write regex captures to parquet arrays.

[mediawiki_dump_tools.git] / test / Wikiq_Unit_Test.py

diff --git a/test/Wikiq_Unit_Test.py b/test/Wikiq_Unit_Test.py
index 14d38f12c2521d1f858ec9fc782e8ac5ad5ca23b..129abfe7dbd397c6c847e77f5be8856d1f5e1ed3 100644 (file)
--- a/test/Wikiq_Unit_Test.py
+++ b/test/Wikiq_Unit_Test.py
@@ -3,7 +3,7 @@ import os
 import subprocess
 from shutil import copyfile
 import pandas as pd
 import subprocess
 from shutil import copyfile
 import pandas as pd

-from pandas.util.testing import assert_frame_equal
+from pandas.testing import assert_frame_equal

 from io import StringIO
 
 # with / without pwr DONE
 from io import StringIO
 
 # with / without pwr DONE


@@ -49,7 +49,7 @@ class Test_Wikipedia(unittest.TestCase):
         # as a test let's make sure that we get equal data frames
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)
         # as a test let's make sure that we get equal data frames
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)

-        assert_frame_equal(test,baseline)
+        assert_frame_equal(test,baseline, check_like=True)

 
     def test_WP_namespaces(self):
         print(os.path.abspath('.'))
 
     def test_WP_namespaces(self):
         print(os.path.abspath('.'))


@@ -71,7 +71,7 @@ class Test_Wikipedia(unittest.TestCase):
         num_wrong_ns = sum(~ test.namespace.isin({0,1}))
         self.assertEqual(num_wrong_ns, 0)
         baseline = pd.read_table(baseline_file)
         num_wrong_ns = sum(~ test.namespace.isin({0,1}))
         self.assertEqual(num_wrong_ns, 0)
         baseline = pd.read_table(baseline_file)

-        assert_frame_equal(test,baseline)
+        assert_frame_equal(test,baseline, check_like=True)

 
     def test_WP_revert_radius(self):
         print(os.path.abspath('.'))
 
     def test_WP_revert_radius(self):
         print(os.path.abspath('.'))


@@ -93,7 +93,7 @@ class Test_Wikipedia(unittest.TestCase):
         num_wrong_ns = sum(~ test.namespace.isin({0,1}))
         self.assertEqual(num_wrong_ns, 0)
         baseline = pd.read_table(baseline_file)
         num_wrong_ns = sum(~ test.namespace.isin({0,1}))
         self.assertEqual(num_wrong_ns, 0)
         baseline = pd.read_table(baseline_file)

-        assert_frame_equal(test,baseline)
+        assert_frame_equal(test,baseline, check_like=True)

 
 
 
 
 
 


@@ -131,7 +131,7 @@ class Test_Basic(unittest.TestCase):
 
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)
 
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)

-        assert_frame_equal(test,baseline)
+        assert_frame_equal(test,baseline, check_like=True)

 
 
     def test_collapse_user(self):
 
 
     def test_collapse_user(self):


@@ -151,7 +151,7 @@ class Test_Basic(unittest.TestCase):
         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)
         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)

-        assert_frame_equal(test,baseline)
+        assert_frame_equal(test,baseline, check_like=True)

 
     def test_pwr_segment(self):
         test_filename =  "persistence_segment_" + self.wikiq_out_name
 
     def test_pwr_segment(self):
         test_filename =  "persistence_segment_" + self.wikiq_out_name


@@ -171,7 +171,7 @@ class Test_Basic(unittest.TestCase):
 
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)
 
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)

-        assert_frame_equal(test,baseline)
+        assert_frame_equal(test,baseline, check_like=True)

 
     def test_pwr_legacy(self):
         test_filename =  "persistence_legacy_" + self.wikiq_out_name
 
     def test_pwr_legacy(self):
         test_filename =  "persistence_legacy_" + self.wikiq_out_name


@@ -191,7 +191,7 @@ class Test_Basic(unittest.TestCase):
 
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)
 
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)

-        assert_frame_equal(test,baseline)
+        assert_frame_equal(test,baseline, check_like=True)

 
     def test_pwr(self):
         test_filename =  "persistence_" + self.wikiq_out_name
 
     def test_pwr(self):
         test_filename =  "persistence_" + self.wikiq_out_name


@@ -211,7 +211,9 @@ class Test_Basic(unittest.TestCase):
 
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)
 
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)

-        assert_frame_equal(test,baseline)
+
+        test = test.reindex(columns=sorted(test.columns))
+        assert_frame_equal(test,baseline, check_like=True)

 
 
     def test_url_encode(self):
 
 
     def test_url_encode(self):


@@ -230,7 +232,9 @@ class Test_Basic(unittest.TestCase):
         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)
         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
         test = pd.read_table(test_file)
         baseline = pd.read_table(baseline_file)

-        assert_frame_equal(test,baseline)
+
+        test = test.reindex(columns=sorted(test.columns))
+        assert_frame_equal(test,baseline, check_like=True)

 
 
 class Test_Malformed(unittest.TestCase):
 
 
 class Test_Malformed(unittest.TestCase):


@@ -241,7 +245,7 @@ class Test_Malformed(unittest.TestCase):
         self.wiki = 'twinpeaks'
         self.wikiq_out_name =  self.wiki + ".tsv"
         self.test_output_dir = os.path.join(".", "test_output")
         self.wiki = 'twinpeaks'
         self.wikiq_out_name =  self.wiki + ".tsv"
         self.test_output_dir = os.path.join(".", "test_output")

-        self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
+        self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name) 

 
         self.infile = "{0}.xml.7z".format(self.wiki)
         self.base_call = "../wikiq {0} -o {1}"
 
         self.infile = "{0}.xml.7z".format(self.wiki)
         self.base_call = "../wikiq {0} -o {1}"


@@ -258,6 +262,82 @@ class Test_Malformed(unittest.TestCase):
         errlines = str(errs).split("\\n")
         self.assertEqual(errlines[-2],'xml.etree.ElementTree.ParseError: no element found: line 1369, column 0')
 
         errlines = str(errs).split("\\n")
         self.assertEqual(errlines[-2],'xml.etree.ElementTree.ParseError: no element found: line 1369, column 0')
 

+
+class Test_Redirect_Ikwiki(unittest.TestCase):
+    def setUp(self):
+        self.wiki = 'ikwiki-20180301-pages-meta-history'
+        wiki_siteinfo = "ikwiki-20191101-siteinfo-namespaces.json.gz"
+        self.input_dir = "dumps"
+        self.wiki_siteinfo = os.path.join(".", self.input_dir, wiki_siteinfo)
+        self.wikiq_out_name =  self.wiki + ".tsv"
+        self.infile = "{0}.xml.bz2".format(self.wiki)    
+
+        self.base_call = "../wikiq {0} --siteinfo {1} -o {2}"
+        self.input_dir = "dumps"
+        self.input_file = os.path.join(".", self.input_dir, self.infile)
+        self.baseline_output_dir = "baseline_output"
+
+        self.test_output_dir = os.path.join(".", "test_output")
+        self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
+
+        self.infile = "{0}.xml.bz2".format(self.wiki)    
+
+    def test_redirect(self):
+
+        call = self.base_call.format(self.input_file, self.wiki_siteinfo, self.test_output_dir)
+        print(call)
+        proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
+        proc.wait()
+
+        test_file = os.path.join("redirect_" + self.wikiq_out_name)
+        copyfile(self.call_output, test_file)
+        baseline_file = os.path.join(".", self.baseline_output_dir, test_file)
+
+        # as a test let's make sure that we get equal data frames
+        test = pd.read_table(os.path.join(self.test_output_dir, test_file))
+
+        baseline = pd.read_table(baseline_file)
+        assert_frame_equal(test,baseline, check_like=True)
+
+
+
+class Test_Redirect_Wikia(unittest.TestCase):
+    def setUp(self):
+        self.wiki = 'pokemonfandomcom_fr-20200215-history'
+        wiki_siteinfo = "pokemonfandomcom_fr-20200215-history.xml.7z"
+        self.input_dir = "dumps"
+        self.wiki_siteinfo = os.path.join(".", self.input_dir, wiki_siteinfo)
+        self.wikiq_out_name =  self.wiki + ".tsv"
+        self.infile = "{0}.xml.7z".format(self.wiki)    
+
+        self.base_call = "../wikiq {0} --siteinfo {1} -o {2}"
+        self.input_dir = "dumps"
+        self.input_file = os.path.join(".", self.input_dir, self.infile)
+        self.baseline_output_dir = "baseline_output"
+
+        self.test_output_dir = os.path.join(".", "test_output")
+        self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
+
+        self.infile = "{0}.xml.7z".format(self.wiki)    
+
+    def test_redirect(self):
+
+        call = self.base_call.format(self.input_file, self.wiki_siteinfo, self.test_output_dir)
+        print(call)
+        proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
+        proc.wait()
+
+        test_file = os.path.join("redirect_" + self.wikiq_out_name)
+        copyfile(self.call_output, test_file)
+        baseline_file = os.path.join(".", self.baseline_output_dir, test_file)
+
+        # as a test let's make sure that we get equal data frames
+        test = pd.read_table(os.path.join(self.test_output_dir, test_file))
+        baseline = pd.read_table(baseline_file)
+        assert_frame_equal(test,baseline, check_like=True)
+
+
+    

 class Test_Stdout(unittest.TestCase):
 
     def setUp(self):
 class Test_Stdout(unittest.TestCase):
 
     def setUp(self):


@@ -273,15 +353,18 @@ class Test_Stdout(unittest.TestCase):
     def test_noargs(self):
 
         call = self.base_call.format(self.input_file)
     def test_noargs(self):
 
         call = self.base_call.format(self.input_file)

+        print(call)

         proc = subprocess.run(call,stdout=subprocess.PIPE,shell=True)
         outs = proc.stdout.decode("utf8")
 
         test_file = "noargs_" + self.wikiq_out_name
         baseline_file = os.path.join(".", self.baseline_output_dir, test_file)
         print(baseline_file)
         proc = subprocess.run(call,stdout=subprocess.PIPE,shell=True)
         outs = proc.stdout.decode("utf8")
 
         test_file = "noargs_" + self.wikiq_out_name
         baseline_file = os.path.join(".", self.baseline_output_dir, test_file)
         print(baseline_file)

-        test = pd.read_table(StringIO(outs))
+        s = StringIO(outs)
+        test = pd.read_table(s)

         baseline = pd.read_table(baseline_file)
         baseline = pd.read_table(baseline_file)

-        assert_frame_equal(test,baseline)
+        assert_frame_equal(test,baseline, check_like=True)
+        

 
 class Test_Regex(unittest.TestCase):
 
 
 class Test_Regex(unittest.TestCase):
 


@@ -334,9 +417,9 @@ class Test_Regex(unittest.TestCase):
 
 
     def test_regex_inputs(self):
 
 
     def test_regex_inputs(self):

-        for input in self.bad_inputs_list:
+        for inpt in self.bad_inputs_list:

             call = self.base_call.format(self.input_file)
             call = self.base_call.format(self.input_file)

-            call = call + " --stdout " + input
+            call = call + " --stdout " + inpt

             print(call)
             proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE,shell=True)
             stdout,stderr = proc.communicate()
             print(call)
             proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE,shell=True)
             stdout,stderr = proc.communicate()


@@ -347,7 +430,7 @@ class Test_Regex(unittest.TestCase):
             self.assertNotEqual(proc.returncode,0)
 
     def test_basic_regex(self):
             self.assertNotEqual(proc.returncode,0)
 
     def test_basic_regex(self):

-        for i, input in enumerate(self.good_inputs_list):
+        for i, inpt in enumerate(self.good_inputs_list):

 
             test_filename = "basic_{0}_{1}.tsv".format(self.wikiq_out_name[:-4], str(i))
             #print(test_filename)
 
             test_filename = "basic_{0}_{1}.tsv".format(self.wikiq_out_name[:-4], str(i))
             #print(test_filename)


@@ -356,7 +439,7 @@ class Test_Regex(unittest.TestCase):
                 os.remove(test_file)
 
             call = self.base_call_outs.format(self.input_file, self.test_output_dir)
                 os.remove(test_file)
 
             call = self.base_call_outs.format(self.input_file, self.test_output_dir)

-            call = call + " " + input
+            call = call + " " + inpt

             print(call)
 
             proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE,shell=True)
             print(call)
 
             proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE,shell=True)


@@ -367,12 +450,12 @@ class Test_Regex(unittest.TestCase):
             
             baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
             baseline = pd.read_table(baseline_file)
             
             baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
             baseline = pd.read_table(baseline_file)

-            assert_frame_equal(test, baseline)
+            assert_frame_equal(test, baseline, check_like=True)

             print(i)
 
 
     def test_capturegroup_regex(self):
             print(i)
 
 
     def test_capturegroup_regex(self):

-        for i, input in enumerate(self.cap_inputs_list):
+        for i, inpt in enumerate(self.cap_inputs_list):

             test_filename = "capturegroup_{0}_{1}.tsv".format(self.wikiq_out_name[:-4], str(i))
             print(test_filename)
             test_file = os.path.join(self.test_output_dir, test_filename)
             test_filename = "capturegroup_{0}_{1}.tsv".format(self.wikiq_out_name[:-4], str(i))
             print(test_filename)
             test_file = os.path.join(self.test_output_dir, test_filename)


@@ -380,7 +463,7 @@ class Test_Regex(unittest.TestCase):
                 os.remove(test_file)
 
             call = self.base_call_outs.format(self.input_file, self.test_output_dir)
                 os.remove(test_file)
 
             call = self.base_call_outs.format(self.input_file, self.test_output_dir)

-            call = call + " " + input
+            call = call + " " + inpt

             print(call)
 
             proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE,shell=True)
             print(call)
 
             proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE,shell=True)


@@ -392,7 +475,7 @@ class Test_Regex(unittest.TestCase):
             
             baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
             baseline = pd.read_table(baseline_file)
             
             baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
             baseline = pd.read_table(baseline_file)

-            assert_frame_equal(test, baseline)
+            assert_frame_equal(test, baseline, check_like=True)

 
 
 if __name__ == '__main__':
 
 
 if __name__ == '__main__':