test/Wikiq_Unit_Test.py

   1 import unittest
   2 import os
   3 import subprocess
   4 from shutil import copyfile
   5 import pandas as pd
   6 from pandas.testing import assert_frame_equal
   7 from io import StringIO
   8
   9 # with / without pwr DONE
  10 # with / without url encode DONE
  11 # with / without collapse user DONE
  12 # with output to sdtout DONE
  13 # note that the persistence radius is 7 by default
  14 # reading various file formats including
  15 #        7z, gz, bz2, xml  DONE
  16 # wikia and wikipedia data DONE
  17 # malformed xmls DONE
  18
  19 class Test_Wikipedia(unittest.TestCase):
  20     def setUp(self):
  21         if not os.path.exists("test_output"):
  22             os.mkdir("test_output")
  23
  24         self.wiki = 'ikwiki-20180301-pages-meta-history'
  25         self.wikiq_out_name =  self.wiki + ".tsv"
  26         self.test_output_dir = os.path.join(".", "test_output")
  27         self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
  28
  29         self.infile = "{0}.xml.bz2".format(self.wiki)
  30         self.base_call = "../wikiq {0} -o {1}"
  31         self.input_dir = "dumps"
  32         self.input_file = os.path.join(".", self.input_dir,self.infile)
  33         self.baseline_output_dir = "baseline_output"
  34
  35     def test_WP_url_encode(self):
  36         test_filename =  "url-encode_" + self.wikiq_out_name
  37         test_file = os.path.join(self.test_output_dir, test_filename)
  38         if os.path.exists(test_file):
  39             os.remove(test_file)
  40
  41         call = self.base_call.format(self.input_file, self.test_output_dir)
  42         call = call + " --url-encode"
  43         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
  44         proc.wait()
  45
  46         copyfile(self.call_output, test_file)
  47         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
  48
  49         # as a test let's make sure that we get equal data frames
  50         test = pd.read_table(test_file)
  51         baseline = pd.read_table(baseline_file)
  52         assert_frame_equal(test,baseline, check_like=True)
  53
  54     def test_WP_namespaces(self):
  55         print(os.path.abspath('.'))
  56         test_filename =  "namespaces_" + self.wikiq_out_name
  57         test_file = os.path.join(self.test_output_dir, test_filename)
  58         if os.path.exists(test_file):
  59             os.remove(test_file)
  60
  61         call = self.base_call.format(self.input_file, self.test_output_dir)
  62         call = call + " -n 0 -n 1"
  63         print(call)
  64         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
  65         proc.wait()
  66         copyfile(self.call_output, test_file)
  67         baseline_file = os.path.join(os.path.abspath("."), self.baseline_output_dir, test_filename)
  68
  69         # as a test let's make sure that we get equal data frames
  70         test = pd.read_table(test_file)
  71         num_wrong_ns = sum(~ test.namespace.isin({0,1}))
  72         self.assertEqual(num_wrong_ns, 0)
  73         baseline = pd.read_table(baseline_file)
  74         assert_frame_equal(test,baseline, check_like=True)
  75
  76     def test_WP_revert_radius(self):
  77         print(os.path.abspath('.'))
  78         test_filename =  "revert_radius_" + self.wikiq_out_name
  79         test_file = os.path.join(self.test_output_dir, test_filename)
  80         if os.path.exists(test_file):
  81             os.remove(test_file)
  82
  83         call = self.base_call.format(self.input_file, self.test_output_dir)
  84         call = call + " -n 0 -n 1 -rr 1"
  85         print(call)
  86         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
  87         proc.wait()
  88         copyfile(self.call_output, test_file)
  89         baseline_file = os.path.join(os.path.abspath("."), self.baseline_output_dir, test_filename)
  90
  91         # as a test let's make sure that we get equal data frames
  92         test = pd.read_table(test_file)
  93         num_wrong_ns = sum(~ test.namespace.isin({0,1}))
  94         self.assertEqual(num_wrong_ns, 0)
  95         baseline = pd.read_table(baseline_file)
  96         assert_frame_equal(test,baseline, check_like=True)
  97
  98
  99
 100 class Test_Basic(unittest.TestCase):
 101
 102     def setUp(self):
 103         if not os.path.exists("test_output"):
 104             os.mkdir("test_output")
 105
 106         self.wiki = 'sailormoon'
 107         self.wikiq_out_name =  self.wiki + ".tsv"
 108         self.test_output_dir = os.path.join(".", "test_output")
 109         self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
 110
 111         self.infile = "{0}.xml.7z".format(self.wiki)
 112         self.base_call = "../wikiq {0} -o {1}"
 113         self.input_dir = "dumps"
 114         self.input_file = os.path.join(".", self.input_dir,self.infile)
 115         self.baseline_output_dir = "baseline_output"
 116
 117     def test_noargs(self):
 118
 119         test_filename =  "noargs_" + self.wikiq_out_name
 120         test_file = os.path.join(self.test_output_dir, test_filename)
 121         if os.path.exists(test_file):
 122             os.remove(test_file)
 123
 124         call = self.base_call.format(self.input_file, self.test_output_dir)
 125         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
 126         proc.wait()
 127
 128         copyfile(self.call_output, test_file)
 129
 130         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
 131
 132         test = pd.read_table(test_file)
 133         baseline = pd.read_table(baseline_file)
 134         assert_frame_equal(test,baseline, check_like=True)
 135
 136
 137     def test_collapse_user(self):
 138         test_filename =  "collapse-user_" + self.wikiq_out_name
 139         test_file = os.path.join(self.test_output_dir, test_filename)
 140         if os.path.exists(test_file):
 141             os.remove(test_file)
 142
 143         call = self.base_call.format(self.input_file, self.test_output_dir)
 144         call = call + " --collapse-user"
 145
 146         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
 147         proc.wait()
 148
 149         copyfile(self.call_output, test_file)
 150
 151         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
 152         test = pd.read_table(test_file)
 153         baseline = pd.read_table(baseline_file)
 154         assert_frame_equal(test,baseline, check_like=True)
 155
 156     def test_pwr_segment(self):
 157         test_filename =  "persistence_segment_" + self.wikiq_out_name
 158         test_file = os.path.join(self.test_output_dir, test_filename)
 159         if os.path.exists(test_file):
 160             os.remove(test_file)
 161
 162         call = self.base_call.format(self.input_file, self.test_output_dir)
 163         call = call + " --persistence segment"
 164         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
 165         proc.wait()
 166
 167
 168         copyfile(self.call_output, test_file)
 169
 170         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
 171
 172         test = pd.read_table(test_file)
 173         baseline = pd.read_table(baseline_file)
 174         assert_frame_equal(test,baseline, check_like=True)
 175
 176     def test_pwr_legacy(self):
 177         test_filename =  "persistence_legacy_" + self.wikiq_out_name
 178         test_file = os.path.join(self.test_output_dir, test_filename)
 179         if os.path.exists(test_file):
 180             os.remove(test_file)
 181
 182         call = self.base_call.format(self.input_file, self.test_output_dir)
 183         call = call + " --persistence legacy"
 184         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
 185         proc.wait()
 186
 187
 188         copyfile(self.call_output, test_file)
 189
 190         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
 191
 192         test = pd.read_table(test_file)
 193         baseline = pd.read_table(baseline_file)
 194         assert_frame_equal(test,baseline, check_like=True)
 195
 196     def test_pwr(self):
 197         test_filename =  "persistence_" + self.wikiq_out_name
 198         test_file = os.path.join(self.test_output_dir, test_filename)
 199         if os.path.exists(test_file):
 200            os.remove(test_file)
 201
 202         call = self.base_call.format(self.input_file, self.test_output_dir)
 203         call = call + " --persistence"
 204         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
 205         proc.wait()
 206
 207
 208         copyfile(self.call_output, test_file)
 209
 210         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
 211
 212         test = pd.read_table(test_file)
 213         baseline = pd.read_table(baseline_file)
 214
 215         test = test.reindex(columns=sorted(test.columns))
 216         assert_frame_equal(test,baseline, check_like=True)
 217
 218
 219     def test_url_encode(self):
 220         test_filename =  "url-encode_" + self.wikiq_out_name
 221
 222         test_file = os.path.join(self.test_output_dir, test_filename)
 223         if os.path.exists(test_file):
 224             os.remove(test_file)
 225
 226         call = self.base_call.format(self.input_file, self.test_output_dir)
 227         call = call + " --url-encode"
 228         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
 229         proc.wait()
 230
 231         copyfile(self.call_output, test_file)
 232         baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
 233         test = pd.read_table(test_file)
 234         baseline = pd.read_table(baseline_file)
 235
 236         test = test.reindex(columns=sorted(test.columns))
 237         assert_frame_equal(test,baseline, check_like=True)
 238
 239
 240 class Test_Malformed(unittest.TestCase):
 241     def setUp(self):
 242         if not os.path.exists("test_output"):
 243             os.mkdir("test_output")
 244
 245         self.wiki = 'twinpeaks'
 246         self.wikiq_out_name =  self.wiki + ".tsv"
 247         self.test_output_dir = os.path.join(".", "test_output")
 248         self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
 249
 250         self.infile = "{0}.xml.7z".format(self.wiki)
 251         self.base_call = "../wikiq {0} -o {1}"
 252         self.input_dir = "dumps"
 253         self.input_file = os.path.join(".", self.input_dir,self.infile)
 254
 255
 256     def test_malformed_noargs(self):
 257
 258         call = self.base_call.format(self.input_file, self.test_output_dir)
 259         proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE, shell=True)
 260         proc.wait()
 261         outs, errs = proc.communicate()
 262         errlines = str(errs).split("\\n")
 263         self.assertEqual(errlines[-2],'xml.etree.ElementTree.ParseError: no element found: line 1369, column 0')
 264
 265
 266 class Test_Redirect_Ikwiki(unittest.TestCase):
 267     def setUp(self):
 268         self.wiki = 'ikwiki-20180301-pages-meta-history'
 269         wiki_siteinfo = "ikwiki-20191101-siteinfo-namespaces.json.gz"
 270         self.input_dir = "dumps"
 271         self.wiki_siteinfo = os.path.join(".", self.input_dir, wiki_siteinfo)
 272         self.wikiq_out_name =  self.wiki + ".tsv"
 273         self.infile = "{0}.xml.bz2".format(self.wiki)
 274
 275         self.base_call = "../wikiq {0} --siteinfo {1} -o {2}"
 276         self.input_dir = "dumps"
 277         self.input_file = os.path.join(".", self.input_dir, self.infile)
 278         self.baseline_output_dir = "baseline_output"
 279
 280         self.test_output_dir = os.path.join(".", "test_output")
 281         self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
 282
 283         self.infile = "{0}.xml.bz2".format(self.wiki)
 284
 285     def test_redirect(self):
 286
 287         call = self.base_call.format(self.input_file, self.wiki_siteinfo, self.test_output_dir)
 288         print(call)
 289         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
 290         proc.wait()
 291
 292         test_file = os.path.join(self.test_output_dir, "redirect_" + self.wikiq_out_name)
 293         copyfile(self.call_output, test_file)
 294         baseline_file = os.path.join(".", self.baseline_output_dir, test_file)
 295
 296         # as a test let's make sure that we get equal data frames
 297         test = pd.read_table(test_file)
 298         baseline = pd.read_table(baseline_file)
 299         assert_frame_equal(test,baseline, check_like=True)
 300
 301
 302
 303 class Test_Redirect_Wikia(unittest.TestCase):
 304     def setUp(self):
 305         self.wiki = 'pokemonfandomcom_fr-20200215-history'
 306         wiki_siteinfo = "pokemonfandomcom_fr-20200215-history.xml.7z"
 307         self.input_dir = "dumps"
 308         self.wiki_siteinfo = os.path.join(".", self.input_dir, wiki_siteinfo)
 309         self.wikiq_out_name =  self.wiki + ".tsv"
 310         self.infile = "{0}.xml.7z".format(self.wiki)
 311
 312         self.base_call = "../wikiq {0} --siteinfo {1} -o {2}"
 313         self.input_dir = "dumps"
 314         self.input_file = os.path.join(".", self.input_dir, self.infile)
 315         self.baseline_output_dir = "baseline_output"
 316
 317         self.test_output_dir = os.path.join(".", "test_output")
 318         self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
 319
 320         self.infile = "{0}.xml.7z".format(self.wiki)
 321
 322     def test_redirect(self):
 323
 324         call = self.base_call.format(self.input_file, self.wiki_siteinfo, self.test_output_dir)
 325         print(call)
 326         proc = subprocess.Popen(call,stdout=subprocess.PIPE,shell=True)
 327         proc.wait()
 328
 329         test_file = os.path.join(self.test_output_dir, "redirect_" + self.wikiq_out_name)
 330         copyfile(self.call_output, test_file)
 331         baseline_file = os.path.join(".", self.baseline_output_dir, test_file)
 332
 333         # as a test let's make sure that we get equal data frames
 334         test = pd.read_table(test_file)
 335         baseline = pd.read_table(baseline_file)
 336         assert_frame_equal(test,baseline, check_like=True)
 337
 338
 339
 340 class Test_Stdout(unittest.TestCase):
 341
 342     def setUp(self):
 343         self.wiki = 'itwiki'
 344         self.wikiq_out_name =  self.wiki + ".tsv"
 345
 346         self.infile = "{0}.xml.7z".format(self.wiki)
 347         self.base_call = "../wikiq {0} --stdout"
 348         self.input_dir = "dumps"
 349         self.input_file = os.path.join(".", self.input_dir,self.infile)
 350         self.baseline_output_dir = "baseline_output"
 351
 352     def test_noargs(self):
 353
 354         call = self.base_call.format(self.input_file)
 355         print(call)
 356         proc = subprocess.run(call,stdout=subprocess.PIPE,shell=True)
 357         outs = proc.stdout.decode("utf8")
 358
 359         test_file = "noargs_" + self.wikiq_out_name
 360         baseline_file = os.path.join(".", self.baseline_output_dir, test_file)
 361         print(baseline_file)
 362         test = pd.read_table(StringIO(outs))
 363         baseline = pd.read_table(baseline_file)
 364         assert_frame_equal(test,baseline, check_like=True)
 365
 366
 367 class Test_Regex(unittest.TestCase):
 368
 369     def setUp(self):
 370         self.wiki = 'regextest'
 371         self.wikiq_out_name = self.wiki + '.tsv'
 372         self.infile = "{0}.xml.bz2".format(self.wiki)
 373
 374         self.input_dir = "dumps"
 375         self.input_file = os.path.join(".", self.input_dir,self.infile)
 376
 377         if not os.path.exists("test_output"):
 378             os.mkdir("test_output")
 379
 380         self.test_output_dir = os.path.join(".", "test_output")
 381         self.call_output = os.path.join(self.test_output_dir, self.wikiq_out_name)
 382         # we have two base calls, one for checking inputs and the other for checking outputs
 383         self.base_call = "../wikiq {0}"
 384         self.base_call_outs = "../wikiq {0} -o {1}"
 385
 386         self.baseline_output_dir = "baseline_output"
 387
 388         # sample inputs for checking that bad inputs get terminated / test_regex_inputs
 389         self.bad_inputs_list = [
 390             #label is missing
 391             "-RP '\\b\\d+\\b'",
 392             #number of reg and number of labels do not match
 393             "-RP 'NPO V' -RP THE -RPl testlabel",
 394             #cp but rp label
 395             "-CP '(Tamil|Li)' -RPl testlabel",
 396             #regex is missing
 397             "-CPl testlabel",
 398             "-RP '\\b\\w{3}\\b' -RPl threeletters -CP '\\b\\w{3}\\b'"
 399         ]
 400
 401         # sample inputs for checking the outcomes of good inputs / test_basic_regex
 402         self.good_inputs_list = [
 403             "-RP '\\b\\d{3}\\b' -RPl threedigits",
 404             "-RP 'TestCase' -RP 'page' -RPl testcases -RPl page_word",
 405             "-CP 'Chevalier' -CPl chev_com -RP 'welcome to Wikipedia' -RPl wiki_welcome -CP 'Warning' -CPl warning",
 406             "-CP 'WP:EVADE' -CPl wp_evade"
 407         ]
 408
 409
 410         self.cap_inputs_list = [
 411             "-RP 'Li Chevalier' -RPl li_cheval -CP '(?P<letter>\\b[a-zA-Z]{3}\\b)|(?P<number>\\b\\d+\\b)|(?P<cat>\\bcat\\b)' -CPl three",
 412             "-CP '(?P<a>\\bTestCaseA\\b)|(?P<b>\\bTestCaseB\\b)|(?P<c>\\bTestCaseC\\b)|(?P<d>\\bTestCaseD\\b)' -CPl testcase -RP '(?P<npov>npov|NPOV)|(?P<neutral>neutral point of view)' -RPl npov"
 413         ]
 414
 415
 416
 417     def test_regex_inputs(self):
 418         for input in self.bad_inputs_list:
 419             call = self.base_call.format(self.input_file)
 420             call = call + " --stdout " + input
 421             print(call)
 422             proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE,shell=True)
 423             stdout,stderr = proc.communicate()
 424             #print(proc.returncode)
 425
 426             # we want to check that the bad inputs were caught and sys.exit is stopping the code
 427             print(stderr.decode("utf-8"))
 428             self.assertNotEqual(proc.returncode,0)
 429
 430     def test_basic_regex(self):
 431         for i, input in enumerate(self.good_inputs_list):
 432
 433             test_filename = "basic_{0}_{1}.tsv".format(self.wikiq_out_name[:-4], str(i))
 434             #print(test_filename)
 435             test_file = os.path.join(self.test_output_dir, test_filename)
 436             if os.path.exists(test_file):
 437                 os.remove(test_file)
 438
 439             call = self.base_call_outs.format(self.input_file, self.test_output_dir)
 440             call = call + " " + input
 441             print(call)
 442
 443             proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE,shell=True)
 444             proc.wait()
 445             copyfile(self.call_output, test_file)
 446
 447             test = pd.read_table(test_file)
 448
 449             baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
 450             baseline = pd.read_table(baseline_file)
 451             assert_frame_equal(test, baseline, check_like=True)
 452             print(i)
 453
 454
 455     def test_capturegroup_regex(self):
 456         for i, input in enumerate(self.cap_inputs_list):
 457             test_filename = "capturegroup_{0}_{1}.tsv".format(self.wikiq_out_name[:-4], str(i))
 458             print(test_filename)
 459             test_file = os.path.join(self.test_output_dir, test_filename)
 460             if os.path.exists(test_file):
 461                 os.remove(test_file)
 462
 463             call = self.base_call_outs.format(self.input_file, self.test_output_dir)
 464             call = call + " " + input
 465             print(call)
 466
 467             proc = subprocess.Popen(call,stdout=subprocess.PIPE,stderr=subprocess.PIPE,shell=True)
 468             proc.wait()
 469
 470             copyfile(self.call_output, test_file)
 471
 472             test = pd.read_table(test_file)
 473
 474             baseline_file = os.path.join(".", self.baseline_output_dir, test_filename)
 475             baseline = pd.read_table(baseline_file)
 476             assert_frame_equal(test, baseline, check_like=True)
 477
 478
 479 if __name__ == '__main__':
 480     unittest.main()