X-Git-Url: https://code.communitydata.science/mediawiki_dump_tools.git/blobdiff_plain/317bafb50db0c9dba34348e4414095cbe900a9ac..1b81b9542d5ecd0a2da90cfefccdb98bc65684a8:/test/Wikiq_Unit_Test.py diff --git a/test/Wikiq_Unit_Test.py b/test/Wikiq_Unit_Test.py index 91ac928..c04bbd8 100644 --- a/test/Wikiq_Unit_Test.py +++ b/test/Wikiq_Unit_Test.py @@ -10,12 +10,145 @@ from io import StringIO # with / without url encode DONE # with / without collapse user DONE # with output to sdtout DONE -# note that the persistence radius is 7 by default +# note that the persistence radius is 7 by default # reading various file formats including # 7z, gz, bz2, xml DONE # wikia and wikipedia data DONE # malformed xmls DONE +class Test_Persistence(unittest.TestCase): + + def setUp(self): + + if not os.path.exists("test_output"): + os.mkdir("test_output") + + self.wiki = 'pwr-test' + 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.infile = "{0}.xml".format(self.wiki) + self.base_call = "../wikiq {0} -o {1}" + self.input_dir = "dumps" + self.input_file = os.path.join(".", self.input_dir,self.infile) + self.baseline_output_dir = "baseline_output" + + def test_sequence_persistence(self): + test_filename = "sequence-" + self.wikiq_out_name + test_file = os.path.join(self.test_output_dir, test_filename) + if os.path.exists(test_file): + os.remove(test_file) + + call = self.base_call.format(self.input_file, self.test_output_dir) + call = call + " --url-encode --persistence sequence" + print(os.path.abspath('.')) + print(call) + proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True) + proc.wait() + + copyfile(self.call_output, test_file) + baseline_file = os.path.join(".", self.baseline_output_dir, test_filename) + + # as a test let's make sure that we get equal data frames + test = pd.read_table(test_file) + self.assertEqual(test['tokens_added'][0],7) + self.assertEqual(test['tokens_added'][1],10) + self.assertEqual(test['tokens_added'][2],0) + self.assertEqual(test['tokens_added'][3],8) + self.assertEqual(test['tokens_added'][4],0) + self.assertEqual(test['tokens_removed'][0],0) + self.assertEqual(test['tokens_removed'][1],0) + self.assertEqual(test['tokens_removed'][2],0) + self.assertEqual(test['tokens_removed'][3],4) + self.assertEqual(test['tokens_removed'][4],0) + self.assertEqual(test['token_revs'][0],8*3) + self.assertEqual(test['token_revs'][1],0) + self.assertEqual(test['token_revs'][2],0) + self.assertEqual(test['token_revs'][3],0) + self.assertEqual(test['token_revs'][4],0) + + baseline = pd.read_table(baseline_file) + assert_frame_equal(test,baseline) + + def test_legacy_persistence(self): + test_filename = "legacy-" + self.wikiq_out_name + test_file = os.path.join(self.test_output_dir, test_filename) + if os.path.exists(test_file): + os.remove(test_file) + + call = self.base_call.format(self.input_file, self.test_output_dir) + call = call + " --url-encode --persistence legacy" + print(os.path.abspath('.')) + print(call) + proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True) + proc.wait() + + copyfile(self.call_output, test_file) + baseline_file = os.path.join(".", self.baseline_output_dir, test_filename) + + # as a test let's make sure that we get equal data frames + test = pd.read_table(test_file) + self.assertEqual(test['tokens_added'][0],7) + self.assertEqual(test['tokens_added'][1],10) + self.assertEqual(test['tokens_added'][2],0) + self.assertEqual(test['tokens_added'][3],11) + self.assertEqual(test['tokens_added'][4],0) + self.assertEqual(test['tokens_removed'][0],0) + self.assertEqual(test['tokens_removed'][1],0) + self.assertEqual(test['tokens_removed'][2],0) + self.assertEqual(test['tokens_removed'][3],7) + self.assertEqual(test['tokens_removed'][4],0) + self.assertEqual(test['token_revs'][0],7*3) + self.assertEqual(test['token_revs'][1],0) + self.assertEqual(test['token_revs'][2],0) + self.assertEqual(test['token_revs'][3],0) + self.assertEqual(test['token_revs'][4],0) + + baseline = pd.read_table(baseline_file) + assert_frame_equal(test,baseline) + + + +class Test_Persistence_Bug(unittest.TestCase): + + def setUp(self): + if not os.path.exists("test_output"): + os.mkdir("test_output") + + self.wiki = 'enwiki-test' + 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.infile = "{0}.xml".format(self.wiki) + self.base_call = "../wikiq {0} -o {1}" + self.input_dir = "dumps" + self.input_file = os.path.join(".", self.input_dir,self.infile) + self.baseline_output_dir = "baseline_output" + + def test_sequence_persistence(self): + test_filename = "sequence-" + self.wikiq_out_name + test_file = os.path.join(self.test_output_dir, test_filename) + if os.path.exists(test_file): + os.remove(test_file) + + call = self.base_call.format(self.input_file, self.test_output_dir) + call = call + " --url-encode --persistence sequence --collapse-user" + print(os.path.abspath('.')) + print(call) + proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True) + proc.wait() + + copyfile(self.call_output, test_file) + baseline_file = os.path.join(".", self.baseline_output_dir, test_filename) + + # 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) + + class Test_Wikipedia(unittest.TestCase): def setUp(self): if not os.path.exists("test_output"): @@ -40,13 +173,15 @@ class Test_Wikipedia(unittest.TestCase): call = self.base_call.format(self.input_file, self.test_output_dir) call = call + " --url-encode" + print(call) proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True) proc.wait() copyfile(self.call_output, test_file) baseline_file = os.path.join(".", self.baseline_output_dir, test_filename) - # as a test let's make sure that we get equal data frames +# 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) @@ -66,10 +201,10 @@ class Test_Wikipedia(unittest.TestCase): copyfile(self.call_output, test_file) baseline_file = os.path.join(os.path.abspath("."), self.baseline_output_dir, test_filename) - # as a test let's make sure that we get equal data frames +# as a test let's make sure that we get equal data frames test = pd.read_table(test_file) num_wrong_ns = sum(~ test.namespace.isin({0,1})) - self.assertEquals(num_wrong_ns, 0) + self.assertEqual(num_wrong_ns, 0) baseline = pd.read_table(baseline_file) assert_frame_equal(test,baseline) @@ -99,6 +234,7 @@ class Test_Basic(unittest.TestCase): os.remove(test_file) call = self.base_call.format(self.input_file, self.test_output_dir) + print(call) proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True) proc.wait() @@ -120,6 +256,7 @@ class Test_Basic(unittest.TestCase): call = self.base_call.format(self.input_file, self.test_output_dir) call = call + " --collapse-user" + print(call) proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True) proc.wait() @@ -138,6 +275,7 @@ class Test_Basic(unittest.TestCase): call = self.base_call.format(self.input_file, self.test_output_dir) call = call + " --persistence segment" + print(call) proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True) proc.wait() @@ -147,6 +285,7 @@ class Test_Basic(unittest.TestCase): baseline_file = os.path.join(".", self.baseline_output_dir, test_filename) test = pd.read_table(test_file) + print(test) baseline = pd.read_table(baseline_file) assert_frame_equal(test,baseline) @@ -158,10 +297,10 @@ class Test_Basic(unittest.TestCase): call = self.base_call.format(self.input_file, self.test_output_dir) call = call + " --persistence legacy" + print(call) proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True) proc.wait() - copyfile(self.call_output, test_file) baseline_file = os.path.join(".", self.baseline_output_dir, test_filename) @@ -178,6 +317,7 @@ class Test_Basic(unittest.TestCase): call = self.base_call.format(self.input_file, self.test_output_dir) call = call + " --persistence" + print(call) proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True) proc.wait() @@ -200,7 +340,9 @@ class Test_Basic(unittest.TestCase): call = self.base_call.format(self.input_file, self.test_output_dir) call = call + " --url-encode" + print(call) proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True) + proc.wait() copyfile(self.call_output, test_file) @@ -229,6 +371,7 @@ class Test_Malformed(unittest.TestCase): def test_malformed_noargs(self): call = self.base_call.format(self.input_file, self.test_output_dir) + print(call) proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE, shell=True) proc.wait() outs, errs = proc.communicate() @@ -250,6 +393,7 @@ class Test_Stdout(unittest.TestCase): 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")