generate unicode strings to test utf-8 handling for all non-IWSLT dat…

…aset tests. (#1599)

test/common/case_utils.py

@@ -1,3 +1,4 @@
+import random
 import os.path
 import tempfile
 import unittest
@@ -53,3 +54,27 @@ def zip_equal(*iterables):
         if sentinel in combo:
             raise ValueError("Iterables have different lengths")
         yield combo
+
+
+def get_random_unicode(length):
+    # taken from https://stackoverflow.com/a/21666621/2883245
+
+    # Update this to include code point ranges to be sampled
+    include_ranges = [
+        (0x0021, 0x0021),
+        (0x0023, 0x0026),
+        (0x0028, 0x007E),
+        (0x00A1, 0x00AC),
+        (0x00AE, 0x00FF),
+        (0x0100, 0x017F),
+        (0x0180, 0x024F),
+        (0x2C60, 0x2C7F),
+        (0x16A0, 0x16F0),
+        (0x0370, 0x0377),
+        (0x037A, 0x037E),
+        (0x0384, 0x038A),
+        (0x038C, 0x038C),
+    ]
+
+    alphabet = [chr(code_point) for current_range in include_ranges for code_point in range(current_range[0], current_range[1] + 1)]
+    return ''.join(random.choice(alphabet) for i in range(length))

test/datasets/test_agnews.py

@@ -1,13 +1,11 @@
 import os
-import random
-import string
 from collections import defaultdict
 from unittest.mock import patch
 
 from parameterized import parameterized
 from torchtext.datasets.ag_news import AG_NEWS
 
-from ..common.case_utils import TempDirMixin, zip_equal
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
@@ -22,12 +20,10 @@ def _get_mock_dataset(root_dir):
     mocked_data = defaultdict(list)
     for file_name in ("train.csv", "test.csv"):
         txt_file = os.path.join(base_dir, file_name)
-        with open(txt_file, "w") as f:
+        with open(txt_file, "w", encoding="utf-8") as f:
             for i in range(5):
                 label = seed % 4 + 1
-                rand_string = " ".join(
-                    random.choice(string.ascii_letters) for i in range(seed)
-                )
+                rand_string = get_random_unicode(seed)
                 dataset_line = (label, f"{rand_string} {rand_string}")
                 # append line to correct dataset split
                 mocked_data[os.path.splitext(file_name)[0]].append(dataset_line)

test/datasets/test_amazonreviews.py

@@ -1,15 +1,13 @@
 import os
-import random
-import string
 import tarfile
 from collections import defaultdict
 from unittest.mock import patch
 
 from torchtext.datasets.amazonreviewfull import AmazonReviewFull
 from torchtext.datasets.amazonreviewpolarity import AmazonReviewPolarity
 
-from ..common.case_utils import TempDirMixin, zip_equal
 from ..common.parameterized_utils import nested_params
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
@@ -26,15 +24,14 @@ def _get_mock_dataset(root_dir, base_dir_name):
     mocked_data = defaultdict(list)
     for file_name in ("train.csv", "test.csv"):
         txt_file = os.path.join(temp_dataset_dir, file_name)
-        with open(txt_file, "w") as f:
+        with open(txt_file, "w", encoding="utf-8") as f:
             for i in range(5):
                 if base_dir_name == AmazonReviewFull.__name__:
                     label = seed % 5 + 1
                 else:
                     label = seed % 2 + 1
-                rand_string = " ".join(
-                    random.choice(string.ascii_letters) for i in range(seed)
-                )
+                label = seed % 2 + 1
+                rand_string = get_random_unicode(seed)
                 dataset_line = (label, f"{rand_string} {rand_string}")
                 # append line to correct dataset split
                 mocked_data[os.path.splitext(file_name)[0]].append(dataset_line)

test/datasets/test_cc100.py

@@ -1,15 +1,13 @@
-import lzma
 import os
-import random
-import string
+import lzma
 from collections import defaultdict
 from unittest.mock import patch
 
 from parameterized import parameterized
 from torchtext.datasets import CC100
 from torchtext.datasets.cc100 import VALID_CODES
 
-from ..common.case_utils import TempDirMixin, zip_equal
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
@@ -26,11 +24,9 @@ def _get_mock_dataset(root_dir):
     for language_code in VALID_CODES:
         file_name = f"{language_code}.txt.xz"
         compressed_file = os.path.join(base_dir, file_name)
-        with lzma.open(compressed_file, "wt") as f:
+        with lzma.open(compressed_file, "wt", encoding="utf-8") as f:
             for i in range(5):
-                rand_string = " ".join(
-                    random.choice(string.ascii_letters) for i in range(seed)
-                )
+                rand_string = get_random_unicode(seed)
                 content = f"{rand_string}\n"
                 f.write(content)
                 mocked_data[language_code].append((language_code, rand_string))

test/datasets/test_conll2000chunking.py

@@ -1,14 +1,12 @@
-import gzip
 import os
-import random
-import string
+import gzip
 from collections import defaultdict
 from unittest.mock import patch
 
 from parameterized import parameterized
 from torchtext.datasets.conll2000chunking import CoNLL2000Chunking
 
-from ..common.case_utils import TempDirMixin, zip_equal
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
@@ -25,17 +23,11 @@ def _get_mock_dataset(root_dir):
     for file_name in ("train.txt", "test.txt"):
         txt_file = os.path.join(temp_dataset_dir, file_name)
         mocked_lines = mocked_data[os.path.splitext(file_name)[0]]
-        with open(txt_file, "w") as f:
+        with open(txt_file, "w", encoding="utf-8") as f:
             for i in range(5):
-                rand_strings = [
-                    random.choice(string.ascii_letters) for i in range(seed)
-                ]
-                rand_label_1 = [
-                    random.choice(string.ascii_letters) for i in range(seed)
-                ]
-                rand_label_2 = [
-                    random.choice(string.ascii_letters) for i in range(seed)
-                ]
+                rand_strings = [get_random_unicode(seed)]
+                rand_label_1 = [get_random_unicode(seed)]
+                rand_label_2 = [get_random_unicode(seed)]
                 # one token per line (each sample ends with an extra \n)
                 for rand_string, label_1, label_2 in zip(
                     rand_strings, rand_label_1, rand_label_2

test/datasets/test_dbpedia.py

@@ -1,14 +1,12 @@
 import os
-import random
-import string
 import tarfile
 from collections import defaultdict
 from unittest.mock import patch
 
 from parameterized import parameterized
 from torchtext.datasets.dbpedia import DBpedia
 
-from ..common.case_utils import TempDirMixin, zip_equal
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
@@ -25,12 +23,10 @@ def _get_mock_dataset(root_dir):
     for file_name in ("train.csv", "test.csv"):
         csv_file = os.path.join(temp_dataset_dir, file_name)
         mocked_lines = mocked_data[os.path.splitext(file_name)[0]]
-        with open(csv_file, "w") as f:
+        with open(csv_file, "w", encoding="utf-8") as f:
             for i in range(5):
                 label = seed % 14 + 1
-                rand_string = " ".join(
-                    random.choice(string.ascii_letters) for i in range(seed)
-                )
+                rand_string = get_random_unicode(seed)
                 dataset_line = (label, rand_string + " " + rand_string)
                 f.write(f'{label},"{rand_string}","{rand_string}"\n')
 

test/datasets/test_enwik9.py

@@ -1,12 +1,10 @@
 import os
-import random
-import string
 import zipfile
 from unittest.mock import patch
 
 from torchtext.datasets.enwik9 import EnWik9
 
-from ..common.case_utils import TempDirMixin, zip_equal
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
@@ -22,13 +20,9 @@ def _get_mock_dataset(root_dir):
     file_name = "enwik9"
     txt_file = os.path.join(temp_dataset_dir, file_name)
     mocked_data = []
-    with open(txt_file, "w") as f:
+    with open(txt_file, "w", encoding="utf-8") as f:
         for i in range(5):
-            rand_string = (
-                "<"
-                + " ".join(random.choice(string.ascii_letters) for i in range(seed))
-                + ">"
-            )
+            rand_string = "<" + get_random_unicode(seed) + ">"
             dataset_line = f"'{rand_string}'"
             f.write(f"'{rand_string}'\n")
 

test/datasets/test_imdb.py

@@ -1,14 +1,12 @@
 import os
-import random
-import string
 import tarfile
 from collections import defaultdict
 from unittest.mock import patch
 
 from parameterized import parameterized
 from torchtext.datasets.imdb import IMDB
 
-from ..common.case_utils import TempDirMixin, zip_equal
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
@@ -34,10 +32,8 @@ def _get_mock_dataset(root_dir):
             label = "neg" if i < 2 else "pos"
             cur_dir = pos_dir if label == "pos" else neg_dir
             txt_file = os.path.join(cur_dir, f"{i}{i}_{i}.txt")
-            with open(txt_file, "w") as f:
-                rand_string = " ".join(
-                    random.choice(string.ascii_letters) for i in range(seed)
-                )
+            with open(txt_file, "w", encoding="utf-8") as f:
+                rand_string = get_random_unicode(seed)
                 dataset_line = (label, rand_string)
                 # append line to correct dataset split
                 mocked_data[split].append(dataset_line)

test/datasets/test_multi30k.py

@@ -1,14 +1,12 @@
 import os
-import random
-import string
 import tarfile
 from collections import defaultdict
 from unittest.mock import patch
 
 from torchtext.datasets import Multi30k
 
-from ..common.case_utils import TempDirMixin, zip_equal
 from ..common.parameterized_utils import nested_params
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
@@ -24,11 +22,9 @@ def _get_mock_dataset(root_dir):
     mocked_data = defaultdict(list)
     for file_name in ("train.de", "train.en", "val.de", "val.en", "test.de", "test.en"):
         txt_file = os.path.join(temp_dataset_dir, file_name)
-        with open(txt_file, "w") as f:
+        with open(txt_file, "w", encoding="utf-8") as f:
             for i in range(5):
-                rand_string = " ".join(
-                    random.choice(string.ascii_letters) for i in range(seed)
-                )
+                rand_string = get_random_unicode(seed)
                 f.write(rand_string + "\n")
                 mocked_data[file_name].append(rand_string)
                 seed += 1

test/datasets/test_penntreebank.py

@@ -1,13 +1,11 @@
 import os
-import random
-import string
 from collections import defaultdict
 from unittest.mock import patch
 
 from parameterized import parameterized
 from torchtext.datasets.penntreebank import PennTreebank
 
-from ..common.case_utils import TempDirMixin, zip_equal
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
@@ -22,11 +20,9 @@ def _get_mock_dataset(root_dir):
     mocked_data = defaultdict(list)
     for file_name in ("ptb.train.txt", "ptb.valid.txt", "ptb.test.txt"):
         txt_file = os.path.join(base_dir, file_name)
-        with open(txt_file, "w") as f:
+        with open(txt_file, "w", encoding="utf-8") as f:
             for i in range(5):
-                rand_string = " ".join(
-                    random.choice(string.ascii_letters) for i in range(seed)
-                )
+                rand_string = get_random_unicode(seed)
                 dataset_line = f"{rand_string}"
                 # append line to correct dataset split
                 split = file_name.replace("ptb.", "").replace(".txt", "")

test/datasets/test_sogounews.py

@@ -1,14 +1,12 @@
 import os
-import random
-import string
 import tarfile
 from collections import defaultdict
 from unittest.mock import patch
 
 from parameterized import parameterized
 from torchtext.datasets.sogounews import SogouNews
 
-from ..common.case_utils import TempDirMixin, zip_equal
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
@@ -24,12 +22,10 @@ def _get_mock_dataset(root_dir):
     mocked_data = defaultdict(list)
     for file_name in ("train.csv", "test.csv"):
         txt_file = os.path.join(temp_dataset_dir, file_name)
-        with open(txt_file, "w") as f:
+        with open(txt_file, "w", encoding="utf-8") as f:
             for i in range(5):
                 label = seed % 5 + 1
-                rand_string = " ".join(
-                    random.choice(string.ascii_letters) for i in range(seed)
-                )
+                rand_string = get_random_unicode(seed)
                 dataset_line = (label, f"{rand_string} {rand_string}")
                 # append line to correct dataset split
                 mocked_data[os.path.splitext(file_name)[0]].append(dataset_line)

test/datasets/test_squads.py

@@ -1,7 +1,5 @@
 import json
 import os
-import random
-import string
 import uuid
 from collections import defaultdict
 from random import randint
@@ -11,13 +9,13 @@
 from torchtext.datasets.squad1 import SQuAD1
 from torchtext.datasets.squad2 import SQuAD2
 
-from ..common.case_utils import TempDirMixin, zip_equal
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
 from ..common.parameterized_utils import nested_params
 from ..common.torchtext_test_case import TorchtextTestCase
 
 
 def _get_mock_json_data():
-    rand_string = " ".join(random.choice(string.ascii_letters) for i in range(10))
+    rand_string = get_random_unicode(10)
     mock_json_data = {
         "data": [
             {
@@ -60,7 +58,7 @@ def _get_mock_dataset(root_dir, base_dir_name):
     mocked_data = defaultdict(list)
     for file_name in file_names:
         txt_file = os.path.join(base_dir, file_name)
-        with open(txt_file, "w") as f:
+        with open(txt_file, "w", encoding="utf-8") as f:
             mock_json_data = _get_mock_json_data()
             f.write(json.dumps(mock_json_data))