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

[mediawiki_dump_tools.git] / wikiq

diff --git a/wikiq b/wikiq
index 7cf4be26e60db633db9d210e7f5ec74addf1f357..a6f079eff7e3c80d04b251949eef8269d3e837ae 100755 (executable)
--- a/wikiq
+++ b/wikiq
@@ -63,6 +63,11 @@ class WikiqPage():
     def __init__(self, page, namespace_map, collapse_user=False):
         self.id = page.id
         self.namespace = page.namespace
     def __init__(self, page, namespace_map, collapse_user=False):
         self.id = page.id
         self.namespace = page.namespace

+        # following mwxml, we assume namespace 0 in cases where
+        # page.namespace is inconsistent with namespace_map
+        if page.namespace not in namespace_map:
+            self.title = page.title
+            page.namespace = 0

         if page.namespace != 0:
             self.title = ':'.join([namespace_map[page.namespace], page.title])
         else:
         if page.namespace != 0:
             self.title = ':'.join([namespace_map[page.namespace], page.title])
         else:


@@ -123,7 +128,7 @@ class WikiqPage():
 
 class WikiqParser():
     
 
 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
         """ 
         Parameters:
            persist : what persistence method to use. Takes a PersistMethod value


@@ -136,7 +141,11 @@ class WikiqParser():
         self.printed_header = False
         self.namespaces = []
         self.urlencode = urlencode
         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
 
     def __get_namespace_from_title(self, title):
         default_ns = None
 


@@ -171,6 +180,13 @@ class WikiqParser():
 
         # Iterate through pages
         for page in dump:
 
         # 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:
             rev_detector = mwreverts.Detector()
 
             if self.persist != PersistMethod.none:


@@ -189,6 +205,8 @@ class WikiqParser():
                     from mw.lib import persistence
                     state = persistence.State()
 
                     from mw.lib import persistence
                     state = persistence.State()
 

+                
+

             # Iterate through a page's revisions
             for rev in page:
 
             # Iterate through a page's revisions
             for rev in page:
 


@@ -197,7 +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 + '"',
                             '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
                             'deleted' : "TRUE" if rev.deleted.text else "FALSE" } 
 
                 # if revisions are deleted, /many/ things will be missing


@@ -368,6 +386,10 @@ parser.add_argument('-p', '--persistence', dest="persist", default=None, const='
 parser.add_argument('-u', '--url-encode', dest="urlencode", action="store_true",
                     help="Output url encoded text strings. This works around some data issues like newlines in editor names. In the future it may be used to output other text data.")
 
 parser.add_argument('-u', '--url-encode', dest="urlencode", action="store_true",
                     help="Output url encoded text strings. This works around some data issues like newlines in editor names. In the future it may be used to output other text data.")
 

+parser.add_argument('-n', '--namespace-include', dest="namespace_filter", type=int, action='append',
+                    help="Id number of namspace to include.")
+
+

 args = parser.parse_args()
 
 # set persistence method
 args = parser.parse_args()
 
 # set persistence method


@@ -381,6 +403,11 @@ elif args.persist == "legacy":
 else:
     persist = PersistMethod.sequence
 
 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)
 if len(args.dumpfiles) > 0:
     for filename in args.dumpfiles:
         input_file = open_input_file(filename)


@@ -400,9 +427,10 @@ if len(args.dumpfiles) > 0:
             output_file = open_output_file(filename)
 
             wikiq = WikiqParser(input_file, output_file, 
             output_file = open_output_file(filename)
 
             wikiq = WikiqParser(input_file, output_file, 

-                            collapse_user=args.collapse_user,
+                                collapse_user=args.collapse_user,

                                 persist=persist,
                                 persist=persist,

-                            urlencode=args.urlencode)
+                                urlencode=args.urlencode,
+                                namespaces = namespaces)

 
 
         wikiq.process()
 
 
         wikiq.process()