Bump google.golang.org/api

This commit is contained in:
Andrea Spacca
2019-03-17 20:19:56 +01:00
parent 8e39b7fa01
commit ec086b4eb3
5432 changed files with 2486664 additions and 231553 deletions

View File

@@ -0,0 +1,17 @@
// Copyright 2017, The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE.md file.
// +build !cmp_debug
package diff
var debug debugger
type debugger struct{}
func (debugger) Begin(_, _ int, f EqualFunc, _, _ *EditScript) EqualFunc {
return f
}
func (debugger) Update() {}
func (debugger) Finish() {}

View File

@@ -0,0 +1,122 @@
// Copyright 2017, The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE.md file.
// +build cmp_debug
package diff
import (
"fmt"
"strings"
"sync"
"time"
)
// The algorithm can be seen running in real-time by enabling debugging:
// go test -tags=cmp_debug -v
//
// Example output:
// === RUN TestDifference/#34
// ┌───────────────────────────────┐
// │ \ · · · · · · · · · · · · · · │
// │ · # · · · · · · · · · · · · · │
// │ · \ · · · · · · · · · · · · · │
// │ · · \ · · · · · · · · · · · · │
// │ · · · X # · · · · · · · · · · │
// │ · · · # \ · · · · · · · · · · │
// │ · · · · · # # · · · · · · · · │
// │ · · · · · # \ · · · · · · · · │
// │ · · · · · · · \ · · · · · · · │
// │ · · · · · · · · \ · · · · · · │
// │ · · · · · · · · · \ · · · · · │
// │ · · · · · · · · · · \ · · # · │
// │ · · · · · · · · · · · \ # # · │
// │ · · · · · · · · · · · # # # · │
// │ · · · · · · · · · · # # # # · │
// │ · · · · · · · · · # # # # # · │
// │ · · · · · · · · · · · · · · \ │
// └───────────────────────────────┘
// [.Y..M.XY......YXYXY.|]
//
// The grid represents the edit-graph where the horizontal axis represents
// list X and the vertical axis represents list Y. The start of the two lists
// is the top-left, while the ends are the bottom-right. The '·' represents
// an unexplored node in the graph. The '\' indicates that the two symbols
// from list X and Y are equal. The 'X' indicates that two symbols are similar
// (but not exactly equal) to each other. The '#' indicates that the two symbols
// are different (and not similar). The algorithm traverses this graph trying to
// make the paths starting in the top-left and the bottom-right connect.
//
// The series of '.', 'X', 'Y', and 'M' characters at the bottom represents
// the currently established path from the forward and reverse searches,
// separated by a '|' character.
const (
updateDelay = 100 * time.Millisecond
finishDelay = 500 * time.Millisecond
ansiTerminal = true // ANSI escape codes used to move terminal cursor
)
var debug debugger
type debugger struct {
sync.Mutex
p1, p2 EditScript
fwdPath, revPath *EditScript
grid []byte
lines int
}
func (dbg *debugger) Begin(nx, ny int, f EqualFunc, p1, p2 *EditScript) EqualFunc {
dbg.Lock()
dbg.fwdPath, dbg.revPath = p1, p2
top := "┌─" + strings.Repeat("──", nx) + "┐\n"
row := "│ " + strings.Repeat("· ", nx) + "│\n"
btm := "└─" + strings.Repeat("──", nx) + "┘\n"
dbg.grid = []byte(top + strings.Repeat(row, ny) + btm)
dbg.lines = strings.Count(dbg.String(), "\n")
fmt.Print(dbg)
// Wrap the EqualFunc so that we can intercept each result.
return func(ix, iy int) (r Result) {
cell := dbg.grid[len(top)+iy*len(row):][len("│ ")+len("· ")*ix:][:len("·")]
for i := range cell {
cell[i] = 0 // Zero out the multiple bytes of UTF-8 middle-dot
}
switch r = f(ix, iy); {
case r.Equal():
cell[0] = '\\'
case r.Similar():
cell[0] = 'X'
default:
cell[0] = '#'
}
return
}
}
func (dbg *debugger) Update() {
dbg.print(updateDelay)
}
func (dbg *debugger) Finish() {
dbg.print(finishDelay)
dbg.Unlock()
}
func (dbg *debugger) String() string {
dbg.p1, dbg.p2 = *dbg.fwdPath, dbg.p2[:0]
for i := len(*dbg.revPath) - 1; i >= 0; i-- {
dbg.p2 = append(dbg.p2, (*dbg.revPath)[i])
}
return fmt.Sprintf("%s[%v|%v]\n\n", dbg.grid, dbg.p1, dbg.p2)
}
func (dbg *debugger) print(d time.Duration) {
if ansiTerminal {
fmt.Printf("\x1b[%dA", dbg.lines) // Reset terminal cursor
}
fmt.Print(dbg)
time.Sleep(d)
}

View File

@@ -0,0 +1,372 @@
// Copyright 2017, The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE.md file.
// Package diff implements an algorithm for producing edit-scripts.
// The edit-script is a sequence of operations needed to transform one list
// of symbols into another (or vice-versa). The edits allowed are insertions,
// deletions, and modifications. The summation of all edits is called the
// Levenshtein distance as this problem is well-known in computer science.
//
// This package prioritizes performance over accuracy. That is, the run time
// is more important than obtaining a minimal Levenshtein distance.
package diff
// EditType represents a single operation within an edit-script.
type EditType uint8
const (
// Identity indicates that a symbol pair is identical in both list X and Y.
Identity EditType = iota
// UniqueX indicates that a symbol only exists in X and not Y.
UniqueX
// UniqueY indicates that a symbol only exists in Y and not X.
UniqueY
// Modified indicates that a symbol pair is a modification of each other.
Modified
)
// EditScript represents the series of differences between two lists.
type EditScript []EditType
// String returns a human-readable string representing the edit-script where
// Identity, UniqueX, UniqueY, and Modified are represented by the
// '.', 'X', 'Y', and 'M' characters, respectively.
func (es EditScript) String() string {
b := make([]byte, len(es))
for i, e := range es {
switch e {
case Identity:
b[i] = '.'
case UniqueX:
b[i] = 'X'
case UniqueY:
b[i] = 'Y'
case Modified:
b[i] = 'M'
default:
panic("invalid edit-type")
}
}
return string(b)
}
// stats returns a histogram of the number of each type of edit operation.
func (es EditScript) stats() (s struct{ NI, NX, NY, NM int }) {
for _, e := range es {
switch e {
case Identity:
s.NI++
case UniqueX:
s.NX++
case UniqueY:
s.NY++
case Modified:
s.NM++
default:
panic("invalid edit-type")
}
}
return
}
// Dist is the Levenshtein distance and is guaranteed to be 0 if and only if
// lists X and Y are equal.
func (es EditScript) Dist() int { return len(es) - es.stats().NI }
// LenX is the length of the X list.
func (es EditScript) LenX() int { return len(es) - es.stats().NY }
// LenY is the length of the Y list.
func (es EditScript) LenY() int { return len(es) - es.stats().NX }
// EqualFunc reports whether the symbols at indexes ix and iy are equal.
// When called by Difference, the index is guaranteed to be within nx and ny.
type EqualFunc func(ix int, iy int) Result
// Result is the result of comparison.
// NumSame is the number of sub-elements that are equal.
// NumDiff is the number of sub-elements that are not equal.
type Result struct{ NumSame, NumDiff int }
// BoolResult returns a Result that is either Equal or not Equal.
func BoolResult(b bool) Result {
if b {
return Result{NumSame: 1} // Equal, Similar
} else {
return Result{NumDiff: 2} // Not Equal, not Similar
}
}
// Equal indicates whether the symbols are equal. Two symbols are equal
// if and only if NumDiff == 0. If Equal, then they are also Similar.
func (r Result) Equal() bool { return r.NumDiff == 0 }
// Similar indicates whether two symbols are similar and may be represented
// by using the Modified type. As a special case, we consider binary comparisons
// (i.e., those that return Result{1, 0} or Result{0, 1}) to be similar.
//
// The exact ratio of NumSame to NumDiff to determine similarity may change.
func (r Result) Similar() bool {
// Use NumSame+1 to offset NumSame so that binary comparisons are similar.
return r.NumSame+1 >= r.NumDiff
}
// Difference reports whether two lists of lengths nx and ny are equal
// given the definition of equality provided as f.
//
// This function returns an edit-script, which is a sequence of operations
// needed to convert one list into the other. The following invariants for
// the edit-script are maintained:
// • eq == (es.Dist()==0)
// • nx == es.LenX()
// • ny == es.LenY()
//
// This algorithm is not guaranteed to be an optimal solution (i.e., one that
// produces an edit-script with a minimal Levenshtein distance). This algorithm
// favors performance over optimality. The exact output is not guaranteed to
// be stable and may change over time.
func Difference(nx, ny int, f EqualFunc) (es EditScript) {
// This algorithm is based on traversing what is known as an "edit-graph".
// See Figure 1 from "An O(ND) Difference Algorithm and Its Variations"
// by Eugene W. Myers. Since D can be as large as N itself, this is
// effectively O(N^2). Unlike the algorithm from that paper, we are not
// interested in the optimal path, but at least some "decent" path.
//
// For example, let X and Y be lists of symbols:
// X = [A B C A B B A]
// Y = [C B A B A C]
//
// The edit-graph can be drawn as the following:
// A B C A B B A
// ┌─────────────┐
// C │_|_|\|_|_|_|_│ 0
// B │_|\|_|_|\|\|_│ 1
// A │\|_|_|\|_|_|\│ 2
// B │_|\|_|_|\|\|_│ 3
// A │\|_|_|\|_|_|\│ 4
// C │ | |\| | | | │ 5
// └─────────────┘ 6
// 0 1 2 3 4 5 6 7
//
// List X is written along the horizontal axis, while list Y is written
// along the vertical axis. At any point on this grid, if the symbol in
// list X matches the corresponding symbol in list Y, then a '\' is drawn.
// The goal of any minimal edit-script algorithm is to find a path from the
// top-left corner to the bottom-right corner, while traveling through the
// fewest horizontal or vertical edges.
// A horizontal edge is equivalent to inserting a symbol from list X.
// A vertical edge is equivalent to inserting a symbol from list Y.
// A diagonal edge is equivalent to a matching symbol between both X and Y.
// Invariants:
// • 0 ≤ fwdPath.X ≤ (fwdFrontier.X, revFrontier.X) ≤ revPath.X ≤ nx
// • 0 ≤ fwdPath.Y ≤ (fwdFrontier.Y, revFrontier.Y) ≤ revPath.Y ≤ ny
//
// In general:
// • fwdFrontier.X < revFrontier.X
// • fwdFrontier.Y < revFrontier.Y
// Unless, it is time for the algorithm to terminate.
fwdPath := path{+1, point{0, 0}, make(EditScript, 0, (nx+ny)/2)}
revPath := path{-1, point{nx, ny}, make(EditScript, 0)}
fwdFrontier := fwdPath.point // Forward search frontier
revFrontier := revPath.point // Reverse search frontier
// Search budget bounds the cost of searching for better paths.
// The longest sequence of non-matching symbols that can be tolerated is
// approximately the square-root of the search budget.
searchBudget := 4 * (nx + ny) // O(n)
// The algorithm below is a greedy, meet-in-the-middle algorithm for
// computing sub-optimal edit-scripts between two lists.
//
// The algorithm is approximately as follows:
// • Searching for differences switches back-and-forth between
// a search that starts at the beginning (the top-left corner), and
// a search that starts at the end (the bottom-right corner). The goal of
// the search is connect with the search from the opposite corner.
// • As we search, we build a path in a greedy manner, where the first
// match seen is added to the path (this is sub-optimal, but provides a
// decent result in practice). When matches are found, we try the next pair
// of symbols in the lists and follow all matches as far as possible.
// • When searching for matches, we search along a diagonal going through
// through the "frontier" point. If no matches are found, we advance the
// frontier towards the opposite corner.
// • This algorithm terminates when either the X coordinates or the
// Y coordinates of the forward and reverse frontier points ever intersect.
//
// This algorithm is correct even if searching only in the forward direction
// or in the reverse direction. We do both because it is commonly observed
// that two lists commonly differ because elements were added to the front
// or end of the other list.
//
// Running the tests with the "cmp_debug" build tag prints a visualization
// of the algorithm running in real-time. This is educational for
// understanding how the algorithm works. See debug_enable.go.
f = debug.Begin(nx, ny, f, &fwdPath.es, &revPath.es)
for {
// Forward search from the beginning.
if fwdFrontier.X >= revFrontier.X || fwdFrontier.Y >= revFrontier.Y || searchBudget == 0 {
break
}
for stop1, stop2, i := false, false, 0; !(stop1 && stop2) && searchBudget > 0; i++ {
// Search in a diagonal pattern for a match.
z := zigzag(i)
p := point{fwdFrontier.X + z, fwdFrontier.Y - z}
switch {
case p.X >= revPath.X || p.Y < fwdPath.Y:
stop1 = true // Hit top-right corner
case p.Y >= revPath.Y || p.X < fwdPath.X:
stop2 = true // Hit bottom-left corner
case f(p.X, p.Y).Equal():
// Match found, so connect the path to this point.
fwdPath.connect(p, f)
fwdPath.append(Identity)
// Follow sequence of matches as far as possible.
for fwdPath.X < revPath.X && fwdPath.Y < revPath.Y {
if !f(fwdPath.X, fwdPath.Y).Equal() {
break
}
fwdPath.append(Identity)
}
fwdFrontier = fwdPath.point
stop1, stop2 = true, true
default:
searchBudget-- // Match not found
}
debug.Update()
}
// Advance the frontier towards reverse point.
if revPath.X-fwdFrontier.X >= revPath.Y-fwdFrontier.Y {
fwdFrontier.X++
} else {
fwdFrontier.Y++
}
// Reverse search from the end.
if fwdFrontier.X >= revFrontier.X || fwdFrontier.Y >= revFrontier.Y || searchBudget == 0 {
break
}
for stop1, stop2, i := false, false, 0; !(stop1 && stop2) && searchBudget > 0; i++ {
// Search in a diagonal pattern for a match.
z := zigzag(i)
p := point{revFrontier.X - z, revFrontier.Y + z}
switch {
case fwdPath.X >= p.X || revPath.Y < p.Y:
stop1 = true // Hit bottom-left corner
case fwdPath.Y >= p.Y || revPath.X < p.X:
stop2 = true // Hit top-right corner
case f(p.X-1, p.Y-1).Equal():
// Match found, so connect the path to this point.
revPath.connect(p, f)
revPath.append(Identity)
// Follow sequence of matches as far as possible.
for fwdPath.X < revPath.X && fwdPath.Y < revPath.Y {
if !f(revPath.X-1, revPath.Y-1).Equal() {
break
}
revPath.append(Identity)
}
revFrontier = revPath.point
stop1, stop2 = true, true
default:
searchBudget-- // Match not found
}
debug.Update()
}
// Advance the frontier towards forward point.
if revFrontier.X-fwdPath.X >= revFrontier.Y-fwdPath.Y {
revFrontier.X--
} else {
revFrontier.Y--
}
}
// Join the forward and reverse paths and then append the reverse path.
fwdPath.connect(revPath.point, f)
for i := len(revPath.es) - 1; i >= 0; i-- {
t := revPath.es[i]
revPath.es = revPath.es[:i]
fwdPath.append(t)
}
debug.Finish()
return fwdPath.es
}
type path struct {
dir int // +1 if forward, -1 if reverse
point // Leading point of the EditScript path
es EditScript
}
// connect appends any necessary Identity, Modified, UniqueX, or UniqueY types
// to the edit-script to connect p.point to dst.
func (p *path) connect(dst point, f EqualFunc) {
if p.dir > 0 {
// Connect in forward direction.
for dst.X > p.X && dst.Y > p.Y {
switch r := f(p.X, p.Y); {
case r.Equal():
p.append(Identity)
case r.Similar():
p.append(Modified)
case dst.X-p.X >= dst.Y-p.Y:
p.append(UniqueX)
default:
p.append(UniqueY)
}
}
for dst.X > p.X {
p.append(UniqueX)
}
for dst.Y > p.Y {
p.append(UniqueY)
}
} else {
// Connect in reverse direction.
for p.X > dst.X && p.Y > dst.Y {
switch r := f(p.X-1, p.Y-1); {
case r.Equal():
p.append(Identity)
case r.Similar():
p.append(Modified)
case p.Y-dst.Y >= p.X-dst.X:
p.append(UniqueY)
default:
p.append(UniqueX)
}
}
for p.X > dst.X {
p.append(UniqueX)
}
for p.Y > dst.Y {
p.append(UniqueY)
}
}
}
func (p *path) append(t EditType) {
p.es = append(p.es, t)
switch t {
case Identity, Modified:
p.add(p.dir, p.dir)
case UniqueX:
p.add(p.dir, 0)
case UniqueY:
p.add(0, p.dir)
}
debug.Update()
}
type point struct{ X, Y int }
func (p *point) add(dx, dy int) { p.X += dx; p.Y += dy }
// zigzag maps a consecutive sequence of integers to a zig-zag sequence.
// [0 1 2 3 4 5 ...] => [0 -1 +1 -2 +2 ...]
func zigzag(x int) int {
if x&1 != 0 {
x = ^x
}
return x >> 1
}

View File

@@ -0,0 +1,444 @@
// Copyright 2017, The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE.md file.
package diff
import (
"fmt"
"math/rand"
"strings"
"testing"
"unicode"
)
func TestDifference(t *testing.T) {
tests := []struct {
// Before passing x and y to Difference, we strip all spaces so that
// they can be used by the test author to indicate a missing symbol
// in one of the lists.
x, y string
want string
}{{
x: "",
y: "",
want: "",
}, {
x: "#",
y: "#",
want: ".",
}, {
x: "##",
y: "# ",
want: ".X",
}, {
x: "a#",
y: "A ",
want: "MX",
}, {
x: "#a",
y: " A",
want: "XM",
}, {
x: "# ",
y: "##",
want: ".Y",
}, {
x: " #",
y: "@#",
want: "Y.",
}, {
x: "@#",
y: " #",
want: "X.",
}, {
x: "##########0123456789",
y: " 0123456789",
want: "XXXXXXXXXX..........",
}, {
x: " 0123456789",
y: "##########0123456789",
want: "YYYYYYYYYY..........",
}, {
x: "#####0123456789#####",
y: " 0123456789 ",
want: "XXXXX..........XXXXX",
}, {
x: " 0123456789 ",
y: "#####0123456789#####",
want: "YYYYY..........YYYYY",
}, {
x: "01234##########56789",
y: "01234 56789",
want: ".....XXXXXXXXXX.....",
}, {
x: "01234 56789",
y: "01234##########56789",
want: ".....YYYYYYYYYY.....",
}, {
x: "0123456789##########",
y: "0123456789 ",
want: "..........XXXXXXXXXX",
}, {
x: "0123456789 ",
y: "0123456789##########",
want: "..........YYYYYYYYYY",
}, {
x: "abcdefghij0123456789",
y: "ABCDEFGHIJ0123456789",
want: "MMMMMMMMMM..........",
}, {
x: "ABCDEFGHIJ0123456789",
y: "abcdefghij0123456789",
want: "MMMMMMMMMM..........",
}, {
x: "01234abcdefghij56789",
y: "01234ABCDEFGHIJ56789",
want: ".....MMMMMMMMMM.....",
}, {
x: "01234ABCDEFGHIJ56789",
y: "01234abcdefghij56789",
want: ".....MMMMMMMMMM.....",
}, {
x: "0123456789abcdefghij",
y: "0123456789ABCDEFGHIJ",
want: "..........MMMMMMMMMM",
}, {
x: "0123456789ABCDEFGHIJ",
y: "0123456789abcdefghij",
want: "..........MMMMMMMMMM",
}, {
x: "ABCDEFGHIJ0123456789 ",
y: " 0123456789abcdefghij",
want: "XXXXXXXXXX..........YYYYYYYYYY",
}, {
x: " 0123456789abcdefghij",
y: "ABCDEFGHIJ0123456789 ",
want: "YYYYYYYYYY..........XXXXXXXXXX",
}, {
x: "ABCDE0123456789 FGHIJ",
y: " 0123456789abcdefghij",
want: "XXXXX..........YYYYYMMMMM",
}, {
x: " 0123456789abcdefghij",
y: "ABCDE0123456789 FGHIJ",
want: "YYYYY..........XXXXXMMMMM",
}, {
x: "ABCDE01234F G H I J 56789 ",
y: " 01234 a b c d e56789fghij",
want: "XXXXX.....XYXYXYXYXY.....YYYYY",
}, {
x: " 01234a b c d e 56789fghij",
y: "ABCDE01234 F G H I J56789 ",
want: "YYYYY.....XYXYXYXYXY.....XXXXX",
}, {
x: "FGHIJ01234ABCDE56789 ",
y: " 01234abcde56789fghij",
want: "XXXXX.....MMMMM.....YYYYY",
}, {
x: " 01234abcde56789fghij",
y: "FGHIJ01234ABCDE56789 ",
want: "YYYYY.....MMMMM.....XXXXX",
}, {
x: "ABCAB BA ",
y: " C BABAC",
want: "XX.X.Y..Y",
}, {
x: "# #### ###",
y: "#y####yy###",
want: ".Y....YY...",
}, {
x: "# #### # ##x#x",
y: "#y####y y## # ",
want: ".Y....YXY..X.X",
}, {
x: "###z#z###### x #",
y: "#y##Z#Z###### yy#",
want: ".Y..M.M......XYY.",
}, {
x: "0 12z3x 456789 x x 0",
y: "0y12Z3 y456789y y y0",
want: ".Y..M.XY......YXYXY.",
}, {
x: "0 2 4 6 8 ..................abXXcdEXF.ghXi",
y: " 1 3 5 7 9..................AB CDE F.GH I",
want: "XYXYXYXYXY..................MMXXMM.X..MMXM",
}, {
x: "I HG.F EDC BA..................9 7 5 3 1 ",
y: "iXhg.FXEdcXXba.................. 8 6 4 2 0",
want: "MYMM..Y.MMYYMM..................XYXYXYXYXY",
}, {
x: "x1234",
y: " 1234",
want: "X....",
}, {
x: "x123x4",
y: " 123 4",
want: "X...X.",
}, {
x: "x1234x56",
y: " 1234 ",
want: "X....XXX",
}, {
x: "x1234xxx56",
y: " 1234 56",
want: "X....XXX..",
}, {
x: ".1234...ab",
y: " 1234 AB",
want: "X....XXXMM",
}, {
x: "x1234xxab.",
y: " 1234 AB ",
want: "X....XXMMX",
}, {
x: " 0123456789",
y: "9012345678 ",
want: "Y.........X",
}, {
x: " 0123456789",
y: "8901234567 ",
want: "YY........XX",
}, {
x: " 0123456789",
y: "7890123456 ",
want: "YYY.......XXX",
}, {
x: " 0123456789",
y: "6789012345 ",
want: "YYYY......XXXX",
}, {
x: "0123456789 ",
y: " 5678901234",
want: "XXXXX.....YYYYY",
}, {
x: "0123456789 ",
y: " 4567890123",
want: "XXXX......YYYY",
}, {
x: "0123456789 ",
y: " 3456789012",
want: "XXX.......YYY",
}, {
x: "0123456789 ",
y: " 2345678901",
want: "XX........YY",
}, {
x: "0123456789 ",
y: " 1234567890",
want: "X.........Y",
}, {
x: "0 1 2 3 45 6 7 8 9 ",
y: " 9 8 7 6 54 3 2 1 0",
want: "XYXYXYXYX.YXYXYXYXY",
}, {
x: "0 1 2345678 9 ",
y: " 6 72 5 819034",
want: "XYXY.XX.XX.Y.YYY",
}, {
x: "F B Q M O I G T L N72X90 E 4S P 651HKRJU DA 83CVZW",
y: " 5 W H XO10R9IV K ZLCTAJ8P3N SEQM4 7 2G6 UBD F ",
want: "XYXYXYXY.YYYY.YXYXY.YYYYYYY.XXXXXY.YY.XYXYY.XXXXXX.Y.XYXXXXXX",
}}
for _, tt := range tests {
t.Run("", func(t *testing.T) {
x := strings.Replace(tt.x, " ", "", -1)
y := strings.Replace(tt.y, " ", "", -1)
es := testStrings(t, x, y)
if got := es.String(); got != tt.want {
t.Errorf("Difference(%s, %s):\ngot %s\nwant %s", x, y, got, tt.want)
}
})
}
}
func TestDifferenceFuzz(t *testing.T) {
tests := []struct{ px, py, pm float32 }{
{px: 0.0, py: 0.0, pm: 0.1},
{px: 0.0, py: 0.1, pm: 0.0},
{px: 0.1, py: 0.0, pm: 0.0},
{px: 0.0, py: 0.1, pm: 0.1},
{px: 0.1, py: 0.0, pm: 0.1},
{px: 0.2, py: 0.2, pm: 0.2},
{px: 0.3, py: 0.1, pm: 0.2},
{px: 0.1, py: 0.3, pm: 0.2},
{px: 0.2, py: 0.2, pm: 0.2},
{px: 0.3, py: 0.3, pm: 0.3},
{px: 0.1, py: 0.1, pm: 0.5},
{px: 0.4, py: 0.1, pm: 0.5},
{px: 0.3, py: 0.2, pm: 0.5},
{px: 0.2, py: 0.3, pm: 0.5},
{px: 0.1, py: 0.4, pm: 0.5},
}
for i, tt := range tests {
t.Run(fmt.Sprintf("P%d", i), func(t *testing.T) {
// Sweep from 1B to 1KiB.
for n := 1; n <= 1024; n <<= 1 {
t.Run(fmt.Sprintf("N%d", n), func(t *testing.T) {
for j := 0; j < 10; j++ {
x, y := generateStrings(n, tt.px, tt.py, tt.pm, int64(j))
testStrings(t, x, y)
}
})
}
})
}
}
func BenchmarkDifference(b *testing.B) {
for n := 1 << 10; n <= 1<<20; n <<= 2 {
b.Run(fmt.Sprintf("N%d", n), func(b *testing.B) {
x, y := generateStrings(n, 0.05, 0.05, 0.10, 0)
b.ReportAllocs()
b.SetBytes(int64(len(x) + len(y)))
for i := 0; i < b.N; i++ {
Difference(len(x), len(y), func(ix, iy int) Result {
return compareByte(x[ix], y[iy])
})
}
})
}
}
func generateStrings(n int, px, py, pm float32, seed int64) (string, string) {
if px+py+pm > 1.0 {
panic("invalid probabilities")
}
py += px
pm += py
b := make([]byte, n)
r := rand.New(rand.NewSource(seed))
r.Read(b)
var x, y []byte
for len(b) > 0 {
switch p := r.Float32(); {
case p < px: // UniqueX
x = append(x, b[0])
case p < py: // UniqueY
y = append(y, b[0])
case p < pm: // Modified
x = append(x, 'A'+(b[0]%26))
y = append(y, 'a'+(b[0]%26))
default: // Identity
x = append(x, b[0])
y = append(y, b[0])
}
b = b[1:]
}
return string(x), string(y)
}
func testStrings(t *testing.T, x, y string) EditScript {
es := Difference(len(x), len(y), func(ix, iy int) Result {
return compareByte(x[ix], y[iy])
})
if es.LenX() != len(x) {
t.Errorf("es.LenX = %d, want %d", es.LenX(), len(x))
}
if es.LenY() != len(y) {
t.Errorf("es.LenY = %d, want %d", es.LenY(), len(y))
}
if !validateScript(x, y, es) {
t.Errorf("invalid edit script: %v", es)
}
return es
}
func validateScript(x, y string, es EditScript) bool {
var bx, by []byte
for _, e := range es {
switch e {
case Identity:
if !compareByte(x[len(bx)], y[len(by)]).Equal() {
return false
}
bx = append(bx, x[len(bx)])
by = append(by, y[len(by)])
case UniqueX:
bx = append(bx, x[len(bx)])
case UniqueY:
by = append(by, y[len(by)])
case Modified:
if !compareByte(x[len(bx)], y[len(by)]).Similar() {
return false
}
bx = append(bx, x[len(bx)])
by = append(by, y[len(by)])
}
}
return string(bx) == x && string(by) == y
}
// compareByte returns a Result where the result is Equal if x == y,
// similar if x and y differ only in casing, and different otherwise.
func compareByte(x, y byte) (r Result) {
switch {
case x == y:
return equalResult // Identity
case unicode.ToUpper(rune(x)) == unicode.ToUpper(rune(y)):
return similarResult // Modified
default:
return differentResult // UniqueX or UniqueY
}
}
var (
equalResult = Result{NumDiff: 0}
similarResult = Result{NumDiff: 1}
differentResult = Result{NumDiff: 2}
)
func TestResult(t *testing.T) {
tests := []struct {
result Result
wantEqual bool
wantSimilar bool
}{
// equalResult is equal since NumDiff == 0, by definition of Equal method.
{equalResult, true, true},
// similarResult is similar since it is a binary result where only one
// element was compared (i.e., Either NumSame==1 or NumDiff==1).
{similarResult, false, true},
// differentResult is different since there are enough differences that
// it isn't even considered similar.
{differentResult, false, false},
// Zero value is always equal.
{Result{NumSame: 0, NumDiff: 0}, true, true},
// Binary comparisons (where NumSame+NumDiff == 1) are always similar.
{Result{NumSame: 1, NumDiff: 0}, true, true},
{Result{NumSame: 0, NumDiff: 1}, false, true},
// More complex ratios. The exact ratio for similarity may change,
// and may require updates to these test cases.
{Result{NumSame: 1, NumDiff: 1}, false, true},
{Result{NumSame: 1, NumDiff: 2}, false, true},
{Result{NumSame: 1, NumDiff: 3}, false, false},
{Result{NumSame: 2, NumDiff: 1}, false, true},
{Result{NumSame: 2, NumDiff: 2}, false, true},
{Result{NumSame: 2, NumDiff: 3}, false, true},
{Result{NumSame: 3, NumDiff: 1}, false, true},
{Result{NumSame: 3, NumDiff: 2}, false, true},
{Result{NumSame: 3, NumDiff: 3}, false, true},
{Result{NumSame: 1000, NumDiff: 0}, true, true},
{Result{NumSame: 1000, NumDiff: 1}, false, true},
{Result{NumSame: 1000, NumDiff: 2}, false, true},
{Result{NumSame: 0, NumDiff: 1000}, false, false},
{Result{NumSame: 1, NumDiff: 1000}, false, false},
{Result{NumSame: 2, NumDiff: 1000}, false, false},
}
for _, tt := range tests {
if got := tt.result.Equal(); got != tt.wantEqual {
t.Errorf("%#v.Equal() = %v, want %v", tt.result, got, tt.wantEqual)
}
if got := tt.result.Similar(); got != tt.wantSimilar {
t.Errorf("%#v.Similar() = %v, want %v", tt.result, got, tt.wantSimilar)
}
}
}