Merge pull request 'Add DedupeStrings to remove repeated values from a string slice' (#2) from clawbot/util:proposal-dedupe-strings into master

Reviewed-on: sneak/util#2

atomicwrite.go

@@ -0,0 +1,56 @@
+package util
+
+import (
+	"os"
+	"path/filepath"
+)
+
+// AtomicWriteFile writes data to path without ever leaving a half-written file
+// behind. It writes to a temporary file in the same directory, flushes it to
+// disk, gives it the mode in perm and renames it over path, so anything reading
+// path sees either the previous contents or the complete new contents and
+// never something in between. If any step fails, the temporary file is removed
+// and path is left as it was.
+//
+// The temporary file has to live in the same directory as path, because a
+// rename across two filesystems is not possible and would not be instant if it
+// were.
+//
+// Two details differ from os.WriteFile. The mode is applied after the file is
+// created, so perm is what the finished file ends up with rather than perm with
+// the process umask taken out of it. And the directory holding the file is not
+// flushed, so a machine that loses power immediately after this returns may
+// come back with the rename undone, even though the data itself was written.
+func AtomicWriteFile(path string, data []byte, perm os.FileMode) error {
+	directory := filepath.Dir(path)
+
+	temporary, err := os.CreateTemp(directory, "."+filepath.Base(path)+".tmp")
+	if err != nil {
+		return err
+	}
+	temporaryPath := temporary.Name()
+
+	// Does nothing once the rename below has succeeded, because by then the
+	// temporary name no longer refers to anything.
+	defer os.Remove(temporaryPath)
+
+	if _, err := temporary.Write(data); err != nil {
+		temporary.Close()
+		return err
+	}
+
+	if err := temporary.Sync(); err != nil {
+		temporary.Close()
+		return err
+	}
+
+	if err := temporary.Close(); err != nil {
+		return err
+	}
+
+	if err := os.Chmod(temporaryPath, perm); err != nil {
+		return err
+	}
+
+	return os.Rename(temporaryPath, path)
+}

atomicwrite_test.go

@@ -0,0 +1,110 @@
+package util
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func TestAtomicWriteFile(t *testing.T) {
+	tests := []struct {
+		name     string
+		existing string
+		data     []byte
+		perm     os.FileMode
+	}{
+		{"a new file", "", []byte("hello"), 0644},
+		{"replacing a shorter file", "old", []byte("a much longer set of contents"), 0644},
+		{"replacing a longer file", "a much longer set of contents", []byte("new"), 0644},
+		{"an empty payload", "something", []byte{}, 0644},
+		{"a nil payload", "something", nil, 0644},
+		{"a private mode", "", []byte("secret"), 0600},
+		{"an executable mode", "", []byte("#!/bin/sh\n"), 0755},
+		{"binary contents", "", []byte{0x00, 0xff, 0x10, 0x00}, 0644},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			directory := t.TempDir()
+			path := filepath.Join(directory, "target")
+
+			if test.existing != "" {
+				if err := os.WriteFile(path, []byte(test.existing), 0666); err != nil {
+					t.Fatalf("could not write the file to be replaced: %v", err)
+				}
+			}
+
+			if err := AtomicWriteFile(path, test.data, test.perm); err != nil {
+				t.Fatalf("did not expect an error, got %v", err)
+			}
+
+			written, err := os.ReadFile(path)
+			if err != nil {
+				t.Fatalf("could not read the file back: %v", err)
+			}
+			if string(written) != string(test.data) {
+				t.Errorf("expected contents %q got %q", string(test.data), string(written))
+			}
+
+			info, err := os.Stat(path)
+			if err != nil {
+				t.Fatalf("could not look at the file: %v", err)
+			}
+			if info.Mode().Perm() != test.perm {
+				t.Errorf("expected mode %v got %v", test.perm, info.Mode().Perm())
+			}
+
+			entries, err := os.ReadDir(directory)
+			if err != nil {
+				t.Fatalf("could not list the directory: %v", err)
+			}
+			if len(entries) != 1 {
+				names := make([]string, 0, len(entries))
+				for _, entry := range entries {
+					names = append(names, entry.Name())
+				}
+				t.Errorf("expected only the target file to be left, got %v", names)
+			}
+		})
+	}
+}
+
+func TestAtomicWriteFileReportsAMissingDirectory(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "not-there", "target")
+
+	if err := AtomicWriteFile(path, []byte("hello"), 0644); err == nil {
+		t.Errorf("expected an error for a directory that does not exist")
+	}
+}
+
+func TestAtomicWriteFileLeavesTheOldFileAloneOnFailure(t *testing.T) {
+	directory := t.TempDir()
+	path := filepath.Join(directory, "target")
+
+	if err := os.WriteFile(path, []byte("original"), 0644); err != nil {
+		t.Fatalf("could not write the file to be replaced: %v", err)
+	}
+
+	// A directory that cannot be written to means the temporary file cannot
+	// be created, so the write has to fail before anything is replaced.
+	if err := os.Chmod(directory, 0500); err != nil {
+		t.Fatalf("could not change the directory mode: %v", err)
+	}
+	defer os.Chmod(directory, 0700)
+
+	if os.Geteuid() == 0 {
+		t.Skip("running as root, which ignores the directory mode")
+	}
+
+	if err := AtomicWriteFile(path, []byte("replacement"), 0644); err == nil {
+		t.Fatalf("expected an error for a directory that cannot be written to")
+	}
+
+	existing, err := os.ReadFile(path)
+	if err != nil {
+		t.Fatalf("could not read the file back: %v", err)
+	}
+	if string(existing) != "original" {
+		t.Errorf("expected the original contents to survive, got %q", string(existing))
+	}
+}

chunk.go

@@ -0,0 +1,26 @@
+package util
+
+// ChunkStrings splits input into consecutive slices of at most size elements.
+// The last slice holds whatever is left over and may be shorter than the
+// others. Each returned slice is a copy, so appending to one of them cannot
+// disturb another. A size below one, or an empty input, returns nil.
+func ChunkStrings(input []string, size int) [][]string {
+	if size < 1 || len(input) == 0 {
+		return nil
+	}
+
+	out := make([][]string, 0, (len(input)+size-1)/size)
+
+	for start := 0; start < len(input); start += size {
+		end := start + size
+		if end > len(input) {
+			end = len(input)
+		}
+
+		chunk := make([]string, end-start)
+		copy(chunk, input[start:end])
+		out = append(out, chunk)
+	}
+
+	return out
+}

chunk_test.go

@@ -0,0 +1,43 @@
+package util
+
+import (
+	"reflect"
+	"testing"
+)
+
+func TestChunkStrings(t *testing.T) {
+	tests := []struct {
+		name     string
+		input    []string
+		size     int
+		expected [][]string
+	}{
+		{"exact multiple", []string{"a", "b", "c", "d"}, 2, [][]string{{"a", "b"}, {"c", "d"}}},
+		{"with a remainder", []string{"a", "b", "c"}, 2, [][]string{{"a", "b"}, {"c"}}},
+		{"size larger than input", []string{"a", "b"}, 5, [][]string{{"a", "b"}}},
+		{"size of one", []string{"a", "b"}, 1, [][]string{{"a"}, {"b"}}},
+		{"size of zero", []string{"a", "b"}, 0, nil},
+		{"negative size", []string{"a", "b"}, -1, nil},
+		{"empty input", []string{}, 2, nil},
+		{"nil input", nil, 2, nil},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			got := ChunkStrings(test.input, test.size)
+			if !reflect.DeepEqual(got, test.expected) {
+				t.Errorf("expected %#v got %#v", test.expected, got)
+			}
+		})
+	}
+}
+
+func TestChunkStringsReturnsCopies(t *testing.T) {
+	input := []string{"a", "b", "c", "d"}
+	chunks := ChunkStrings(input, 2)
+	chunks[0][0] = "changed"
+
+	if input[0] != "a" {
+		t.Errorf("expected the input to be unchanged, got %#v", input)
+	}
+}

dedupe.go

@@ -0,0 +1,23 @@
+package util
+
+// DedupeStrings returns a new slice holding each value of input once, in the
+// order in which each value first appears. The input slice is not modified.
+// A nil input returns nil.
+func DedupeStrings(input []string) []string {
+	if input == nil {
+		return nil
+	}
+
+	seen := make(map[string]struct{}, len(input))
+	out := make([]string, 0, len(input))
+
+	for _, value := range input {
+		if _, ok := seen[value]; ok {
+			continue
+		}
+		seen[value] = struct{}{}
+		out = append(out, value)
+	}
+
+	return out
+}

dedupe_test.go

@@ -0,0 +1,40 @@
+package util
+
+import (
+	"reflect"
+	"testing"
+)
+
+func TestDedupeStrings(t *testing.T) {
+	tests := []struct {
+		name     string
+		input    []string
+		expected []string
+	}{
+		{"nil stays nil", nil, nil},
+		{"empty slice", []string{}, []string{}},
+		{"nothing repeated", []string{"a", "b", "c"}, []string{"a", "b", "c"}},
+		{"all one value", []string{"a", "a", "a"}, []string{"a"}},
+		{"repeats next to each other", []string{"a", "a", "b", "b"}, []string{"a", "b"}},
+		{"repeats far apart", []string{"a", "b", "a", "c", "b"}, []string{"a", "b", "c"}},
+		{"empty string is a value", []string{"", "a", ""}, []string{"", "a"}},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			got := DedupeStrings(test.input)
+			if !reflect.DeepEqual(got, test.expected) {
+				t.Errorf("expected %#v got %#v", test.expected, got)
+			}
+		})
+	}
+}
+
+func TestDedupeStringsLeavesInputAlone(t *testing.T) {
+	input := []string{"b", "a", "b"}
+	DedupeStrings(input)
+
+	if !reflect.DeepEqual(input, []string{"b", "a", "b"}) {
+		t.Errorf("expected the input to be unchanged, got %#v", input)
+	}
+}

exists.go

@@ -0,0 +1,31 @@
+package util
+
+import "os"
+
+// FileExists reports whether there is a regular file at path. A directory, a
+// device node or a socket gives false, and so does anything the process is not
+// allowed to look at, since from the caller's point of view there is no usable
+// file there either way.
+//
+// Symbolic links are followed, so a link pointing at a regular file gives true
+// and a link pointing at nothing gives false.
+func FileExists(path string) bool {
+	info, err := os.Stat(path)
+	if err != nil {
+		return false
+	}
+
+	return info.Mode().IsRegular()
+}
+
+// DirExists reports whether there is a directory at path. As with FileExists,
+// anything that cannot be looked at gives false, and symbolic links are
+// followed.
+func DirExists(path string) bool {
+	info, err := os.Stat(path)
+	if err != nil {
+		return false
+	}
+
+	return info.IsDir()
+}

exists_test.go

@@ -0,0 +1,57 @@
+package util
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func TestFileExistsAndDirExists(t *testing.T) {
+	root := t.TempDir()
+
+	file := filepath.Join(root, "a-file")
+	if err := os.WriteFile(file, []byte("contents"), 0644); err != nil {
+		t.Fatalf("could not write the test file: %v", err)
+	}
+
+	directory := filepath.Join(root, "a-directory")
+	if err := os.Mkdir(directory, 0755); err != nil {
+		t.Fatalf("could not make the test directory: %v", err)
+	}
+
+	linkToFile := filepath.Join(root, "link-to-file")
+	if err := os.Symlink(file, linkToFile); err != nil {
+		t.Fatalf("could not make the test link: %v", err)
+	}
+
+	brokenLink := filepath.Join(root, "broken-link")
+	if err := os.Symlink(filepath.Join(root, "nothing-here"), brokenLink); err != nil {
+		t.Fatalf("could not make the broken test link: %v", err)
+	}
+
+	tests := []struct {
+		name          string
+		path          string
+		expectedFile  bool
+		expectedIsDir bool
+	}{
+		{"a regular file", file, true, false},
+		{"a directory", directory, false, true},
+		{"a link to a file", linkToFile, true, false},
+		{"a link to nothing", brokenLink, false, false},
+		{"a path that is not there", filepath.Join(root, "missing"), false, false},
+		{"a path whose parent is not there", filepath.Join(root, "missing", "deeper"), false, false},
+		{"the empty path", "", false, false},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			if got := FileExists(test.path); got != test.expectedFile {
+				t.Errorf("FileExists: expected %v got %v", test.expectedFile, got)
+			}
+			if got := DirExists(test.path); got != test.expectedIsDir {
+				t.Errorf("DirExists: expected %v got %v", test.expectedIsDir, got)
+			}
+		})
+	}
+}

humanbytes.go

@@ -0,0 +1,38 @@
+package util
+
+import (
+	"fmt"
+	"math"
+)
+
+const bytesPerUnit = 1024
+
+var byteUnits = []string{"KiB", "MiB", "GiB", "TiB", "PiB", "EiB"}
+
+// HumanBytes writes a byte count the way a person would read it, using powers
+// of 1024: "0 B", "1023 B", "1.0 KiB", "1.5 MiB", "16.0 EiB". Counts below 1024
+// are given as whole bytes, and anything larger gets one digit after the
+// decimal point.
+func HumanBytes(bytes uint64) string {
+	if bytes < bytesPerUnit {
+		return fmt.Sprintf("%d B", bytes)
+	}
+
+	value := float64(bytes)
+	unit := -1
+
+	for value >= bytesPerUnit && unit < len(byteUnits)-1 {
+		value /= bytesPerUnit
+		unit++
+	}
+
+	// Rounding happens after the unit has been chosen, so a count just short
+	// of the next threshold would otherwise come out as "1024.0 KiB" rather
+	// than "1.0 MiB".
+	if unit < len(byteUnits)-1 && math.Round(value*10)/10 >= bytesPerUnit {
+		value /= bytesPerUnit
+		unit++
+	}
+
+	return fmt.Sprintf("%.1f %s", value, byteUnits[unit])
+}

humanbytes_test.go

@@ -0,0 +1,33 @@
+package util
+
+import "testing"
+
+func TestHumanBytes(t *testing.T) {
+	tests := []struct {
+		name     string
+		input    uint64
+		expected string
+	}{
+		{"zero", 0, "0 B"},
+		{"one byte", 1, "1 B"},
+		{"just below a kibibyte", 1023, "1023 B"},
+		{"exactly a kibibyte", 1024, "1.0 KiB"},
+		{"half a kibibyte more", 1536, "1.5 KiB"},
+		{"rounds up into the next unit", 1048575, "1.0 MiB"},
+		{"exactly a mebibyte", 1 << 20, "1.0 MiB"},
+		{"a gibibyte", 1 << 30, "1.0 GiB"},
+		{"a tebibyte", 1 << 40, "1.0 TiB"},
+		{"a pebibyte", 1 << 50, "1.0 PiB"},
+		{"an exbibyte", 1 << 60, "1.0 EiB"},
+		{"the largest count there is", ^uint64(0), "16.0 EiB"},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			got := HumanBytes(test.input)
+			if got != test.expected {
+				t.Errorf("expected %q got %q", test.expected, got)
+			}
+		})
+	}
+}

random.go

@@ -0,0 +1,28 @@
+package util
+
+import (
+	"crypto/rand"
+	"encoding/hex"
+	"errors"
+)
+
+// RandomHexString returns byteLength random bytes from the operating system's
+// random source, written as lowercase hexadecimal, so the returned string is
+// twice as long as byteLength. The bytes come from crypto/rand, which means the
+// result is fit for session tokens, temporary filenames and anything else a
+// stranger should not be able to guess.
+//
+// An error comes back only for a negative length or if the random source
+// itself fails, which does not happen on a working system.
+func RandomHexString(byteLength int) (string, error) {
+	if byteLength < 0 {
+		return "", errors.New("byte length cannot be negative")
+	}
+
+	buffer := make([]byte, byteLength)
+	if _, err := rand.Read(buffer); err != nil {
+		return "", err
+	}
+
+	return hex.EncodeToString(buffer), nil
+}

random_test.go

@@ -0,0 +1,57 @@
+package util
+
+import (
+	"encoding/hex"
+	"testing"
+)
+
+func TestRandomHexStringLength(t *testing.T) {
+	tests := []struct {
+		name           string
+		byteLength     int
+		expectedLength int
+	}{
+		{"no bytes at all", 0, 0},
+		{"one byte", 1, 2},
+		{"eight bytes", 8, 16},
+		{"sixteen bytes", 16, 32},
+		{"thirty-two bytes", 32, 64},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			got, err := RandomHexString(test.byteLength)
+			if err != nil {
+				t.Fatalf("did not expect an error, got %v", err)
+			}
+			if len(got) != test.expectedLength {
+				t.Errorf("expected a string of %d characters, got %d (%q)", test.expectedLength, len(got), got)
+			}
+			if _, err := hex.DecodeString(got); err != nil {
+				t.Errorf("expected valid hexadecimal, got %q: %v", got, err)
+			}
+		})
+	}
+}
+
+func TestRandomHexStringRejectsNegativeLength(t *testing.T) {
+	got, err := RandomHexString(-1)
+	if err == nil {
+		t.Errorf("expected an error for a negative length, got %q", got)
+	}
+}
+
+func TestRandomHexStringDiffersBetweenCalls(t *testing.T) {
+	seen := make(map[string]struct{})
+
+	for i := 0; i < 100; i++ {
+		value, err := RandomHexString(16)
+		if err != nil {
+			t.Fatalf("did not expect an error, got %v", err)
+		}
+		if _, repeated := seen[value]; repeated {
+			t.Fatalf("got the same string twice: %q", value)
+		}
+		seen[value] = struct{}{}
+	}
+}

sha256file.go

@@ -0,0 +1,30 @@
+package util
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"io"
+	"os"
+)
+
+// SHA256File returns the SHA-256 digest of the contents of the file at path,
+// written as lowercase hexadecimal, in the same form as the sha256sum command
+// prints. The file is copied through the hash in chunks rather than read into
+// memory, so its size does not matter.
+//
+// The error is whatever went wrong opening or reading the file, passed along
+// unchanged so a caller can test it with os.IsNotExist and the like.
+func SHA256File(path string) (string, error) {
+	file, err := os.Open(path)
+	if err != nil {
+		return "", err
+	}
+	defer file.Close()
+
+	digest := sha256.New()
+	if _, err := io.Copy(digest, file); err != nil {
+		return "", err
+	}
+
+	return hex.EncodeToString(digest.Sum(nil)), nil
+}

sha256file_test.go

@@ -0,0 +1,76 @@
+package util
+
+import (
+	"bytes"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+func TestSHA256File(t *testing.T) {
+	tests := []struct {
+		name     string
+		contents []byte
+		expected string
+	}{
+		{
+			"an empty file",
+			[]byte{},
+			"e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+		},
+		{
+			"a short file",
+			[]byte("hello world\n"),
+			"a948904f2f0f479b8f8197694b30184b0d2ed1c1cd2a1ec0fb85d299a192a447",
+		},
+		{
+			"binary contents",
+			[]byte{0x00, 0x01, 0x02, 0x03},
+			"054edec1d0211f624fed0cbca9d4f9400b0e491c43742af2c5b0abebf0c990d8",
+		},
+		{
+			"larger than one read buffer",
+			bytes.Repeat([]byte("a"), 100000),
+			"6d1cf22d7cc09b085dfc25ee1a1f3ae0265804c607bc2074ad253bcc82fd81ee",
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			path := filepath.Join(t.TempDir(), "file")
+			if err := os.WriteFile(path, test.contents, 0644); err != nil {
+				t.Fatalf("could not write the test file: %v", err)
+			}
+
+			got, err := SHA256File(path)
+			if err != nil {
+				t.Fatalf("did not expect an error, got %v", err)
+			}
+			if got != test.expected {
+				t.Errorf("expected %q got %q", test.expected, got)
+			}
+			if got != strings.ToLower(got) {
+				t.Errorf("expected lowercase hexadecimal, got %q", got)
+			}
+		})
+	}
+}
+
+func TestSHA256FileReportsAMissingFile(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "not-there")
+
+	got, err := SHA256File(path)
+	if err == nil {
+		t.Fatalf("expected an error for a file that does not exist, got %q", got)
+	}
+	if !os.IsNotExist(err) {
+		t.Errorf("expected an error that os.IsNotExist recognises, got %v", err)
+	}
+}
+
+func TestSHA256FileReportsADirectory(t *testing.T) {
+	if _, err := SHA256File(t.TempDir()); err == nil {
+		t.Errorf("expected an error for a directory")
+	}
+}

slugify.go

@@ -0,0 +1,36 @@
+package util
+
+import "strings"
+
+// Slugify returns input as lowercase ASCII letters, digits and hyphens, which
+// is safe to put in a URL path, a filename or a fragment identifier. Every run
+// of other characters becomes a single hyphen, and the result never starts or
+// ends with one.
+//
+// Characters outside ASCII are treated as separators rather than being folded
+// to their nearest ASCII relative, because doing that properly needs a Unicode
+// table that this package does not carry. Text with no ASCII letters or digits
+// in it therefore slugifies to an empty string.
+func Slugify(input string) string {
+	var out strings.Builder
+	pendingHyphen := false
+
+	for _, character := range strings.ToLower(input) {
+		if (character >= 'a' && character <= 'z') || (character >= '0' && character <= '9') {
+			if pendingHyphen {
+				out.WriteByte('-')
+				pendingHyphen = false
+			}
+			out.WriteRune(character)
+			continue
+		}
+
+		// Remember that a separator was seen, but only write a hyphen once
+		// another usable character turns up, so nothing trails off the end.
+		if out.Len() > 0 {
+			pendingHyphen = true
+		}
+	}
+
+	return out.String()
+}

slugify_test.go

@@ -0,0 +1,34 @@
+package util
+
+import "testing"
+
+func TestSlugify(t *testing.T) {
+	tests := []struct {
+		name     string
+		input    string
+		expected string
+	}{
+		{"already a slug", "hello-world", "hello-world"},
+		{"mixed case", "Hello World", "hello-world"},
+		{"punctuation", "What's new?", "what-s-new"},
+		{"run of separators", "one   ---   two", "one-two"},
+		{"leading separators", "///one", "one"},
+		{"trailing separators", "one///", "one"},
+		{"separators at both ends", "  --one two--  ", "one-two"},
+		{"digits are kept", "Go 1.14 release", "go-1-14-release"},
+		{"underscores are separators", "some_name_here", "some-name-here"},
+		{"nothing usable", "!!! ???", ""},
+		{"empty string", "", ""},
+		{"outside ascii", "Grüße, Welt", "gr-e-welt"},
+		{"only characters outside ascii", "日本語", ""},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			got := Slugify(test.input)
+			if got != test.expected {
+				t.Errorf("expected %q got %q", test.expected, got)
+			}
+		})
+	}
+}

truncate.go

@@ -0,0 +1,21 @@
+package util
+
+// TruncateString returns at most max characters from the start of s. It counts
+// characters rather than bytes, so a multi-byte character is never cut in half
+// and the result is always valid text. A max of zero or less returns an empty
+// string, and a string already at or below the limit is returned unchanged.
+func TruncateString(s string, max int) string {
+	if max <= 0 {
+		return ""
+	}
+
+	count := 0
+	for offset := range s {
+		if count == max {
+			return s[:offset]
+		}
+		count++
+	}
+
+	return s
+}

truncate_test.go

@@ -0,0 +1,41 @@
+package util
+
+import "testing"
+
+func TestTruncateString(t *testing.T) {
+	tests := []struct {
+		name     string
+		input    string
+		max      int
+		expected string
+	}{
+		{"shorter than the limit", "abc", 5, "abc"},
+		{"exactly at the limit", "abc", 3, "abc"},
+		{"longer than the limit", "abcdef", 3, "abc"},
+		{"max of one", "abc", 1, "a"},
+		{"max of zero", "abc", 0, ""},
+		{"negative max", "abc", -1, ""},
+		{"empty string", "", 3, ""},
+		{"multi-byte characters kept whole", "héllo", 3, "hél"},
+		{"multi-byte characters below the limit", "héllo", 99, "héllo"},
+		{"characters outside the basic plane", "a😀b", 2, "a😀"},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			got := TruncateString(test.input, test.max)
+			if got != test.expected {
+				t.Errorf("expected %q got %q", test.expected, got)
+			}
+		})
+	}
+}
+
+func TestTruncateStringCountsCharactersNotBytes(t *testing.T) {
+	// Each of these characters is two bytes long, so a byte-based cut at
+	// three would leave half a character behind.
+	got := TruncateString("äöü", 3)
+	if got != "äöü" {
+		t.Errorf("expected %q got %q", "äöü", got)
+	}
+}