add todos to readme

[mediawiki_dump_tools.git] / wikiq

diff --git a/wikiq b/wikiq
index 03b3147446b5af284efc01d6d2435e0ee1ba9c5c..1a951e8c16228867a1a2a0e8dac43712e0eada4e 100755 (executable)
--- a/wikiq
+++ b/wikiq
@@ -8,7 +8,7 @@ import argparse
 import sys
 import os, os.path
 import re
 import sys
 import os, os.path
 import re

-from datetime import datetime
+from datetime import datetime,timezone

 
 from subprocess import Popen, PIPE
 from collections import deque
 
 from subprocess import Popen, PIPE
 from collections import deque


@@ -25,7 +25,8 @@ PERSISTENCE_RADIUS=7
 from deltas import SequenceMatcher
 from deltas import SegmentMatcher
 
 from deltas import SequenceMatcher
 from deltas import SegmentMatcher
 

-from dataclasses import dataclass
+import dataclasses as dc
+from dataclasses import dataclass, make_dataclass

 import pyarrow as pa
 import pyarrow.parquet as pq
 
 import pyarrow as pa
 import pyarrow.parquet as pq
 


@@ -39,7 +40,6 @@ 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
 class WikiqIterator():
     def __init__(self, fh, collapse_user=False):
         self.fh = fh


@@ -141,6 +141,15 @@ class RegexPair(object):
         if self.has_groups:
             self.capture_groups = list(self.pattern.groupindex.keys())
             
         if self.has_groups:
             self.capture_groups = list(self.pattern.groupindex.keys())
             

+    def get_pyarrow_fields(self):
+        if self.has_groups:
+            fields = [pa.field(self._make_key(cap_group),pa.list_(pa.string()))
+                      for cap_group in self.capture_groups]
+        else:
+            fields = [pa.field(self.label, pa.list_(pa.string()))]
+
+        return fields
+

     def _make_key(self, cap_group):
         return ("{}_{}".format(self.label, cap_group))
 
     def _make_key(self, cap_group):
         return ("{}_{}".format(self.label, cap_group))
 


@@ -187,8 +196,8 @@ class RegexPair(object):
                     temp_dict[self.label] = None
 
         # update rev_data with our new columns
                     temp_dict[self.label] = None
 
         # update rev_data with our new columns

-        for k, v in temp_dict:
-            rev_data.setattr(k,v)
+        for k, v in temp_dict.items():
+            setattr(rev_data, k, v)

 
         return rev_data
 
 
         return rev_data
 


@@ -208,33 +217,32 @@ class RevDataBase():
     minor: bool = None
     editor: str = None
     anon: bool = None
     minor: bool = None
     editor: str = None
     anon: bool = None

-    collapsed_revs:int = None

 
 

+    urlencode = False

     pa_schema_fields = [
     pa_schema_fields = [

-        pa.field("revid", pa.int64),
+        pa.field("revid", pa.int64()),

         pa.field("date_time",pa.timestamp('ms')),
         pa.field("articleid",pa.int64()),
         pa.field("editorid",pa.int64()),
         pa.field("title",pa.string()),
         pa.field("namespace",pa.int32()),
         pa.field("date_time",pa.timestamp('ms')),
         pa.field("articleid",pa.int64()),
         pa.field("editorid",pa.int64()),
         pa.field("title",pa.string()),
         pa.field("namespace",pa.int32()),

-        pa.field("deleted",pa.binary()),
+        pa.field("deleted",pa.bool_()),

         pa.field("test_chars",pa.int32()),
         pa.field("test_chars",pa.int32()),

-        pa.field("revert",pa.binary()),
+        pa.field("revert",pa.bool_()),

         pa.field("reverteds",pa.list_(pa.int64())),
         pa.field("sha1",pa.string()),
         pa.field("reverteds",pa.list_(pa.int64())),
         pa.field("sha1",pa.string()),

-        pa.field("minor",pa.binary()),
+        pa.field("minor",pa.bool_()),

         pa.field("editor",pa.string()),
         pa.field("editor",pa.string()),

-        pa.field("anon",pa.binary())
+        pa.field("anon",pa.bool_())

     ]
 
     def to_pyarrow(self):
     ]
 
     def to_pyarrow(self):

-        return pa.array(self.astuple(), map(self.pa_schema_fields, pa.field.type))
-        
+        return dc.astuple(self)

 
     def to_tsv_row(self):
         
         row = []
 
     def to_tsv_row(self):
         
         row = []

-        for f in self.fields():
+        for f in dc.fields(self):

             val = getattr(self, f.name)
             if getattr(self, f.name) is None:
                 row.append("")
             val = getattr(self, f.name)
             if getattr(self, f.name) is None:
                 row.append("")


@@ -246,43 +254,32 @@ class RevDataBase():
 
             elif f.name in {'editor','title'}:
                 s = '"' + val + '"'
 
             elif f.name in {'editor','title'}:
                 s = '"' + val + '"'

-                if f.name in TO_ENCODE:
-                    row.append(quote(str(val)))
+                if self.urlencode and f.name in TO_ENCODE:
+                    row.append(quote(str(s)))
+                else:
+                    row.append(s)

 
             elif f.type == list[int]:
                 row.append('"' + ",".join([str(x) for x in val]) + '"')
 
             elif f.type == str:
 
             elif f.type == list[int]:
                 row.append('"' + ",".join([str(x) for x in val]) + '"')
 
             elif f.type == str:

-                if f.name in TO_ENCODE:
+                if self.urlencode and f.name in TO_ENCODE:

                     row.append(quote(str(val)))
                     row.append(quote(str(val)))

+                else:
+                    row.append(val)

             else:
                 row.append(val)
 
             else:
                 row.append(val)
 

-        return '\t'.join(row)
-    
-    # def __init__(revid: int,
-    #              date_time: datetime,
-    #              articleid: int,
-    #              editorid: int,
-    #              title: str,
-    #              namespace: int,
-    #              deleted: bool,
-    #              test_chars: int,
-    #              revert: bool,
-    #              reverteds: list[bool],
-    #              sha1: str,
-    #              minor: bool,
-    #              editor: str,
-    #              anon: bool):
+        return '\t'.join(map(str,row))

 
 

-        
+    def header_row(self):
+        return '\t'.join(map(lambda f: f.name, dc.fields(self)))

 
 @dataclass()
 class RevDataCollapse(RevDataBase):
     collapsed_revs:int = None
     pa_collapsed_revs_schema = pa.field('collapsed_revs',pa.int64())
 
 @dataclass()
 class RevDataCollapse(RevDataBase):
     collapsed_revs:int = None
     pa_collapsed_revs_schema = pa.field('collapsed_revs',pa.int64())

-    pa_schema_fields = RevDataBase.pa_schema_fields + pa_collapsed_revs_schema
-    pa_schema = pa.schema(pa_schema_fields)    
+    pa_schema_fields = RevDataBase.pa_schema_fields + [pa_collapsed_revs_schema]

 
 @dataclass()
 class RevDataPersistence(RevDataBase):
 
 @dataclass()
 class RevDataPersistence(RevDataBase):


@@ -290,17 +287,18 @@ class RevDataPersistence(RevDataBase):
     tokens_added:int = None
     tokens_removed:int = None
     tokens_window:int = None
     tokens_added:int = None
     tokens_removed:int = None
     tokens_window:int = None

+

     pa_persistence_schema_fields = [
     pa_persistence_schema_fields = [

-        pa.field(token_revs, pa.int64()),
-        pa.field(tokens_added, pa.int64()),
-        pa.field(tokens_removed, pa.int64()),
-        pa.tokens_window, pa.int64()]
+        pa.field("token_revs", pa.int64()),
+        pa.field("tokens_added", pa.int64()),
+        pa.field("tokens_removed", pa.int64()),
+        pa.field("tokens_window", pa.int64())]

         
     pa_schema_fields = RevDataBase.pa_schema_fields  + pa_persistence_schema_fields
 
 @dataclass()
 class RevDataCollapsePersistence(RevDataCollapse, RevDataPersistence):
         
     pa_schema_fields = RevDataBase.pa_schema_fields  + pa_persistence_schema_fields
 
 @dataclass()
 class RevDataCollapsePersistence(RevDataCollapse, RevDataPersistence):

-    pa_scehma_fields = RevDataCollapse.pa_schema_fields + RevDataPersistence.pa_persistence_schema_fields
+    pa_schema_fields = RevDataCollapse.pa_schema_fields + RevDataPersistence.pa_persistence_schema_fields

 
 class WikiqParser():
     def __init__(self, input_file, output_file, regex_match_revision, regex_match_comment, regex_revision_label, regex_comment_label, collapse_user=False, persist=None, urlencode=False, namespaces = None, revert_radius=15, output_parquet=True, parquet_buffer_size=2000):
 
 class WikiqParser():
     def __init__(self, input_file, output_file, regex_match_revision, regex_match_comment, regex_revision_label, regex_comment_label, collapse_user=False, persist=None, urlencode=False, namespaces = None, revert_radius=15, output_parquet=True, parquet_buffer_size=2000):


@@ -316,9 +314,6 @@ class WikiqParser():
         self.urlencode = urlencode
         self.revert_radius = revert_radius
         
         self.urlencode = urlencode
         self.revert_radius = revert_radius
         

-        self.output_buffer = []
-        self.output_buffer_size = output_buffer_size
-

         if namespaces is not None:
             self.namespace_filter = set(namespaces)
         else:
         if namespaces is not None:
             self.namespace_filter = set(namespaces)
         else:


@@ -338,32 +333,39 @@ class WikiqParser():
         else:
             revdata_type = RevDataBase
 
         else:
             revdata_type = RevDataBase
 

-        regex_fields = [(field.name, list[str]), for field in self.regex_schemas]
-        self.revdata_type = dataclasses.make_dataclass('RevData_Parser',
-                                                       fields=map(regex_fields,
-                                                                  lambda pa_field: (pa_field.name,
-                                                                                    list[string],
-                                                                                    field(default=None))),
-                                                       bases=(revdata_type))
+        regex_fields = [(field.name, list[str], dc.field(default=None)) for field in self.regex_schemas]
+
+        self.revdata_type = make_dataclass('RevData_Parser',
+                                           fields=regex_fields,
+                                           bases=(revdata_type,))

         
         

-        self.revdata_type.pa_schema_fields = revdata_type.pa_schema_fields + regex_fields
+        self.revdata_type.pa_schema_fields = revdata_type.pa_schema_fields + self.regex_schemas

                         
                         

+        self.revdata_type.urlencode = self.urlencode
+

         if output_parquet is True:
             self.output_parquet = True
             self.pq_writer = None
             self.output_file = output_file
         if output_parquet is True:
             self.output_parquet = True
             self.pq_writer = None
             self.output_file = output_file

+            self.parquet_buffer = []
+            self.parquet_buffer_size = parquet_buffer_size

         else:
         else:

-            self.output_file = open(output_file,'w')
-
+            self.print_header = True
+            if output_file == sys.stdout:
+                
+                self.output_file = output_file
+            else:
+                self.output_file = open(output_file,'w')
+            self.output_parquet = False

 
     def make_matchmake_pairs(self, patterns, labels):
         if (patterns is not None and labels is not None) and \
            (len(patterns) == len(labels)):
             result = []
             for pattern, label in zip(patterns, labels):
 
     def make_matchmake_pairs(self, patterns, labels):
         if (patterns is not None and labels is not None) and \
            (len(patterns) == len(labels)):
             result = []
             for pattern, label in zip(patterns, labels):

-                result.append(RegexPair(pattern, label))
-                self.regex_schemas.append(pa.field(label, pa.list_(pa.string())))
-
+                rp = RegexPair(pattern, label)
+                result.append(rp)
+                self.regex_schemas = self.regex_schemas + rp.get_pyarrow_fields()

             return result
         elif (patterns is None and labels is None):
             return []
             return result
         elif (patterns is None and labels is None):
             return []


@@ -432,7 +434,7 @@ class WikiqParser():
 
             if self.persist != PersistMethod.none:
                 window = deque(maxlen=PERSISTENCE_RADIUS)
 
             if self.persist != PersistMethod.none:
                 window = deque(maxlen=PERSISTENCE_RADIUS)

-
+                

                 if self.persist == PersistMethod.sequence:
                     state = mwpersistence.DiffState(SequenceMatcher(tokenizer = wikitext_split),
                                                     revert_radius=PERSISTENCE_RADIUS)
                 if self.persist == PersistMethod.sequence:
                     state = mwpersistence.DiffState(SequenceMatcher(tokenizer = wikitext_split),
                                                     revert_radius=PERSISTENCE_RADIUS)


@@ -450,11 +452,12 @@ class WikiqParser():
             for rev in page:
                 
                 rev_data = self.revdata_type(revid = rev.id,
             for rev in page:
                 
                 rev_data = self.revdata_type(revid = rev.id,

-                                             date_time = rev.timestamp,
+                                             date_time = datetime.fromtimestamp(rev.timestamp.unix(), tz=timezone.utc),

                                              articleid = page.id,
                                              editorid = "" if rev.deleted.user == True or rev.user.id is None else rev.user.id,
                                              title =  page.title,
                                              articleid = page.id,
                                              editorid = "" if rev.deleted.user == True or rev.user.id is None else rev.user.id,
                                              title =  page.title,

-                                             deleted = rev.deleted.text
+                                             deleted = rev.deleted.text,
+                                             namespace = namespace

                                              )
 
                 rev_data = self.matchmake(rev, rev_data)
                                              )
 
                 rev_data = self.matchmake(rev, rev_data)


@@ -468,7 +471,6 @@ class WikiqParser():
                     if rev.sha1:
                         text_sha1 = rev.sha1
                     else:
                     if rev.sha1:
                         text_sha1 = rev.sha1
                     else:

-

                         text_sha1 = sha1(bytes(rev.text, "utf8")).hexdigest()
                     
                     rev_data.sha1 = text_sha1
                         text_sha1 = sha1(bytes(rev.text, "utf8")).hexdigest()
                     
                     rev_data.sha1 = text_sha1


@@ -477,10 +479,13 @@ class WikiqParser():
                     rev_data.text_chars = len(rev.text)
 
                     # generate revert data
                     rev_data.text_chars = len(rev.text)
 
                     # generate revert data

-                    rev_data.revert = rev_detector.process(text_sha1, rev.id)
+                    revert = rev_detector.process(text_sha1, rev.id)

                     
                     if revert:
                     
                     if revert:

+                        rev_data.revert = True

                         rev_data.reverteds = revert.reverteds
                         rev_data.reverteds = revert.reverteds

+                    else:
+                        rev_data.revert = False

 
                 # if the fact that the edit was minor can be hidden, this might be an issue
                 rev_data.minor = rev.minor
 
                 # if the fact that the edit was minor can be hidden, this might be an issue
                 rev_data.minor = rev.minor


@@ -488,8 +493,8 @@ class WikiqParser():
                 if not rev.deleted.user:
                     # wrap user-defined editors in quotes for fread
                     rev_data.editor = rev.user.text 
                 if not rev.deleted.user:
                     # wrap user-defined editors in quotes for fread
                     rev_data.editor = rev.user.text 

-                    rev_data.anon = rev.user.id == None
-                    
+                    rev_data.anon = rev.user.id is None
+                

                 #if re.match(r'^#redirect \[\[.*\]\]', rev.text, re.I):
                 #    redirect = True
                 #else:
                 #if re.match(r'^#redirect \[\[.*\]\]', rev.text, re.I):
                 #    redirect = True
                 #else:


@@ -501,8 +506,8 @@ class WikiqParser():
                 if self.collapse_user:
                     rev_data.collapsed_revs = rev.collapsed_revs
 
                 if self.collapse_user:
                     rev_data.collapsed_revs = rev.collapsed_revs
 

+                # get the 

                 if self.persist != PersistMethod.none:
                 if self.persist != PersistMethod.none:

-

                     if not rev.deleted.text:
 
                         if self.persist != PersistMethod.legacy:
                     if not rev.deleted.text:
 
                         if self.persist != PersistMethod.legacy:


@@ -518,12 +523,12 @@ class WikiqParser():
                             
                             num_token_revs, num_tokens = calculate_persistence(old_tokens_added)
 
                             
                             num_token_revs, num_tokens = calculate_persistence(old_tokens_added)
 

-                            rev_data.token_revs = num_token_revs
-                            rev_data.tokens_added = num_tokens
-                            rev_data.tokens_removed = len(old_tokens_removed)
-                            rev_data.tokens_window = PERSISTENCE_RADIUS-1
+                            old_rev_data.token_revs = num_token_revs
+                            old_rev_data.tokens_added = num_tokens
+                            old_rev_data.tokens_removed = len(old_tokens_removed)
+                            old_rev_data.tokens_window = PERSISTENCE_RADIUS-1

 
 

-                            self.print_rev_data(rev_data)
+                            self.print_rev_data(old_rev_data)

 
                 else:
                     self.print_rev_data(rev_data)
 
                 else:
                     self.print_rev_data(rev_data)


@@ -556,25 +561,41 @@ class WikiqParser():
             self.pq_writer.close()
 
         else:
             self.pq_writer.close()
 
         else:

-            output_file.close()
+            self.output_file.close()

 
 
     def write_parquet_row(self, rev_data):
         padata = rev_data.to_pyarrow()
 
 
     def write_parquet_row(self, rev_data):
         padata = rev_data.to_pyarrow()

-        self.output_buffer.append(padata)
+        self.parquet_buffer.append(padata)

 
 

-        if len(self.output_buffer) >= self.output_buffer_size:
+        if len(self.parquet_buffer) >= self.parquet_buffer_size:

             self.flush_parquet_buffer()
 
 
     def flush_parquet_buffer(self):
             self.flush_parquet_buffer()
 
 
     def flush_parquet_buffer(self):

-        outtable = pa.table.concat_arrays(self.output_buffer)                        
+        schema = pa.schema(self.revdata_type.pa_schema_fields)
+
+        def row_to_col(rg, types):
+            cols = []
+            first = rg[0]
+            for col in first:
+                cols.append([col])
+
+            for row in rg[1:]:
+                for j in range(len(cols)):
+                    cols[j].append(row[j])
+
+            arrays = []
+            for col, typ in zip(cols, types):
+                arrays.append(pa.array(col, typ))
+            return arrays
+
+        outtable = pa.Table.from_arrays(row_to_col(self.parquet_buffer, schema.types), schema=schema)

         if self.pq_writer is None:
         if self.pq_writer is None:

-            schema = pa.schema(self.revdata_type.pa_schema_field)

             self.pq_writer = pq.ParquetWriter(self.output_file, schema, flavor='spark')
 
         self.pq_writer.write_table(outtable)
             self.pq_writer = pq.ParquetWriter(self.output_file, schema, flavor='spark')
 
         self.pq_writer.write_table(outtable)

-        self.output_buffer = []
+        self.parquet_buffer = []

         
     def print_rev_data(self, rev_data):
         if self.output_parquet is False:
         
     def print_rev_data(self, rev_data):
         if self.output_parquet is False:


@@ -585,15 +606,13 @@ class WikiqParser():
         printfunc(rev_data)
 
     def write_tsv_row(self, rev_data):
         printfunc(rev_data)
 
     def write_tsv_row(self, rev_data):

-        
-        self.output_buffer.append(rev_data.to_tsv_line())
-
-        if len(self.output_buffer) >= self.output_buffer_size:
-            self.flush_tsv_buffer()
+        if self.print_header:
+            print(rev_data.header_row(), file=self.output_file)
+            self.print_header = False

 
 

+        line = rev_data.to_tsv_row()
+        print(line, file=self.output_file)

 
 

-    def flush_tsv_buffer():
-        if self.output_header:

 
 def open_input_file(input_filename):
     if re.match(r'.*\.7z$', input_filename):
 
 def open_input_file(input_filename):
     if re.match(r'.*\.7z$', input_filename):


@@ -724,7 +743,6 @@ if len(args.dumpfiles) > 0:
                             regex_comment_label = args.regex_comment_label,
                             output_parquet=output_parquet)
 
                             regex_comment_label = args.regex_comment_label,
                             output_parquet=output_parquet)
 

-        print(wikiq.output_parquet)

         wikiq.process()
 
         # close things 
         wikiq.process()
 
         # close things