Merge branch 'advanced_persistence' of code.communitydata.cc:mediawiki_dump_tools into advanced_persistence

This commit is contained in:
Nathan TeBlunthuis 2018-08-23 18:27:09 -07:00
commit d93769c21f

32
wikiq
View File

@ -128,7 +128,7 @@ class WikiqPage():
class WikiqParser():
def __init__(self, input_file, output_file, collapse_user=False, persist=None, urlencode=False):
def __init__(self, input_file, output_file, collapse_user=False, persist=None, urlencode=False, namespaces = None):
"""
Parameters:
persist : what persistence method to use. Takes a PersistMethod value
@ -141,7 +141,11 @@ class WikiqParser():
self.printed_header = False
self.namespaces = []
self.urlencode = urlencode
if namespaces is not None:
self.namespace_filter = set(namespaces)
else:
self.namespace_filter = None
def __get_namespace_from_title(self, title):
default_ns = None
@ -176,6 +180,13 @@ class WikiqParser():
# Iterate through pages
for page in dump:
namespace = page.namespace if page.namespace is not None else self.__get_namespace_from_title(page.title)
# skip namespaces not in the filter
if self.namespace_filter is not None:
if namespace in self.namespace_filter:
continue
rev_detector = mwreverts.Detector()
if self.persist != PersistMethod.none:
@ -194,6 +205,8 @@ class WikiqParser():
from mw.lib import persistence
state = persistence.State()
# Iterate through a page's revisions
for rev in page:
@ -202,7 +215,7 @@ class WikiqParser():
'articleid' : page.id,
'editor_id' : "" if rev.deleted.user == True or rev.user.id is None else rev.user.id,
'title' : '"' + page.title + '"',
'namespace' : page.namespace if page.namespace is not None else self.__get_namespace_from_title(page.title),
'namespace' : namespace,
'deleted' : "TRUE" if rev.deleted.text else "FALSE" }
# if revisions are deleted, /many/ things will be missing
@ -378,7 +391,7 @@ parser.add_argument('-n', '--namespace-include', dest="namespace_filter", type=i
args = parser.parse_args()
print(args)
# set persistence method
if args.persist is None:
@ -390,6 +403,11 @@ elif args.persist == "legacy":
else:
persist = PersistMethod.sequence
if args.namespace_filter is not None:
namespaces = [int(ns) for ns in args.namespace_filter.split(',')]
else:
namespaces = None
if len(args.dumpfiles) > 0:
for filename in args.dumpfiles:
input_file = open_input_file(filename)
@ -409,9 +427,11 @@ if len(args.dumpfiles) > 0:
output_file = open_output_file(filename)
wikiq = WikiqParser(input_file, output_file,
collapse_user=args.collapse_user,
collapse_user=args.collapse_user,
persist=persist,
urlencode=args.urlencode)
urlencode=args.urlencode,
namespaces = namespaces)
wikiq.process()