bellingcat/snscrape
1
0
Fork 0
mirror of https://github.com/bellingcat/snscrape.git synced 2026-06-12 20:38:29 +03:00
Code Issues Packages Projects Releases Wiki Activity

Fix post order on VK: reinsert pinned post at the correct location in the stream

Browse Source
This commit is contained in:
JustAnotherArchivist
2020-09-12 00:03:29 +00:00
parent 2b5444f89e
commit f8efe98608
1 changed files with 33 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

44
snscrape/modules/vkontakte.py
View File

@@ -51,15 +51,20 @@ class VKontakteUserScraper(snscrape.base.Scraper):
self._initialPage = None self._initialPage = None
self._initialPageSoup = None self._initialPageSoup = None
def _post_div_to_item(self, post):
url = urllib.parse.urljoin(self._baseUrl, post.find('a', class_ = 'post_link')['href'])
assert url.startswith('https://vk.com/wall') and '_' in url and url[-1] != '_' and url.rsplit('_', 1)[1].strip('0123456789') == ''
dateSpan = post.find('div', class_ = 'post_date').find('span', class_ = 'rel_date')
textDiv = post.find('div', class_ = 'wall_post_text')
return VKontaktePost(
url = url,
date = datetime.datetime.fromtimestamp(int(dateSpan['time']), datetime.timezone.utc) if 'time' in dateSpan else None,
content = textDiv.text if textDiv else None,
)
def _soup_to_items(self, soup): def _soup_to_items(self, soup):
for post in soup.find_all('div', class_ = 'post'): for post in soup.find_all('div', class_ = 'post'):
dateSpan = post.find('div', class_ = 'post_date').find('span', class_ = 'rel_date') yield self._post_div_to_item(post)
textDiv = post.find('div', class_ = 'wall_post_text')
yield VKontaktePost(
url = urllib.parse.urljoin(self._baseUrl, post.find('a', class_ = 'post_link')['href']),
date = datetime.datetime.fromtimestamp(int(dateSpan['time']), datetime.timezone.utc) if 'time' in dateSpan else None,
content = textDiv.text if textDiv else None,
)
def _initial_page(self): def _initial_page(self):
if self._initialPage is None: if self._initialPage is None:
@@ -92,13 +97,23 @@ class VKontakteUserScraper(snscrape.base.Scraper):
logger.info('Wall has no posts') logger.info('Wall has no posts')
return return
ownerID = newestPost.attrs['data-post-id'].split('_')[0] ownerID = newestPost.attrs['data-post-id'].split('_')[0]
# If there is a pinned post, we need its ID for the pagination requests # If there is a pinned post, we need its ID for the pagination requests; we also need to keep the post around so it can be inserted into the stream at the right point
if 'post_fixed' in newestPost.attrs['class']: if 'post_fixed' in newestPost.attrs['class']:
fixedPostID = newestPost.attrs['id'].split('_')[1] fixedPostID = int(newestPost.attrs['id'].split('_')[1])
fixedPost = self._post_div_to_item(newestPost)
else: else:
fixedPostID = '' fixedPostID = ''
fixedPost = None
yield from self._soup_to_items(soup) lastPostID = float('infinity')
for item in self._soup_to_items(soup.find(id = 'page_wall_posts')):
postID = int(item.url.rsplit('_', 1)[1])
if postID < lastPostID:
if fixedPost is not None and fixedPostID > postID:
yield fixedPost
fixedPost = None
yield item
lastPostID = postID
headers = self._headers.copy() headers = self._headers.copy()
headers['X-Requested-With'] = 'XMLHttpRequest' headers['X-Requested-With'] = 'XMLHttpRequest'
@@ -119,7 +134,14 @@ class VKontakteUserScraper(snscrape.base.Scraper):
if not posts.startswith('<div id="post'): if not posts.startswith('<div id="post'):
raise snscrape.base.ScraperException(f'Got an unknown response: {posts[:200]!r}...') raise snscrape.base.ScraperException(f'Got an unknown response: {posts[:200]!r}...')
soup = bs4.BeautifulSoup(posts, 'lxml') soup = bs4.BeautifulSoup(posts, 'lxml')
yield from self._soup_to_items(soup) for item in self._soup_to_items(soup):
postID = int(item.url.rsplit('_', 1)[1])
if postID < lastPostID:
if fixedPost is not None and fixedPostID > postID:
yield fixedPost
fixedPost = None
yield item
lastPostID = postID
def _get_entity(self): def _get_entity(self):
r, soup = self._initial_page() r, soup = self._initial_page()