Fix *chan tags extraction
This commit is contained in:
@@ -129,7 +129,7 @@ internal abstract class ChanParser(source: MangaSource) : MangaParser(source) {
|
||||
|
||||
override suspend fun getTags(): Set<MangaTag> {
|
||||
val domain = getDomain()
|
||||
val doc = context.httpGet("https://$domain/catalog").parseHtml()
|
||||
val doc = context.httpGet("https://$domain/mostfavorites&sort=manga").parseHtml()
|
||||
val root = doc.body().selectFirst("div.main_fon")?.getElementById("side")
|
||||
?.select("ul")?.last() ?: throw ParseException("Cannot find root")
|
||||
return root.select("li.sidetag").mapToSet { li ->
|
||||
|
||||
Reference in New Issue
Block a user