checking in work to deepen migration to new mediawikiutils

[mediawiki_dump_tools.git] / wikiq

diff --git a/wikiq b/wikiq
index ae6b40f07970b08975afbd545827c09b5509d08a..f8e5fd616aeeb094258c2523a6daa39b8064c437 100755 (executable)
--- a/wikiq
+++ b/wikiq
@@ -1,9 +1,9 @@
-#!/usr/bin/env python3
+ #!/usr/bin/env python3

 
 # 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
 
 # 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
 import argparse
 import sys
 import os, os.path


@@ -13,16 +13,63 @@ from subprocess import Popen, PIPE
 from collections import deque
 from hashlib import sha1
 
 from collections import deque
 from hashlib import sha1
 

-from mwxml import Dump
+from mwxml import Dump, Page

 
 from deltas.tokenizers import wikitext_split
 
 from deltas.tokenizers import wikitext_split

+from mwdiffs.utilities import dump2diffs

 import mwpersistence
 import mwpersistence

+from mwpersistence.state import Version, apply_opdocs, apply_operations, persist_revision_once
+
+from mwpersistence import Token
+from  mwpersistence.utilities import diffs2persistence 

 import mwreverts
 from urllib.parse import quote
 import mwreverts
 from urllib.parse import quote

-TO_ENCODE = ('title', 'editor')
-PERSISTENCE_RADIUS=7
+

 from deltas import SequenceMatcher
 from deltas import SegmentMatcher
 from deltas import SequenceMatcher
 from deltas import SegmentMatcher

+TO_ENCODE = ('title', 'editor')
+PERSISTENCE_RADIUS=7
+
+# this is a simple override of mwpersistence.DiffState that doesn't do anything special for reverts. 
+class WikiqDiffState(mwpersistence.DiffState):
+    def _update(self, text=None, checksum=None, opdocs=None, revision=None):
+        if checksum is None:
+            if text is None:
+                raise TypeError("Either 'text' or 'checksum' must be " +
+                                "specified.")
+            else:
+                checksum = sha1(bytes(text, 'utf8')).hexdigest()
+
+        current_version = Version()
+
+        # the main difference we have is that we don't do anything special for reverts
+        if opdocs is not None:
+            transition = apply_opdocs(opdocs, self.last.tokens or [])
+            current_version.tokens, _, _ = transition
+        else:
+            # NOTICE: HEAVY COMPUTATION HERE!!!
+            #
+            # Diffs usually run in O(n^2) -- O(n^3) time and most
+            # tokenizers produce a lot of tokens.
+            if self.diff_processor is None:
+                raise RuntimeError("DiffState cannot process raw text " +
+                                   "without a diff_engine specified.")
+            operations, _, current_tokens = \
+                self.diff_processor.process(text, token_class=Token)
+
+            transition = apply_operations(operations,
+                                          self.last.tokens or [],
+                                          current_tokens)
+            current_version.tokens, _, _ = transition
+
+        # Record persistence
+        persist_revision_once(current_version.tokens, revision)
+
+        # Update last version
+        self.last = current_version
+
+        # Return the tranisitoned state
+        return transition

 
 class PersistMethod:
     none = 0
 
 class PersistMethod:
     none = 0


@@ -34,97 +81,150 @@ def calculate_persistence(tokens_added):
     return(sum([(len(x.revisions)-1) for x in tokens_added]),
            len(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
-        self.collapse_user = collapse_user
-        self.mwiterator = Dump.from_file(self.fh)
-        self.namespace_map = { ns.id : ns.name for ns in
-                               self.mwiterator.site_info.namespaces }
-        self.__pages = self.load_pages()
-
-    def load_pages(self):
-        for page in self.mwiterator:
-            yield WikiqPage(page,
-                            namespace_map = self.namespace_map,
-                            collapse_user=self.collapse_user)
+class WikiqIterator(Dump):
+
+    @classmethod
+    def from_file(cls, fh, collapse_user = False):
+        cls = super(WikiqIterator, cls).from_file(fh)
+        cls.fh = fh
+        cls.collapse_user = collapse_user
+        cls.namespace_map = { ns.id : ns.name for ns in
+                               cls.site_info.namespaces }
+        return cls
+
+    @classmethod
+    def process_item(cls, item_element, namespace_map, collapse_user = False):
+        if item_element.tag == "page":
+            return WikiqPage.from_element(item_element, namespace_map, collapse_user)
+        elif item_element.tag == "logitem":
+            return LogItem.from_element(item_element, namespace_map)
+        else:
+            raise MalformedXML("Expected to see <page> or <logitem>.  " +
+                               "Instead saw <{0}>".format(item_element.tag))

 
 

-    def __iter__(self):
-        return self.__pages
+class WikiqPage(Page):
+    __slots__ = ('id', 'title', 'namespace', 'redirect',
+                 'restrictions','collapse_user')
+        
+    @classmethod
+    def from_element(cls, item_element, namespace_map, collapse_user = False):
+        cls.prev_rev = None

 
 

-    def __next__(self):
-        return next(self._pages)
+        inv_namespace_map = {ns.id:name for name,ns in namespace_map.items()}
+        
+        cls = super(WikiqPage, cls).from_element(item_element, namespace_map)

 
 

-class WikiqPage():
-    __slots__ = ('id', 'title', 'namespace', 'redirect',
-                 'restrictions', 'mwpage', '__revisions',
-                 'collapse_user')
-    
-    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
         # 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])
+        # this undoes the "correction" of the namespace in mwxml
+        
+        if cls.namespace not in inv_namespace_map:
+            cls.namespace = 0
+        if cls.namespace != 0:
+            cls.title = ':'.join([inv_namespace_map[cls.namespace], cls.title])
+
+        cls.collapse_user = collapse_user
+        cls.revisions = cls._Page__revisions
+        return cls
+
+    @staticmethod
+    def _correct_sha(rev_data):
+
+        if rev_data.deleted.text:
+            rev_data.text = ""
+            rev_data.text_chars = 0
+            rev_data.sha1 = ""
+            rev_data.revert = ""
+            rev_data.reverteds = ""
+

         else:
         else:

-            self.title = page.title
-        self.restrictions = page.restrictions
-        self.collapse_user = collapse_user
-        self.mwpage = page
-        self.__revisions = self.rev_list()
-
-    def rev_list(self):
-        # Outline for how we want to handle collapse_user=True
-        # iteration   rev.user   prev_rev.user   add prev_rev?
-        #         0          A            None           Never
-        #         1          A               A           False
-        #         2          B               A            True
-        #         3          A               B            True
-        #         4          A               A           False
-        # Post-loop                          A          Always
-        for i, rev in enumerate(self.mwpage):
-            # never yield the first time
-            if i == 0:
-                if self.collapse_user: 
-                    collapsed_revs = 1
-                    rev.collapsed_revs = collapsed_revs
+            if rev_data.text is None :
+                rev_data.text = ""
+                
+        rev_data.text_chars = len(rev_data.text)

 
 

-            else:
-                if self.collapse_user:
-                    # yield if this is the last edit in a seq by a user and reset
-                    # also yield if we do know who the user is
-
-                    if rev.deleted.user or prev_rev.deleted.user:
-                        yield prev_rev
-                        collapsed_revs = 1
-                        rev.collapsed_revs = collapsed_revs
-
-                    elif not rev.user.text == prev_rev.user.text:
-                        yield prev_rev
-                        collapsed_revs = 1
-                        rev.collapsed_revs = collapsed_revs
-                    # otherwise, add one to the counter
-                    else:
-                        collapsed_revs += 1
-                        rev.collapsed_revs = collapsed_revs
-                # if collapse_user is false, we always yield
-                else:
-                    yield prev_rev
+        if hasattr(rev_data,"sha1") and rev_data.sha1 is not None:
+            text_sha1 = rev_data.sha1
+
+        else:
+            text_sha1 = sha1(bytes(rev_data.text, "utf8")).hexdigest()
+
+        rev_data.sha1 = text_sha1
+
+        return rev_data 

 
 

+    # Outline for how we want to handle collapse_user=True
+    # iteration   rev.user   prev_rev.user   add prev_rev?
+    #         0          A            None           Never
+    #         1          A               A           False
+    #         2          B               A            True
+    #         3          A               B            True
+    #         4          A               A           False
+    # Post-loop                          A          Always
+    def __find_next_revision(self):
+
+        if self.prev_rev is None:
+            prev_rev = WikiqPage._correct_sha(next(self.revisions))
+            self.prev_rev = prev_rev
+        else:
+            prev_rev = self.prev_rev
+
+        if self.collapse_user: 
+            collapsed_revs = 1
+            rev.collapsed_revs = collapsed_revs
+
+        for rev in self.revisions:
+            rev = WikiqPage._correct_sha(rev)
+            if self.collapse_user:
+                # yield if this is the last edit in a seq by a user and reset
+                # also yield if we do know who the user is
+
+                if rev.deleted.user or prev_rev.deleted.user:
+                    self.prev_rev = rev
+                    if prev_rev is not None:
+                        prev_rev.collapsed_revs = collapsed_revs
+                        return prev_rev
+
+                elif not rev.user.text == prev_rev.user.text:
+                    self.prev_rev = rev
+                    if prev_rev is not None:
+                        prev_rev.collapsed_revs = collapsed_revs
+                        return prev_rev
+
+                # otherwise, add one to the counter
+                else:
+                    collapsed_revs += 1
+                    rev.collapsed_revs = collapsed_revs
+                # if collapse_user is false, we always yield
+            else:
+                self.prev_rev = rev
+                if prev_rev is not None:
+                    return prev_rev

             prev_rev = rev
 
             prev_rev = rev
 

-        # also yield the final time
-        yield prev_rev
+        self.prev_rev = None
+
+        if self.collapse_user:
+            prev_rev.collapsed_revs = collapsed_revs
+        return prev_rev

 
 

-    def __iter__(self):
-        return self.__revisions

 
     def __next__(self):
 
     def __next__(self):

-        return next(self.__revisions)
+        revision = self.__find_next_revision()
+        revision.page = self
+        return revision
+
+    def __iter__(self):
+        while(True):
+            revision = self.__find_next_revision()
+            revision.page = self
+            yield revision
+
+    # def __iter__(self):
+    #     return self.__revisions
+
+    # def __next__(self):
+    #     return next(self.__revisions)

 
 class WikiqParser():
     
 
 class WikiqParser():
     


@@ -133,7 +233,6 @@ class WikiqParser():
         Parameters:
            persist : what persistence method to use. Takes a PersistMethod value
         """
         Parameters:
            persist : what persistence method to use. Takes a PersistMethod value
         """

-

         self.input_file = input_file
         self.output_file = output_file
         self.collapse_user = collapse_user
         self.input_file = input_file
         self.output_file = output_file
         self.collapse_user = collapse_user


@@ -146,46 +245,54 @@ class WikiqParser():
         else:
             self.namespace_filter = None
 
         else:
             self.namespace_filter = None
 

-    def __get_namespace_from_title(self, title):
-        default_ns = None
-
-        for ns in self.namespaces:
-            # skip if the namespace is not defined
-            if ns == None:
-                default_ns = self.namespaces[ns]
-                continue
-
-            if title.startswith(ns + ":"):
-                return self.namespaces[ns]
-
-        # if we've made it this far with no matches, we return the default namespace
-        return default_ns
-
-    def process(self):
-

         # create a regex that creates the output filename
         # output_filename = re.sub(r'^.*/(enwiki\-\d+)\-.*p(\d+)p.*$',
         #                         r'output/wikiq-\1-\2.tsv',
         #                         input_filename)
 
         # Construct dump file iterator
         # create a regex that creates the output filename
         # output_filename = re.sub(r'^.*/(enwiki\-\d+)\-.*p(\d+)p.*$',
         #                         r'output/wikiq-\1-\2.tsv',
         #                         input_filename)
 
         # Construct dump file iterator

-        dump = WikiqIterator(self.input_file, collapse_user=self.collapse_user)
+        self.dump = WikiqIterator.from_file(self.input_file, self.collapse_user)
+      
+        self.diff_engine = None
+
+        if self.persist == PersistMethod.sequence:  
+            self.diff_engine = SequenceMatcher(tokenizer = wikitext_split)
+
+        if self.persist == PersistMethod.segment:
+            self.diff_engine = SegmentMatcher(tokenizer = wikitext_split)
+
+    # def __get_namespace_from_title(self, title):
+    #     default_ns = None
+
+    #     for ns in self.namespaces:
+    #         # skip if the namespace is not defined
+    #         if ns == None:
+    #             default_ns = self.namespaces[ns]
+    #             continue

 
 

-        # extract list of namspaces
-        self.namespaces = {ns.name : ns.id for ns in dump.mwiterator.site_info.namespaces}
+    #         if title.startswith(ns + ":"):
+    #             return self.namespaces[ns]

 
 

+    #     # if we've made it this far with no matches, we return the default namespace
+    #     return default_ns
+
+    # def _set_namespace(self, rev_docs):
+        
+    #     for rev_data in rev_docs:
+    #         if 'namespace' not in rev_data['page']:
+    #             namespace = self.__get_namespace_from_title(page['title'])
+    #             rev_data['page']['namespace'] = namespace
+    #         yield rev_data
+
+    def process(self):

         page_count = 0
         rev_count = 0
 
         page_count = 0
         rev_count = 0
 

+        for page in self.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 not in self.namespace_filter:
-                    continue
+            # skip pages not in the namespaces we want
+            if self.namespace_filter is not None and page.namespace not in self.namespace_filter:
+                continue

 
             rev_detector = mwreverts.Detector()
 
 
             rev_detector = mwreverts.Detector()
 


@@ -193,27 +300,25 @@ class WikiqParser():
                 window = deque(maxlen=PERSISTENCE_RADIUS)
 
                 if self.persist == PersistMethod.sequence:
                 window = deque(maxlen=PERSISTENCE_RADIUS)
 
                 if self.persist == PersistMethod.sequence:

-                    state = mwpersistence.DiffState(SequenceMatcher(tokenizer = wikitext_split),
+                    state = WikiqDiffState(SequenceMatcher(tokenizer = wikitext_split),

                                                     revert_radius=PERSISTENCE_RADIUS)
 
                 elif self.persist == PersistMethod.segment:
                                                     revert_radius=PERSISTENCE_RADIUS)
 
                 elif self.persist == PersistMethod.segment:

-                    state = mwpersistence.DiffState(SegmentMatcher(tokenizer = wikitext_split),
+                    state = WikiqDiffState(SegmentMatcher(tokenizer = wikitext_split),

                                                     revert_radius=PERSISTENCE_RADIUS)
 
                                                     revert_radius=PERSISTENCE_RADIUS)
 

-                # self.persist == PersistMethod.legacy

                 else:
                     from mw.lib import persistence
                     state = persistence.State()
 
             # Iterate through a page's revisions
             for rev in page:
                 else:
                     from mw.lib import persistence
                     state = persistence.State()
 
             # Iterate through a page's revisions
             for rev in page:

-

                 rev_data = {'revid' : rev.id,
                             'date_time' : rev.timestamp.strftime('%Y-%m-%d %H:%M:%S'),
                             'articleid' : page.id,
                             'editor_id' : "" if rev.deleted.user == True or rev.user.id is None else rev.user.id,
                             'title' : '"' + page.title + '"',
                 rev_data = {'revid' : rev.id,
                             'date_time' : rev.timestamp.strftime('%Y-%m-%d %H:%M:%S'),
                             'articleid' : page.id,
                             'editor_id' : "" if rev.deleted.user == True or rev.user.id is None else rev.user.id,
                             'title' : '"' + page.title + '"',

-                            'namespace' : namespace,
+                            'namespace' : page.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


@@ -270,15 +375,15 @@ class WikiqParser():
                 #TODO missing: additions_size deletions_size
                 
                 # if collapse user was on, lets run that
                 #TODO missing: additions_size deletions_size
                 
                 # if collapse user was on, lets run that

-                if self.collapse_user:
-                    rev_data['collapsed_revs'] = rev.collapsed_revs
+                # if self.collapse_user:
+                #     rev_data.collapsed_revs = rev.collapsed_revs

 
                 if self.persist != PersistMethod.none:
                     if rev.deleted.text:
                         for k in ["token_revs", "tokens_added", "tokens_removed", "tokens_window"]:
                             old_rev_data[k] = None
                     else:
 
                 if self.persist != PersistMethod.none:
                     if rev.deleted.text:
                         for k in ["token_revs", "tokens_added", "tokens_removed", "tokens_window"]:
                             old_rev_data[k] = None
                     else:

-
+ 

                         if self.persist != PersistMethod.legacy:
                             _, tokens_added, tokens_removed = state.update(rev.text, rev.id)
 
                         if self.persist != PersistMethod.legacy:
                             _, tokens_added, tokens_removed = state.update(rev.text, rev.id)