[bugfix] Fix unicode-unaware word boundary check in hashtags (#1049)

* [bugfix] Fix unicode-unaware word boundary check in hashtag regex

Go `\b` does not care for Unicode, and without lookahead, the workarounds got
very ugly. So I replaced the regex with a parser.

The parser runs in O(n) time and performance should not be affected.

* [bugfix] Add back hashtag max length and add tests for it
This commit is contained in:
ugla 2022-11-15 16:05:34 +01:00 committed by GitHub
parent fece7fa706
commit 52109776f6
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
4 changed files with 146 additions and 45 deletions

View file

@ -47,7 +47,6 @@ const (
const (
maximumUsernameLength = 64
maximumEmojiShortcodeLength = 30
maximumHashtagLength = 30
)
var (
@ -66,17 +65,11 @@ var (
// such as @whatever_user@example.org, returning whatever_user and example.org (without the @ symbols)
MentionName = regexp.MustCompile(mentionName)
// mention regex can be played around with here: https://regex101.com/r/G1oGR0/1
// mention regex can be played around with here: https://regex101.com/r/P0vpYG/1
mentionFinder = `(?:^|\s)(@\w+(?:@[a-zA-Z0-9_\-\.]+)?)`
// MentionFinder extracts mentions from a piece of text.
MentionFinder = regexp.MustCompile(mentionFinder)
// hashtag regex can be played with here: https://regex101.com/r/bpyGlj/1
hashtagFinder = fmt.Sprintf(`(?:^|\s)(?:#*)(#[\p{L}\p{N}]{1,%d})(?:#|\b)`, maximumHashtagLength)
// HashtagFinder finds possible hashtags in a string.
// It returns just the string part of the hashtag, not the # symbol.
HashtagFinder = regexp.MustCompile(hashtagFinder)
emojiShortcode = fmt.Sprintf(`\w{2,%d}`, maximumEmojiShortcodeLength)
// EmojiShortcode validates an emoji name.
EmojiShortcode = regexp.MustCompile(fmt.Sprintf("^%s$", emojiShortcode))

View file

@ -27,36 +27,46 @@ import (
"github.com/superseriousbusiness/gotosocial/internal/gtsmodel"
"github.com/superseriousbusiness/gotosocial/internal/log"
"github.com/superseriousbusiness/gotosocial/internal/regexes"
"github.com/superseriousbusiness/gotosocial/internal/util"
)
func (f *formatter) ReplaceTags(ctx context.Context, in string, tags []*gtsmodel.Tag) string {
return regexes.ReplaceAllStringFunc(regexes.HashtagFinder, in, func(match string, buf *bytes.Buffer) string {
// we have a match
matchTrimmed := strings.TrimSpace(match)
tagAsEntered := matchTrimmed[1:]
spans := util.FindHashtagSpansInText(in)
if len(spans) == 0 {
return in
}
var b strings.Builder
i := 0
spans:
for _, t := range spans {
b.WriteString(in[i:t.First])
i = t.Second
tagAsEntered := in[t.First+1 : t.Second]
// check through the tags to find what we're matching
for _, tag := range tags {
if strings.EqualFold(tagAsEntered, tag.Name) {
// Add any dropped space from match
if unicode.IsSpace(rune(match[0])) {
buf.WriteByte(match[0])
}
// replace the #tag with the formatted tag content
// `<a href="tag.URL" class="mention hashtag" rel="tag">#<span>tagAsEntered</span></a>
buf.WriteString(`<a href="`)
buf.WriteString(tag.URL)
buf.WriteString(`" class="mention hashtag" rel="tag">#<span>`)
buf.WriteString(tagAsEntered)
buf.WriteString(`</span></a>`)
return buf.String()
b.WriteString(`<a href="`)
b.WriteString(tag.URL)
b.WriteString(`" class="mention hashtag" rel="tag">#<span>`)
b.WriteString(tagAsEntered)
b.WriteString(`</span></a>`)
continue spans
}
}
// the match wasn't in the list of tags for whatever reason, so just return the match as we found it so nothing changes
return match
})
b.WriteString(in[t.First:t.Second])
}
// Get the last bits.
i = spans[len(spans)-1].Second
b.WriteString(in[i:])
return b.String()
}
func (f *formatter) ReplaceMentions(ctx context.Context, in string, mentions []*gtsmodel.Mention) string {

View file

@ -19,11 +19,16 @@
package util
import (
"strings"
"unicode"
"unicode/utf8"
"github.com/superseriousbusiness/gotosocial/internal/regexes"
)
const (
maximumHashtagLength = 30
)
// DeriveMentionNamesFromText takes a plaintext (ie., not html-formatted) text,
// and applies a regex to it to return a deduplicated list of account names
// mentioned in that text, in the format "@user@example.org" or "@username" for
@ -36,16 +41,71 @@ func DeriveMentionNamesFromText(text string) []string {
return UniqueStrings(mentionedAccounts)
}
// DeriveHashtagsFromText takes a plaintext (ie., not html-formatted) text,
// and applies a regex to it to return a deduplicated list of hashtags
// used in that text, without the leading #. The case of the returned
// tags will be lowered, for consistency.
func DeriveHashtagsFromText(text string) []string {
tags := []string{}
for _, m := range regexes.HashtagFinder.FindAllStringSubmatch(text, -1) {
tags = append(tags, strings.TrimPrefix(m[1], "#"))
type Pair[A, B any] struct {
First A
Second B
}
// Byte index in original string
// `First` includes `#`.
type Span = Pair[int, int]
// Takes a plaintext (ie., not HTML-formatted) text,
// and returns a slice of unique hashtags.
func DeriveHashtagsFromText(text string) []string {
tagsMap := make(map[string]bool)
tags := []string{}
for _, v := range FindHashtagSpansInText(text) {
t := text[v.First+1 : v.Second]
if _, value := tagsMap[t]; !value {
tagsMap[t] = true
tags = append(tags, t)
}
}
return tags
}
// Takes a plaintext (ie., not HTML-formatted) text,
// and returns a list of pairs of indices into the original string, where
// hashtags are located.
func FindHashtagSpansInText(text string) []Span {
tags := []Span{}
start := 0
// Keep one rune of lookbehind.
prev := ' '
inTag := false
for i, r := range text {
if r == '#' && isHashtagBoundary(prev) {
// Start of hashtag.
inTag = true
start = i
} else if inTag && !isPermittedInHashtag(r) && !isHashtagBoundary(r) {
// Inside the hashtag, but it was a phoney, gottem.
inTag = false
} else if inTag && isHashtagBoundary(r) {
// End of hashtag.
inTag = false
appendTag(&tags, text, start, i)
} else if irl := i + utf8.RuneLen(r); inTag && irl == len(text) {
// End of text.
appendTag(&tags, text, start, irl)
}
prev = r
}
return tags
}
func appendTag(tags *[]Span, text string, start int, end int) {
l := end - start - 1
// This check could be moved out into the parsing loop if necessary!
if 0 < l && l <= maximumHashtagLength {
*tags = append(*tags, Span{First: start, Second: end})
}
return UniqueStrings(tags)
}
// DeriveEmojisFromText takes a plaintext (ie., not html-formatted) text,
@ -58,3 +118,17 @@ func DeriveEmojisFromText(text string) []string {
}
return UniqueStrings(emojis)
}
func isPermittedInHashtag(r rune) bool {
return unicode.IsLetter(r) || unicode.IsNumber(r)
}
// Decides where to break before or after a hashtag.
func isHashtagBoundary(r rune) bool {
return r == '#' || // `###lol` should work
unicode.IsSpace(r) || // All kinds of Unicode whitespace.
unicode.IsControl(r) || // All kinds of control characters, like tab.
// Most kinds of punctuation except "Pc" ("Punctuation, connecting", like `_`).
// But `someurl/#fragment` should not match, neither should HTML entities like `&#35;`.
('/' != r && '&' != r && !unicode.Is(unicode.Categories["Pc"], r) && unicode.IsPunct(r))
}

View file

@ -77,26 +77,50 @@ func (suite *StatusTestSuite) TestDeriveHashtagsOK() {
# testing this one shouldn't work
#thisshouldwork
#thisshouldwork #dupe #dupe!! #dupe
here's a link with a fragment: https://example.org/whatever#ahhh
here's another link with a fragment: https://example.org/whatever/#ahhh
#ThisShouldAlsoWork #not_this_though
(#ThisShouldAlsoWork) #not_this_though
#111111 thisalsoshouldn'twork#### ##
#alimentación, #saúde
#alimentación, #saúde, #lävistää, #ö, #
#ThisOneIsThirtyOneCharactersLon... ...ng
#ThisOneIsThirteyCharactersLong
`
tags := util.DeriveHashtagsFromText(statusText)
assert.Len(suite.T(), tags, 7)
assert.Len(suite.T(), tags, 12)
assert.Equal(suite.T(), "testing123", tags[0])
assert.Equal(suite.T(), "also", tags[1])
assert.Equal(suite.T(), "thisshouldwork", tags[2])
assert.Equal(suite.T(), "ThisShouldAlsoWork", tags[3])
assert.Equal(suite.T(), "111111", tags[4])
assert.Equal(suite.T(), "alimentación", tags[5])
assert.Equal(suite.T(), "saúde", tags[6])
assert.Equal(suite.T(), "dupe", tags[3])
assert.Equal(suite.T(), "ThisShouldAlsoWork", tags[4])
assert.Equal(suite.T(), "111111", tags[5])
assert.Equal(suite.T(), "alimentación", tags[6])
assert.Equal(suite.T(), "saúde", tags[7])
assert.Equal(suite.T(), "lävistää", tags[8])
assert.Equal(suite.T(), "ö", tags[9])
assert.Equal(suite.T(), "네", tags[10])
assert.Equal(suite.T(), "ThisOneIsThirteyCharactersLong", tags[11])
statusText = `#올빼미 hej`
tags = util.DeriveHashtagsFromText(statusText)
assert.Equal(suite.T(), "올빼미", tags[0])
}
func (suite *StatusTestSuite) TestHashtagSpansOK() {
statusText := `#0 #3 #8aa`
spans := util.FindHashtagSpansInText(statusText)
assert.Equal(suite.T(), 0, spans[0].First)
assert.Equal(suite.T(), 2, spans[0].Second)
assert.Equal(suite.T(), 3, spans[1].First)
assert.Equal(suite.T(), 5, spans[1].Second)
assert.Equal(suite.T(), 8, spans[2].First)
assert.Equal(suite.T(), 12, spans[2].Second)
}
func (suite *StatusTestSuite) TestDeriveEmojiOK() {
@ -127,7 +151,7 @@ Here's some normal text with an :emoji: at the end
func (suite *StatusTestSuite) TestDeriveMultiple() {
statusText := `Another test @foss_satan@fossbros-anonymous.io
#Hashtag
#HashTag
Text`
@ -139,7 +163,7 @@ func (suite *StatusTestSuite) TestDeriveMultiple() {
assert.Equal(suite.T(), "@foss_satan@fossbros-anonymous.io", ms[0])
assert.Len(suite.T(), hs, 1)
assert.Equal(suite.T(), "Hashtag", hs[0])
assert.Contains(suite.T(), hs, "HashTag")
assert.Len(suite.T(), es, 0)
}