History and Version Control for Dashboard Updates

A simple version control system for dashboards. Closes #1504. Goals 1. To create a new dashboard version every time a dashboard is saved. 2. To allow users to view all versions of a given dashboard. 3. To allow users to rollback to a previous version of a dashboard. 4. To allow users to compare two versions of a dashboard. Usage Navigate to a dashboard, and click the settings cog. From there, click the "Changelog" button to be brought to the Changelog view. In this view, a table containing each version of a dashboard can be seen. Each entry in the table represents a dashboard version. A selectable checkbox, the version number, date created, name of the user who created that version, and commit message is shown in the table, along with a button that allows a user to restore to a previous version of that dashboard. If a user wants to restore to a previous version of their dashboard, they can do so by clicking the previously mentioned button. If a user wants to compare two different versions of a dashboard, they can do so by clicking the checkbox of two different dashboard versions, then clicking the "Compare versions" button located below the dashboard. From there, the user is brought to a view showing a summary of the dashboard differences. Each summarized change contains a link that can be clicked to take the user a JSON diff highlighting the changes line by line. Overview of Changes Backend Changes - A `dashboard_version` table was created to store each dashboard version, along with a dashboard version model and structs to represent the queries and commands necessary for the dashboard version API methods. - API endpoints were created to support working with dashboard versions. - Methods were added to create, update, read, and destroy dashboard versions in the database. - Logic was added to compute the diff between two versions, and display it to the user. - The dashboard migration logic was updated to save a "Version 1" of each existing dashboard in the database. Frontend Changes - New views - Methods to pull JSON and HTML from endpoints New API Endpoints Each endpoint requires the authorization header to be sent in the format, ``` Authorization: Bearer <jwt> ``` where `<jwt>` is a JSON web token obtained from the Grafana admin panel. `GET "/api/dashboards/db/:dashboardId/versions?orderBy=<string>&limit=<int>&start=<int>"` Get all dashboard versions for the given dashboard ID. Accepts three URL parameters: - `orderBy` String to order the results by. Possible values are `version`, `created`, `created_by`, `message`. Default is `versions`. Ordering is always in descending order. - `limit` Maximum number of results to return - `start` Position in results to start from `GET "/api/dashboards/db/:dashboardId/versions/:id"` Get an individual dashboard version by ID, for the given dashboard ID. `POST "/api/dashboards/db/:dashboardId/restore"` Restore to the given dashboard version. Post body is of content-type `application/json`, and must contain. ```json { "dashboardId": <int>, "version": <int> } ``` `GET "/api/dashboards/db/:dashboardId/compare/:versionA...:versionB"` Compare two dashboard versions by ID for the given dashboard ID, returning a JSON delta formatted representation of the diff. The URL format follows what GitHub does. For example, visiting [/api/dashboards/db/18/compare/22...33](http://ec2-54-80-139-44.compute-1.amazonaws.com:3000/api/dashboards/db/18/compare/22...33) will return the diff between versions 22 and 33 for the dashboard ID 18. Dependencies Added - The Go package [gojsondiff](https://github.com/yudai/gojsondiff) was added and vendored.
2017-05-24 19:14:39 -04:00
parent 59f3cca135
commit b6e46c9eb8
60 changed files with 7843 additions and 84 deletions
@@ -0,0 +1,20 @@
+Copyright (c) 2012-2016 The go-diff Authors. All rights reserved.
+
+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.
+
@@ -0,0 +1,46 @@
+// Copyright (c) 2012-2016 The go-diff authors. All rights reserved.
+// https://github.com/sergi/go-diff
+// See the included LICENSE file for license details.
+//
+// go-diff is a Go implementation of Google's Diff, Match, and Patch library
+// Original library is Copyright (c) 2006 Google Inc.
+// http://code.google.com/p/google-diff-match-patch/
+
+// Package diffmatchpatch offers robust algorithms to perform the operations required for synchronizing plain text.
+package diffmatchpatch
+
+import (
+	"time"
+)
+
+// DiffMatchPatch holds the configuration for diff-match-patch operations.
+type DiffMatchPatch struct {
+	// Number of seconds to map a diff before giving up (0 for infinity).
+	DiffTimeout time.Duration
+	// Cost of an empty edit operation in terms of edit characters.
+	DiffEditCost int
+	// How far to search for a match (0 = exact location, 1000+ = broad match). A match this many characters away from the expected location will add 1.0 to the score (0.0 is a perfect match).
+	MatchDistance int
+	// When deleting a large block of text (over ~64 characters), how close do the contents have to be to match the expected contents. (0.0 = perfection, 1.0 = very loose).  Note that MatchThreshold controls how closely the end points of a delete need to match.
+	PatchDeleteThreshold float64
+	// Chunk size for context length.
+	PatchMargin int
+	// The number of bits in an int.
+	MatchMaxBits int
+	// At what point is no match declared (0.0 = perfection, 1.0 = very loose).
+	MatchThreshold float64
+}
+
+// New creates a new DiffMatchPatch object with default parameters.
+func New() *DiffMatchPatch {
+	// Defaults.
+	return &DiffMatchPatch{
+		DiffTimeout:          time.Second,
+		DiffEditCost:         4,
+		MatchThreshold:       0.5,
+		MatchDistance:        1000,
+		PatchDeleteThreshold: 0.5,
+		PatchMargin:          4,
+		MatchMaxBits:         32,
+	}
+}
@@ -0,0 +1,160 @@
+// Copyright (c) 2012-2016 The go-diff authors. All rights reserved.
+// https://github.com/sergi/go-diff
+// See the included LICENSE file for license details.
+//
+// go-diff is a Go implementation of Google's Diff, Match, and Patch library
+// Original library is Copyright (c) 2006 Google Inc.
+// http://code.google.com/p/google-diff-match-patch/
+
+package diffmatchpatch
+
+import (
+	"math"
+)
+
+// MatchMain locates the best instance of 'pattern' in 'text' near 'loc'.
+// Returns -1 if no match found.
+func (dmp *DiffMatchPatch) MatchMain(text, pattern string, loc int) int {
+	// Check for null inputs not needed since null can't be passed in C#.
+
+	loc = int(math.Max(0, math.Min(float64(loc), float64(len(text)))))
+	if text == pattern {
+		// Shortcut (potentially not guaranteed by the algorithm)
+		return 0
+	} else if len(text) == 0 {
+		// Nothing to match.
+		return -1
+	} else if loc+len(pattern) <= len(text) && text[loc:loc+len(pattern)] == pattern {
+		// Perfect match at the perfect spot!  (Includes case of null pattern)
+		return loc
+	}
+	// Do a fuzzy compare.
+	return dmp.MatchBitap(text, pattern, loc)
+}
+
+// MatchBitap locates the best instance of 'pattern' in 'text' near 'loc' using the Bitap algorithm.
+// Returns -1 if no match was found.
+func (dmp *DiffMatchPatch) MatchBitap(text, pattern string, loc int) int {
+	// Initialise the alphabet.
+	s := dmp.MatchAlphabet(pattern)
+
+	// Highest score beyond which we give up.
+	scoreThreshold := dmp.MatchThreshold
+	// Is there a nearby exact match? (speedup)
+	bestLoc := indexOf(text, pattern, loc)
+	if bestLoc != -1 {
+		scoreThreshold = math.Min(dmp.matchBitapScore(0, bestLoc, loc,
+			pattern), scoreThreshold)
+		// What about in the other direction? (speedup)
+		bestLoc = lastIndexOf(text, pattern, loc+len(pattern))
+		if bestLoc != -1 {
+			scoreThreshold = math.Min(dmp.matchBitapScore(0, bestLoc, loc,
+				pattern), scoreThreshold)
+		}
+	}
+
+	// Initialise the bit arrays.
+	matchmask := 1 << uint((len(pattern) - 1))
+	bestLoc = -1
+
+	var binMin, binMid int
+	binMax := len(pattern) + len(text)
+	lastRd := []int{}
+	for d := 0; d < len(pattern); d++ {
+		// Scan for the best match; each iteration allows for one more error. Run a binary search to determine how far from 'loc' we can stray at this error level.
+		binMin = 0
+		binMid = binMax
+		for binMin < binMid {
+			if dmp.matchBitapScore(d, loc+binMid, loc, pattern) <= scoreThreshold {
+				binMin = binMid
+			} else {
+				binMax = binMid
+			}
+			binMid = (binMax-binMin)/2 + binMin
+		}
+		// Use the result from this iteration as the maximum for the next.
+		binMax = binMid
+		start := int(math.Max(1, float64(loc-binMid+1)))
+		finish := int(math.Min(float64(loc+binMid), float64(len(text))) + float64(len(pattern)))
+
+		rd := make([]int, finish+2)
+		rd[finish+1] = (1 << uint(d)) - 1
+
+		for j := finish; j >= start; j-- {
+			var charMatch int
+			if len(text) <= j-1 {
+				// Out of range.
+				charMatch = 0
+			} else if _, ok := s[text[j-1]]; !ok {
+				charMatch = 0
+			} else {
+				charMatch = s[text[j-1]]
+			}
+
+			if d == 0 {
+				// First pass: exact match.
+				rd[j] = ((rd[j+1] << 1) | 1) & charMatch
+			} else {
+				// Subsequent passes: fuzzy match.
+				rd[j] = ((rd[j+1]<<1)|1)&charMatch | (((lastRd[j+1] | lastRd[j]) << 1) | 1) | lastRd[j+1]
+			}
+			if (rd[j] & matchmask) != 0 {
+				score := dmp.matchBitapScore(d, j-1, loc, pattern)
+				// This match will almost certainly be better than any existing match.  But check anyway.
+				if score <= scoreThreshold {
+					// Told you so.
+					scoreThreshold = score
+					bestLoc = j - 1
+					if bestLoc > loc {
+						// When passing loc, don't exceed our current distance from loc.
+						start = int(math.Max(1, float64(2*loc-bestLoc)))
+					} else {
+						// Already passed loc, downhill from here on in.
+						break
+					}
+				}
+			}
+		}
+		if dmp.matchBitapScore(d+1, loc, loc, pattern) > scoreThreshold {
+			// No hope for a (better) match at greater error levels.
+			break
+		}
+		lastRd = rd
+	}
+	return bestLoc
+}
+
+// matchBitapScore computes and returns the score for a match with e errors and x location.
+func (dmp *DiffMatchPatch) matchBitapScore(e, x, loc int, pattern string) float64 {
+	accuracy := float64(e) / float64(len(pattern))
+	proximity := math.Abs(float64(loc - x))
+	if dmp.MatchDistance == 0 {
+		// Dodge divide by zero error.
+		if proximity == 0 {
+			return accuracy
+		}
+
+		return 1.0
+	}
+	return accuracy + (proximity / float64(dmp.MatchDistance))
+}
+
+// MatchAlphabet initialises the alphabet for the Bitap algorithm.
+func (dmp *DiffMatchPatch) MatchAlphabet(pattern string) map[byte]int {
+	s := map[byte]int{}
+	charPattern := []byte(pattern)
+	for _, c := range charPattern {
+		_, ok := s[c]
+		if !ok {
+			s[c] = 0
+		}
+	}
+	i := 0
+
+	for _, c := range charPattern {
+		value := s[c] | int(uint(1)<<uint((len(pattern)-i-1)))
+		s[c] = value
+		i++
+	}
+	return s
+}
@@ -0,0 +1,23 @@
+// Copyright (c) 2012-2016 The go-diff authors. All rights reserved.
+// https://github.com/sergi/go-diff
+// See the included LICENSE file for license details.
+//
+// go-diff is a Go implementation of Google's Diff, Match, and Patch library
+// Original library is Copyright (c) 2006 Google Inc.
+// http://code.google.com/p/google-diff-match-patch/
+
+package diffmatchpatch
+
+func min(x, y int) int {
+	if x < y {
+		return x
+	}
+	return y
+}
+
+func max(x, y int) int {
+	if x > y {
+		return x
+	}
+	return y
+}
@@ -0,0 +1,556 @@
+// Copyright (c) 2012-2016 The go-diff authors. All rights reserved.
+// https://github.com/sergi/go-diff
+// See the included LICENSE file for license details.
+//
+// go-diff is a Go implementation of Google's Diff, Match, and Patch library
+// Original library is Copyright (c) 2006 Google Inc.
+// http://code.google.com/p/google-diff-match-patch/
+
+package diffmatchpatch
+
+import (
+	"bytes"
+	"errors"
+	"math"
+	"net/url"
+	"regexp"
+	"strconv"
+	"strings"
+)
+
+// Patch represents one patch operation.
+type Patch struct {
+	diffs   []Diff
+	start1  int
+	start2  int
+	length1 int
+	length2 int
+}
+
+// String emulates GNU diff's format.
+// Header: @@ -382,8 +481,9 @@
+// Indicies are printed as 1-based, not 0-based.
+func (p *Patch) String() string {
+	var coords1, coords2 string
+
+	if p.length1 == 0 {
+		coords1 = strconv.Itoa(p.start1) + ",0"
+	} else if p.length1 == 1 {
+		coords1 = strconv.Itoa(p.start1 + 1)
+	} else {
+		coords1 = strconv.Itoa(p.start1+1) + "," + strconv.Itoa(p.length1)
+	}
+
+	if p.length2 == 0 {
+		coords2 = strconv.Itoa(p.start2) + ",0"
+	} else if p.length2 == 1 {
+		coords2 = strconv.Itoa(p.start2 + 1)
+	} else {
+		coords2 = strconv.Itoa(p.start2+1) + "," + strconv.Itoa(p.length2)
+	}
+
+	var text bytes.Buffer
+	_, _ = text.WriteString("@@ -" + coords1 + " +" + coords2 + " @@\n")
+
+	// Escape the body of the patch with %xx notation.
+	for _, aDiff := range p.diffs {
+		switch aDiff.Type {
+		case DiffInsert:
+			_, _ = text.WriteString("+")
+		case DiffDelete:
+			_, _ = text.WriteString("-")
+		case DiffEqual:
+			_, _ = text.WriteString(" ")
+		}
+
+		_, _ = text.WriteString(strings.Replace(url.QueryEscape(aDiff.Text), "+", " ", -1))
+		_, _ = text.WriteString("\n")
+	}
+
+	return unescaper.Replace(text.String())
+}
+
+// PatchAddContext increases the context until it is unique, but doesn't let the pattern expand beyond MatchMaxBits.
+func (dmp *DiffMatchPatch) PatchAddContext(patch Patch, text string) Patch {
+	if len(text) == 0 {
+		return patch
+	}
+
+	pattern := text[patch.start2 : patch.start2+patch.length1]
+	padding := 0
+
+	// Look for the first and last matches of pattern in text.  If two different matches are found, increase the pattern length.
+	for strings.Index(text, pattern) != strings.LastIndex(text, pattern) &&
+		len(pattern) < dmp.MatchMaxBits-2*dmp.PatchMargin {
+		padding += dmp.PatchMargin
+		maxStart := max(0, patch.start2-padding)
+		minEnd := min(len(text), patch.start2+patch.length1+padding)
+		pattern = text[maxStart:minEnd]
+	}
+	// Add one chunk for good luck.
+	padding += dmp.PatchMargin
+
+	// Add the prefix.
+	prefix := text[max(0, patch.start2-padding):patch.start2]
+	if len(prefix) != 0 {
+		patch.diffs = append([]Diff{Diff{DiffEqual, prefix}}, patch.diffs...)
+	}
+	// Add the suffix.
+	suffix := text[patch.start2+patch.length1 : min(len(text), patch.start2+patch.length1+padding)]
+	if len(suffix) != 0 {
+		patch.diffs = append(patch.diffs, Diff{DiffEqual, suffix})
+	}
+
+	// Roll back the start points.
+	patch.start1 -= len(prefix)
+	patch.start2 -= len(prefix)
+	// Extend the lengths.
+	patch.length1 += len(prefix) + len(suffix)
+	patch.length2 += len(prefix) + len(suffix)
+
+	return patch
+}
+
+// PatchMake computes a list of patches.
+func (dmp *DiffMatchPatch) PatchMake(opt ...interface{}) []Patch {
+	if len(opt) == 1 {
+		diffs, _ := opt[0].([]Diff)
+		text1 := dmp.DiffText1(diffs)
+		return dmp.PatchMake(text1, diffs)
+	} else if len(opt) == 2 {
+		text1 := opt[0].(string)
+		switch t := opt[1].(type) {
+		case string:
+			diffs := dmp.DiffMain(text1, t, true)
+			if len(diffs) > 2 {
+				diffs = dmp.DiffCleanupSemantic(diffs)
+				diffs = dmp.DiffCleanupEfficiency(diffs)
+			}
+			return dmp.PatchMake(text1, diffs)
+		case []Diff:
+			return dmp.patchMake2(text1, t)
+		}
+	} else if len(opt) == 3 {
+		return dmp.PatchMake(opt[0], opt[2])
+	}
+	return []Patch{}
+}
+
+// patchMake2 computes a list of patches to turn text1 into text2.
+// text2 is not provided, diffs are the delta between text1 and text2.
+func (dmp *DiffMatchPatch) patchMake2(text1 string, diffs []Diff) []Patch {
+	// Check for null inputs not needed since null can't be passed in C#.
+	patches := []Patch{}
+	if len(diffs) == 0 {
+		return patches // Get rid of the null case.
+	}
+
+	patch := Patch{}
+	charCount1 := 0 // Number of characters into the text1 string.
+	charCount2 := 0 // Number of characters into the text2 string.
+	// Start with text1 (prepatchText) and apply the diffs until we arrive at text2 (postpatchText). We recreate the patches one by one to determine context info.
+	prepatchText := text1
+	postpatchText := text1
+
+	for i, aDiff := range diffs {
+		if len(patch.diffs) == 0 && aDiff.Type != DiffEqual {
+			// A new patch starts here.
+			patch.start1 = charCount1
+			patch.start2 = charCount2
+		}
+
+		switch aDiff.Type {
+		case DiffInsert:
+			patch.diffs = append(patch.diffs, aDiff)
+			patch.length2 += len(aDiff.Text)
+			postpatchText = postpatchText[:charCount2] +
+				aDiff.Text + postpatchText[charCount2:]
+		case DiffDelete:
+			patch.length1 += len(aDiff.Text)
+			patch.diffs = append(patch.diffs, aDiff)
+			postpatchText = postpatchText[:charCount2] + postpatchText[charCount2+len(aDiff.Text):]
+		case DiffEqual:
+			if len(aDiff.Text) <= 2*dmp.PatchMargin &&
+				len(patch.diffs) != 0 && i != len(diffs)-1 {
+				// Small equality inside a patch.
+				patch.diffs = append(patch.diffs, aDiff)
+				patch.length1 += len(aDiff.Text)
+				patch.length2 += len(aDiff.Text)
+			}
+			if len(aDiff.Text) >= 2*dmp.PatchMargin {
+				// Time for a new patch.
+				if len(patch.diffs) != 0 {
+					patch = dmp.PatchAddContext(patch, prepatchText)
+					patches = append(patches, patch)
+					patch = Patch{}
+					// Unlike Unidiff, our patch lists have a rolling context. http://code.google.com/p/google-diff-match-patch/wiki/Unidiff Update prepatch text & pos to reflect the application of the just completed patch.
+					prepatchText = postpatchText
+					charCount1 = charCount2
+				}
+			}
+		}
+
+		// Update the current character count.
+		if aDiff.Type != DiffInsert {
+			charCount1 += len(aDiff.Text)
+		}
+		if aDiff.Type != DiffDelete {
+			charCount2 += len(aDiff.Text)
+		}
+	}
+
+	// Pick up the leftover patch if not empty.
+	if len(patch.diffs) != 0 {
+		patch = dmp.PatchAddContext(patch, prepatchText)
+		patches = append(patches, patch)
+	}
+
+	return patches
+}
+
+// PatchDeepCopy returns an array that is identical to a given an array of patches.
+func (dmp *DiffMatchPatch) PatchDeepCopy(patches []Patch) []Patch {
+	patchesCopy := []Patch{}
+	for _, aPatch := range patches {
+		patchCopy := Patch{}
+		for _, aDiff := range aPatch.diffs {
+			patchCopy.diffs = append(patchCopy.diffs, Diff{
+				aDiff.Type,
+				aDiff.Text,
+			})
+		}
+		patchCopy.start1 = aPatch.start1
+		patchCopy.start2 = aPatch.start2
+		patchCopy.length1 = aPatch.length1
+		patchCopy.length2 = aPatch.length2
+		patchesCopy = append(patchesCopy, patchCopy)
+	}
+	return patchesCopy
+}
+
+// PatchApply merges a set of patches onto the text.  Returns a patched text, as well as an array of true/false values indicating which patches were applied.
+func (dmp *DiffMatchPatch) PatchApply(patches []Patch, text string) (string, []bool) {
+	if len(patches) == 0 {
+		return text, []bool{}
+	}
+
+	// Deep copy the patches so that no changes are made to originals.
+	patches = dmp.PatchDeepCopy(patches)
+
+	nullPadding := dmp.PatchAddPadding(patches)
+	text = nullPadding + text + nullPadding
+	patches = dmp.PatchSplitMax(patches)
+
+	x := 0
+	// delta keeps track of the offset between the expected and actual location of the previous patch.  If there are patches expected at positions 10 and 20, but the first patch was found at 12, delta is 2 and the second patch has an effective expected position of 22.
+	delta := 0
+	results := make([]bool, len(patches))
+	for _, aPatch := range patches {
+		expectedLoc := aPatch.start2 + delta
+		text1 := dmp.DiffText1(aPatch.diffs)
+		var startLoc int
+		endLoc := -1
+		if len(text1) > dmp.MatchMaxBits {
+			// PatchSplitMax will only provide an oversized pattern in the case of a monster delete.
+			startLoc = dmp.MatchMain(text, text1[:dmp.MatchMaxBits], expectedLoc)
+			if startLoc != -1 {
+				endLoc = dmp.MatchMain(text,
+					text1[len(text1)-dmp.MatchMaxBits:], expectedLoc+len(text1)-dmp.MatchMaxBits)
+				if endLoc == -1 || startLoc >= endLoc {
+					// Can't find valid trailing context.  Drop this patch.
+					startLoc = -1
+				}
+			}
+		} else {
+			startLoc = dmp.MatchMain(text, text1, expectedLoc)
+		}
+		if startLoc == -1 {
+			// No match found.  :(
+			results[x] = false
+			// Subtract the delta for this failed patch from subsequent patches.
+			delta -= aPatch.length2 - aPatch.length1
+		} else {
+			// Found a match.  :)
+			results[x] = true
+			delta = startLoc - expectedLoc
+			var text2 string
+			if endLoc == -1 {
+				text2 = text[startLoc:int(math.Min(float64(startLoc+len(text1)), float64(len(text))))]
+			} else {
+				text2 = text[startLoc:int(math.Min(float64(endLoc+dmp.MatchMaxBits), float64(len(text))))]
+			}
+			if text1 == text2 {
+				// Perfect match, just shove the Replacement text in.
+				text = text[:startLoc] + dmp.DiffText2(aPatch.diffs) + text[startLoc+len(text1):]
+			} else {
+				// Imperfect match.  Run a diff to get a framework of equivalent indices.
+				diffs := dmp.DiffMain(text1, text2, false)
+				if len(text1) > dmp.MatchMaxBits && float64(dmp.DiffLevenshtein(diffs))/float64(len(text1)) > dmp.PatchDeleteThreshold {
+					// The end points match, but the content is unacceptably bad.
+					results[x] = false
+				} else {
+					diffs = dmp.DiffCleanupSemanticLossless(diffs)
+					index1 := 0
+					for _, aDiff := range aPatch.diffs {
+						if aDiff.Type != DiffEqual {
+							index2 := dmp.DiffXIndex(diffs, index1)
+							if aDiff.Type == DiffInsert {
+								// Insertion
+								text = text[:startLoc+index2] + aDiff.Text + text[startLoc+index2:]
+							} else if aDiff.Type == DiffDelete {
+								// Deletion
+								startIndex := startLoc + index2
+								text = text[:startIndex] +
+									text[startIndex+dmp.DiffXIndex(diffs, index1+len(aDiff.Text))-index2:]
+							}
+						}
+						if aDiff.Type != DiffDelete {
+							index1 += len(aDiff.Text)
+						}
+					}
+				}
+			}
+		}
+		x++
+	}
+	// Strip the padding off.
+	text = text[len(nullPadding) : len(nullPadding)+(len(text)-2*len(nullPadding))]
+	return text, results
+}
+
+// PatchAddPadding adds some padding on text start and end so that edges can match something.
+// Intended to be called only from within patchApply.
+func (dmp *DiffMatchPatch) PatchAddPadding(patches []Patch) string {
+	paddingLength := dmp.PatchMargin
+	nullPadding := ""
+	for x := 1; x <= paddingLength; x++ {
+		nullPadding += string(x)
+	}
+
+	// Bump all the patches forward.
+	for i := range patches {
+		patches[i].start1 += paddingLength
+		patches[i].start2 += paddingLength
+	}
+
+	// Add some padding on start of first diff.
+	if len(patches[0].diffs) == 0 || patches[0].diffs[0].Type != DiffEqual {
+		// Add nullPadding equality.
+		patches[0].diffs = append([]Diff{Diff{DiffEqual, nullPadding}}, patches[0].diffs...)
+		patches[0].start1 -= paddingLength // Should be 0.
+		patches[0].start2 -= paddingLength // Should be 0.
+		patches[0].length1 += paddingLength
+		patches[0].length2 += paddingLength
+	} else if paddingLength > len(patches[0].diffs[0].Text) {
+		// Grow first equality.
+		extraLength := paddingLength - len(patches[0].diffs[0].Text)
+		patches[0].diffs[0].Text = nullPadding[len(patches[0].diffs[0].Text):] + patches[0].diffs[0].Text
+		patches[0].start1 -= extraLength
+		patches[0].start2 -= extraLength
+		patches[0].length1 += extraLength
+		patches[0].length2 += extraLength
+	}
+
+	// Add some padding on end of last diff.
+	last := len(patches) - 1
+	if len(patches[last].diffs) == 0 || patches[last].diffs[len(patches[last].diffs)-1].Type != DiffEqual {
+		// Add nullPadding equality.
+		patches[last].diffs = append(patches[last].diffs, Diff{DiffEqual, nullPadding})
+		patches[last].length1 += paddingLength
+		patches[last].length2 += paddingLength
+	} else if paddingLength > len(patches[last].diffs[len(patches[last].diffs)-1].Text) {
+		// Grow last equality.
+		lastDiff := patches[last].diffs[len(patches[last].diffs)-1]
+		extraLength := paddingLength - len(lastDiff.Text)
+		patches[last].diffs[len(patches[last].diffs)-1].Text += nullPadding[:extraLength]
+		patches[last].length1 += extraLength
+		patches[last].length2 += extraLength
+	}
+
+	return nullPadding
+}
+
+// PatchSplitMax looks through the patches and breaks up any which are longer than the maximum limit of the match algorithm.
+// Intended to be called only from within patchApply.
+func (dmp *DiffMatchPatch) PatchSplitMax(patches []Patch) []Patch {
+	patchSize := dmp.MatchMaxBits
+	for x := 0; x < len(patches); x++ {
+		if patches[x].length1 <= patchSize {
+			continue
+		}
+		bigpatch := patches[x]
+		// Remove the big old patch.
+		patches = append(patches[:x], patches[x+1:]...)
+		x--
+
+		start1 := bigpatch.start1
+		start2 := bigpatch.start2
+		precontext := ""
+		for len(bigpatch.diffs) != 0 {
+			// Create one of several smaller patches.
+			patch := Patch{}
+			empty := true
+			patch.start1 = start1 - len(precontext)
+			patch.start2 = start2 - len(precontext)
+			if len(precontext) != 0 {
+				patch.length1 = len(precontext)
+				patch.length2 = len(precontext)
+				patch.diffs = append(patch.diffs, Diff{DiffEqual, precontext})
+			}
+			for len(bigpatch.diffs) != 0 && patch.length1 < patchSize-dmp.PatchMargin {
+				diffType := bigpatch.diffs[0].Type
+				diffText := bigpatch.diffs[0].Text
+				if diffType == DiffInsert {
+					// Insertions are harmless.
+					patch.length2 += len(diffText)
+					start2 += len(diffText)
+					patch.diffs = append(patch.diffs, bigpatch.diffs[0])
+					bigpatch.diffs = bigpatch.diffs[1:]
+					empty = false
+				} else if diffType == DiffDelete && len(patch.diffs) == 1 && patch.diffs[0].Type == DiffEqual && len(diffText) > 2*patchSize {
+					// This is a large deletion.  Let it pass in one chunk.
+					patch.length1 += len(diffText)
+					start1 += len(diffText)
+					empty = false
+					patch.diffs = append(patch.diffs, Diff{diffType, diffText})
+					bigpatch.diffs = bigpatch.diffs[1:]
+				} else {
+					// Deletion or equality.  Only take as much as we can stomach.
+					diffText = diffText[:min(len(diffText), patchSize-patch.length1-dmp.PatchMargin)]
+
+					patch.length1 += len(diffText)
+					start1 += len(diffText)
+					if diffType == DiffEqual {
+						patch.length2 += len(diffText)
+						start2 += len(diffText)
+					} else {
+						empty = false
+					}
+					patch.diffs = append(patch.diffs, Diff{diffType, diffText})
+					if diffText == bigpatch.diffs[0].Text {
+						bigpatch.diffs = bigpatch.diffs[1:]
+					} else {
+						bigpatch.diffs[0].Text =
+							bigpatch.diffs[0].Text[len(diffText):]
+					}
+				}
+			}
+			// Compute the head context for the next patch.
+			precontext = dmp.DiffText2(patch.diffs)
+			precontext = precontext[max(0, len(precontext)-dmp.PatchMargin):]
+
+			postcontext := ""
+			// Append the end context for this patch.
+			if len(dmp.DiffText1(bigpatch.diffs)) > dmp.PatchMargin {
+				postcontext = dmp.DiffText1(bigpatch.diffs)[:dmp.PatchMargin]
+			} else {
+				postcontext = dmp.DiffText1(bigpatch.diffs)
+			}
+
+			if len(postcontext) != 0 {
+				patch.length1 += len(postcontext)
+				patch.length2 += len(postcontext)
+				if len(patch.diffs) != 0 && patch.diffs[len(patch.diffs)-1].Type == DiffEqual {
+					patch.diffs[len(patch.diffs)-1].Text += postcontext
+				} else {
+					patch.diffs = append(patch.diffs, Diff{DiffEqual, postcontext})
+				}
+			}
+			if !empty {
+				x++
+				patches = append(patches[:x], append([]Patch{patch}, patches[x:]...)...)
+			}
+		}
+	}
+	return patches
+}
+
+// PatchToText takes a list of patches and returns a textual representation.
+func (dmp *DiffMatchPatch) PatchToText(patches []Patch) string {
+	var text bytes.Buffer
+	for _, aPatch := range patches {
+		_, _ = text.WriteString(aPatch.String())
+	}
+	return text.String()
+}
+
+// PatchFromText parses a textual representation of patches and returns a List of Patch objects.
+func (dmp *DiffMatchPatch) PatchFromText(textline string) ([]Patch, error) {
+	patches := []Patch{}
+	if len(textline) == 0 {
+		return patches, nil
+	}
+	text := strings.Split(textline, "\n")
+	textPointer := 0
+	patchHeader := regexp.MustCompile("^@@ -(\\d+),?(\\d*) \\+(\\d+),?(\\d*) @@$")
+
+	var patch Patch
+	var sign uint8
+	var line string
+	for textPointer < len(text) {
+
+		if !patchHeader.MatchString(text[textPointer]) {
+			return patches, errors.New("Invalid patch string: " + text[textPointer])
+		}
+
+		patch = Patch{}
+		m := patchHeader.FindStringSubmatch(text[textPointer])
+
+		patch.start1, _ = strconv.Atoi(m[1])
+		if len(m[2]) == 0 {
+			patch.start1--
+			patch.length1 = 1
+		} else if m[2] == "0" {
+			patch.length1 = 0
+		} else {
+			patch.start1--
+			patch.length1, _ = strconv.Atoi(m[2])
+		}
+
+		patch.start2, _ = strconv.Atoi(m[3])
+
+		if len(m[4]) == 0 {
+			patch.start2--
+			patch.length2 = 1
+		} else if m[4] == "0" {
+			patch.length2 = 0
+		} else {
+			patch.start2--
+			patch.length2, _ = strconv.Atoi(m[4])
+		}
+		textPointer++
+
+		for textPointer < len(text) {
+			if len(text[textPointer]) > 0 {
+				sign = text[textPointer][0]
+			} else {
+				textPointer++
+				continue
+			}
+
+			line = text[textPointer][1:]
+			line = strings.Replace(line, "+", "%2b", -1)
+			line, _ = url.QueryUnescape(line)
+			if sign == '-' {
+				// Deletion.
+				patch.diffs = append(patch.diffs, Diff{DiffDelete, line})
+			} else if sign == '+' {
+				// Insertion.
+				patch.diffs = append(patch.diffs, Diff{DiffInsert, line})
+			} else if sign == ' ' {
+				// Minor equality.
+				patch.diffs = append(patch.diffs, Diff{DiffEqual, line})
+			} else if sign == '@' {
+				// Start of next patch.
+				break
+			} else {
+				// WTF?
+				return patches, errors.New("Invalid patch mode '" + string(sign) + "' in: " + string(line))
+			}
+			textPointer++
+		}
+
+		patches = append(patches, patch)
+	}
+	return patches, nil
+}
@@ -0,0 +1,88 @@
+// Copyright (c) 2012-2016 The go-diff authors. All rights reserved.
+// https://github.com/sergi/go-diff
+// See the included LICENSE file for license details.
+//
+// go-diff is a Go implementation of Google's Diff, Match, and Patch library
+// Original library is Copyright (c) 2006 Google Inc.
+// http://code.google.com/p/google-diff-match-patch/
+
+package diffmatchpatch
+
+import (
+	"strings"
+	"unicode/utf8"
+)
+
+// unescaper unescapes selected chars for compatibility with JavaScript's encodeURI.
+// In speed critical applications this could be dropped since the receiving application will certainly decode these fine. Note that this function is case-sensitive.  Thus "%3F" would not be unescaped.  But this is ok because it is only called with the output of HttpUtility.UrlEncode which returns lowercase hex. Example: "%3f" -> "?", "%24" -> "$", etc.
+var unescaper = strings.NewReplacer(
+	"%21", "!", "%7E", "~", "%27", "'",
+	"%28", "(", "%29", ")", "%3B", ";",
+	"%2F", "/", "%3F", "?", "%3A", ":",
+	"%40", "@", "%26", "&", "%3D", "=",
+	"%2B", "+", "%24", "$", "%2C", ",", "%23", "#", "%2A", "*")
+
+// indexOf returns the first index of pattern in str, starting at str[i].
+func indexOf(str string, pattern string, i int) int {
+	if i > len(str)-1 {
+		return -1
+	}
+	if i <= 0 {
+		return strings.Index(str, pattern)
+	}
+	ind := strings.Index(str[i:], pattern)
+	if ind == -1 {
+		return -1
+	}
+	return ind + i
+}
+
+// lastIndexOf returns the last index of pattern in str, starting at str[i].
+func lastIndexOf(str string, pattern string, i int) int {
+	if i < 0 {
+		return -1
+	}
+	if i >= len(str) {
+		return strings.LastIndex(str, pattern)
+	}
+	_, size := utf8.DecodeRuneInString(str[i:])
+	return strings.LastIndex(str[:i+size], pattern)
+}
+
+// runesIndexOf returns the index of pattern in target, starting at target[i].
+func runesIndexOf(target, pattern []rune, i int) int {
+	if i > len(target)-1 {
+		return -1
+	}
+	if i <= 0 {
+		return runesIndex(target, pattern)
+	}
+	ind := runesIndex(target[i:], pattern)
+	if ind == -1 {
+		return -1
+	}
+	return ind + i
+}
+
+func runesEqual(r1, r2 []rune) bool {
+	if len(r1) != len(r2) {
+		return false
+	}
+	for i, c := range r1 {
+		if c != r2[i] {
+			return false
+		}
+	}
+	return true
+}
+
+// runesIndex is the equivalent of strings.Index for rune slices.
+func runesIndex(r1, r2 []rune) int {
+	last := len(r1) - len(r2)
+	for i := 0; i <= last; i++ {
+		if runesEqual(r1[i:i+len(r2)], r2) {
+			return i
+		}
+	}
+	return -1
+}
@@ -0,0 +1,145 @@
+The MIT License (MIT)
+
+Copyright (c) 2015 Iwasaki Yudai
+
+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.
+
+============================================================================
+
+This repository is build with following third party libraries. Thank you!
+
+## go-diff - https://github.com/sergi/go-diff
+
+Copyright (c) 2012 Sergi Mansilla <sergi.mansilla@gmail.com>
+
+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.
+
+
+## golcs - https://github.com/yudai/golcs
+
+The MIT License (MIT)
+
+Copyright (c) 2015 Iwasaki Yudai
+
+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.
+
+
+## cli.go - https://github.com/urfave/cli
+
+Copyright (C) 2013 Jeremy Saenz
+All Rights Reserved.
+
+MIT LICENSE
+
+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.
+
+
+## ginkgo - https://github.com/onsi/ginkgo
+
+Copyright (c) 2013-2014 Onsi Fakhouri
+
+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.
+
+
+# gomega - https://github.com/onsi/gomega
+
+Copyright (c) 2013-2014 Onsi Fakhouri
+
+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.
@@ -0,0 +1,2 @@
+test:
+	if [ `go fmt $(go list ./... | grep -v /vendor/) | wc -l` -gt 0 ]; then echo "go fmt error"; exit 1; fi
@@ -0,0 +1,157 @@
+# Go JSON Diff (and Patch)
+
+[![Wercker](https://app.wercker.com/status/00d70daaf40ce277fd4f10290f097b9d/s/master)][wercker]
+[![GoDoc](https://godoc.org/github.com/yudai/gojsondiff?status.svg)][godoc]
+[![MIT License](http://img.shields.io/badge/license-MIT-blue.svg)][license]
+
+[wercker]: https://app.wercker.com/project/bykey/00d70daaf40ce277fd4f10290f097b9d
+[godoc]: https://godoc.org/github.com/yudai/gojsondiff
+[license]: https://github.com/yudai/gojsondiff/blob/master/LICENSE
+
+## How to use
+
+### Installation
+
+```sh
+go get github.com/yudai/gojsondiff
+```
+
+### Comparing two JSON strings
+
+See `jd/main.go` for how to use this library.
+
+
+## CLI tool
+
+This repository contains a package that you can use as a CLI tool.
+
+### Installation
+
+```sh
+go get github.com/yudai/gojsondiff/jd
+```
+
+### Usage
+
+#### Diff
+
+Just give two json files to the `jd` command:
+
+```sh
+jd one.json another.json
+```
+
+Outputs would be something like:
+
+```diff
+ {
+   "arr": [
+     0: "arr0",
+     1: 21,
+     2: {
+       "num": 1,
+-      "str": "pek3f"
+      "str": "changed"
+     },
+     3: [
+       0: 0,
+-      1: "1"
+      1: "changed"
+     ]
+   ],
+   "bool": true,
+   "num_float": 39.39,
+   "num_int": 13,
+   "obj": {
+     "arr": [
+       0: 17,
+       1: "str",
+       2: {
+-        "str": "eafeb"
+        "str": "changed"
+       }
+     ],
+    "new": "added",
+-    "num": 19,
+     "obj": {
+-      "num": 14,
+      "num": 9999
+-      "str": "efj3"
+      "str": "changed"
+     },
+     "str": "bcded"
+   },
+   "str": "abcde"
+ }
+```
+
+When you prefer the delta foramt of [jsondiffpatch](https://github.com/benjamine/jsondiffpatch), add the `-f delta` option.
+
+```sh
+jd -f delta one.json another.json
+```
+
+This command shows:
+
+```json
+{
+  "arr": {
+    "2": {
+      "str": [
+        "pek3f",
+        "changed"
+      ]
+    },
+    "3": {
+      "1": [
+        "1",
+        "changed"
+      ],
+      "_t": "a"
+    },
+    "_t": "a"
+  },
+  "obj": {
+    "arr": {
+      "2": {
+        "str": [
+          "eafeb",
+          "changed"
+        ]
+      },
+      "_t": "a"
+    },
+    "new": [
+      "added"
+    ],
+    "num": [
+      19,
+      0,
+      0
+    ],
+    "obj": {
+      "num": [
+        14,
+        9999
+      ],
+      "str": [
+        "efj3",
+        "changed"
+      ]
+    }
+  }
+}
+```
+
+#### Patch
+
+Give a diff file in the delta format and the JSON file to the `jp` command.
+
+```sh
+jp diff.delta one.json
+```
+
+
+## License
+
+MIT License (see `LICENSE` for detail)
@@ -0,0 +1,461 @@
+package gojsondiff
+
+import (
+	"errors"
+	dmp "github.com/sergi/go-diff/diffmatchpatch"
+	"reflect"
+	"strconv"
+)
+
+// A Delta represents an atomic difference between two JSON objects.
+type Delta interface {
+	// Similarity calculates the similarity of the Delta values.
+	// The return value is normalized from 0 to 1,
+	// 0 is completely different and 1 is they are same
+	Similarity() (similarity float64)
+}
+
+// To cache the calculated similarity,
+// concrete Deltas can use similariter and similarityCache
+type similariter interface {
+	similarity() (similarity float64)
+}
+
+type similarityCache struct {
+	similariter
+	value float64
+}
+
+func newSimilarityCache(sim similariter) similarityCache {
+	cache := similarityCache{similariter: sim, value: -1}
+	return cache
+}
+
+func (cache similarityCache) Similarity() (similarity float64) {
+	if cache.value < 0 {
+		cache.value = cache.similariter.similarity()
+	}
+	return cache.value
+}
+
+// A Position represents the position of a Delta in an object or an array.
+type Position interface {
+	// String returns the position as a string
+	String() (name string)
+
+	// CompareTo returns a true if the Position is smaller than another Position.
+	// This function is used to sort Positions by the sort package.
+	CompareTo(another Position) bool
+}
+
+// A Name is a Postition with a string, which means the delta is in an object.
+type Name string
+
+func (n Name) String() (name string) {
+	return string(n)
+}
+
+func (n Name) CompareTo(another Position) bool {
+	return n < another.(Name)
+}
+
+// A Index is a Position with an int value, which means the Delta is in an Array.
+type Index int
+
+func (i Index) String() (name string) {
+	return strconv.Itoa(int(i))
+}
+
+func (i Index) CompareTo(another Position) bool {
+	return i < another.(Index)
+}
+
+// A PreDelta is a Delta that has a position of the left side JSON object.
+// Deltas implements this interface should be applies before PostDeltas.
+type PreDelta interface {
+	// PrePosition returns the Position.
+	PrePosition() Position
+
+	// PreApply applies the delta to object.
+	PreApply(object interface{}) interface{}
+}
+
+type preDelta struct{ Position }
+
+func (i preDelta) PrePosition() Position {
+	return Position(i.Position)
+}
+
+type preDeltas []PreDelta
+
+// for sorting
+func (s preDeltas) Len() int {
+	return len(s)
+}
+
+// for sorting
+func (s preDeltas) Swap(i, j int) {
+	s[i], s[j] = s[j], s[i]
+}
+
+// for sorting
+func (s preDeltas) Less(i, j int) bool {
+	return !s[i].PrePosition().CompareTo(s[j].PrePosition())
+}
+
+// A PreDelta is a Delta that has a position of the right side JSON object.
+// Deltas implements this interface should be applies after PreDeltas.
+type PostDelta interface {
+	// PostPosition returns the Position.
+	PostPosition() Position
+
+	// PostApply applies the delta to object.
+	PostApply(object interface{}) interface{}
+}
+
+type postDelta struct{ Position }
+
+func (i postDelta) PostPosition() Position {
+	return Position(i.Position)
+}
+
+type postDeltas []PostDelta
+
+// for sorting
+func (s postDeltas) Len() int {
+	return len(s)
+}
+
+// for sorting
+func (s postDeltas) Swap(i, j int) {
+	s[i], s[j] = s[j], s[i]
+}
+
+// for sorting
+func (s postDeltas) Less(i, j int) bool {
+	return s[i].PostPosition().CompareTo(s[j].PostPosition())
+}
+
+// An Object is a Delta that represents an object of JSON
+type Object struct {
+	postDelta
+	similarityCache
+
+	// Deltas holds internal Deltas
+	Deltas []Delta
+}
+
+// NewObject returns an Object
+func NewObject(position Position, deltas []Delta) *Object {
+	d := Object{postDelta: postDelta{position}, Deltas: deltas}
+	d.similarityCache = newSimilarityCache(&d)
+	return &d
+}
+
+func (d *Object) PostApply(object interface{}) interface{} {
+	switch object.(type) {
+	case map[string]interface{}:
+		o := object.(map[string]interface{})
+		n := string(d.PostPosition().(Name))
+		o[n] = applyDeltas(d.Deltas, o[n])
+	case []interface{}:
+		o := object.([]interface{})
+		n := int(d.PostPosition().(Index))
+		o[n] = applyDeltas(d.Deltas, o[n])
+	}
+	return object
+}
+
+func (d *Object) similarity() (similarity float64) {
+	similarity = deltasSimilarity(d.Deltas)
+	return
+}
+
+// An Array is a Delta that represents an array of JSON
+type Array struct {
+	postDelta
+	similarityCache
+
+	// Deltas holds internal Deltas
+	Deltas []Delta
+}
+
+// NewArray returns an Array
+func NewArray(position Position, deltas []Delta) *Array {
+	d := Array{postDelta: postDelta{position}, Deltas: deltas}
+	d.similarityCache = newSimilarityCache(&d)
+	return &d
+}
+
+func (d *Array) PostApply(object interface{}) interface{} {
+	switch object.(type) {
+	case map[string]interface{}:
+		o := object.(map[string]interface{})
+		n := string(d.PostPosition().(Name))
+		o[n] = applyDeltas(d.Deltas, o[n])
+	case []interface{}:
+		o := object.([]interface{})
+		n := int(d.PostPosition().(Index))
+		o[n] = applyDeltas(d.Deltas, o[n])
+	}
+	return object
+}
+
+func (d *Array) similarity() (similarity float64) {
+	similarity = deltasSimilarity(d.Deltas)
+	return
+}
+
+// An Added represents a new added field of an object or an array
+type Added struct {
+	postDelta
+	similarityCache
+
+	// Values holds the added value
+	Value interface{}
+}
+
+// NewAdded returns a new Added
+func NewAdded(position Position, value interface{}) *Added {
+	d := Added{postDelta: postDelta{position}, Value: value}
+	return &d
+}
+
+func (d *Added) PostApply(object interface{}) interface{} {
+	switch object.(type) {
+	case map[string]interface{}:
+		object.(map[string]interface{})[string(d.PostPosition().(Name))] = d.Value
+	case []interface{}:
+		i := int(d.PostPosition().(Index))
+		o := object.([]interface{})
+		if i < len(o) {
+			o = append(o, 0) //dummy
+			copy(o[i+1:], o[i:])
+			o[i] = d.Value
+			object = o
+		} else {
+			object = append(o, d.Value)
+		}
+	}
+
+	return object
+}
+
+func (d *Added) similarity() (similarity float64) {
+	return 0
+}
+
+// A Modified represents a field whose value is changed.
+type Modified struct {
+	postDelta
+	similarityCache
+
+	// The value before modification
+	OldValue interface{}
+
+	// The value after modification
+	NewValue interface{}
+}
+
+// NewModified returns a Modified
+func NewModified(position Position, oldValue, newValue interface{}) *Modified {
+	d := Modified{
+		postDelta: postDelta{position},
+		OldValue:  oldValue,
+		NewValue:  newValue,
+	}
+	d.similarityCache = newSimilarityCache(&d)
+	return &d
+
+}
+
+func (d *Modified) PostApply(object interface{}) interface{} {
+	switch object.(type) {
+	case map[string]interface{}:
+		// TODO check old value
+		object.(map[string]interface{})[string(d.PostPosition().(Name))] = d.NewValue
+	case []interface{}:
+		object.([]interface{})[int(d.PostPosition().(Index))] = d.NewValue
+	}
+	return object
+}
+
+func (d *Modified) similarity() (similarity float64) {
+	similarity += 0.3 // at least, they are at the same position
+	if reflect.TypeOf(d.OldValue) == reflect.TypeOf(d.NewValue) {
+		similarity += 0.3 // types are same
+
+		switch d.OldValue.(type) {
+		case string:
+			similarity += 0.4 * stringSimilarity(d.OldValue.(string), d.NewValue.(string))
+		case float64:
+			ratio := d.OldValue.(float64) / d.NewValue.(float64)
+			if ratio > 1 {
+				ratio = 1 / ratio
+			}
+			similarity += 0.4 * ratio
+		}
+	}
+	return
+}
+
+// A TextDiff represents a Modified with TextDiff between the old and the new values.
+type TextDiff struct {
+	Modified
+
+	// Diff string
+	Diff []dmp.Patch
+}
+
+// NewTextDiff returns
+func NewTextDiff(position Position, diff []dmp.Patch, oldValue, newValue interface{}) *TextDiff {
+	d := TextDiff{
+		Modified: *NewModified(position, oldValue, newValue),
+		Diff:     diff,
+	}
+	return &d
+}
+
+func (d *TextDiff) PostApply(object interface{}) interface{} {
+	switch object.(type) {
+	case map[string]interface{}:
+		o := object.(map[string]interface{})
+		i := string(d.PostPosition().(Name))
+		// TODO error
+		d.OldValue = o[i]
+		// TODO error
+		d.patch()
+		o[i] = d.NewValue
+	case []interface{}:
+		o := object.([]interface{})
+		i := d.PostPosition().(Index)
+		d.OldValue = o[i]
+		// TODO error
+		d.patch()
+		o[i] = d.NewValue
+	}
+	return object
+}
+
+func (d *TextDiff) patch() error {
+	if d.OldValue == nil {
+		return errors.New("Old Value is not set")
+	}
+	patcher := dmp.New()
+	patched, successes := patcher.PatchApply(d.Diff, d.OldValue.(string))
+	for _, success := range successes {
+		if !success {
+			return errors.New("Failed to apply a patch")
+		}
+	}
+	d.NewValue = patched
+	return nil
+}
+
+func (d *TextDiff) DiffString() string {
+	dmp := dmp.New()
+	return dmp.PatchToText(d.Diff)
+}
+
+// A Delted represents deleted field or index of an Object or an Array.
+type Deleted struct {
+	preDelta
+
+	// The value deleted
+	Value interface{}
+}
+
+// NewDeleted returns a Deleted
+func NewDeleted(position Position, value interface{}) *Deleted {
+	d := Deleted{
+		preDelta: preDelta{position},
+		Value:    value,
+	}
+	return &d
+
+}
+
+func (d *Deleted) PreApply(object interface{}) interface{} {
+	switch object.(type) {
+	case map[string]interface{}:
+		// TODO check old value
+		delete(object.(map[string]interface{}), string(d.PrePosition().(Name)))
+	case []interface{}:
+		i := int(d.PrePosition().(Index))
+		o := object.([]interface{})
+		object = append(o[:i], o[i+1:]...)
+	}
+	return object
+}
+
+func (d Deleted) Similarity() (similarity float64) {
+	return 0
+}
+
+// A Moved represents field that is moved, which means the index or name is
+// changed. Note that, in this library, assigning a Moved and a Modified to
+// a single position is not allowed. For the compatibility with jsondiffpatch,
+// the Moved in this library can hold the old and new value in it.
+type Moved struct {
+	preDelta
+	postDelta
+	similarityCache
+	// The value before moving
+	Value interface{}
+	// The delta applied after moving (for compatibility)
+	Delta interface{}
+}
+
+func NewMoved(oldPosition Position, newPosition Position, value interface{}, delta Delta) *Moved {
+	d := Moved{
+		preDelta:  preDelta{oldPosition},
+		postDelta: postDelta{newPosition},
+		Value:     value,
+		Delta:     delta,
+	}
+	d.similarityCache = newSimilarityCache(&d)
+	return &d
+}
+
+func (d *Moved) PreApply(object interface{}) interface{} {
+	switch object.(type) {
+	case map[string]interface{}:
+		//not supported
+	case []interface{}:
+		i := int(d.PrePosition().(Index))
+		o := object.([]interface{})
+		d.Value = o[i]
+		object = append(o[:i], o[i+1:]...)
+	}
+	return object
+}
+
+func (d *Moved) PostApply(object interface{}) interface{} {
+	switch object.(type) {
+	case map[string]interface{}:
+		//not supported
+	case []interface{}:
+		i := int(d.PostPosition().(Index))
+		o := object.([]interface{})
+		o = append(o, 0) //dummy
+		copy(o[i+1:], o[i:])
+		o[i] = d.Value
+		object = o
+	}
+
+	if d.Delta != nil {
+		d.Delta.(PostDelta).PostApply(object)
+	}
+
+	return object
+}
+
+func (d *Moved) similarity() (similarity float64) {
+	similarity = 0.6 // as type and contens are same
+	ratio := float64(d.PrePosition().(Index)) / float64(d.PostPosition().(Index))
+	if ratio > 1 {
+		ratio = 1 / ratio
+	}
+	similarity += 0.4 * ratio
+	return
+}
@@ -0,0 +1,370 @@
+package formatter
+
+import (
+	"bytes"
+	"errors"
+	"fmt"
+	"sort"
+
+	diff "github.com/yudai/gojsondiff"
+)
+
+func NewAsciiFormatter(left interface{}, config AsciiFormatterConfig) *AsciiFormatter {
+	return &AsciiFormatter{
+		left:   left,
+		config: config,
+	}
+}
+
+type AsciiFormatter struct {
+	left    interface{}
+	config  AsciiFormatterConfig
+	buffer  *bytes.Buffer
+	path    []string
+	size    []int
+	inArray []bool
+	line    *AsciiLine
+}
+
+type AsciiFormatterConfig struct {
+	ShowArrayIndex bool
+	Coloring       bool
+}
+
+var AsciiFormatterDefaultConfig = AsciiFormatterConfig{}
+
+type AsciiLine struct {
+	marker string
+	indent int
+	buffer *bytes.Buffer
+}
+
+func (f *AsciiFormatter) Format(diff diff.Diff) (result string, err error) {
+	f.buffer = bytes.NewBuffer([]byte{})
+	f.path = []string{}
+	f.size = []int{}
+	f.inArray = []bool{}
+
+	if v, ok := f.left.(map[string]interface{}); ok {
+		f.formatObject(v, diff)
+	} else if v, ok := f.left.([]interface{}); ok {
+		f.formatArray(v, diff)
+	} else {
+		return "", fmt.Errorf("expected map[string]interface{} or []interface{}, got %T",
+			f.left)
+	}
+
+	return f.buffer.String(), nil
+}
+
+func (f *AsciiFormatter) formatObject(left map[string]interface{}, df diff.Diff) {
+	f.addLineWith(AsciiSame, "{")
+	f.push("ROOT", len(left), false)
+	f.processObject(left, df.Deltas())
+	f.pop()
+	f.addLineWith(AsciiSame, "}")
+}
+
+func (f *AsciiFormatter) formatArray(left []interface{}, df diff.Diff) {
+	f.addLineWith(AsciiSame, "[")
+	f.push("ROOT", len(left), true)
+	f.processArray(left, df.Deltas())
+	f.pop()
+	f.addLineWith(AsciiSame, "]")
+}
+
+func (f *AsciiFormatter) processArray(array []interface{}, deltas []diff.Delta) error {
+	patchedIndex := 0
+	for index, value := range array {
+		f.processItem(value, deltas, diff.Index(index))
+		patchedIndex++
+	}
+
+	// additional Added
+	for _, delta := range deltas {
+		switch delta.(type) {
+		case *diff.Added:
+			d := delta.(*diff.Added)
+			// skip items already processed
+			if int(d.Position.(diff.Index)) < len(array) {
+				continue
+			}
+			f.printRecursive(d.Position.String(), d.Value, AsciiAdded)
+		}
+	}
+
+	return nil
+}
+
+func (f *AsciiFormatter) processObject(object map[string]interface{}, deltas []diff.Delta) error {
+	names := sortedKeys(object)
+	for _, name := range names {
+		value := object[name]
+		f.processItem(value, deltas, diff.Name(name))
+	}
+
+	// Added
+	for _, delta := range deltas {
+		switch delta.(type) {
+		case *diff.Added:
+			d := delta.(*diff.Added)
+			f.printRecursive(d.Position.String(), d.Value, AsciiAdded)
+		}
+	}
+
+	return nil
+}
+
+func (f *AsciiFormatter) processItem(value interface{}, deltas []diff.Delta, position diff.Position) error {
+	matchedDeltas := f.searchDeltas(deltas, position)
+	positionStr := position.String()
+	if len(matchedDeltas) > 0 {
+		for _, matchedDelta := range matchedDeltas {
+
+			switch matchedDelta.(type) {
+			case *diff.Object:
+				d := matchedDelta.(*diff.Object)
+				switch value.(type) {
+				case map[string]interface{}:
+					//ok
+				default:
+					return errors.New("Type mismatch")
+				}
+				o := value.(map[string]interface{})
+
+				f.newLine(AsciiSame)
+				f.printKey(positionStr)
+				f.print("{")
+				f.closeLine()
+				f.push(positionStr, len(o), false)
+				f.processObject(o, d.Deltas)
+				f.pop()
+				f.newLine(AsciiSame)
+				f.print("}")
+				f.printComma()
+				f.closeLine()
+
+			case *diff.Array:
+				d := matchedDelta.(*diff.Array)
+				switch value.(type) {
+				case []interface{}:
+					//ok
+				default:
+					return errors.New("Type mismatch")
+				}
+				a := value.([]interface{})
+
+				f.newLine(AsciiSame)
+				f.printKey(positionStr)
+				f.print("[")
+				f.closeLine()
+				f.push(positionStr, len(a), true)
+				f.processArray(a, d.Deltas)
+				f.pop()
+				f.newLine(AsciiSame)
+				f.print("]")
+				f.printComma()
+				f.closeLine()
+
+			case *diff.Added:
+				d := matchedDelta.(*diff.Added)
+				f.printRecursive(positionStr, d.Value, AsciiAdded)
+				f.size[len(f.size)-1]++
+
+			case *diff.Modified:
+				d := matchedDelta.(*diff.Modified)
+				savedSize := f.size[len(f.size)-1]
+				f.printRecursive(positionStr, d.OldValue, AsciiDeleted)
+				f.size[len(f.size)-1] = savedSize
+				f.printRecursive(positionStr, d.NewValue, AsciiAdded)
+
+			case *diff.TextDiff:
+				savedSize := f.size[len(f.size)-1]
+				d := matchedDelta.(*diff.TextDiff)
+				f.printRecursive(positionStr, d.OldValue, AsciiDeleted)
+				f.size[len(f.size)-1] = savedSize
+				f.printRecursive(positionStr, d.NewValue, AsciiAdded)
+
+			case *diff.Deleted:
+				d := matchedDelta.(*diff.Deleted)
+				f.printRecursive(positionStr, d.Value, AsciiDeleted)
+
+			default:
+				return errors.New("Unknown Delta type detected")
+			}
+
+		}
+	} else {
+		f.printRecursive(positionStr, value, AsciiSame)
+	}
+
+	return nil
+}
+
+func (f *AsciiFormatter) searchDeltas(deltas []diff.Delta, postion diff.Position) (results []diff.Delta) {
+	results = make([]diff.Delta, 0)
+	for _, delta := range deltas {
+		switch delta.(type) {
+		case diff.PostDelta:
+			if delta.(diff.PostDelta).PostPosition() == postion {
+				results = append(results, delta)
+			}
+		case diff.PreDelta:
+			if delta.(diff.PreDelta).PrePosition() == postion {
+				results = append(results, delta)
+			}
+		default:
+			panic("heh")
+		}
+	}
+	return
+}
+
+const (
+	AsciiSame    = " "
+	AsciiAdded   = "+"
+	AsciiDeleted = "-"
+)
+
+var AsciiStyles = map[string]string{
+	AsciiAdded:   "30;42",
+	AsciiDeleted: "30;41",
+}
+
+func (f *AsciiFormatter) push(name string, size int, array bool) {
+	f.path = append(f.path, name)
+	f.size = append(f.size, size)
+	f.inArray = append(f.inArray, array)
+}
+
+func (f *AsciiFormatter) pop() {
+	f.path = f.path[0 : len(f.path)-1]
+	f.size = f.size[0 : len(f.size)-1]
+	f.inArray = f.inArray[0 : len(f.inArray)-1]
+}
+
+func (f *AsciiFormatter) addLineWith(marker string, value string) {
+	f.line = &AsciiLine{
+		marker: marker,
+		indent: len(f.path),
+		buffer: bytes.NewBufferString(value),
+	}
+	f.closeLine()
+}
+
+func (f *AsciiFormatter) newLine(marker string) {
+	f.line = &AsciiLine{
+		marker: marker,
+		indent: len(f.path),
+		buffer: bytes.NewBuffer([]byte{}),
+	}
+}
+
+func (f *AsciiFormatter) closeLine() {
+	style, ok := AsciiStyles[f.line.marker]
+	if f.config.Coloring && ok {
+		f.buffer.WriteString("\x1b[" + style + "m")
+	}
+
+	f.buffer.WriteString(f.line.marker)
+	for n := 0; n < f.line.indent; n++ {
+		f.buffer.WriteString("  ")
+	}
+	f.buffer.Write(f.line.buffer.Bytes())
+
+	if f.config.Coloring && ok {
+		f.buffer.WriteString("\x1b[0m")
+	}
+
+	f.buffer.WriteRune('\n')
+}
+
+func (f *AsciiFormatter) printKey(name string) {
+	if !f.inArray[len(f.inArray)-1] {
+		fmt.Fprintf(f.line.buffer, `"%s": `, name)
+	} else if f.config.ShowArrayIndex {
+		fmt.Fprintf(f.line.buffer, `%s: `, name)
+	}
+}
+
+func (f *AsciiFormatter) printComma() {
+	f.size[len(f.size)-1]--
+	if f.size[len(f.size)-1] > 0 {
+		f.line.buffer.WriteRune(',')
+	}
+}
+
+func (f *AsciiFormatter) printValue(value interface{}) {
+	switch value.(type) {
+	case string:
+		fmt.Fprintf(f.line.buffer, `"%s"`, value)
+	case nil:
+		f.line.buffer.WriteString("null")
+	default:
+		fmt.Fprintf(f.line.buffer, `%#v`, value)
+	}
+}
+
+func (f *AsciiFormatter) print(a string) {
+	f.line.buffer.WriteString(a)
+}
+
+func (f *AsciiFormatter) printRecursive(name string, value interface{}, marker string) {
+	switch value.(type) {
+	case map[string]interface{}:
+		f.newLine(marker)
+		f.printKey(name)
+		f.print("{")
+		f.closeLine()
+
+		m := value.(map[string]interface{})
+		size := len(m)
+		f.push(name, size, false)
+
+		keys := sortedKeys(m)
+		for _, key := range keys {
+			f.printRecursive(key, m[key], marker)
+		}
+		f.pop()
+
+		f.newLine(marker)
+		f.print("}")
+		f.printComma()
+		f.closeLine()
+
+	case []interface{}:
+		f.newLine(marker)
+		f.printKey(name)
+		f.print("[")
+		f.closeLine()
+
+		s := value.([]interface{})
+		size := len(s)
+		f.push("", size, true)
+		for _, item := range s {
+			f.printRecursive("", item, marker)
+		}
+		f.pop()
+
+		f.newLine(marker)
+		f.print("]")
+		f.printComma()
+		f.closeLine()
+
+	default:
+		f.newLine(marker)
+		f.printKey(name)
+		f.printValue(value)
+		f.printComma()
+		f.closeLine()
+	}
+}
+
+func sortedKeys(m map[string]interface{}) (keys []string) {
+	keys = make([]string, 0, len(m))
+	for key, _ := range m {
+		keys = append(keys, key)
+	}
+	sort.Strings(keys)
+	return
+}
@@ -0,0 +1,124 @@
+package formatter
+
+import (
+	"encoding/json"
+	"errors"
+	"fmt"
+
+	diff "github.com/yudai/gojsondiff"
+)
+
+const (
+	DeltaDelete   = 0
+	DeltaTextDiff = 2
+	DeltaMove     = 3
+)
+
+func NewDeltaFormatter() *DeltaFormatter {
+	return &DeltaFormatter{
+		PrintIndent: true,
+	}
+}
+
+type DeltaFormatter struct {
+	PrintIndent bool
+}
+
+func (f *DeltaFormatter) Format(diff diff.Diff) (result string, err error) {
+	jsonObject, err := f.formatObject(diff.Deltas())
+	if err != nil {
+		return "", err
+	}
+	var resultBytes []byte
+	if f.PrintIndent {
+		resultBytes, err = json.MarshalIndent(jsonObject, "", "  ")
+	} else {
+		resultBytes, err = json.Marshal(jsonObject)
+	}
+	if err != nil {
+		return "", err
+	}
+
+	return string(resultBytes) + "\n", nil
+}
+
+func (f *DeltaFormatter) FormatAsJson(diff diff.Diff) (json map[string]interface{}, err error) {
+	return f.formatObject(diff.Deltas())
+}
+
+func (f *DeltaFormatter) formatObject(deltas []diff.Delta) (deltaJson map[string]interface{}, err error) {
+	deltaJson = map[string]interface{}{}
+	for _, delta := range deltas {
+		switch delta.(type) {
+		case *diff.Object:
+			d := delta.(*diff.Object)
+			deltaJson[d.Position.String()], err = f.formatObject(d.Deltas)
+			if err != nil {
+				return nil, err
+			}
+		case *diff.Array:
+			d := delta.(*diff.Array)
+			deltaJson[d.Position.String()], err = f.formatArray(d.Deltas)
+			if err != nil {
+				return nil, err
+			}
+		case *diff.Added:
+			d := delta.(*diff.Added)
+			deltaJson[d.PostPosition().String()] = []interface{}{d.Value}
+		case *diff.Modified:
+			d := delta.(*diff.Modified)
+			deltaJson[d.PostPosition().String()] = []interface{}{d.OldValue, d.NewValue}
+		case *diff.TextDiff:
+			d := delta.(*diff.TextDiff)
+			deltaJson[d.PostPosition().String()] = []interface{}{d.DiffString(), 0, DeltaTextDiff}
+		case *diff.Deleted:
+			d := delta.(*diff.Deleted)
+			deltaJson[d.PrePosition().String()] = []interface{}{d.Value, 0, DeltaDelete}
+		case *diff.Moved:
+			return nil, errors.New("Delta type 'Move' is not supported in objects")
+		default:
+			return nil, errors.New(fmt.Sprintf("Unknown Delta type detected: %#v", delta))
+		}
+	}
+	return
+}
+
+func (f *DeltaFormatter) formatArray(deltas []diff.Delta) (deltaJson map[string]interface{}, err error) {
+	deltaJson = map[string]interface{}{
+		"_t": "a",
+	}
+	for _, delta := range deltas {
+		switch delta.(type) {
+		case *diff.Object:
+			d := delta.(*diff.Object)
+			deltaJson[d.Position.String()], err = f.formatObject(d.Deltas)
+			if err != nil {
+				return nil, err
+			}
+		case *diff.Array:
+			d := delta.(*diff.Array)
+			deltaJson[d.Position.String()], err = f.formatArray(d.Deltas)
+			if err != nil {
+				return nil, err
+			}
+		case *diff.Added:
+			d := delta.(*diff.Added)
+			deltaJson[d.PostPosition().String()] = []interface{}{d.Value}
+		case *diff.Modified:
+			d := delta.(*diff.Modified)
+			deltaJson[d.PostPosition().String()] = []interface{}{d.OldValue, d.NewValue}
+		case *diff.TextDiff:
+			d := delta.(*diff.TextDiff)
+			deltaJson[d.PostPosition().String()] = []interface{}{d.DiffString(), 0, DeltaTextDiff}
+		case *diff.Deleted:
+			d := delta.(*diff.Deleted)
+			deltaJson["_"+d.PrePosition().String()] = []interface{}{d.Value, 0, DeltaDelete}
+		case *diff.Moved:
+			d := delta.(*diff.Moved)
+			deltaJson["_"+d.PrePosition().String()] = []interface{}{"", d.PostPosition(), DeltaMove}
+		default:
+			return nil, errors.New(fmt.Sprintf("Unknown Delta type detected: %#v", delta))
+		}
+	}
+	return
+}
@@ -0,0 +1,426 @@
+// Package gojsondiff implements "Diff" that compares two JSON objects and
+// generates Deltas that describes differences between them. The package also
+// provides "Patch" that apply Deltas to a JSON object.
+package gojsondiff
+
+import (
+	"container/list"
+	"encoding/json"
+	"reflect"
+	"sort"
+
+	dmp "github.com/sergi/go-diff/diffmatchpatch"
+	"github.com/yudai/golcs"
+)
+
+// A Diff holds deltas generated by a Differ
+type Diff interface {
+	// Deltas returns Deltas that describe differences between two JSON objects
+	Deltas() []Delta
+	// Modified returnes true if Diff has at least one Delta.
+	Modified() bool
+}
+
+type diff struct {
+	deltas []Delta
+}
+
+func (diff *diff) Deltas() []Delta {
+	return diff.deltas
+}
+
+func (diff *diff) Modified() bool {
+	return len(diff.deltas) > 0
+}
+
+// A Differ conmapres JSON objects and apply patches
+type Differ struct {
+	textDiffMinimumLength int
+}
+
+// New returns new Differ with default configuration
+func New() *Differ {
+	return &Differ{
+		textDiffMinimumLength: 30,
+	}
+}
+
+// Compare compares two JSON strings as []bytes and return a Diff object.
+func (differ *Differ) Compare(
+	left []byte,
+	right []byte,
+) (Diff, error) {
+	var leftMap, rightMap map[string]interface{}
+	err := json.Unmarshal(left, &leftMap)
+	if err != nil {
+		return nil, err
+	}
+
+	err = json.Unmarshal(right, &rightMap)
+	if err != nil {
+		return nil, err
+	}
+	return differ.CompareObjects(leftMap, rightMap), nil
+}
+
+// CompareObjects compares two JSON object as map[string]interface{}
+// and return a Diff object.
+func (differ *Differ) CompareObjects(
+	left map[string]interface{},
+	right map[string]interface{},
+) Diff {
+	deltas := differ.compareMaps(left, right)
+	return &diff{deltas: deltas}
+}
+
+// CompareArrays compares two JSON arrays as []interface{}
+// and return a Diff object.
+func (differ *Differ) CompareArrays(
+	left []interface{},
+	right []interface{},
+) Diff {
+	deltas := differ.compareArrays(left, right)
+	return &diff{deltas: deltas}
+}
+
+func (differ *Differ) compareMaps(
+	left map[string]interface{},
+	right map[string]interface{},
+) (deltas []Delta) {
+	deltas = make([]Delta, 0)
+
+	names := sortedKeys(left) // stabilize delta order
+	for _, name := range names {
+		if rightValue, ok := right[name]; ok {
+			same, delta := differ.compareValues(Name(name), left[name], rightValue)
+			if !same {
+				deltas = append(deltas, delta)
+			}
+		} else {
+			deltas = append(deltas, NewDeleted(Name(name), left[name]))
+		}
+	}
+
+	names = sortedKeys(right) // stabilize delta order
+	for _, name := range names {
+		if _, ok := left[name]; !ok {
+			deltas = append(deltas, NewAdded(Name(name), right[name]))
+		}
+	}
+
+	return deltas
+}
+
+// ApplyPatch applies a Diff to an JSON object. This method is destructive.
+func (differ *Differ) ApplyPatch(json map[string]interface{}, patch Diff) {
+	applyDeltas(patch.Deltas(), json)
+}
+
+type maybe struct {
+	index    int
+	lcsIndex int
+	item     interface{}
+}
+
+func (differ *Differ) compareArrays(
+	left []interface{},
+	right []interface{},
+) (deltas []Delta) {
+	deltas = make([]Delta, 0)
+	// LCS index pairs
+	lcsPairs := lcs.New(left, right).IndexPairs()
+
+	// list up items not in LCS, they are maybe deleted
+	maybeDeleted := list.New() // but maybe moved or modified
+	lcsI := 0
+	for i, leftValue := range left {
+		if lcsI < len(lcsPairs) && lcsPairs[lcsI].Left == i {
+			lcsI++
+		} else {
+			maybeDeleted.PushBack(maybe{index: i, lcsIndex: lcsI, item: leftValue})
+		}
+	}
+
+	// list up items not in LCS, they are maybe Added
+	maybeAdded := list.New() // but maybe moved or modified
+	lcsI = 0
+	for i, rightValue := range right {
+		if lcsI < len(lcsPairs) && lcsPairs[lcsI].Right == i {
+			lcsI++
+		} else {
+			maybeAdded.PushBack(maybe{index: i, lcsIndex: lcsI, item: rightValue})
+		}
+	}
+
+	// find moved items
+	var delNext *list.Element // for prefetch to remove item in iteration
+	for delCandidate := maybeDeleted.Front(); delCandidate != nil; delCandidate = delNext {
+		delCan := delCandidate.Value.(maybe)
+		delNext = delCandidate.Next()
+
+		for addCandidate := maybeAdded.Front(); addCandidate != nil; addCandidate = addCandidate.Next() {
+			addCan := addCandidate.Value.(maybe)
+			if reflect.DeepEqual(delCan.item, addCan.item) {
+				deltas = append(deltas, NewMoved(Index(delCan.index), Index(addCan.index), delCan.item, nil))
+				maybeAdded.Remove(addCandidate)
+				maybeDeleted.Remove(delCandidate)
+				break
+			}
+		}
+	}
+
+	// find modified or add+del
+	prevIndexDel := 0
+	prevIndexAdd := 0
+	delElement := maybeDeleted.Front()
+	addElement := maybeAdded.Front()
+	for i := 0; i <= len(lcsPairs); i++ { // not "< len(lcsPairs)"
+		var lcsPair lcs.IndexPair
+		var delSize, addSize int
+		if i < len(lcsPairs) {
+			lcsPair = lcsPairs[i]
+			delSize = lcsPair.Left - prevIndexDel - 1
+			addSize = lcsPair.Right - prevIndexAdd - 1
+			prevIndexDel = lcsPair.Left
+			prevIndexAdd = lcsPair.Right
+		}
+
+		var delSlice []maybe
+		if delSize > 0 {
+			delSlice = make([]maybe, 0, delSize)
+		} else {
+			delSlice = make([]maybe, 0, maybeDeleted.Len())
+		}
+		for ; delElement != nil; delElement = delElement.Next() {
+			d := delElement.Value.(maybe)
+			if d.lcsIndex != i {
+				break
+			}
+			delSlice = append(delSlice, d)
+		}
+
+		var addSlice []maybe
+		if addSize > 0 {
+			addSlice = make([]maybe, 0, addSize)
+		} else {
+			addSlice = make([]maybe, 0, maybeAdded.Len())
+		}
+		for ; addElement != nil; addElement = addElement.Next() {
+			a := addElement.Value.(maybe)
+			if a.lcsIndex != i {
+				break
+			}
+			addSlice = append(addSlice, a)
+		}
+
+		if len(delSlice) > 0 && len(addSlice) > 0 {
+			var bestDeltas []Delta
+			bestDeltas, delSlice, addSlice = differ.maximizeSimilarities(delSlice, addSlice)
+			for _, delta := range bestDeltas {
+				deltas = append(deltas, delta)
+			}
+		}
+
+		for _, del := range delSlice {
+			deltas = append(deltas, NewDeleted(Index(del.index), del.item))
+		}
+		for _, add := range addSlice {
+			deltas = append(deltas, NewAdded(Index(add.index), add.item))
+		}
+	}
+
+	return deltas
+}
+
+func (differ *Differ) compareValues(
+	position Position,
+	left interface{},
+	right interface{},
+) (same bool, delta Delta) {
+	if reflect.TypeOf(left) != reflect.TypeOf(right) {
+		return false, NewModified(position, left, right)
+	}
+
+	switch left.(type) {
+
+	case map[string]interface{}:
+		l := left.(map[string]interface{})
+		childDeltas := differ.compareMaps(l, right.(map[string]interface{}))
+		if len(childDeltas) > 0 {
+			return false, NewObject(position, childDeltas)
+		}
+
+	case []interface{}:
+		l := left.([]interface{})
+		childDeltas := differ.compareArrays(l, right.([]interface{}))
+
+		if len(childDeltas) > 0 {
+			return false, NewArray(position, childDeltas)
+		}
+
+	default:
+		if !reflect.DeepEqual(left, right) {
+
+			if reflect.ValueOf(left).Kind() == reflect.String &&
+				reflect.ValueOf(right).Kind() == reflect.String &&
+				differ.textDiffMinimumLength <= len(left.(string)) {
+
+				textDiff := dmp.New()
+				patchs := textDiff.PatchMake(left.(string), right.(string))
+				return false, NewTextDiff(position, patchs, left, right)
+
+			} else {
+				return false, NewModified(position, left, right)
+			}
+		}
+	}
+
+	return true, nil
+}
+
+func applyDeltas(deltas []Delta, object interface{}) interface{} {
+	preDeltas := make(preDeltas, 0)
+	for _, delta := range deltas {
+		switch delta.(type) {
+		case PreDelta:
+			preDeltas = append(preDeltas, delta.(PreDelta))
+		}
+	}
+	sort.Sort(preDeltas)
+	for _, delta := range preDeltas {
+		object = delta.PreApply(object)
+	}
+
+	postDeltas := make(postDeltas, 0, len(deltas)-len(preDeltas))
+	for _, delta := range deltas {
+		switch delta.(type) {
+		case PostDelta:
+			postDeltas = append(postDeltas, delta.(PostDelta))
+		}
+	}
+	sort.Sort(postDeltas)
+
+	for _, delta := range postDeltas {
+		object = delta.PostApply(object)
+	}
+
+	return object
+}
+
+func (differ *Differ) maximizeSimilarities(left []maybe, right []maybe) (resultDeltas []Delta, freeLeft, freeRight []maybe) {
+	deltaTable := make([][]Delta, len(left))
+	for i := 0; i < len(left); i++ {
+		deltaTable[i] = make([]Delta, len(right))
+	}
+	for i, leftValue := range left {
+		for j, rightValue := range right {
+			_, delta := differ.compareValues(Index(rightValue.index), leftValue.item, rightValue.item)
+			deltaTable[i][j] = delta
+		}
+	}
+
+	sizeX := len(left) + 1 // margins for both sides
+	sizeY := len(right) + 1
+
+	// fill out with similarities
+	dpTable := make([][]float64, sizeX)
+	for i := 0; i < sizeX; i++ {
+		dpTable[i] = make([]float64, sizeY)
+	}
+	for x := sizeX - 2; x >= 0; x-- {
+		for y := sizeY - 2; y >= 0; y-- {
+			prevX := dpTable[x+1][y]
+			prevY := dpTable[x][y+1]
+			score := deltaTable[x][y].Similarity() + dpTable[x+1][y+1]
+
+			dpTable[x][y] = max(prevX, prevY, score)
+		}
+	}
+
+	minLength := len(left)
+	if minLength > len(right) {
+		minLength = len(right)
+	}
+	maxInvalidLength := minLength - 1
+
+	freeLeft = make([]maybe, 0, len(left)-minLength)
+	freeRight = make([]maybe, 0, len(right)-minLength)
+
+	resultDeltas = make([]Delta, 0, minLength)
+	var x, y int
+	for x, y = 0, 0; x <= sizeX-2 && y <= sizeY-2; {
+		current := dpTable[x][y]
+		nextX := dpTable[x+1][y]
+		nextY := dpTable[x][y+1]
+
+		xValidLength := len(left) - maxInvalidLength + y
+		yValidLength := len(right) - maxInvalidLength + x
+
+		if x+1 < xValidLength && current == nextX {
+			freeLeft = append(freeLeft, left[x])
+			x++
+		} else if y+1 < yValidLength && current == nextY {
+			freeRight = append(freeRight, right[y])
+			y++
+		} else {
+			resultDeltas = append(resultDeltas, deltaTable[x][y])
+			x++
+			y++
+		}
+	}
+	for ; x < sizeX-1; x++ {
+		freeLeft = append(freeLeft, left[x-1])
+	}
+	for ; y < sizeY-1; y++ {
+		freeRight = append(freeRight, right[y-1])
+	}
+
+	return resultDeltas, freeLeft, freeRight
+}
+
+func deltasSimilarity(deltas []Delta) (similarity float64) {
+	for _, delta := range deltas {
+		similarity += delta.Similarity()
+	}
+	similarity = similarity / float64(len(deltas))
+	return
+}
+
+func stringSimilarity(left, right string) (similarity float64) {
+	matchingLength := float64(
+		lcs.New(
+			stringToInterfaceSlice(left),
+			stringToInterfaceSlice(right),
+		).Length(),
+	)
+	similarity =
+		(matchingLength / float64(len(left))) * (matchingLength / float64(len(right)))
+	return
+}
+
+func stringToInterfaceSlice(str string) []interface{} {
+	s := make([]interface{}, len(str))
+	for i, v := range str {
+		s[i] = v
+	}
+	return s
+}
+
+func sortedKeys(m map[string]interface{}) (keys []string) {
+	keys = make([]string, 0, len(m))
+	for key, _ := range m {
+		keys = append(keys, key)
+	}
+	sort.Strings(keys)
+	return
+}
+
+func max(first float64, rest ...float64) (max float64) {
+	max = first
+	for _, value := range rest {
+		if max < value {
+			max = value
+		}
+	}
+	return max
+}
@@ -0,0 +1,131 @@
+package gojsondiff
+
+import (
+	"encoding/json"
+	"errors"
+	dmp "github.com/sergi/go-diff/diffmatchpatch"
+	"io"
+	"strconv"
+)
+
+type Unmarshaller struct {
+}
+
+func NewUnmarshaller() *Unmarshaller {
+	return &Unmarshaller{}
+}
+
+func (um *Unmarshaller) UnmarshalBytes(diffBytes []byte) (Diff, error) {
+	var diffObj map[string]interface{}
+	json.Unmarshal(diffBytes, &diffObj)
+	return um.UnmarshalObject(diffObj)
+}
+
+func (um *Unmarshaller) UnmarshalString(diffString string) (Diff, error) {
+	return um.UnmarshalBytes([]byte(diffString))
+}
+
+func (um *Unmarshaller) UnmarshalReader(diffReader io.Reader) (Diff, error) {
+	var diffBytes []byte
+	io.ReadFull(diffReader, diffBytes)
+	return um.UnmarshalBytes(diffBytes)
+}
+
+func (um *Unmarshaller) UnmarshalObject(diffObj map[string]interface{}) (Diff, error) {
+	result, err := process(Name(""), diffObj)
+	if err != nil {
+		return nil, err
+	}
+	return &diff{deltas: result.(*Object).Deltas}, nil
+}
+
+func process(position Position, object interface{}) (Delta, error) {
+	var delta Delta
+	switch object.(type) {
+	case map[string]interface{}:
+		o := object.(map[string]interface{})
+		if isArray, typed := o["_t"]; typed && isArray == "a" {
+			deltas := make([]Delta, 0, len(o))
+			for name, value := range o {
+				if name == "_t" {
+					continue
+				}
+
+				normalizedName := name
+				if normalizedName[0] == '_' {
+					normalizedName = name[1:]
+				}
+				index, err := strconv.Atoi(normalizedName)
+				if err != nil {
+					return nil, err
+				}
+
+				childDelta, err := process(Index(index), value)
+				if err != nil {
+					return nil, err
+				}
+
+				deltas = append(deltas, childDelta)
+			}
+
+			for _, d := range deltas {
+				switch d.(type) {
+				case *Moved:
+					moved := d.(*Moved)
+
+					var dd interface{}
+					var i int
+					for i, dd = range deltas {
+						switch dd.(type) {
+						case *Moved:
+						case PostDelta:
+							pd := dd.(PostDelta)
+							if moved.PostPosition() == pd.PostPosition() {
+								moved.Delta = pd
+								deltas = append(deltas[:i], deltas[i+1:]...)
+							}
+						}
+					}
+				}
+			}
+
+			delta = NewArray(position, deltas)
+		} else {
+			deltas := make([]Delta, 0, len(o))
+			for name, value := range o {
+				childDelta, err := process(Name(name), value)
+				if err != nil {
+					return nil, err
+				}
+				deltas = append(deltas, childDelta)
+			}
+			delta = NewObject(position, deltas)
+		}
+	case []interface{}:
+		o := object.([]interface{})
+		switch len(o) {
+		case 1:
+			delta = NewAdded(position, o[0])
+		case 2:
+			delta = NewModified(position, o[0], o[1])
+		case 3:
+			switch o[2] {
+			case float64(0):
+				delta = NewDeleted(position, o[0])
+			case float64(2):
+				dmp := dmp.New()
+				patches, err := dmp.PatchFromText(o[0].(string))
+				if err != nil {
+					return nil, err
+				}
+				delta = NewTextDiff(position, patches, nil, nil)
+			case float64(3):
+				delta = NewMoved(position, Index(int(o[1].(float64))), nil, nil)
+			default:
+				return nil, errors.New("Unknown delta type")
+			}
+		}
+	}
+
+	return delta, nil
+}
@@ -0,0 +1,8 @@
+box: golang:1.6.3
+
+build:
+  steps:
+    - setup-go-workspace
+    - script:
+        name: test
+        code: make test
@@ -0,0 +1,73 @@
+The MIT License (MIT)
+
+Copyright (c) 2015 Iwasaki Yudai
+
+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.
+
+============================================================================
+
+This repository is build with following third party libraries. Thank you!
+
+
+## ginkgo - https://github.com/onsi/ginkgo
+
+Copyright (c) 2013-2014 Onsi Fakhouri
+
+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.
+
+
+# gomega - https://github.com/onsi/gomega
+
+Copyright (c) 2013-2014 Onsi Fakhouri
+
+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.
@@ -0,0 +1,54 @@
+# Go Longest Common Subsequence (LCS)
+
+A package to calculate [LCS](http://en.wikipedia.org/wiki/Longest_common_subsequence_problem) of slices.
+
+## Usage
+
+```sh
+go get github.com/yudai/golcs
+```
+
+```go
+import " github.com/yudai/golcs"
+
+left = []interface{}{1, 2, 5, 3, 1, 1, 5, 8, 3}
+right = []interface{}{1, 2, 3, 3, 4, 4, 5, 1, 6}
+
+lcs := golcs.New(left, right)
+
+lcs.Values()     // LCS values       => []interface{}{1, 2, 5, 1}
+lcs.IndexPairs() // Matched indices  => [{Left: 0, Right: 0}, {Left: 1, Right: 1}, {Left: 2, Right: 6}, {Left: 4, Right: 7}]
+lcs.Length()     // Matched length   => 4
+
+lcs.Table()      // Memo table
+```
+
+All the methods of `Lcs` cache their return values. For example, the memo table is calculated only once and reused when `Values()`, `Length()` and other methods are called.
+
+
+## FAQ
+
+### How can I give `[]byte` values to `Lcs()` as its arguments?
+
+As `[]interface{}` is incompatible with `[]othertype` like `[]byte`, you need to create a `[]interface{}` slice and copy the values in your `[]byte` slice into it. Unfortunately, Go doesn't provide any mesure to cast a slice into `[]interface{}` with zero cost. Your copy costs O(n).
+
+```go
+leftBytes := []byte("TGAGTA")
+left = make([]interface{}, len(leftBytes))
+for i, v := range leftBytes {
+	left[i] = v
+}
+
+rightBytes := []byte("GATA")
+right = make([]interface{}, len(rightBytes))
+for i, v := range rightBytes {
+	right[i] = v
+}
+
+lcs.New(left, right)
+```
+
+
+## LICENSE
+
+The MIT license (See `LICENSE` for detail)
@@ -0,0 +1,130 @@
+package lcs
+
+import (
+	"reflect"
+)
+
+type Lcs interface {
+	Values() (values []interface{})
+	IndexPairs() (pairs []IndexPair)
+	Length() (length int)
+	Left() (leftValues []interface{})
+	Right() (righttValues []interface{})
+}
+
+type IndexPair struct {
+	Left  int
+	Right int
+}
+
+type lcs struct {
+	left  []interface{}
+	right []interface{}
+	/* for caching */
+	table      [][]int
+	indexPairs []IndexPair
+	values     []interface{}
+}
+
+func New(left, right []interface{}) Lcs {
+	return &lcs{
+		left:       left,
+		right:      right,
+		table:      nil,
+		indexPairs: nil,
+		values:     nil,
+	}
+}
+
+func (lcs *lcs) Table() (table [][]int) {
+	if lcs.table != nil {
+		return lcs.table
+	}
+
+	sizeX := len(lcs.left) + 1
+	sizeY := len(lcs.right) + 1
+
+	table = make([][]int, sizeX)
+	for x := 0; x < sizeX; x++ {
+		table[x] = make([]int, sizeY)
+	}
+
+	for y := 1; y < sizeY; y++ {
+		for x := 1; x < sizeX; x++ {
+			increment := 0
+			if reflect.DeepEqual(lcs.left[x-1], lcs.right[y-1]) {
+				increment = 1
+			}
+			table[x][y] = max(table[x-1][y-1]+increment, table[x-1][y], table[x][y-1])
+		}
+	}
+
+	lcs.table = table
+	return
+}
+
+func (lcs *lcs) Length() (length int) {
+	length = lcs.Table()[len(lcs.left)][len(lcs.right)]
+	return
+}
+
+func (lcs *lcs) IndexPairs() (pairs []IndexPair) {
+	if lcs.indexPairs != nil {
+		return lcs.indexPairs
+	}
+
+	table := lcs.Table()
+	pairs = make([]IndexPair, table[len(table)-1][len(table[0])-1])
+
+	for x, y := len(lcs.left), len(lcs.right); x > 0 && y > 0; {
+		if reflect.DeepEqual(lcs.left[x-1], lcs.right[y-1]) {
+			pairs[table[x][y]-1] = IndexPair{Left: x - 1, Right: y - 1}
+			x--
+			y--
+		} else {
+			if table[x-1][y] >= table[x][y-1] {
+				x--
+			} else {
+				y--
+			}
+		}
+	}
+
+	lcs.indexPairs = pairs
+
+	return
+}
+
+func (lcs *lcs) Values() (values []interface{}) {
+	if lcs.values != nil {
+		return lcs.values
+	}
+
+	pairs := lcs.IndexPairs()
+	values = make([]interface{}, len(pairs))
+	for i, pair := range pairs {
+		values[i] = lcs.left[pair.Left]
+	}
+	lcs.values = values
+	return
+}
+
+func (lcs *lcs) Left() (leftValues []interface{}) {
+	leftValues = lcs.left
+	return
+}
+
+func (lcs *lcs) Right() (rightValues []interface{}) {
+	rightValues = lcs.right
+	return
+}
+
+func max(first int, rest ...int) (max int) {
+	max = first
+	for _, value := range rest {
+		if value > max {
+			max = value
+		}
+	}
+	return
+}