Files
forge/modules/util/truncate.go
zeekayandHanzo Dev 9a07b4d089 legal: Hanzo AI, Inc. copyright above the upstream notices
2978 files. The fork is 2026-07-15 (rebrand: gitea.dev -> github.com/hanzoai/git),
so 2026 is the year, and "Copyright 2026 Hanzo AI, Inc." was already the form
used by the handful of files that had ours.

Shape is now:

	// Copyright 2026 Hanzo AI, Inc. All rights reserved.
	// Copyright 2015 The Gogs Authors. All rights reserved.
	// Copyright 2019 The Gitea Authors. All rights reserved.
	// SPDX-License-Identifier: MIT

Ours on top, theirs retained verbatim. Retained because MIT requires it, and
separately because they are true — those authors did write that code.

Not renamed to "The Git Authors": Git is a different project under a different
license, and putting its name here would attribute this code to people who have
nothing to do with it. Adding our line is the fix; overwriting theirs would have
been both a licence violation and a false statement.

gofmt still reports import-order drift in some of these files. It predates this
change (verified against a stash) and is left alone rather than burying a
one-line-per-file legal change under 3000 files of unrelated reordering.

Build clean.

Co-authored-by: Hanzo Dev <[email protected]>
2026-07-27 13:49:52 -07:00

134 lines
4.3 KiB
Go

// Copyright 2026 Hanzo AI, Inc. All rights reserved.
// Copyright 2021 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
package util
import (
"strings"
"unicode"
"unicode/utf8"
)
// in UTF8 "…" is 3 bytes so doesn't really gain us anything...
const (
utf8Ellipsis = "…"
asciiEllipsis = "..."
)
func IsLikelyEllipsisLeftPart(s string) bool {
return strings.HasSuffix(s, utf8Ellipsis) || strings.HasSuffix(s, asciiEllipsis)
}
func ellipsisDisplayGuessWidth(r rune) int {
// To make the truncated string as long as possible,
// CJK/emoji chars are considered as 2-ASCII width but not 3-4 bytes width.
// Here we only make the best guess (better than counting them in bytes),
// it's impossible to 100% correctly determine the width of a rune without a real font and render.
//
// ATTENTION: the guessed width can't be zero, more details in ellipsisDisplayString's comment
if r <= 255 {
return 1
}
switch {
case r == '\u3000': /* ideographic (CJK) characters, still use 2 */
return 2
case unicode.Is(unicode.M, r), /* (Mark) */
unicode.Is(unicode.Cf, r), /* (Other, format) */
unicode.Is(unicode.Cs, r), /* (Other, surrogate) */
unicode.Is(unicode.Z /* (Space) */, r):
return 1
default:
return 2
}
}
// EllipsisDisplayString returns a truncated short string for display purpose.
// The length is the approximate number of ASCII-width in the string (CJK/emoji are 2-ASCII width)
// It appends "…" or "..." at the end of truncated string.
// It guarantees the length of the returned runes doesn't exceed the limit.
func EllipsisDisplayString(str string, limit int) string {
s, _, _, _ := ellipsisDisplayString(str, limit, ellipsisDisplayGuessWidth)
return s
}
// EllipsisDisplayStringX works like EllipsisDisplayString while it also returns the right part
func EllipsisDisplayStringX(str string, limit int) (left, right string) {
return ellipsisDisplayStringX(str, limit, ellipsisDisplayGuessWidth)
}
func ellipsisDisplayStringX(str string, limit int, widthGuess func(rune) int) (left, right string) {
left, offset, truncated, encounterInvalid := ellipsisDisplayString(str, limit, widthGuess)
if truncated {
right = str[offset:]
r, _ := utf8.DecodeRune(UnsafeStringToBytes(right))
encounterInvalid = encounterInvalid || r == utf8.RuneError
ellipsis := utf8Ellipsis
if encounterInvalid {
ellipsis = asciiEllipsis
}
right = ellipsis + right
}
return left, right
}
func ellipsisDisplayString(str string, limit int, widthGuess func(rune) int) (res string, offset int, truncated, encounterInvalid bool) {
if len(str) <= limit {
return str, len(str), false, false
}
// To future maintainers: this logic must guarantee that the length of the returned runes doesn't exceed the limit,
// because the returned string will also be used as database value. UTF-8 VARCHAR(10) could store 10 rune characters,
// So each rune must be countered as at least 1 width.
// Even if there are some special Unicode characters (zero-width, combining, etc.), they should NEVER be counted as zero.
pos, used := 0, 0
for i, r := range str {
encounterInvalid = encounterInvalid || r == utf8.RuneError
pos = i
runeWidth := widthGuess(r)
if used+runeWidth+3 > limit {
break
}
used += runeWidth
offset += utf8.RuneLen(r)
}
// if the remaining are fewer than 3 runes, then maybe we could add them, no need to ellipse
if len(str)-pos <= 12 {
var nextCnt, nextWidth int
for _, r := range str[pos:] {
if nextCnt >= 4 {
break
}
nextWidth += widthGuess(r)
nextCnt++
}
if nextCnt <= 3 && used+nextWidth <= limit {
return str, len(str), false, false
}
}
if limit < 3 {
// if the limit is so small, do not add ellipsis
return str[:offset], offset, true, false
}
ellipsis := utf8Ellipsis
if encounterInvalid {
ellipsis = asciiEllipsis
}
return str[:offset] + ellipsis, offset, true, encounterInvalid
}
func EllipsisTruncateRunes(str string, limit int) (left, right string) {
return ellipsisDisplayStringX(str, limit, func(r rune) int { return 1 })
}
// TruncateRunes returns a truncated string with given rune limit,
// it returns input string if its rune length doesn't exceed the limit.
func TruncateRunes(str string, limit int) string {
if utf8.RuneCountInString(str) < limit {
return str
}
return string([]rune(str)[:limit])
}