@ -1013,7 +1013,7 @@ class YoutubeDL(object):
for key , value in extra_info . items ( ) :
info_dict . setdefault ( key , value )
def extract_info ( self , url , download = True , ie_key = None , info_dict= None , extra_info= { } ,
def extract_info ( self , url , download = True , ie_key = None , extra_info= { } ,
process = True , force_generic_extractor = False ) :
'''
Returns a list with a dictionary for each video we find .
@ -1049,7 +1049,7 @@ class YoutubeDL(object):
self . to_screen ( " [ %s ] %s : has already been recorded in archive " % (
ie_key , temp_id ) )
break
return self . __extract_info ( url , ie , download , extra_info , process , info_dict )
return self . __extract_info ( url , ie , download , extra_info , process )
else :
self . report_error ( ' no suitable InfoExtractor for URL %s ' % url )
@ -1076,7 +1076,7 @@ class YoutubeDL(object):
return wrapper
@__handle_extraction_exceptions
def __extract_info ( self , url , ie , download , extra_info , process , info_dict ):
def __extract_info ( self , url , ie , download , extra_info , process ):
ie_result = ie . extract ( url )
if ie_result is None : # Finished already (backwards compatibility; listformats and friends should be moved here)
return
@ -1086,11 +1086,6 @@ class YoutubeDL(object):
' _type ' : ' compat_list ' ,
' entries ' : ie_result ,
}
if info_dict :
if info_dict . get ( ' id ' ) :
ie_result [ ' id ' ] = info_dict [ ' id ' ]
if info_dict . get ( ' title ' ) :
ie_result [ ' title ' ] = info_dict [ ' title ' ]
self . add_default_extra_info ( ie_result , ie , url )
if process :
return self . process_ie_result ( ie_result , download , extra_info )
@ -1130,7 +1125,7 @@ class YoutubeDL(object):
# We have to add extra_info to the results because it may be
# contained in a playlist
return self . extract_info ( ie_result [ ' url ' ] ,
download , info_dict = ie_result ,
download ,
ie_key = ie_result . get ( ' ie_key ' ) ,
extra_info = extra_info )
elif result_type == ' url_transparent ' :