]> code.communitydata.science - mediawiki_dump_tools.git/blob - test/Wikiq_Unit_Test.py
migrate to mwpersistence. this fixes many issues. We preserve legacy persistence...
[mediawiki_dump_tools.git] / test / Wikiq_Unit_Test.py
1 import unittest
2 import os
3 import sys
4 import subprocess
5 from shutil import copyfile
6 import pdb
7
8 # with / without pwr DONE
9 # with / without url encode DONE
10 # with / without collapse user DONE
11 # with output to sdtout DONE
12 # note that the persistence radius is 7 by default
13 # reading various file formats including
14 #        7z, gz, bz2, xml  DONE
15 # wikia and wikipedia data DONE
16 # malformed xmls DONE
17
18 class Test_Wikipedia(unittest.TestCase):
19     def setUp(self):
20         if not os.path.exists("test_output"):
21             os.mkdir("test_output")
22
23         self.wiki = 'ikwiki-20180301-pages-meta-history'
24         self.wikiq_out_name =  self.wiki + ".tsv"
25         self.test_output_dir = os.path.join(".", "test_output")
26         self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
27
28         self.infile = "{0}.xml.bz2".format(self.wiki)    
29         self.base_call = "../wikiq {0} -o {1}"
30         self.input_dir = "dumps"
31         self.input_file = os.path.join(".", self.input_dir,self.infile)
32         self.baseline_output_dir = "baseline_output"
33
34     def test_WP_url_encode(self):
35         test_filename =  "url-encode_" + self.wikiq_out_name
36         test_file = os.path.join(self.test_output_dir, test_filename)
37         if os.path.exists(test_file):
38             os.remove(test_file)
39         
40         call = self.base_call.format(self.input_file, self.test_output_dir)
41         call = call + " --url-encode"
42         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
43         proc.wait()
44
45         copyfile(self.call_output, test_file)
46         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
47
48         test_lines = open(test_file)
49         baseline_lines = open(baseline_file)
50         for test, baseline in zip(test_lines, baseline_lines):
51             self.assertEqual(test,baseline)
52
53         test_lines.close()
54         baseline_lines.close()
55
56
57 class Test_Basic(unittest.TestCase):
58
59     def setUp(self):
60         if not os.path.exists("test_output"):
61             os.mkdir("test_output")
62
63         self.wiki = 'sailormoon'
64         self.wikiq_out_name =  self.wiki + ".tsv"
65         self.test_output_dir = os.path.join(".", "test_output")
66         self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
67
68         self.infile = "{0}.xml.7z".format(self.wiki)
69         self.base_call = "../wikiq {0} -o {1}"
70         self.input_dir = "dumps"
71         self.input_file = os.path.join(".", self.input_dir,self.infile)
72         self.baseline_output_dir = "baseline_output"
73
74     def test_noargs(self):
75
76         test_filename =  "noargs_" + self.wikiq_out_name
77         test_file = os.path.join(self.test_output_dir, test_filename)
78         if os.path.exists(test_file):
79             os.remove(test_file)
80         
81         call = self.base_call.format(self.input_file, self.test_output_dir)
82         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
83         proc.wait()
84
85         copyfile(self.call_output, test_file)
86
87         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
88
89         test_lines = open(test_file)
90         baseline_lines = open(baseline_file)
91         for test, baseline in zip(test_lines, baseline_lines):
92             self.assertEqual(test, baseline)
93
94         test_lines.close()
95         baseline_lines.close()
96
97     def test_collapse_user(self):
98         test_filename =  "collapse-user_" + self.wikiq_out_name
99         test_file = os.path.join(self.test_output_dir, test_filename)
100         if os.path.exists(test_file):
101             os.remove(test_file)
102         
103         call = self.base_call.format(self.input_file, self.test_output_dir)
104         call = call + " --collapse-user"
105
106         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
107         proc.wait()
108
109         copyfile(self.call_output, test_file)
110
111         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
112
113         test_lines = open(test_file)
114         baseline_lines = open(baseline_file)
115         for test, baseline in zip(test_lines, baseline_lines):
116             self.assertEqual(test,baseline)
117
118         test_lines.close()
119         baseline_lines.close()
120
121     def test_pwr_legacy(self):
122         test_filename =  "persistence_legacy_" + self.wikiq_out_name
123         test_file = os.path.join(self.test_output_dir, test_filename)
124         if os.path.exists(test_file):
125             os.remove(test_file)
126         
127         call = self.base_call.format(self.input_file, self.test_output_dir)
128         call = call + " --persistence-legacy"
129         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
130         proc.wait()
131
132
133         copyfile(self.call_output, test_file)
134
135         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
136
137         test_lines = open(test_file)
138         baseline_lines = open(baseline_file)
139         for test, baseline in zip(test_lines, baseline_lines):
140             self.assertEqual(test,baseline)
141
142         test_lines.close()
143         baseline_lines.close()
144
145     def test_url_encode(self):
146         test_filename =  "url-encode_" + self.wikiq_out_name
147
148         test_file = os.path.join(self.test_output_dir, test_filename)
149         if os.path.exists(test_file):
150             os.remove(test_file)
151         
152         call = self.base_call.format(self.input_file, self.test_output_dir)
153         call = call + " --url-encode"
154         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
155         proc.wait()
156
157         copyfile(self.call_output, test_file)
158         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
159
160         test_lines = open(test_file)
161         baseline_lines = open(baseline_file)
162         for test, baseline in zip(test_lines, baseline_lines):
163             self.assertEqual(test,baseline)
164
165         test_lines.close()
166         baseline_lines.close()
167
168 class Test_Malformed(unittest.TestCase):
169
170     def setUp(self):
171         if not os.path.exists("test_output"):
172             os.mkdir("test_output")
173
174         self.wiki = 'twinpeaks'
175         self.wikiq_out_name =  self.wiki + ".tsv"
176         self.test_output_dir = os.path.join(".", "test_output")
177         self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
178
179         self.infile = "{0}.xml.7z".format(self.wiki)
180         self.base_call = "../wikiq {0} -o {1}"
181         self.input_dir = "dumps"
182         self.input_file = os.path.join(".", self.input_dir,self.infile)
183
184
185     def test_malformed_noargs(self):
186
187         call = self.base_call.format(self.input_file, self.test_output_dir)
188         proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE, shell=True)
189         proc.wait()
190         outs, errs = proc.communicate()
191         errlines = str(errs).split("\\n")
192         self.assertEqual(errlines[-2],'xml.etree.ElementTree.ParseError: no element found: line 1369, column 0')
193
194 class Test_Stdout(unittest.TestCase):
195
196     def setUp(self):
197         self.wiki = 'sailormoon'
198         self.wikiq_out_name =  self.wiki + ".tsv"
199
200         self.infile = "{0}.xml.7z".format(self.wiki)
201         self.base_call = "../wikiq {0} --stdout"
202         self.input_dir = "dumps"
203         self.input_file = os.path.join(".", self.input_dir,self.infile)
204         self.baseline_output_dir = "baseline_output"
205
206     def test_noargs(self):
207
208         call = self.base_call.format(self.input_file)
209         proc = subprocess.run(call,stdout=subprocess.PIPE,shell=True)
210         outs = proc.stdout.decode('utf-8')
211         
212         test_file = "noargs_" + self.wikiq_out_name
213         baseline_file = os.path.join(".", self.baseline_output_dir, test_file)
214
215         test_lines = outs.splitlines(True)
216         baseline_lines = open(baseline_file)
217         for test, baseline in zip(test_lines, baseline_lines):
218             self.assertEqual(test,baseline)
219
220     def test_persistence(self):
221
222         call = self.base_call.format(self.input_file) + " --persistence"
223         proc = subprocess.run(call,stdout=subprocess.PIPE,shell=True)
224         outs = proc.stdout.decode('utf-8')
225         
226         test_file = "persistence_" + self.wikiq_out_name
227         baseline_file = os.path.join(".", self.baseline_output_dir, test_file)
228
229         test_lines = outs.splitlines(True)
230         baseline_lines = open(baseline_file)
231         for test, baseline in zip(test_lines, baseline_lines):
232             self.assertEqual(test,baseline)
233
234         # test_file =  "noargs_" + self.wikiq_out_name
235         # copyfile(self.call_output, os.path.join(self.test_output_dir, test_file))
236
237         # baseline_file = os.path.join(".", self.baseline_output_dir, test_file)
238
239         # test_lines = open(os.path.join(self.test_output_dir,test_file))
240         # baseline_lines = open(baseline_file)
241         # for test, baseline in zip(test_lines, baseline_lines):
242         #     self.assertEqual(test,baseline)
243         # test_lines.close()
244         # baseline_lines.close()
245
246         
247 if __name__ == '__main__':
248     unittest.main()

Community Data Science Collective || Want to submit a patch?