Refine code

wikipedia.py

@@ -24,7 +24,7 @@ import xml.etree.cElementTree as etree
 from urllib.parse import quote
 import mwparserfromhell
 from tqdm import tqdm
-
+import multiprocess
 import datasets
 
 logger = datasets.logging.get_logger(__name__)
@@ -1025,16 +1025,35 @@ class Wikipedia(datasets.GeneratorBasedBuilder):
 
         print("Parsing and cleaning Wikipedia examples")
 
-        def parse_and_clean(filepath):
-            content = _extract_content(filepath)
-            for obj in content:
-                yield _clean_content(obj, language=language)
+        result_q = multiprocess.Queue()
+        processes = []
+
+        with multiprocess.Manager() as manager:
+            for filepath in filepaths:
+                def parse_and_clean():
+                    content = _extract_content(filepath)
+                    for obj in tqdm(content):
+                        result_q.put(_clean_content(obj, language=language))
+                    result_q.put(ProcessDone())
+
+                p = multiprocess.Process(target=parse_and_clean, args=(examples,))
+                p.start()
+                processes.append(p)
+
+        n = len(filepaths)
+        complete = 0
+        while complete < n:
+            obj = result_q.get()
+            if isinstance(obj, ProcessDone):
+                complete += 1
+                continue
+            if obj is None:
+                continue
 
-        for filepath in tqdm(filepaths):
-            for item in parse_and_clean(filepath):
-                if item is not None:
-                    yield item
+            yield obj
 
+        for p in processes:
+            p.join()
 
 def _parse_and_clean_wikicode(raw_content, parser, language):
     """Strips formatting and unwanted sections from raw page content."""