mirror of
https://codeberg.org/forgejo/forgejo.git
synced 2024-11-14 06:39:40 +01:00
e9e2a9cdcc
* Allow square brackets in external issue patterns * Added false test cases for checklist elements
567 lines
17 KiB
Go
567 lines
17 KiB
Go
// Copyright 2017 The Gitea Authors. All rights reserved.
|
|
// Use of this source code is governed by a MIT-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
package markup
|
|
|
|
import (
|
|
"bytes"
|
|
"fmt"
|
|
"io"
|
|
"net/url"
|
|
"path"
|
|
"path/filepath"
|
|
"regexp"
|
|
"strings"
|
|
|
|
"code.gitea.io/gitea/modules/base"
|
|
"code.gitea.io/gitea/modules/log"
|
|
"code.gitea.io/gitea/modules/setting"
|
|
|
|
"github.com/Unknwon/com"
|
|
"golang.org/x/net/html"
|
|
)
|
|
|
|
// Issue name styles
|
|
const (
|
|
IssueNameStyleNumeric = "numeric"
|
|
IssueNameStyleAlphanumeric = "alphanumeric"
|
|
)
|
|
|
|
var (
|
|
// NOTE: All below regex matching do not perform any extra validation.
|
|
// Thus a link is produced even if the linked entity does not exist.
|
|
// While fast, this is also incorrect and lead to false positives.
|
|
// TODO: fix invalid linking issue
|
|
|
|
// MentionPattern matches string that mentions someone, e.g. @Unknwon
|
|
MentionPattern = regexp.MustCompile(`(\s|^|\W)@[0-9a-zA-Z-_\.]+`)
|
|
|
|
// IssueNumericPattern matches string that references to a numeric issue, e.g. #1287
|
|
IssueNumericPattern = regexp.MustCompile(`( |^|\(|\[)#[0-9]+\b`)
|
|
// IssueAlphanumericPattern matches string that references to an alphanumeric issue, e.g. ABC-1234
|
|
IssueAlphanumericPattern = regexp.MustCompile(`( |^|\(|\[)[A-Z]{1,10}-[1-9][0-9]*\b`)
|
|
// CrossReferenceIssueNumericPattern matches string that references a numeric issue in a different repository
|
|
// e.g. gogits/gogs#12345
|
|
CrossReferenceIssueNumericPattern = regexp.MustCompile(`( |^)[0-9a-zA-Z-_\.]+/[0-9a-zA-Z-_\.]+#[0-9]+\b`)
|
|
|
|
// Sha1CurrentPattern matches string that represents a commit SHA, e.g. d8a994ef243349f321568f9e36d5c3f444b99cae
|
|
// Although SHA1 hashes are 40 chars long, the regex matches the hash from 7 to 40 chars in length
|
|
// so that abbreviated hash links can be used as well. This matches git and github useability.
|
|
Sha1CurrentPattern = regexp.MustCompile(`(?:^|\s|\()([0-9a-f]{7,40})\b`)
|
|
|
|
// ShortLinkPattern matches short but difficult to parse [[name|link|arg=test]] syntax
|
|
ShortLinkPattern = regexp.MustCompile(`(\[\[.*?\]\]\w*)`)
|
|
|
|
// AnySHA1Pattern allows to split url containing SHA into parts
|
|
AnySHA1Pattern = regexp.MustCompile(`(http\S*)://(\S+)/(\S+)/(\S+)/(\S+)/([0-9a-f]{40})(?:/?([^#\s]+)?(?:#(\S+))?)?`)
|
|
|
|
validLinksPattern = regexp.MustCompile(`^[a-z][\w-]+://`)
|
|
)
|
|
|
|
// regexp for full links to issues/pulls
|
|
var issueFullPattern *regexp.Regexp
|
|
|
|
// IsLink reports whether link fits valid format.
|
|
func IsLink(link []byte) bool {
|
|
return isLink(link)
|
|
}
|
|
|
|
// isLink reports whether link fits valid format.
|
|
func isLink(link []byte) bool {
|
|
return validLinksPattern.Match(link)
|
|
}
|
|
|
|
func getIssueFullPattern() *regexp.Regexp {
|
|
if issueFullPattern == nil {
|
|
appURL := setting.AppURL
|
|
if len(appURL) > 0 && appURL[len(appURL)-1] != '/' {
|
|
appURL += "/"
|
|
}
|
|
issueFullPattern = regexp.MustCompile(appURL +
|
|
`\w+/\w+/(?:issues|pulls)/((?:\w{1,10}-)?[1-9][0-9]*)([\?|#]\S+.(\S+)?)?\b`)
|
|
}
|
|
return issueFullPattern
|
|
}
|
|
|
|
// FindAllMentions matches mention patterns in given content
|
|
// and returns a list of found user names without @ prefix.
|
|
func FindAllMentions(content string) []string {
|
|
mentions := MentionPattern.FindAllString(content, -1)
|
|
for i := range mentions {
|
|
mentions[i] = mentions[i][strings.Index(mentions[i], "@")+1:] // Strip @ character
|
|
}
|
|
return mentions
|
|
}
|
|
|
|
// cutoutVerbosePrefix cutouts URL prefix including sub-path to
|
|
// return a clean unified string of request URL path.
|
|
func cutoutVerbosePrefix(prefix string) string {
|
|
if len(prefix) == 0 || prefix[0] != '/' {
|
|
return prefix
|
|
}
|
|
count := 0
|
|
for i := 0; i < len(prefix); i++ {
|
|
if prefix[i] == '/' {
|
|
count++
|
|
}
|
|
if count >= 3+setting.AppSubURLDepth {
|
|
return prefix[:i]
|
|
}
|
|
}
|
|
return prefix
|
|
}
|
|
|
|
// URLJoin joins url components, like path.Join, but preserving contents
|
|
func URLJoin(base string, elems ...string) string {
|
|
u, err := url.Parse(base)
|
|
if err != nil {
|
|
log.Error(4, "URLJoin: Invalid base URL %s", base)
|
|
return ""
|
|
}
|
|
joinArgs := make([]string, 0, len(elems)+1)
|
|
joinArgs = append(joinArgs, u.Path)
|
|
joinArgs = append(joinArgs, elems...)
|
|
u.Path = path.Join(joinArgs...)
|
|
return u.String()
|
|
}
|
|
|
|
// RenderIssueIndexPatternOptions options for RenderIssueIndexPattern function
|
|
type RenderIssueIndexPatternOptions struct {
|
|
// url to which non-special formatting should be linked. If empty,
|
|
// no such links will be added
|
|
DefaultURL string
|
|
URLPrefix string
|
|
Metas map[string]string
|
|
}
|
|
|
|
// addText add text to the given buffer, adding a link to the default url
|
|
// if appropriate
|
|
func (opts RenderIssueIndexPatternOptions) addText(text []byte, buf *bytes.Buffer) {
|
|
if len(text) == 0 {
|
|
return
|
|
} else if len(opts.DefaultURL) == 0 {
|
|
buf.Write(text)
|
|
return
|
|
}
|
|
buf.WriteString(`<a rel="nofollow" href="`)
|
|
buf.WriteString(opts.DefaultURL)
|
|
buf.WriteString(`">`)
|
|
buf.Write(text)
|
|
buf.WriteString(`</a>`)
|
|
}
|
|
|
|
// RenderIssueIndexPattern renders issue indexes to corresponding links.
|
|
func RenderIssueIndexPattern(rawBytes []byte, opts RenderIssueIndexPatternOptions) []byte {
|
|
opts.URLPrefix = cutoutVerbosePrefix(opts.URLPrefix)
|
|
|
|
pattern := IssueNumericPattern
|
|
if opts.Metas["style"] == IssueNameStyleAlphanumeric {
|
|
pattern = IssueAlphanumericPattern
|
|
}
|
|
|
|
var buf bytes.Buffer
|
|
remainder := rawBytes
|
|
for {
|
|
indices := pattern.FindIndex(remainder)
|
|
if indices == nil || len(indices) < 2 {
|
|
opts.addText(remainder, &buf)
|
|
return buf.Bytes()
|
|
}
|
|
startIndex := indices[0]
|
|
endIndex := indices[1]
|
|
opts.addText(remainder[:startIndex], &buf)
|
|
if remainder[startIndex] == '(' || remainder[startIndex] == ' ' {
|
|
buf.WriteByte(remainder[startIndex])
|
|
startIndex++
|
|
}
|
|
if opts.Metas == nil {
|
|
buf.WriteString(`<a href="`)
|
|
buf.WriteString(URLJoin(
|
|
opts.URLPrefix, "issues", string(remainder[startIndex+1:endIndex])))
|
|
buf.WriteString(`">`)
|
|
buf.Write(remainder[startIndex:endIndex])
|
|
buf.WriteString(`</a>`)
|
|
} else {
|
|
// Support for external issue tracker
|
|
buf.WriteString(`<a href="`)
|
|
if opts.Metas["style"] == IssueNameStyleAlphanumeric {
|
|
opts.Metas["index"] = string(remainder[startIndex:endIndex])
|
|
} else {
|
|
opts.Metas["index"] = string(remainder[startIndex+1 : endIndex])
|
|
}
|
|
buf.WriteString(com.Expand(opts.Metas["format"], opts.Metas))
|
|
buf.WriteString(`">`)
|
|
buf.Write(remainder[startIndex:endIndex])
|
|
buf.WriteString(`</a>`)
|
|
}
|
|
if endIndex < len(remainder) &&
|
|
(remainder[endIndex] == ')' || remainder[endIndex] == ' ') {
|
|
buf.WriteByte(remainder[endIndex])
|
|
endIndex++
|
|
}
|
|
remainder = remainder[endIndex:]
|
|
}
|
|
}
|
|
|
|
// IsSameDomain checks if given url string has the same hostname as current Gitea instance
|
|
func IsSameDomain(s string) bool {
|
|
if strings.HasPrefix(s, "/") {
|
|
return true
|
|
}
|
|
if uapp, err := url.Parse(setting.AppURL); err == nil {
|
|
if u, err := url.Parse(s); err == nil {
|
|
return u.Host == uapp.Host
|
|
}
|
|
return false
|
|
}
|
|
return false
|
|
}
|
|
|
|
// renderFullSha1Pattern renders SHA containing URLs
|
|
func renderFullSha1Pattern(rawBytes []byte, urlPrefix string) []byte {
|
|
ms := AnySHA1Pattern.FindAllSubmatch(rawBytes, -1)
|
|
for _, m := range ms {
|
|
all := m[0]
|
|
protocol := string(m[1])
|
|
paths := string(m[2])
|
|
path := protocol + "://" + paths
|
|
author := string(m[3])
|
|
repoName := string(m[4])
|
|
path = URLJoin(path, author, repoName)
|
|
ltype := "src"
|
|
itemType := m[5]
|
|
if IsSameDomain(paths) {
|
|
ltype = string(itemType)
|
|
} else if string(itemType) == "commit" {
|
|
ltype = "commit"
|
|
}
|
|
sha := m[6]
|
|
var subtree string
|
|
if len(m) > 7 && len(m[7]) > 0 {
|
|
subtree = string(m[7])
|
|
}
|
|
var line []byte
|
|
if len(m) > 8 && len(m[8]) > 0 {
|
|
line = m[8]
|
|
}
|
|
urlSuffix := ""
|
|
text := base.ShortSha(string(sha))
|
|
if subtree != "" {
|
|
urlSuffix = "/" + subtree
|
|
text += urlSuffix
|
|
}
|
|
if line != nil {
|
|
value := string(line)
|
|
urlSuffix += "#"
|
|
urlSuffix += value
|
|
text += " ("
|
|
text += value
|
|
text += ")"
|
|
}
|
|
rawBytes = bytes.Replace(rawBytes, all, []byte(fmt.Sprintf(
|
|
`<a href="%s">%s</a>`, URLJoin(path, ltype, string(sha))+urlSuffix, text)), -1)
|
|
}
|
|
return rawBytes
|
|
}
|
|
|
|
// RenderFullIssuePattern renders issues-like URLs
|
|
func RenderFullIssuePattern(rawBytes []byte) []byte {
|
|
ms := getIssueFullPattern().FindAllSubmatch(rawBytes, -1)
|
|
for _, m := range ms {
|
|
all := m[0]
|
|
id := string(m[1])
|
|
text := "#" + id
|
|
// TODO if m[2] is not nil, then link is to a comment,
|
|
// and we should indicate that in the text somehow
|
|
rawBytes = bytes.Replace(rawBytes, all, []byte(fmt.Sprintf(
|
|
`<a href="%s">%s</a>`, string(all), text)), -1)
|
|
}
|
|
return rawBytes
|
|
}
|
|
|
|
func firstIndexOfByte(sl []byte, target byte) int {
|
|
for i := 0; i < len(sl); i++ {
|
|
if sl[i] == target {
|
|
return i
|
|
}
|
|
}
|
|
return -1
|
|
}
|
|
|
|
func lastIndexOfByte(sl []byte, target byte) int {
|
|
for i := len(sl) - 1; i >= 0; i-- {
|
|
if sl[i] == target {
|
|
return i
|
|
}
|
|
}
|
|
return -1
|
|
}
|
|
|
|
// RenderShortLinks processes [[syntax]]
|
|
//
|
|
// noLink flag disables making link tags when set to true
|
|
// so this function just replaces the whole [[...]] with the content text
|
|
//
|
|
// isWikiMarkdown is a flag to choose linking url prefix
|
|
func RenderShortLinks(rawBytes []byte, urlPrefix string, noLink bool, isWikiMarkdown bool) []byte {
|
|
ms := ShortLinkPattern.FindAll(rawBytes, -1)
|
|
for _, m := range ms {
|
|
orig := bytes.TrimSpace(m)
|
|
m = orig[2:]
|
|
tailPos := lastIndexOfByte(m, ']') + 1
|
|
tail := []byte{}
|
|
if tailPos < len(m) {
|
|
tail = m[tailPos:]
|
|
m = m[:tailPos-1]
|
|
}
|
|
m = m[:len(m)-2]
|
|
props := map[string]string{}
|
|
|
|
// MediaWiki uses [[link|text]], while GitHub uses [[text|link]]
|
|
// It makes page handling terrible, but we prefer GitHub syntax
|
|
// And fall back to MediaWiki only when it is obvious from the look
|
|
// Of text and link contents
|
|
sl := bytes.Split(m, []byte("|"))
|
|
for _, v := range sl {
|
|
switch bytes.Count(v, []byte("=")) {
|
|
|
|
// Piped args without = sign, these are mandatory arguments
|
|
case 0:
|
|
{
|
|
sv := string(v)
|
|
if props["name"] == "" {
|
|
if isLink(v) {
|
|
// If we clearly see it is a link, we save it so
|
|
|
|
// But first we need to ensure, that if both mandatory args provided
|
|
// look like links, we stick to GitHub syntax
|
|
if props["link"] != "" {
|
|
props["name"] = props["link"]
|
|
}
|
|
|
|
props["link"] = strings.TrimSpace(sv)
|
|
} else {
|
|
props["name"] = sv
|
|
}
|
|
} else {
|
|
props["link"] = strings.TrimSpace(sv)
|
|
}
|
|
}
|
|
|
|
// Piped args with = sign, these are optional arguments
|
|
case 1:
|
|
{
|
|
sep := firstIndexOfByte(v, '=')
|
|
key, val := string(v[:sep]), html.UnescapeString(string(v[sep+1:]))
|
|
lastCharIndex := len(val) - 1
|
|
if (val[0] == '"' || val[0] == '\'') && (val[lastCharIndex] == '"' || val[lastCharIndex] == '\'') {
|
|
val = val[1:lastCharIndex]
|
|
}
|
|
props[key] = val
|
|
}
|
|
}
|
|
}
|
|
|
|
var name string
|
|
var link string
|
|
if props["link"] != "" {
|
|
link = props["link"]
|
|
} else if props["name"] != "" {
|
|
link = props["name"]
|
|
}
|
|
if props["title"] != "" {
|
|
name = props["title"]
|
|
} else if props["name"] != "" {
|
|
name = props["name"]
|
|
} else {
|
|
name = link
|
|
}
|
|
|
|
name += string(tail)
|
|
image := false
|
|
ext := filepath.Ext(string(link))
|
|
if ext != "" {
|
|
switch ext {
|
|
case ".jpg", ".jpeg", ".png", ".tif", ".tiff", ".webp", ".gif", ".bmp", ".ico", ".svg":
|
|
{
|
|
image = true
|
|
}
|
|
}
|
|
}
|
|
absoluteLink := isLink([]byte(link))
|
|
if !absoluteLink {
|
|
link = strings.Replace(link, " ", "+", -1)
|
|
}
|
|
if image {
|
|
if !absoluteLink {
|
|
if IsSameDomain(urlPrefix) {
|
|
urlPrefix = strings.Replace(urlPrefix, "/src/", "/raw/", 1)
|
|
}
|
|
if isWikiMarkdown {
|
|
link = URLJoin("wiki", "raw", link)
|
|
}
|
|
link = URLJoin(urlPrefix, link)
|
|
}
|
|
title := props["title"]
|
|
if title == "" {
|
|
title = props["alt"]
|
|
}
|
|
if title == "" {
|
|
title = path.Base(string(name))
|
|
}
|
|
alt := props["alt"]
|
|
if alt == "" {
|
|
alt = name
|
|
}
|
|
if alt != "" {
|
|
alt = `alt="` + alt + `"`
|
|
}
|
|
name = fmt.Sprintf(`<img src="%s" %s title="%s" />`, link, alt, title)
|
|
} else if !absoluteLink {
|
|
if isWikiMarkdown {
|
|
link = URLJoin("wiki", link)
|
|
}
|
|
link = URLJoin(urlPrefix, link)
|
|
}
|
|
if noLink {
|
|
rawBytes = bytes.Replace(rawBytes, orig, []byte(name), -1)
|
|
} else {
|
|
rawBytes = bytes.Replace(rawBytes, orig,
|
|
[]byte(fmt.Sprintf(`<a href="%s">%s</a>`, link, name)), -1)
|
|
}
|
|
}
|
|
return rawBytes
|
|
}
|
|
|
|
// RenderCrossReferenceIssueIndexPattern renders issue indexes from other repositories to corresponding links.
|
|
func RenderCrossReferenceIssueIndexPattern(rawBytes []byte, urlPrefix string, metas map[string]string) []byte {
|
|
ms := CrossReferenceIssueNumericPattern.FindAll(rawBytes, -1)
|
|
for _, m := range ms {
|
|
if m[0] == ' ' || m[0] == '(' {
|
|
m = m[1:] // ignore leading space or opening parentheses
|
|
}
|
|
|
|
repo := string(bytes.Split(m, []byte("#"))[0])
|
|
issue := string(bytes.Split(m, []byte("#"))[1])
|
|
|
|
link := fmt.Sprintf(`<a href="%s">%s</a>`, URLJoin(setting.AppURL, repo, "issues", issue), m)
|
|
rawBytes = bytes.Replace(rawBytes, m, []byte(link), 1)
|
|
}
|
|
return rawBytes
|
|
}
|
|
|
|
// renderSha1CurrentPattern renders SHA1 strings to corresponding links that assumes in the same repository.
|
|
func renderSha1CurrentPattern(rawBytes []byte, urlPrefix string) []byte {
|
|
ms := Sha1CurrentPattern.FindAllSubmatch(rawBytes, -1)
|
|
for _, m := range ms {
|
|
hash := m[1]
|
|
// The regex does not lie, it matches the hash pattern.
|
|
// However, a regex cannot know if a hash actually exists or not.
|
|
// We could assume that a SHA1 hash should probably contain alphas AND numerics
|
|
// but that is not always the case.
|
|
// Although unlikely, deadbeef and 1234567 are valid short forms of SHA1 hash
|
|
// as used by git and github for linking and thus we have to do similar.
|
|
rawBytes = bytes.Replace(rawBytes, hash, []byte(fmt.Sprintf(
|
|
`<a href="%s">%s</a>`, URLJoin(urlPrefix, "commit", string(hash)), base.ShortSha(string(hash)))), -1)
|
|
}
|
|
return rawBytes
|
|
}
|
|
|
|
// RenderSpecialLink renders mentions, indexes and SHA1 strings to corresponding links.
|
|
func RenderSpecialLink(rawBytes []byte, urlPrefix string, metas map[string]string, isWikiMarkdown bool) []byte {
|
|
ms := MentionPattern.FindAll(rawBytes, -1)
|
|
for _, m := range ms {
|
|
m = m[bytes.Index(m, []byte("@")):]
|
|
rawBytes = bytes.Replace(rawBytes, m,
|
|
[]byte(fmt.Sprintf(`<a href="%s">%s</a>`, URLJoin(setting.AppURL, string(m[1:])), m)), -1)
|
|
}
|
|
|
|
rawBytes = RenderFullIssuePattern(rawBytes)
|
|
rawBytes = RenderShortLinks(rawBytes, urlPrefix, false, isWikiMarkdown)
|
|
rawBytes = RenderIssueIndexPattern(rawBytes, RenderIssueIndexPatternOptions{
|
|
URLPrefix: urlPrefix,
|
|
Metas: metas,
|
|
})
|
|
rawBytes = RenderCrossReferenceIssueIndexPattern(rawBytes, urlPrefix, metas)
|
|
rawBytes = renderFullSha1Pattern(rawBytes, urlPrefix)
|
|
rawBytes = renderSha1CurrentPattern(rawBytes, urlPrefix)
|
|
return rawBytes
|
|
}
|
|
|
|
var (
|
|
leftAngleBracket = []byte("</")
|
|
rightAngleBracket = []byte(">")
|
|
)
|
|
|
|
var noEndTags = []string{"img", "input", "br", "hr"}
|
|
|
|
// PostProcess treats different types of HTML differently,
|
|
// and only renders special links for plain text blocks.
|
|
func PostProcess(rawHTML []byte, urlPrefix string, metas map[string]string, isWikiMarkdown bool) []byte {
|
|
startTags := make([]string, 0, 5)
|
|
var buf bytes.Buffer
|
|
tokenizer := html.NewTokenizer(bytes.NewReader(rawHTML))
|
|
|
|
OUTER_LOOP:
|
|
for html.ErrorToken != tokenizer.Next() {
|
|
token := tokenizer.Token()
|
|
switch token.Type {
|
|
case html.TextToken:
|
|
buf.Write(RenderSpecialLink([]byte(token.String()), urlPrefix, metas, isWikiMarkdown))
|
|
|
|
case html.StartTagToken:
|
|
buf.WriteString(token.String())
|
|
tagName := token.Data
|
|
// If this is an excluded tag, we skip processing all output until a close tag is encountered.
|
|
if strings.EqualFold("a", tagName) || strings.EqualFold("code", tagName) || strings.EqualFold("pre", tagName) {
|
|
stackNum := 1
|
|
for html.ErrorToken != tokenizer.Next() {
|
|
token = tokenizer.Token()
|
|
|
|
// Copy the token to the output verbatim
|
|
buf.Write(RenderShortLinks([]byte(token.String()), urlPrefix, true, isWikiMarkdown))
|
|
|
|
if token.Type == html.StartTagToken && !com.IsSliceContainsStr(noEndTags, token.Data) {
|
|
stackNum++
|
|
}
|
|
|
|
// If this is the close tag to the outer-most, we are done
|
|
if token.Type == html.EndTagToken {
|
|
stackNum--
|
|
|
|
if stackNum <= 0 && strings.EqualFold(tagName, token.Data) {
|
|
break
|
|
}
|
|
}
|
|
}
|
|
continue OUTER_LOOP
|
|
}
|
|
|
|
if !com.IsSliceContainsStr(noEndTags, tagName) {
|
|
startTags = append(startTags, tagName)
|
|
}
|
|
|
|
case html.EndTagToken:
|
|
if len(startTags) == 0 {
|
|
buf.WriteString(token.String())
|
|
break
|
|
}
|
|
|
|
buf.Write(leftAngleBracket)
|
|
buf.WriteString(startTags[len(startTags)-1])
|
|
buf.Write(rightAngleBracket)
|
|
startTags = startTags[:len(startTags)-1]
|
|
default:
|
|
buf.WriteString(token.String())
|
|
}
|
|
}
|
|
|
|
if io.EOF == tokenizer.Err() {
|
|
return buf.Bytes()
|
|
}
|
|
|
|
// If we are not at the end of the input, then some other parsing error has occurred,
|
|
// so return the input verbatim.
|
|
return rawHTML
|
|
}
|