Get Parquet test working

This requires some data smoothing to get read_table and read_parquet
DataFrames to look close enough, but the test now passes and validates
that the data match.

Signed-off-by: Will Beason <willbeason@gmail.com>

test/Wikiq_Unit_Test.py

@@ -1,12 +1,18 @@
+import math
 import unittest
 import os
 import subprocess
 from shutil import copyfile
+
+import numpy as np
 import pandas as pd
+from pandas import DataFrame
+from pandas._testing import assert_series_equal
 from pandas.testing import assert_frame_equal
 from io import StringIO
 import tracemalloc
 from typing import Final
+from datetime import datetime
 
 # Make references to files and wikiq relative to this file, not to the current working directory.
 TEST_DIR: Final[str] = os.path.dirname(os.path.realpath(__file__))
@@ -20,6 +26,7 @@ SAILORMOON: Final[str] = "sailormoon"
 TWINPEAKS: Final[str] = "twinpeaks"
 REGEXTEST: Final[str] = "regextest"
 
+
 def setup():
     tracemalloc.start()
 
@@ -41,6 +48,7 @@ class WikiqTester:
                  case_name: str | None = None,
                  suffix: str | None = None,
                  in_compression: str = "bz2",
+                 baseline_format: str = "tsv",
                  out_format: str = "tsv",
                  ):
         self.input_file = os.path.join(TEST_DIR, "dumps", "{0}.xml.{1}".format(wiki, in_compression))
@@ -51,14 +59,16 @@ class WikiqTester:
             self.output_dir = "{0}.parquet".format(TEST_OUTPUT_DIR)
 
         if suffix is None:
+            self.wikiq_baseline_name = "{0}.{1}".format(wiki, baseline_format)
             self.wikiq_out_name = "{0}.{1}".format(wiki, out_format)
         else:
+            self.wikiq_baseline_name = "{0}_{1}.{2}".format(wiki, suffix, baseline_format)
             self.wikiq_out_name = "{0}_{1}.{2}".format(wiki, suffix, out_format)
         self.call_output = os.path.join(self.output_dir, "{0}.{1}".format(wiki, out_format))
 
         # If case_name is unset, there are no relevant baseline or test files.
         if case_name is not None:
-            self.baseline_file = os.path.join(BASELINE_DIR, "{0}_{1}".format(case_name, self.wikiq_out_name))
+            self.baseline_file = os.path.join(BASELINE_DIR, "{0}_{1}".format(case_name, self.wikiq_baseline_name))
             self.test_file = os.path.join(self.output_dir, "{0}_{1}".format(case_name, self.wikiq_out_name))
             if os.path.exists(self.test_file):
                 os.remove(self.test_file)
@@ -78,6 +88,7 @@ class WikiqTester:
         print(call)
         return subprocess.check_output(call, stderr=subprocess.PIPE, shell=True)
 
+
 # with / without pwr DONE
 # with / without url encode DONE
 # with / without collapse user DONE
@@ -103,7 +114,6 @@ class WikiqTestCase(unittest.TestCase):
         baseline = pd.read_table(tester.baseline_file)
         assert_frame_equal(test, baseline, check_like=True)
 
-
     def test_WP_url_encode(self):
         tester = WikiqTester(IKWIKI, "url-encode")
 
@@ -338,7 +348,7 @@ class WikiqTestCase(unittest.TestCase):
             assert_frame_equal(test, baseline, check_like=True)
 
     def test_parquet(self):
-        tester = WikiqTester(IKWIKI, "parquet", out_format="parquet")
+        tester = WikiqTester(IKWIKI, "noargs", out_format="parquet")
 
         try:
             tester.call_wikiq()
@@ -348,9 +358,34 @@ class WikiqTestCase(unittest.TestCase):
         copyfile(tester.call_output, tester.test_file)
 
         # as a test let's make sure that we get equal data frames
-        test = pd.read_parquet(tester.test_file)
+        test: DataFrame = pd.read_parquet(tester.test_file)
-        baseline = pd.read_table(tester.baseline_file)
+        # test = test.drop(['reverteds'], axis=1)
-        assert_frame_equal(test, baseline, check_like=True)
+
+        baseline: DataFrame = pd.read_table(tester.baseline_file)
+
+        # Pandas does not read timestamps as the desired datetime type.
+        baseline['date_time'] = pd.to_datetime(baseline['date_time'])
+        # Split strings to the arrays of reverted IDs so they can be compared.
+        baseline['revert'] = baseline['revert'].replace(np.nan, None)
+        baseline['reverteds'] = [None if i is np.nan else [int(j) for j in str(i).split(",")] for i in baseline['reverteds']]
+        baseline['sha1'] = baseline['sha1'].replace(np.nan, None)
+        baseline['editor'] = baseline['editor'].replace(np.nan, None)
+        baseline['anon'] = baseline['anon'].replace(np.nan, None)
+
+        for index, row in baseline.iterrows():
+            if row['editorid'] is None or test['editorid'][index] is None:
+                if row['editorid'] != test['editorid'][index]:
+                    print(row['revid'], ":", row['editorid'], "!=", test['editorid'][index])
+
+        for col in baseline.columns:
+            try:
+                assert_series_equal(test[col], baseline[col], check_like=True, check_dtype=False)
+            except ValueError as exc:
+                print(f"Error comparing column {col}")
+                self.fail(exc)
+
+        # assert_frame_equal(test, baseline, check_like=True, check_dtype=False)
+
 
 if __name__ == '__main__':
     unittest.main()

wikiq

@@ -270,7 +270,7 @@ class RevDataBase:
         pa.field("title", pa.string()),
         pa.field("namespace", pa.int32()),
         pa.field("deleted", pa.bool_()),
-        pa.field("test_chars", pa.int32()),
+        pa.field("text_chars", pa.int32()),
         pa.field("revert", pa.bool_()),
         pa.field("reverteds", pa.list_(pa.int64())),
         pa.field("sha1", pa.string()),