset[1]
+				b += 2 - diff
+				e += 2 - diff
+				b = Max(b, 2)
+				if b < e {
+					offsets[idx] = Offset{b, e}
+				}
+			}
+			text = append([]rune(".."), text...)
+		}
+	}
+
+	sort.Sort(ByOrder(offsets))
+	index := 0
+	for _, offset := range offsets {
+		b := Max(index, offset[0])
+		e := Max(index, offset[1])
+		C.CPrint(col1, bold, string(text[index:b]))
+		C.CPrint(col2, bold, string(text[b:e]))
+		index = e
+	}
+	if index < len(text) {
+		C.CPrint(col1, bold, string(text[index:]))
+	}
+}
+
+func (t *Terminal) printAll() {
+	t.printList()
+	t.printInfo()
+	t.printPrompt()
+}
+
+func (t *Terminal) refresh() {
+	t.placeCursor()
+	C.Refresh()
+}
+
+func (t *Terminal) delChar() bool {
+	if len(t.input) > 0 && t.cx < len(t.input) {
+		t.input = append(t.input[:t.cx], t.input[t.cx+1:]...)
+		return true
+	}
+	return false
+}
+
+func findLastMatch(pattern string, str string) int {
+	rx, err := regexp.Compile(pattern)
+	if err != nil {
+		return -1
+	}
+	locs := rx.FindAllStringIndex(str, -1)
+	if locs == nil {
+		return -1
+	}
+	return locs[len(locs)-1][0]
+}
+
+func findFirstMatch(pattern string, str string) int {
+	rx, err := regexp.Compile(pattern)
+	if err != nil {
+		return -1
+	}
+	loc := rx.FindStringIndex(str)
+	if loc == nil {
+		return -1
+	}
+	return loc[0]
+}
+
+func copySlice(slice []rune) []rune {
+	ret := make([]rune, len(slice))
+	copy(ret, slice)
+	return ret
+}
+
+func (t *Terminal) rubout(pattern string) {
+	pcx := t.cx
+	after := t.input[t.cx:]
+	t.cx = findLastMatch(pattern, string(t.input[:t.cx])) + 1
+	t.yanked = copySlice(t.input[t.cx:pcx])
+	t.input = append(t.input[:t.cx], after...)
+}
+
+func (t *Terminal) Loop() {
+	{ // Late initialization
+		t.mutex.Lock()
+		t.initFunc()
+		t.printInfo()
+		t.printPrompt()
+		t.refresh()
+		t.mutex.Unlock()
+	}
+
+	go func() {
+		for {
+			t.reqBox.Wait(func(events *Events) {
+				defer events.Clear()
+				t.mutex.Lock()
+				for req := range *events {
+					switch req {
+					case REQ_PROMPT:
+						t.printPrompt()
+					case REQ_INFO:
+						t.printInfo()
+					case REQ_LIST:
+						t.printList()
+					case REQ_REDRAW:
+						C.Clear()
+						t.printAll()
+					case REQ_CLOSE:
+						C.Close()
+						t.output()
+						os.Exit(0)
+					case REQ_QUIT:
+						C.Close()
+						os.Exit(1)
+					}
+				}
+				t.mutex.Unlock()
+			})
+			t.refresh()
+		}
+	}()
+
+	looping := true
+	for looping {
+		event := C.GetChar()
+
+		t.mutex.Lock()
+		previousInput := t.input
+		events := []EventType{REQ_PROMPT}
+		toggle := func() {
+			item := t.list[t.listIndex(t.cy)]
+			if _, found := t.selected[item.text]; !found {
+				t.selected[item.text] = item.origText
+			} else {
+				delete(t.selected, item.text)
+			}
+		}
+		req := func(evts ...EventType) {
+			for _, event := range evts {
+				events = append(events, event)
+				if event == REQ_CLOSE || event == REQ_QUIT {
+					looping = false
+				}
+			}
+		}
+		switch event.Type {
+		case C.INVALID:
+			continue
+		case C.CTRL_A:
+			t.cx = 0
+		case C.CTRL_B:
+			if t.cx > 0 {
+				t.cx -= 1
+			}
+		case C.CTRL_C, C.CTRL_G, C.CTRL_Q, C.ESC:
+			req(REQ_QUIT)
+		case C.CTRL_D:
+			if !t.delChar() && t.cx == 0 {
+				req(REQ_QUIT)
+			}
+		case C.CTRL_E:
+			t.cx = len(t.input)
+		case C.CTRL_F:
+			if t.cx < len(t.input) {
+				t.cx += 1
+			}
+		case C.CTRL_H:
+			if t.cx > 0 {
+				t.input = append(t.input[:t.cx-1], t.input[t.cx:]...)
+				t.cx -= 1
+			}
+		case C.TAB:
+			if t.multi && len(t.list) > 0 {
+				toggle()
+				t.vmove(-1)
+				req(REQ_LIST, REQ_INFO)
+			}
+		case C.BTAB:
+			if t.multi && len(t.list) > 0 {
+				toggle()
+				t.vmove(1)
+				req(REQ_LIST, REQ_INFO)
+			}
+		case C.CTRL_J, C.CTRL_N:
+			t.vmove(-1)
+			req(REQ_LIST)
+		case C.CTRL_K, C.CTRL_P:
+			t.vmove(1)
+			req(REQ_LIST)
+		case C.CTRL_M:
+			req(REQ_CLOSE)
+		case C.CTRL_L:
+			req(REQ_REDRAW)
+		case C.CTRL_U:
+			if t.cx > 0 {
+				t.yanked = copySlice(t.input[:t.cx])
+				t.input = t.input[t.cx:]
+				t.cx = 0
+			}
+		case C.CTRL_W:
+			if t.cx > 0 {
+				t.rubout("\\s\\S")
+			}
+		case C.ALT_BS:
+			if t.cx > 0 {
+				t.rubout("[^[:alnum:]][[:alnum:]]")
+			}
+		case C.CTRL_Y:
+			t.input = append(append(t.input[:t.cx], t.yanked...), t.input[t.cx:]...)
+			t.cx += len(t.yanked)
+		case C.DEL:
+			t.delChar()
+		case C.PGUP:
+			t.vmove(maxItems() - 1)
+			req(REQ_LIST)
+		case C.PGDN:
+			t.vmove(-(maxItems() - 1))
+			req(REQ_LIST)
+		case C.ALT_B:
+			t.cx = findLastMatch("[^[:alnum:]][[:alnum:]]", string(t.input[:t.cx])) + 1
+		case C.ALT_F:
+			t.cx += findFirstMatch("[[:alnum:]][^[:alnum:]]|(.$)", string(t.input[t.cx:])) + 1
+		case C.ALT_D:
+			ncx := t.cx +
+				findFirstMatch("[[:alnum:]][^[:alnum:]]|(.$)", string(t.input[t.cx:])) + 1
+			if ncx > t.cx {
+				t.yanked = copySlice(t.input[t.cx:ncx])
+				t.input = append(t.input[:t.cx], t.input[ncx:]...)
+			}
+		case C.RUNE:
+			prefix := copySlice(t.input[:t.cx])
+			t.input = append(append(prefix, event.Char), t.input[t.cx:]...)
+			t.cx += 1
+		case C.MOUSE:
+			me := event.MouseEvent
+			mx, my := Min(len(t.input), Max(0, me.X-len(t.prompt))), me.Y
+			if !t.reverse {
+				my = C.MaxY() - my - 1
+			}
+			if me.S != 0 {
+				// Scroll
+				if me.Mod {
+					toggle()
+				}
+				t.vmove(me.S)
+				req(REQ_LIST)
+			} else if me.Double {
+				// Double-click
+				if my >= 2 {
+					t.cy = my - 2
+					req(REQ_CLOSE)
+				}
+			} else if me.Down {
+				if my == 0 && mx >= 0 {
+					// Prompt
+					t.cx = mx
+					req(REQ_PROMPT)
+				} else if my >= 2 {
+					// List
+					t.cy = my - 2
+					if me.Mod {
+						toggle()
+					}
+					req(REQ_LIST)
+				}
+			}
+		}
+		changed := string(previousInput) != string(t.input)
+		t.mutex.Unlock() // Must be unlocked before touching reqBox
+
+		if changed {
+			t.eventBox.Set(EVT_SEARCH_NEW, nil)
+		}
+		for _, event := range events {
+			t.reqBox.Set(event, nil)
+		}
+	}
+}
+
+func (t *Terminal) constrain() {
+	count := len(t.list)
+	height := C.MaxY() - 2
+	diffpos := t.cy - t.offset
+
+	t.cy = Max(0, Min(t.cy, count-1))
+
+	if t.cy > t.offset+(height-1) {
+		// Ceil
+		t.offset = t.cy - (height - 1)
+	} else if t.offset > t.cy {
+		// Floor
+		t.offset = t.cy
+	}
+
+	// Adjustment
+	if count-t.offset < height {
+		t.offset = Max(0, count-height)
+		t.cy = Max(0, Min(t.offset+diffpos, count-1))
+	}
+}
+
+func (t *Terminal) vmove(o int) {
+	if t.reverse {
+		t.cy -= o
+	} else {
+		t.cy += o
+	}
+}
+
+func maxItems() int {
+	return C.MaxY() - 2
+}
diff --git a/src/tokenizer.go b/src/tokenizer.go
new file mode 100644
index 00000000..c187529b
--- /dev/null
+++ b/src/tokenizer.go
@@ -0,0 +1,194 @@
+package fzf
+
+import (
+	"regexp"
+	"strconv"
+	"strings"
+)
+
+const RANGE_ELLIPSIS = 0
+
+type Range struct {
+	begin int
+	end   int
+}
+
+type Transformed struct {
+	whole *string
+	parts []Token
+}
+
+type Token struct {
+	text         *string
+	prefixLength int
+}
+
+func ParseRange(str *string) (Range, bool) {
+	if (*str) == ".." {
+		return Range{RANGE_ELLIPSIS, RANGE_ELLIPSIS}, true
+	} else if strings.HasPrefix(*str, "..") {
+		end, err := strconv.Atoi((*str)[2:])
+		if err != nil || end == 0 {
+			return Range{}, false
+		} else {
+			return Range{RANGE_ELLIPSIS, end}, true
+		}
+	} else if strings.HasSuffix(*str, "..") {
+		begin, err := strconv.Atoi((*str)[:len(*str)-2])
+		if err != nil || begin == 0 {
+			return Range{}, false
+		} else {
+			return Range{begin, RANGE_ELLIPSIS}, true
+		}
+	} else if strings.Contains(*str, "..") {
+		ns := strings.Split(*str, "..")
+		if len(ns) != 2 {
+			return Range{}, false
+		}
+		begin, err1 := strconv.Atoi(ns[0])
+		end, err2 := strconv.Atoi(ns[1])
+		if err1 != nil || err2 != nil {
+			return Range{}, false
+		}
+		return Range{begin, end}, true
+	}
+
+	n, err := strconv.Atoi(*str)
+	if err != nil || n == 0 {
+		return Range{}, false
+	}
+	return Range{n, n}, true
+}
+
+func withPrefixLengths(tokens []string, begin int) []Token {
+	ret := make([]Token, len(tokens))
+
+	prefixLength := begin
+	for idx, token := range tokens {
+		// Need to define a new local variable instead of the reused token to take
+		// the pointer to it
+		str := token
+		ret[idx] = Token{text: &str, prefixLength: prefixLength}
+		prefixLength += len([]rune(token))
+	}
+	return ret
+}
+
+const (
+	AWK_NIL = iota
+	AWK_BLACK
+	AWK_WHITE
+)
+
+func awkTokenizer(input *string) ([]string, int) {
+	// 9, 32
+	ret := []string{}
+	str := []rune{}
+	prefixLength := 0
+	state := AWK_NIL
+	for _, r := range []rune(*input) {
+		white := r == 9 || r == 32
+		switch state {
+		case AWK_NIL:
+			if white {
+				prefixLength++
+			} else {
+				state = AWK_BLACK
+				str = append(str, r)
+			}
+		case AWK_BLACK:
+			str = append(str, r)
+			if white {
+				state = AWK_WHITE
+			}
+		case AWK_WHITE:
+			if white {
+				str = append(str, r)
+			} else {
+				ret = append(ret, string(str))
+				state = AWK_BLACK
+				str = []rune{r}
+			}
+		}
+	}
+	if len(str) > 0 {
+		ret = append(ret, string(str))
+	}
+	return ret, prefixLength
+}
+
+func Tokenize(str *string, delimiter *regexp.Regexp) []Token {
+	prefixLength := 0
+	if delimiter == nil {
+		// AWK-style (\S+\s*)
+		tokens, prefixLength := awkTokenizer(str)
+		return withPrefixLengths(tokens, prefixLength)
+	} else {
+		tokens := delimiter.FindAllString(*str, -1)
+		return withPrefixLengths(tokens, prefixLength)
+	}
+}
+
+func joinTokens(tokens []Token) string {
+	ret := ""
+	for _, token := range tokens {
+		ret += *token.text
+	}
+	return ret
+}
+
+func Transform(tokens []Token, withNth []Range) *Transformed {
+	transTokens := make([]Token, len(withNth))
+	numTokens := len(tokens)
+	whole := ""
+	for idx, r := range withNth {
+		part := ""
+		minIdx := 0
+		if r.begin == r.end {
+			idx := r.begin
+			if idx == RANGE_ELLIPSIS {
+				part += joinTokens(tokens)
+			} else {
+				if idx < 0 {
+					idx += numTokens + 1
+				}
+				if idx >= 1 && idx <= numTokens {
+					minIdx = idx - 1
+					part += *tokens[idx-1].text
+				}
+			}
+		} else {
+			var begin, end int
+			if r.begin == RANGE_ELLIPSIS { // ..N
+				begin, end = 1, r.end
+				if end < 0 {
+					end += numTokens + 1
+				}
+			} else if r.end == RANGE_ELLIPSIS { // N..
+				begin, end = r.begin, numTokens
+				if begin < 0 {
+					begin += numTokens + 1
+				}
+			} else {
+				begin, end = r.begin, r.end
+				if begin < 0 {
+					begin += numTokens + 1
+				}
+				if end < 0 {
+					end += numTokens + 1
+				}
+			}
+			minIdx = Max(0, begin-1)
+			for idx := begin; idx <= end; idx++ {
+				if idx >= 1 && idx <= numTokens {
+					part += *tokens[idx-1].text
+				}
+			}
+		}
+		whole += part
+		transTokens[idx] = Token{&part, tokens[minIdx].prefixLength}
+	}
+	return &Transformed{
+		whole: &whole,
+		parts: transTokens}
+}
diff --git a/src/tokenizer_test.go b/src/tokenizer_test.go
new file mode 100644
index 00000000..ed77efe9
--- /dev/null
+++ b/src/tokenizer_test.go
@@ -0,0 +1,97 @@
+package fzf
+
+import "testing"
+
+func TestParseRange(t *testing.T) {
+	{
+		i := ".."
+		r, _ := ParseRange(&i)
+		if r.begin != RANGE_ELLIPSIS || r.end != RANGE_ELLIPSIS {
+			t.Errorf("%s", r)
+		}
+	}
+	{
+		i := "3.."
+		r, _ := ParseRange(&i)
+		if r.begin != 3 || r.end != RANGE_ELLIPSIS {
+			t.Errorf("%s", r)
+		}
+	}
+	{
+		i := "3..5"
+		r, _ := ParseRange(&i)
+		if r.begin != 3 || r.end != 5 {
+			t.Errorf("%s", r)
+		}
+	}
+	{
+		i := "-3..-5"
+		r, _ := ParseRange(&i)
+		if r.begin != -3 || r.end != -5 {
+			t.Errorf("%s", r)
+		}
+	}
+	{
+		i := "3"
+		r, _ := ParseRange(&i)
+		if r.begin != 3 || r.end != 3 {
+			t.Errorf("%s", r)
+		}
+	}
+}
+
+func TestTokenize(t *testing.T) {
+	// AWK-style
+	input := "  abc:  def:  ghi  "
+	tokens := Tokenize(&input, nil)
+	if *tokens[0].text != "abc:  " || tokens[0].prefixLength != 2 {
+		t.Errorf("%s", tokens)
+	}
+
+	// With delimiter
+	tokens = Tokenize(&input, delimiterRegexp(":"))
+	if *tokens[0].text != "  abc:" || tokens[0].prefixLength != 0 {
+		t.Errorf("%s", tokens)
+	}
+}
+
+func TestTransform(t *testing.T) {
+	input := "  abc:  def:  ghi:  jkl"
+	{
+		tokens := Tokenize(&input, nil)
+		{
+			ranges := splitNth("1,2,3")
+			tx := Transform(tokens, ranges)
+			if *tx.whole != "abc:  def:  ghi:  " {
+				t.Errorf("%s", *tx)
+			}
+		}
+		{
+			ranges := splitNth("1..2,3,2..,1")
+			tx := Transform(tokens, ranges)
+			if *tx.whole != "abc:  def:  ghi:  def:  ghi:  jklabc:  " ||
+				len(tx.parts) != 4 ||
+				*tx.parts[0].text != "abc:  def:  " || tx.parts[0].prefixLength != 2 ||
+				*tx.parts[1].text != "ghi:  " || tx.parts[1].prefixLength != 14 ||
+				*tx.parts[2].text != "def:  ghi:  jkl" || tx.parts[2].prefixLength != 8 ||
+				*tx.parts[3].text != "abc:  " || tx.parts[3].prefixLength != 2 {
+				t.Errorf("%s", *tx)
+			}
+		}
+	}
+	{
+		tokens := Tokenize(&input, delimiterRegexp(":"))
+		{
+			ranges := splitNth("1..2,3,2..,1")
+			tx := Transform(tokens, ranges)
+			if *tx.whole != "  abc:  def:  ghi:  def:  ghi:  jkl  abc:" ||
+				len(tx.parts) != 4 ||
+				*tx.parts[0].text != "  abc:  def:" || tx.parts[0].prefixLength != 0 ||
+				*tx.parts[1].text != "  ghi:" || tx.parts[1].prefixLength != 12 ||
+				*tx.parts[2].text != "  def:  ghi:  jkl" || tx.parts[2].prefixLength != 6 ||
+				*tx.parts[3].text != "  abc:" || tx.parts[3].prefixLength != 0 {
+				t.Errorf("%s", *tx)
+			}
+		}
+	}
+}
diff --git a/src/util.go b/src/util.go
new file mode 100644
index 00000000..2144e54e
--- /dev/null
+++ b/src/util.go
@@ -0,0 +1,21 @@
+package fzf
+
+func Max(first int, items ...int) int {
+	max := first
+	for _, item := range items {
+		if item > max {
+			max = item
+		}
+	}
+	return max
+}
+
+func Min(first int, items ...int) int {
+	min := first
+	for _, item := range items {
+		if item < min {
+			min = item
+		}
+	}
+	return min
+}
diff --git a/src/util_test.go b/src/util_test.go
new file mode 100644
index 00000000..814b42ce
--- /dev/null
+++ b/src/util_test.go
@@ -0,0 +1,18 @@
+package fzf
+
+import "testing"
+
+func TestMax(t *testing.T) {
+	if Max(-2, 5, 1, 4, 3) != 5 {
+		t.Error("Invalid result")
+	}
+}
+
+func TestMin(t *testing.T) {
+	if Min(2, -3) != -3 {
+		t.Error("Invalid result")
+	}
+	if Min(-2, 5, 1, 4, 3) != -2 {
+		t.Error("Invalid result")
+	}
+}