sort: add Find function

For golang/go#50340

Change-Id: I3b4d278affc8e7ec706db8c9777f7a8c8ce7441d
Reviewed-on: https://go-review.googlesource.com/c/go/+/396514
Reviewed-by: Ian Lance Taylor <iant@golang.org>
Trust: Cherry Mui <cherryyz@google.com>
Run-TryBot: Russ Cox <rsc@golang.org>
Auto-Submit: Russ Cox <rsc@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>

3 files changed, 148 insertions, 0 deletions

diff --git a/api/next/50340.txt b/api/next/50340.txt
new file mode 100644
index 0000000000..211392cd25
--- /dev/null
+++ b/api/next/50340.txt
@@ -0,0 +1 @@
+pkg sort, func Find(int, func(int) int) (int, bool) #50340
diff --git a/src/sort/search.go b/src/sort/search.go
index 601557a94b..434349416e 100644
--- a/src/sort/search.go
+++ b/src/sort/search.go
@@ -72,6 +72,47 @@ func Search(n int, f func(int) bool) int {
 	return i
 }
 
+// Find uses binary search to find and return the smallest index i in [0, n)
+// at which cmp(i) <= 0. If there is no such index i, Find returns i = n.
+// The found result is true if i < n and cmp(i) == 0.
+// Find calls cmp(i) only for i in the range [0, n).
+//
+// To permit binary search, Find requires that cmp(i) > 0 for a leading
+// prefix of the range, cmp(i) == 0 in the middle, and cmp(i) < 0 for
+// the final suffix of the range. (Each subrange could be empty.)
+// The usual way to establish this condition is to interpret cmp(i)
+// as a comparison of a desired target value t against entry i in an
+// underlying indexed data structure x, returning <0, 0, and >0
+// when t < x[i], t == x[i], and t > x[i], respectively.
+//
+// For example, to look for a particular string in a sorted, random-access
+// list of strings:
+//    i, found := sort.Find(x.Len(), func(i int) int {
+//        return strings.Compare(target, x.At(i))
+//    })
+//    if found {
+//        fmt.Printf("found %s at entry %d\n", target, i)
+//    } else {
+//        fmt.Printf("%s not found, would insert at %d", target, i)
+//    }
+func Find(n int, cmp func(int) int) (i int, found bool) {
+	// The invariants here are similar to the ones in Search.
+	// Define cmp(-1) > 0 and cmp(n) <= 0
+	// Invariant: cmp(i-1) > 0, cmp(j) <= 0
+	i, j := 0, n
+	for i < j {
+		h := int(uint(i+j) >> 1) // avoid overflow when computing h
+		// i ≤ h < j
+		if cmp(h) > 0 {
+			i = h + 1 // preserves cmp(i-1) > 0
+		} else {
+			j = h // preserves cmp(j) <= 0
+		}
+	}
+	// i == j, cmp(i-1) > 0 and cmp(j) <= 0
+	return i, i < n && cmp(i) == 0
+}
+
 // Convenience wrappers for common cases.
 
 // SearchInts searches for x in a sorted slice of ints and returns the index
diff --git a/src/sort/search_test.go b/src/sort/search_test.go
index f06897ee21..49813eaecb 100644
--- a/src/sort/search_test.go
+++ b/src/sort/search_test.go
@@ -7,6 +7,7 @@ package sort_test
 import (
 	"runtime"
 	. "sort"
+	stringspkg "strings"
 	"testing"
 )
 
@@ -57,6 +58,80 @@ func TestSearch(t *testing.T) {
 	}
 }
 
+func TestFind(t *testing.T) {
+	str1 := []string{"foo"}
+	str2 := []string{"ab", "ca"}
+	str3 := []string{"mo", "qo", "vo"}
+	str4 := []string{"ab", "ad", "ca", "xy"}
+
+	// slice with repeating elements
+	strRepeats := []string{"ba", "ca", "da", "da", "da", "ka", "ma", "ma", "ta"}
+
+	// slice with all element equal
+	strSame := []string{"xx", "xx", "xx"}
+
+	tests := []struct {
+		data      []string
+		target    string
+		wantPos   int
+		wantFound bool
+	}{
+		{[]string{}, "foo", 0, false},
+		{[]string{}, "", 0, false},
+
+		{str1, "foo", 0, true},
+		{str1, "bar", 0, false},
+		{str1, "zx", 1, false},
+
+		{str2, "aa", 0, false},
+		{str2, "ab", 0, true},
+		{str2, "ad", 1, false},
+		{str2, "ca", 1, true},
+		{str2, "ra", 2, false},
+
+		{str3, "bb", 0, false},
+		{str3, "mo", 0, true},
+		{str3, "nb", 1, false},
+		{str3, "qo", 1, true},
+		{str3, "tr", 2, false},
+		{str3, "vo", 2, true},
+		{str3, "xr", 3, false},
+
+		{str4, "aa", 0, false},
+		{str4, "ab", 0, true},
+		{str4, "ac", 1, false},
+		{str4, "ad", 1, true},
+		{str4, "ax", 2, false},
+		{str4, "ca", 2, true},
+		{str4, "cc", 3, false},
+		{str4, "dd", 3, false},
+		{str4, "xy", 3, true},
+		{str4, "zz", 4, false},
+
+		{strRepeats, "da", 2, true},
+		{strRepeats, "db", 5, false},
+		{strRepeats, "ma", 6, true},
+		{strRepeats, "mb", 8, false},
+
+		{strSame, "xx", 0, true},
+		{strSame, "ab", 0, false},
+		{strSame, "zz", 3, false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.target, func(t *testing.T) {
+			cmp := func(i int) int {
+				return stringspkg.Compare(tt.target, tt.data[i])
+			}
+
+			pos, found := Find(len(tt.data), cmp)
+			if pos != tt.wantPos || found != tt.wantFound {
+				t.Errorf("Find got (%v, %v), want (%v, %v)", pos, found, tt.wantPos, tt.wantFound)
+			}
+		})
+	}
+}
+
 // log2 computes the binary logarithm of x, rounded up to the next integer.
 // (log2(0) == 0, log2(1) == 0, log2(2) == 1, log2(3) == 2, etc.)
 func log2(x int) int {
@@ -158,3 +233,34 @@ func TestSearchExhaustive(t *testing.T) {
 		}
 	}
 }
+
+// Abstract exhaustive test for Find.
+func TestFindExhaustive(t *testing.T) {
+	// Test Find for different sequence sizes and search targets.
+	// For each size, we have a (unmaterialized) sequence of integers:
+	//   2,4...size*2
+	// And we're looking for every possible integer between 1 and size*2 + 1.
+	for size := 0; size <= 100; size++ {
+		for x := 1; x <= size*2+1; x++ {
+			var wantFound bool
+			var wantPos int
+
+			cmp := func(i int) int {
+				// Encodes the unmaterialized sequence with elem[i] == (i+1)*2
+				return x - (i+1)*2
+			}
+			pos, found := Find(size, cmp)
+
+			if x%2 == 0 {
+				wantPos = x/2 - 1
+				wantFound = true
+			} else {
+				wantPos = x / 2
+				wantFound = false
+			}
+			if found != wantFound || pos != wantPos {
+				t.Errorf("Find(%d, %d): got (%v, %v), want (%v, %v)", size, x, pos, found, wantPos, wantFound)
+			}
+		}
+	}
+}