Polish multiprocessing code

wikipedia.py

@@ -16,7 +16,6 @@
 # Lint as: python3
 """Wikipedia dataset containing cleaned articles of all languages."""
 
-
 import bz2
 import codecs
 import json
@@ -24,15 +23,13 @@ import re
 import xml.etree.cElementTree as etree
 from urllib.parse import quote
 import mwparserfromhell
-from multiprocess import Process, Manager
+import multiprocessing
 from tqdm import tqdm
 
 import datasets
 
-
 logger = datasets.logging.get_logger(__name__)
 
-
 _CITATION = """\
 @ONLINE {wikidump,
     author = {Wikimedia Foundation},
@@ -878,7 +875,6 @@ CAT_ALIASES = {
 _BASE_URL_TMPL = "https://dumps.wikimedia.org/{lang}wiki/{date}/"
 _INFO_FILE = "dumpstatus.json"
 
-
 _VERSION = datasets.Version("2.0.0", "")
 
 
@@ -907,6 +903,9 @@ class WikipediaConfig(datasets.BuilderConfig):
 _DATE = "20220301"
 
 
+class ProcessDone: pass
+
+
 class Wikipedia(datasets.GeneratorBasedBuilder):
     """Wikipedia dataset."""
 
@@ -953,7 +952,7 @@ class Wikipedia(datasets.GeneratorBasedBuilder):
             dump_info = json.load(f)
         multistream_dump_info = dump_info["jobs"]["articlesmultistreamdump"]
         assert (
-            multistream_dump_info["status"] == "done"
+                multistream_dump_info["status"] == "done"
         ), "Specified dump (%s) multistream status is not 'done': %s" % (
             _base_url(lang),
             multistream_dump_info["status"],
@@ -979,7 +978,6 @@ class Wikipedia(datasets.GeneratorBasedBuilder):
 
     def _generate_examples(self, filepaths, language, no_labels=False):
 
-
         def _extract_content(filepath):
             """Extracts article content from a single WikiMedia XML file."""
             logger.info("generating examples from = %s", filepath)
@@ -1029,26 +1027,31 @@ class Wikipedia(datasets.GeneratorBasedBuilder):
             return id_, {"id": id_, "url": url, "title": title, "text": text}
 
         print("Parsing and cleaning Wikipedia examples")
-        with Manager() as manager:
-            examples = manager.list()
-            processes = []
-            for filepath in filepaths:
-                def parse_and_clean(examples):
-                    content = _extract_content(filepath)
-                    for obj in tqdm(content):
-                        examples.append(_clean_content(obj, language=language))
-                p = Process(target=parse_and_clean, args=(examples,))
-                p.start()
-                processes.append(p)
-
-            for p in processes:
-                p.join()
-
-            print("Parsed and cleaned Wikipedia examples")
-
-            for example in examples:
-                if example is not None:
-                    yield example
+
+        examples = multiprocessing.Queue()
+
+        def parse_and_clean(filepath):
+            content = _extract_content(filepath)
+            for obj in tqdm(content):
+                examples.put(_clean_content(obj, language=language))
+            examples.put(ProcessDone())
+
+        with multiprocessing.Pool() as pool:
+            result = pool.map_async(parse_and_clean, filepaths)
+            n = len(filepaths)
+            complete = 0
+            while complete != n:
+                item = examples.get()
+                if isinstance(item, ProcessDone):
+                    complete += 1
+                    continue
+
+                if examples is None:
+                    continue
+
+                yield examples
+
+            result.wait()
 
 
 def _parse_and_clean_wikicode(raw_content, parser, language):