[issuu] unescape HTML entities

This commit is contained in:
Mike Fährmann
2025-02-02 18:33:18 +01:00
parent 6c9b20fe45
commit 5807daa19a
2 changed files with 33 additions and 2 deletions

View File

@@ -30,8 +30,8 @@ class IssuuPublicationExtractor(IssuuBase, GalleryExtractor):
def metadata(self, page): def metadata(self, page):
pos = page.rindex('id="initial-data"') pos = page.rindex('id="initial-data"')
data = util.json_loads(text.rextract( data = util.json_loads(text.unescape(text.rextract(
page, '<script data-json="', '"', pos)[0].replace("&quot;", '"')) page, '<script data-json="', '"', pos)[0]))
doc = data["initialDocumentData"]["document"] doc = data["initialDocumentData"]["document"]
doc["date"] = text.parse_datetime( doc["date"] = text.parse_datetime(

View File

@@ -34,6 +34,37 @@ __tests__ = (
"num" : int, "num" : int,
}, },
{
"#url" : "https://issuu.com/foodhome1955/docs/fh_winter2025-issuu-011625",
"#comment" : "HTML escapes",
"#class" : issuu.IssuuPublicationExtractor,
"#count" : 84,
"document": {
"access" : "PUBLIC",
"date" : "dt:2025-01-17 00:00:00",
"description" : "Santa Barbara's Lifestyle Magazine",
"documentName" : "fh_winter2025-issuu-011625",
"isDocumentGated" : False,
"originalPublishDateInISOString": "2025-01-17T00:00:00.000Z",
"pageCount" : 84,
"publicationId" : "b89e35d4bd2201c7ecd871160fe000fa",
"revisionId" : "250117005419",
"title" : "Food & Home Winter 2025",
"username" : "foodhome1955",
"contentRating" : {
"isAdsafe" : True,
"isExplicit" : False,
"isReviewed" : True,
},
"path" : {
"documentName": "fh_winter2025-issuu-011625",
"type" : "user",
"username" : "foodhome1955",
},
},
},
{ {
"#url" : "https://issuu.com/issuu", "#url" : "https://issuu.com/issuu",
"#class" : issuu.IssuuUserExtractor, "#class" : issuu.IssuuUserExtractor,