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 WorkOnURL( self, gallery_token_name, gallery_seed_log, file_seeds_callable, status_hook, title_hook, network_job_factory, network_job_presentation_context_factory, file_import_options, 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
added_new_gallery_pages = False
stop_reason = ''
try:
gallery_url = self.url
url_for_child_referral = gallery_url
( url_type, match_name, can_parse, cannot_parse_reason ) = HG.client_controller.network_engine.domain_manager.GetURLParseCapability( gallery_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( 'Cannot parse {}: {}'.format( match_name, cannot_parse_reason) )
( url_to_check, parser ) = HG.client_controller.network_engine.domain_manager.GetURLToFetchAndParser( gallery_url )
status_hook( 'downloading gallery page' )
if self._referral_url is not None and self._referral_url != url_to_check:
referral_url = self._referral_url
elif gallery_url != url_to_check:
referral_url = gallery_url
else:
referral_url = None
network_job = network_job_factory( 'GET', url_to_check, referral_url = referral_url )
network_job.SetGalleryToken( gallery_token_name )
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()
parsing_text = network_job.GetContentText()
actual_fetched_url = network_job.GetActualFetchedURL()
do_parse = True
if actual_fetched_url != url_to_check:
( url_type, match_name, can_parse, cannot_parse_reason ) = HG.client_controller.network_engine.domain_manager.GetURLParseCapability( actual_fetched_url )
if url_type == HC.URL_TYPE_GALLERY:
if can_parse:
gallery_url = actual_fetched_url
url_for_child_referral = gallery_url
( url_to_check, parser ) = HG.client_controller.network_engine.domain_manager.GetURLToFetchAndParser( gallery_url )
else:
do_parse = False
status = CC.STATUS_ERROR
note = 'Could not parse {}: {}'.format( match_name, cannot_parse_reason )
else:
do_parse = False
from hydrus.client.importing import ClientImportFileSeeds
file_seed = ClientImportFileSeeds.FileSeed( ClientImportFileSeeds.FILE_SEED_TYPE_URL, actual_fetched_url )
file_seed.SetReferralURL( url_for_child_referral )
file_seeds = [ file_seed ]
file_seeds_callable( ( file_seed, ) )
status = CC.STATUS_SUCCESSFUL_AND_NEW
note = 'was redirected to a non-gallery url, which has been queued as a file import'
if do_parse:
parsing_context = {}
parsing_context[ 'gallery_url' ] = gallery_url
parsing_context[ 'url' ] = url_to_check
parsing_context[ 'post_index' ] = '0'
all_parse_results = parser.Parse( parsing_context, parsing_text )
if len( all_parse_results ) == 0:
raise HydrusExceptions.VetoException( 'The parser found nothing in the document!' )
file_seeds = ClientImporting.ConvertAllParseResultsToFileSeeds( all_parse_results, url_for_child_referral, file_import_options )
title = ClientParsing.GetTitleFromAllParseResults( all_parse_results )
if title is not None:
title_hook( title )
for file_seed in file_seeds:
file_seed.SetExternalFilterableTags( self._external_filterable_tags )
file_seed.SetExternalAdditionalServiceKeysToTags( self._external_additional_service_keys_to_tags )
num_urls_total = len( file_seeds )
( num_urls_added, num_urls_already_in_file_seed_cache, can_search_for_more_files, stop_reason ) = file_seeds_callable( file_seeds )
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_search_for_more_files:
note += ' - ' + stop_reason
if parser.CanOnlyGenerateGalleryURLs() or self._force_next_page_url_generation:
can_add_more_gallery_urls = True
else:
# 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_added > 0 and can_search_for_more_files
flattened_results = list( itertools.chain.from_iterable( all_parse_results ) )
sub_gallery_urls = ClientParsing.GetURLsFromParseResults( flattened_results, ( HC.URL_TYPE_SUB_GALLERY, ), only_get_top_priority = True )
sub_gallery_urls = HydrusData.DedupeList( sub_gallery_urls )
new_sub_gallery_urls = [ sub_gallery_url for sub_gallery_url in sub_gallery_urls if sub_gallery_url not in gallery_urls_seen_before ]
num_new_sub_gallery_urls = len( new_sub_gallery_urls )
if num_new_sub_gallery_urls > 0:
sub_gallery_seeds = [ GallerySeed( sub_gallery_url ) for sub_gallery_url in new_sub_gallery_urls ]
for sub_gallery_seed in sub_gallery_seeds:
sub_gallery_seed.SetRunToken( self._run_token )
sub_gallery_seed.SetExternalFilterableTags( self._external_filterable_tags )
sub_gallery_seed.SetExternalAdditionalServiceKeysToTags( self._external_additional_service_keys_to_tags )
gallery_seed_log.AddGallerySeeds( sub_gallery_seeds )
added_new_gallery_pages = True
gallery_urls_seen_before.update( sub_gallery_urls )
note += ' - {} sub-gallery urls found'.format( HydrusData.ToHumanInt( num_new_sub_gallery_urls ) )
if self._can_generate_more_pages and can_add_more_gallery_urls:
next_page_urls = ClientParsing.GetURLsFromParseResults( flattened_results, ( HC.URL_TYPE_NEXT, ), only_get_top_priority = True )
if self.url in next_page_urls:
next_page_urls.remove( self.url )
if url_to_check in next_page_urls:
next_page_urls.remove( url_to_check )
if len( next_page_urls ) > 0:
next_page_generation_phrase = ' next gallery pages found'
else:
# we have failed to parse a next page url, but we would still like one, so let's see if the url match can provide one
url_class = HG.client_controller.network_engine.domain_manager.GetURLClass( url_to_check )
if url_class is not None and url_class.CanGenerateNextGalleryPage():
try:
next_page_url = url_class.GetNextGalleryPage( url_to_check )
next_page_urls = [ next_page_url ]
except Exception as e:
note += ' - Attempted to generate a next gallery page url, but failed!'
note += os.linesep
note += traceback.format_exc()
next_page_generation_phrase = ' next gallery pages extrapolated from url class'
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 ]
for next_gallery_seed in next_gallery_seeds:
next_gallery_seed.SetRunToken( self._run_token )
next_gallery_seed.SetReferralURL( url_for_child_referral )
next_gallery_seed.SetExternalFilterableTags( self._external_filterable_tags )
next_gallery_seed.SetExternalAdditionalServiceKeysToTags( self._external_additional_service_keys_to_tags )
gallery_seed_log.AddGallerySeeds( next_gallery_seeds )
added_new_gallery_pages = True
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_page_generation_phrase
else:
note += ' - ' + HydrusData.ToHumanInt( num_new_next_page_urls ) + next_page_generation_phrase + ', 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_page_generation_phrase + ', 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 = str( e )
self.SetStatus( status, note = note )
if isinstance( e, HydrusExceptions.CancelledException ):
status_hook( 'cancelled!' )
time.sleep( 2 )
except HydrusExceptions.InsufficientCredentialsException:
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
finally:
gallery_seed_log.NotifyGallerySeedsUpdated( ( self, ) )
return ( num_urls_added, num_urls_already_in_file_seed_cache, num_urls_total, result_404, added_new_gallery_pages, stop_reason )
