The third attempt to implement unionRunRunInPlace

This commit is contained in:
Kuba Podgórski 2020-06-18 15:27:31 +02:00
parent eddbb7d0b0
commit 443f2d8f7c
2 changed files with 333 additions and 545 deletions

View file

@ -4726,116 +4726,186 @@ func unionRunRunInPlace(a, b *Container) *Container {
return a
}
// unionInterval16InPlace merges two slice of intervals in place (in a).
// The main concept is to go value by value (instead of interval by interval)
// and count `.start` and `.last` points.
// If we get the `state == 0` it means we just built a new interval (`val`),
// and we can set it in `a` at the possition `off`
func unionInterval16InPlace(a, b []interval16) ([]interval16, int32) {
n := int32(0)
an, bn := len(a), len(b)
i, j, k := 0, 0, 0
var (
// ai - index of a, aii - subindex (0: a[ai].start, 1: a[ai].last).
ai, aii int = 0, 0
// bi - index of b, bii - subindex (0: b[bi].start, 1: b[bi].last).
bi, bii int = 0, 0
// Offset of a - next available index to set.
off int = 0
// Value to set/append to a at off
val interval16
// Current state - state equals 0 means we are clear (out of intervals)
// When we start a new interval we add +1 when we get out of interval we add -1.
state int
// subindex (ii) to state mapping
// .start: [0] -> 1
// .last: [1] -> -1
iiMap = [2]int{1, -1}
// If fromB is equal 2 it means that both val.start and val.last come from b,
// so we need to extend a, first
fromB int8
// eval functions evaluates global state and value
eval = func(arr [2]uint16, ii int, onlyB bool) {
if state == 0 && ii == 0 {
// we are clear and start a new interval
val.start = arr[ii]
if onlyB {
fromB++
}
}
state += iiMap[ii]
if state == 0 {
// we just got out of interval
// ii == 1
val.last = arr[ii]
if onlyB {
fromB++
}
}
}
// eval2 function is a special variant for eval function
// it's only used when two interval endings are equal, e.g.:
// a: ------------------|
// b: -----------|
// the most important part is to change the global for both endings
// before we check if we're getting out of interval and start the new one.
eval2 = func(arr [2]uint16, i1, i2 int) {
if state == 0 && (i1 == 0 || i2 == 0) {
// we are clear and start a new interval
val.start = arr[i1]
}
state += iiMap[i1]
state += iiMap[i2]
if state == 0 {
// (i1 == 1 || i2 == 1)
// we just got out of interval
val.last = arr[i1]
}
}
)
for {
var (
v interval16
local *interval16
)
// av, bv reflects a[ai] and b[bi] intervals as an array,
// so we can internally iterate over values (points).
var av, bv [2]uint16
if i < an && j < bn {
if a[i].start <= b[j].start {
v = a[i]
i++
} else {
v = b[j]
j++
if ai < an && bi < bn {
av[0], av[1] = a[ai].start, a[ai].last
bv[0], bv[1] = b[bi].start, b[bi].last
if av[aii] < bv[bii] {
// a: |-------------------
// b: |-------------------
eval(av, aii, false)
aii++
} else if av[aii] == bv[bii] {
// a: |-------------------
// b: |-------------------
// or
// a: ------------------|
// b: |-------------------
// or
// a: ------------------|
// b: |------------|
// ...
eval2(av, aii, bii)
aii++
bii++
} else { // bv[bii] < av[aii]
// a: |-------------------
// b: |-------------------
eval(bv, bii, true)
bii++
}
} else if ai < an { // only a left
av[0], av[1] = a[ai].start, a[ai].last
eval(av, aii, false)
aii++
} else if bi < bn { // only b left
bv[0], bv[1] = b[bi].start, b[bi].last
eval(bv, bii, false)
bii++
} else {
if i < an {
v = a[i]
i++
} else if j < bn {
v = b[j]
j++
} else {
break
break
}
if state == 0 {
if fromB == 2 {
// val.start and val.last come from b, so we need to extend a, first
a = append(a, interval16{})
copy(a[off+1:], a[off:])
ai++
an++
}
fromB = 0
a, off = appendInterval16At(a, val, off)
n += int32(val.last-val.start) + 1
}
reeval:
if k == an {
// append what's left and try to optimize it later
a = append(a, v)
an++
continue
if aii == 2 {
// move to the next a's interval
aii = 0
ai++
}
// locally unioned interval
if local == nil {
local = &a[k]
}
if v.last < local.start {
// [---- v ----][---- local ----]
// prepend v into a
a = append(a, interval16{})
copy(a[i+1:], a[i:])
a[i] = v
an++
i++
continue
}
if v.start > local.last {
// [---- local ----][---- v ----]
// check if we can optimize run and combine neighbours
if int32(v.start-local.last) <= 1 {
local.last = v.last
continue
}
// set already unioned interval and reevaluate v
a[k] = *local
k++
n += int32(local.last-local.start) + 1
local = &v
goto reeval
}
if v.start < local.start {
if v.last >= local.start && v.last <= local.last {
// [---- local ----]
// [---- v ----]
local.start = v.start
} else if v.last > local.last {
// [- local -]
// [---- v ----]
*local = v
}
} else if v.start <= local.last {
if v.last <= local.last {
// [---- local ----]
// [-- v --]
// this assignment overwrites the first available position to set - a[k]
// it's done to "swallow" many nested intervals, e.g.:
// [a1] [a2] [a3] [a4]
// [------ b ------]
// will give us:
// [ ----- a1 -----] [a4]
a[k] = *local
} else if v.last > local.last {
// [---- local ----]
// [---- v ----]
local.last = v.last
}
if bii == 2 {
// move to the next b's interval
bii = 0
bi++
}
}
if len(a) > 0 {
// count the latest interval
n += int32(a[k].last-a[k].start) + 1
a = a[:k+1]
a = a[:off]
}
return a, n
}
// appendInterval16At appends or sets val in a at off position
// The function returns modified a ([]interval16) and new offset (off)
func appendInterval16At(a []interval16, val interval16, off int) ([]interval16, int) {
if off > 0 && int32(val.start)-int32(a[off-1].last) <= 1 {
a[off-1].last = val.last
return a, off
}
if off == len(a) {
a = append(a, val)
off++
return a, off
}
a[off] = val
off++
return a, off
}
func difference(a, b *Container) *Container {
if a.N() == 0 || b.N() == maxContainerVal+1 {
return nil

View file

@ -722,467 +722,185 @@ func TestUnionMixed(t *testing.T) {
}
func TestUnionInterval16InPlace(t *testing.T) {
t.Run("single overlap", func(t *testing.T) {
a := []interval16{interval16{1, 10}, interval16{21, 28}}
b := []interval16{interval16{8, 12}}
bb := make([]interval16, len(b))
copy(bb, b)
tests := []struct {
name string
a []interval16
b []interval16
expected []interval16
expectedN int32
}{
{
name: "firstBitUnset lastBitSet",
a: []interval16{interval16{1, 10}},
b: []interval16{interval16{10, 10}},
expected: []interval16{interval16{1, 10}},
expectedN: 10,
},
{
name: "single overlap",
a: []interval16{interval16{1, 10}, interval16{21, 28}},
b: []interval16{interval16{8, 12}},
expected: []interval16{interval16{1, 12}, interval16{21, 28}},
expectedN: 20,
},
{
name: "nested intervals",
a: []interval16{interval16{3, 13}, interval16{17, 20}},
b: []interval16{interval16{1, 4}, interval16{6, 7}, interval16{8, 9}, interval16{10, 11}, interval16{14, 17}},
expected: []interval16{interval16{1, 20}},
expectedN: 20,
},
{
name: "no overlap",
a: []interval16{interval16{3, 4}, interval16{7, 8}},
b: []interval16{interval16{1, 2}, interval16{5, 6}, interval16{9, 10}},
expected: []interval16{interval16{1, 10}},
expectedN: 10,
},
{
name: "b in a",
a: []interval16{interval16{1, 10}},
b: []interval16{interval16{5, 7}},
expected: []interval16{interval16{1, 10}},
expectedN: 10,
},
{
name: "a eq b",
a: []interval16{interval16{1, 10}},
b: []interval16{interval16{1, 10}},
expected: []interval16{interval16{1, 10}},
expectedN: 10,
},
{
name: "a in b",
a: []interval16{interval16{5, 7}},
b: []interval16{interval16{1, 10}},
expected: []interval16{interval16{1, 10}},
expectedN: 10,
},
{
name: "a ahead b",
a: []interval16{interval16{1, 2}, interval16{3, 4}, interval16{5, 7}},
b: []interval16{interval16{10, 11}, interval16{12, 13}, interval16{14, 15}},
expected: []interval16{interval16{1, 7}, interval16{10, 15}},
expectedN: 13,
},
{
name: "b ahead a",
a: []interval16{interval16{10, 11}, interval16{12, 13}, interval16{14, 15}},
b: []interval16{interval16{1, 2}, interval16{3, 4}, interval16{5, 7}},
expected: []interval16{interval16{1, 7}, interval16{10, 15}},
expectedN: 13,
},
{
name: "empty a and b",
a: []interval16{},
b: []interval16{},
expected: []interval16{},
expectedN: 0,
},
{
name: "empty a",
a: []interval16{},
b: []interval16{interval16{1, 2}, interval16{3, 4}, interval16{5, 7}},
expected: []interval16{interval16{1, 7}},
expectedN: 7,
},
{
name: "empty b",
a: []interval16{interval16{1, 2}, interval16{3, 4}, interval16{5, 7}},
b: []interval16{},
expected: []interval16{interval16{1, 7}},
expectedN: 7,
},
{
name: "single a",
a: []interval16{interval16{1, 2}},
b: []interval16{},
expected: []interval16{interval16{1, 2}},
expectedN: 2,
},
{
name: "single b",
a: []interval16{},
b: []interval16{interval16{1, 2}},
expected: []interval16{interval16{1, 2}},
expectedN: 2,
},
{
name: "single a single b",
a: []interval16{interval16{3, 4}},
b: []interval16{interval16{1, 2}},
expected: []interval16{interval16{1, 4}},
expectedN: 4,
},
{
name: "oddBitsSet lastBitUnset",
a: []interval16{interval16{1, 1}, interval16{3, 3}, interval16{5, 5}},
b: []interval16{interval16{0, 4}},
expected: []interval16{interval16{0, 5}},
expectedN: 6,
},
{
name: "all bits",
a: []interval16{interval16{1, 1}, interval16{3, 3}, interval16{5, 5}},
b: []interval16{interval16{0, 0}, interval16{2, 2}, interval16{4, 4}},
expected: []interval16{interval16{0, 5}},
expectedN: 6,
},
{
name: "short a long b",
a: []interval16{interval16{5, 5}, interval16{7, 7}, interval16{9, 10}, interval16{12, 12}, interval16{15, 17}, interval16{19, 20}},
b: []interval16{interval16{1, 10}, interval16{12, 12}, interval16{14, 18}},
expected: []interval16{interval16{1, 10}, interval16{12, 12}, interval16{14, 20}},
expectedN: 18,
},
{
name: "common endings",
a: []interval16{interval16{1, 5}, interval16{15, 20}, interval16{25, 35}},
b: []interval16{interval16{1, 10}, interval16{15, 20}, interval16{30, 35}},
expected: []interval16{interval16{1, 10}, interval16{15, 20}, interval16{25, 35}},
expectedN: 27,
},
{
name: "common endings and overlap",
a: []interval16{interval16{1, 5}, interval16{10, 15}},
b: []interval16{interval16{5, 10}, interval16{12, 17}},
expected: []interval16{interval16{1, 17}},
expectedN: 17,
},
{
name: "no common endings and overlap",
a: []interval16{interval16{5, 10}, interval16{12, 17}},
b: []interval16{interval16{0, 11}, interval16{15, 20}},
expected: []interval16{interval16{0, 20}},
expectedN: 21,
},
}
runs, n := unionInterval16InPlace(a, b)
for _, tc := range tests {
t.Run(tc.name, func(t *testing.T) {
bb := make([]interval16, len(tc.b))
copy(bb, tc.b)
expected := []interval16{interval16{1, 12}, interval16{21, 28}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
runs, n := unionInterval16InPlace(tc.a, tc.b)
for i, v := range tc.expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", tc.expected, runs)
}
}
}
const expectedN = 20
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
if n != tc.expectedN {
t.Fatalf("N expected: %d, got: %d", tc.expectedN, n)
}
}
})
t.Run("nested intervals", func(t *testing.T) {
a := []interval16{interval16{3, 13}, interval16{17, 20}}
b := []interval16{interval16{1, 4}, interval16{6, 7}, interval16{8, 9}, interval16{10, 11}, interval16{14, 17}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 20}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
for i, v := range bb {
if tc.b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, tc.b)
}
}
}
const expectedN = 20
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("no overlap", func(t *testing.T) {
a := []interval16{interval16{3, 4}, interval16{7, 8}}
b := []interval16{interval16{1, 2}, interval16{5, 6}, interval16{9, 10}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 10}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 10
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("b in a", func(t *testing.T) {
a := []interval16{interval16{1, 10}}
b := []interval16{interval16{5, 7}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 10}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 10
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("a eq b", func(t *testing.T) {
a := []interval16{interval16{1, 10}}
b := []interval16{interval16{1, 10}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 10}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 10
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("a in b", func(t *testing.T) {
a := []interval16{interval16{5, 7}}
b := []interval16{interval16{1, 10}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 10}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 10
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("a ahead b", func(t *testing.T) {
a := []interval16{interval16{1, 2}, interval16{3, 4}, interval16{5, 7}}
b := []interval16{interval16{10, 11}, interval16{12, 13}, interval16{14, 15}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{
interval16{1, 7},
interval16{10, 15},
}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 13
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("b ahead a", func(t *testing.T) {
a := []interval16{interval16{10, 11}, interval16{12, 13}, interval16{14, 15}}
b := []interval16{interval16{1, 2}, interval16{3, 4}, interval16{5, 7}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{
interval16{1, 7},
interval16{10, 15},
}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 13
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("empty a and b", func(t *testing.T) {
a := []interval16{}
b := []interval16{}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 0
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("empty a", func(t *testing.T) {
a := []interval16{}
b := []interval16{interval16{1, 2}, interval16{3, 4}, interval16{5, 7}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 7}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 7
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("empty b", func(t *testing.T) {
a := []interval16{interval16{1, 2}, interval16{3, 4}, interval16{5, 7}}
b := []interval16{}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 7}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 7
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("single a", func(t *testing.T) {
a := []interval16{interval16{1, 2}}
b := []interval16{}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 2}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 2
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("single b", func(t *testing.T) {
a := []interval16{}
b := []interval16{interval16{1, 2}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 2}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 2
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("single a single b", func(t *testing.T) {
a := []interval16{interval16{3, 4}}
b := []interval16{interval16{1, 2}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 4}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 4
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("oddBitsSet lastBitUnset", func(t *testing.T) {
a := []interval16{interval16{1, 1}, interval16{3, 3}, interval16{5, 5}}
b := []interval16{interval16{0, 4}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{0, 5}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 6
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("all bits", func(t *testing.T) {
a := []interval16{interval16{1, 1}, interval16{3, 3}, interval16{5, 5}}
b := []interval16{interval16{0, 0}, interval16{2, 2}, interval16{4, 4}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{0, 5}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 6
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
t.Run("short a long b", func(t *testing.T) {
a := []interval16{interval16{5, 5}, interval16{7, 7}, interval16{9, 10}, interval16{12, 12}, interval16{15, 17}, interval16{19, 20}}
b := []interval16{interval16{1, 10}, interval16{12, 12}, interval16{14, 18}}
bb := make([]interval16, len(b))
copy(bb, b)
runs, n := unionInterval16InPlace(a, b)
expected := []interval16{interval16{1, 10}, interval16{12, 12}, interval16{14, 20}}
for i, v := range expected {
if runs[i] != v {
t.Fatalf("runs expected: %+v, got: %+v", expected, runs)
}
}
const expectedN = 18
if n != expectedN {
t.Fatalf("N expected: %d, got: %d", expectedN, n)
}
for i, v := range bb {
if b[i] != v {
t.Fatalf("b changed - runs expected: %+v, got: %+v", bb, b)
}
}
})
})
}
}
func TestIntersectMixed(t *testing.T) {