def ConvertAllParseResultsToFileSeeds(all_parse_results, source_url,
file_import_options):
file_seeds = []
seen_urls = set()
for parse_results in all_parse_results:
parsed_urls = ClientParsing.GetURLsFromParseResults(
parse_results, (HC.URL_TYPE_DESIRED, ), only_get_top_priority=True)
parsed_urls = HydrusData.DedupeList(parsed_urls)
parsed_urls = [url for url in parsed_urls if url not in seen_urls]
seen_urls.update(parsed_urls)
# note we do this recursively due to parse_results being appropriate only for these urls--don't move this out again, or tags will be messed up
for url in parsed_urls:
file_seed = ClientImportFileSeeds.FileSeed(
ClientImportFileSeeds.FILE_SEED_TYPE_URL, url)
file_seed.SetReferralURL(source_url)
file_seed.AddParseResults(parse_results, file_import_options)
file_seeds.append(file_seed)
return file_seeds
def ForceLogins(self, domains_to_login):
with self._lock:
self._domains_to_login.extend(domains_to_login)
self._domains_to_login = HydrusData.DedupeList(
self._domains_to_login)
def WorkOnURL( self, gallery_seed_log, file_seed_cache, status_hook, title_hook, network_job_factory, network_job_presentation_context_factory, file_import_options, max_new_urls_allowed = None, gallery_urls_seen_before = None ):
if gallery_urls_seen_before is None:
gallery_urls_seen_before = set()
gallery_urls_seen_before.add( self.url )
# maybe something like 'append urls' vs 'reverse-prepend' for subs or something
# should also take--and populate--a set of urls we have seen this 'run', so we can bomb out if next_gallery_url ends up in some loop
num_urls_added = 0
num_urls_already_in_file_seed_cache = 0
num_urls_total = 0
result_404 = False
try:
( url_type, match_name, can_parse ) = HG.client_controller.network_engine.domain_manager.GetURLParseCapability( self.url )
if url_type not in ( HC.URL_TYPE_GALLERY, HC.URL_TYPE_WATCHABLE ):
raise HydrusExceptions.VetoException( 'Did not recognise this as a gallery or watchable URL!' )
if not can_parse:
raise HydrusExceptions.VetoException( 'Did not have a parser for this URL!' )
( url_to_check, parser ) = HG.client_controller.network_engine.domain_manager.GetURLToFetchAndParser( self.url )
status_hook( 'downloading page' )
if self._referral_url not in ( self.url, url_to_check ):
referral_url = self._referral_url
else:
referral_url = None
network_job = network_job_factory( 'GET', url_to_check, referral_url = referral_url )
network_job.OverrideBandwidth( 30 )
HG.client_controller.network_engine.AddJob( network_job )
with network_job_presentation_context_factory( network_job ) as njpc:
network_job.WaitUntilDone()
data = network_job.GetContent()
parsing_context = {}
parsing_context[ 'gallery_url' ] = self.url
parsing_context[ 'url' ] = url_to_check
all_parse_results = parser.Parse( parsing_context, data )
if len( all_parse_results ) == 0:
raise HydrusExceptions.VetoException( 'Could not parse any data!' )
title = ClientParsing.GetTitleFromAllParseResults( all_parse_results )
if title is not None:
title_hook( title )
( num_urls_added, num_urls_already_in_file_seed_cache, num_urls_total ) = ClientImporting.UpdateFileSeedCacheWithAllParseResults( file_seed_cache, all_parse_results, self.url, max_new_urls_allowed )
if max_new_urls_allowed is None:
can_add_more_file_urls = True
else:
can_add_more_file_urls = num_urls_added < max_new_urls_allowed
status = CC.STATUS_SUCCESSFUL_AND_NEW
note = HydrusData.ToHumanInt( num_urls_added ) + ' new urls found'
if num_urls_already_in_file_seed_cache > 0:
note += ' (' + HydrusData.ToHumanInt( num_urls_already_in_file_seed_cache ) + ' of page already in)'
if not can_add_more_file_urls:
note += ' - hit file limit'
# only keep searching if we found any files, otherwise this could be a blank results page with another stub page
can_add_more_gallery_urls = num_urls_total > 0 and can_add_more_file_urls
if self._can_generate_more_pages and can_add_more_gallery_urls:
flattened_results = list( itertools.chain.from_iterable( all_parse_results ) )
next_page_urls = ClientParsing.GetURLsFromParseResults( flattened_results, ( HC.URL_TYPE_NEXT, ), only_get_top_priority = True )
if len( next_page_urls ) > 0:
next_page_urls = HydrusData.DedupeList( next_page_urls )
new_next_page_urls = [ next_page_url for next_page_url in next_page_urls if next_page_url not in gallery_urls_seen_before ]
duplicate_next_page_urls = gallery_urls_seen_before.intersection( new_next_page_urls )
num_new_next_page_urls = len( new_next_page_urls )
num_dupe_next_page_urls = len( duplicate_next_page_urls )
if num_new_next_page_urls > 0:
next_gallery_seeds = [ GallerySeed( next_page_url ) for next_page_url in new_next_page_urls ]
gallery_seed_log.AddGallerySeeds( next_gallery_seeds )
gallery_urls_seen_before.update( new_next_page_urls )
if num_dupe_next_page_urls == 0:
note += ' - ' + HydrusData.ToHumanInt( num_new_next_page_urls ) + ' next gallery pages found'
else:
note += ' - ' + HydrusData.ToHumanInt( num_new_next_page_urls ) + ' next gallery pages found, but ' + HydrusData.ToHumanInt( num_dupe_next_page_urls ) + ' had already been visited this run and were not added'
else:
note += ' - ' + HydrusData.ToHumanInt( num_dupe_next_page_urls ) + ' next gallery pages found, but they had already been visited this run and were not added'
self.SetStatus( status, note = note )
except HydrusExceptions.ShutdownException:
pass
except HydrusExceptions.VetoException as e:
status = CC.STATUS_VETOED
note = HydrusData.ToUnicode( e )
self.SetStatus( status, note = note )
if isinstance( e, HydrusExceptions.CancelledException ):
status_hook( 'cancelled!' )
time.sleep( 2 )
except HydrusExceptions.ForbiddenException:
status = CC.STATUS_VETOED
note = '403'
self.SetStatus( status, note = note )
status_hook( '403' )
time.sleep( 2 )
result_404 = True
except HydrusExceptions.NotFoundException:
status = CC.STATUS_VETOED
note = '404'
self.SetStatus( status, note = note )
status_hook( '404' )
time.sleep( 2 )
result_404 = True
except Exception as e:
status = CC.STATUS_ERROR
self.SetStatus( status, exception = e )
status_hook( 'error!' )
time.sleep( 3 )
if isinstance( e, HydrusExceptions.NetworkException ): # so the larger queue can set a delaywork or whatever
raise
gallery_seed_log.NotifyGallerySeedsUpdated( ( self, ) )
return ( num_urls_added, num_urls_already_in_file_seed_cache, num_urls_total, result_404 )
