migrate to mwpersistence. this fixes many issues. We preserve legacy persistence...

[mediawiki_dump_tools.git] / wikiq

diff --git a/wikiq b/wikiq
index 7a2f8e45ca20d61464e042704f675f214fd0be09..9260f35710d029e99f2c4bffe470aa000e7f1615 100755 (executable)
--- a/wikiq
+++ b/wikiq
@@ -3,7 +3,7 @@
 # original wikiq headers are: title articleid revid date_time anon
 # editor editor_id minor text_size text_entropy text_md5 reversion
 # additions_size deletions_size
-
+import pdb
 import argparse
 import sys
 import os, os.path
@@ -14,16 +14,20 @@ from collections import deque
 from hashlib import sha1
 
 from mw.xml_dump import Iterator
-from mw.lib import persistence
-from mw.lib import reverts
+
+from deltas.tokenizers import wikitext_split
+import mwpersistence
+import mwreverts
 from urllib.parse import quote
 TO_ENCODE = ('title', 'editor')
 PERSISTENCE_RADIUS=7
+from deltas import SequenceMatcher
 
 def calculate_persistence(tokens_added):
     return(sum([(len(x.revisions)-1) for x in tokens_added]),
            len(tokens_added))
 
+
 class WikiqIterator():
     def __init__(self, fh, collapse_user=False):
         self.fh = fh
@@ -101,12 +105,13 @@ class WikiqPage():
 class WikiqParser():
 
 
-    def __init__(self, input_file, output_file, collapse_user=False, persist=False, urlencode=False):
+    def __init__(self, input_file, output_file, collapse_user=False, persist=False, urlencode=False, persist_legacy=False):
         
         self.input_file = input_file
         self.output_file = output_file
         self.collapse_user = collapse_user
         self.persist = persist
+        self.persist_legacy = persist_legacy
         self.printed_header = False
         self.namespaces = []
         self.urlencode = urlencode
@@ -141,13 +146,22 @@ class WikiqParser():
 
         page_count = 0
         rev_count = 0
+
+
         # Iterate through pages
         for page in dump:
-            if self.persist:
-                state = persistence.State()
+            rev_detector = mwreverts.Detector()
+
+            if self.persist or self.persist_legacy:
                 window = deque(maxlen=PERSISTENCE_RADIUS)
 
-            rev_detector = reverts.Detector()
+                if not self.persist_legacy:
+                    state = mwpersistence.DiffState(SequenceMatcher(tokenizer = wikitext_split),
+                                                    revert_radius=PERSISTENCE_RADIUS)
+
+                else:
+                    from mw.lib import persistence
+                    state = persistence.State()
 
             # Iterate through a page's revisions
             for rev in page:
@@ -181,6 +195,7 @@ class WikiqParser():
                
                     # generate revert data
                     revert = rev_detector.process(text_sha1, rev.id)
+                    
                     if revert:
                         rev_data['revert'] = "TRUE"
                         rev_data['reverteds'] = '"' + ",".join([str(x) for x in revert.reverteds]) + '"'
@@ -211,12 +226,18 @@ class WikiqParser():
                 if self.collapse_user:
                     rev_data['collapsed_revs'] = rev.collapsed_revs
 
-                if self.persist:
+                if self.persist or self.persist_legacy:
                     if rev.text.deleted:
                         for k in ["token_revs", "tokens_added", "tokens_removed", "tokens_window"]:
                             old_rev_data[k] = None
                     else:
-                        _, tokens_added, tokens_removed = state.process(rev.text, rev.id, text_sha1)
+
+                        if not self.persist_legacy:
+                            _, tokens_added, tokens_removed = state.update(rev.text, rev.id)
+
+                        else:
+                            _, tokens_added, tokens_removed = state.process(rev.text, rev.id,text_sha1)
+                            
                         window.append((rev.id, rev_data, tokens_added, tokens_removed))
                         
                         if len(window) == PERSISTENCE_RADIUS:
@@ -236,7 +257,7 @@ class WikiqParser():
 
                 rev_count += 1
 
-            if self.persist:
+            if self.persist or self.persist_legacy:
                 # print out metadata for the last RADIUS revisions
                 for i, item in enumerate(window):
                     # if the window was full, we've already printed item 0
@@ -277,7 +298,7 @@ def open_input_file(input_filename):
     elif re.match(r'.*\.gz$', input_filename):
         cmd = ["zcat", input_filename] 
     elif re.match(r'.*\.bz2$', input_filename):
-        cmd = ["zcat", input_filename] 
+        cmd = ["bzcat", "-dk", input_filename] 
 
     try:
         input_file = Popen(cmd, stdout=PIPE).stdout
@@ -316,6 +337,9 @@ parser.add_argument('-p', '--persistence', dest="persist", action="store_true",
 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('--persistence-legacy', dest="persist_legacy", action="store_true",
+                    help="Legacy behavior for persistence calculation. 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.")
+
 args = parser.parse_args()
 
 if len(args.dumpfiles) > 0:
@@ -339,6 +363,7 @@ if len(args.dumpfiles) > 0:
         wikiq = WikiqParser(input_file, output_file, 
                             collapse_user=args.collapse_user,
                             persist=args.persist,
+                            persist_legacy=args.persist_legacy,
                             urlencode=args.urlencode)
 
 
@@ -351,6 +376,7 @@ else:
     wikiq = WikiqParser(sys.stdin, sys.stdout,
                         collapse_user=args.collapse_user,
                         persist=args.persist,
+                        persist_legacy=args.persist_legacy,
                         urlencode=args.urlencode)
     wikiq.process()