| 1 |
#!/usr/bin/python3 |
| 2 |
# -*- coding: utf-8 -*- |
| 3 |
|
| 4 |
import functools |
| 5 |
import unittest |
| 6 |
import tempfile |
| 7 |
from os import devnull |
| 8 |
from os.path import join as pjoin |
| 9 |
from os.path import abspath, dirname |
| 10 |
from subprocess import check_call, Popen, PIPE, CalledProcessError |
| 11 |
from sys import stderr |
| 12 |
|
| 13 |
|
| 14 |
import signal |
| 15 |
|
| 16 |
|
| 17 |
# Utilities |
| 18 |
def tmp(contents): |
| 19 |
t = tempfile.NamedTemporaryFile(mode='w', delete=False) |
| 20 |
t.write(contents) |
| 21 |
return t.name |
| 22 |
|
| 23 |
|
| 24 |
def rel(fn): |
| 25 |
return abspath(pjoin(dirname(abspath(__file__)), fn)) |
| 26 |
|
| 27 |
|
| 28 |
DEBUG = False |
| 29 |
APERTIUM_TAGGER = rel("../../apertium/apertium-tagger") |
| 30 |
APERTIUM_FILTER_AMBIGUITY = rel("../../apertium/apertium-filter-ambiguity") |
| 31 |
|
| 32 |
|
| 33 |
def check_output(*popenargs, **kwargs): |
| 34 |
# Essentially a copypasted version of check_output with input backported |
| 35 |
# for Python 3.2 Can be significantly abridged with Python 3.5's run(...) |
| 36 |
# or Python 3.3's check_output. Please remove and use stdlib version when |
| 37 |
# Python 3.2/Wheezy is deprecated. |
| 38 |
if 'stdout' in kwargs: |
| 39 |
raise ValueError('stdout argument not allowed, it will be overridden.') |
| 40 |
if 'input' in kwargs: |
| 41 |
if 'stdin' in kwargs: |
| 42 |
raise ValueError('stdin and input arguments may not both be used.') |
| 43 |
inputdata = kwargs['input'] |
| 44 |
del kwargs['input'] |
| 45 |
kwargs['stdin'] = PIPE |
| 46 |
else: |
| 47 |
inputdata = None |
| 48 |
with Popen(*popenargs, stdout=PIPE, **kwargs) as process: |
| 49 |
try: |
| 50 |
out, unused_err = process.communicate(inputdata) |
| 51 |
except: |
| 52 |
process.kill() |
| 53 |
process.wait() |
| 54 |
raise |
| 55 |
retcode = process.poll() |
| 56 |
if retcode: |
| 57 |
raise CalledProcessError(retcode, process.args, output=out) |
| 58 |
return out |
| 59 |
|
| 60 |
|
| 61 |
def check_stderr(*popenargs, **kwargs): |
| 62 |
# Essentially a copypasted version of check_output. |
| 63 |
# Can be significantly abridged with Python 3.5's run(...) |
| 64 |
if 'stderr' in kwargs: |
| 65 |
raise ValueError('stderr argument not allowed, it will be overridden.') |
| 66 |
if 'input' in kwargs: |
| 67 |
if 'stdin' in kwargs: |
| 68 |
raise ValueError('stdin and input arguments may not both be used.') |
| 69 |
inputdata = kwargs['input'] |
| 70 |
del kwargs['input'] |
| 71 |
kwargs['stdin'] = PIPE |
| 72 |
else: |
| 73 |
inputdata = None |
| 74 |
with Popen(*popenargs, stderr=PIPE, **kwargs) as process: |
| 75 |
try: |
| 76 |
unused_output, err = process.communicate(inputdata) |
| 77 |
except: |
| 78 |
process.kill() |
| 79 |
process.wait() |
| 80 |
raise |
| 81 |
retcode = process.poll() |
| 82 |
if retcode: |
| 83 |
raise CalledProcessError(retcode, process.args, output=err) |
| 84 |
return err |
| 85 |
|
| 86 |
|
| 87 |
def trace_dec(f): |
| 88 |
@functools.wraps(f) |
| 89 |
def inner(*args, **kwargs): |
| 90 |
if len(args) > 0 and DEBUG: |
| 91 |
print("run " + " ".join(args[0])) |
| 92 |
return f(*args, **kwargs) |
| 93 |
return inner |
| 94 |
|
| 95 |
|
| 96 |
def trace_plus_unicode(f): |
| 97 |
return functools.partial(trace_dec(f), universal_newlines=True) |
| 98 |
|
| 99 |
|
| 100 |
check_call = trace_plus_unicode(check_call) |
| 101 |
check_output = trace_plus_unicode(check_output) |
| 102 |
check_stderr = trace_plus_unicode(check_stderr) |
| 103 |
|
| 104 |
# Test files |
| 105 |
DIC = """ |
| 106 |
^the/the<det><def><sp>$ |
| 107 |
^books/book<n><pl>/book<vblex><pri><p3><sg>$ |
| 108 |
^has/have<vbhaver><pres><p3><sg>$ |
| 109 |
^booked/book<vblex><pp>/book<vblex><past>$ |
| 110 |
^close/close<adj><sint>/close<n><sg>/close<vblex><inf>/close<vblex><pres>/close<vblex><imp>$ |
| 111 |
^cat/cat<n><sg>$ |
| 112 |
^room/room<n><sg>$ |
| 113 |
^red/red<adj><sint>$ |
| 114 |
^./.<sent>$ |
| 115 |
""".strip() |
| 116 |
|
| 117 |
TSX = """ |
| 118 |
<?xml version="1.0" encoding="utf-8"?> |
| 119 |
<tagger name="test"> |
| 120 |
<tagset> |
| 121 |
<def-label name="DET" closed="true"> |
| 122 |
<tags-item tags="det.*"/> |
| 123 |
<tags-item tags="det.*.*"/> |
| 124 |
</def-label> |
| 125 |
<def-label name="VERB"> |
| 126 |
<tags-item tags="vblex.*"/> |
| 127 |
<tags-item tags="vbhaver.*"/> |
| 128 |
</def-label> |
| 129 |
<def-label name="NOUN"> |
| 130 |
<tags-item tags="n.*"/> |
| 131 |
</def-label> |
| 132 |
<def-label name="ADJ"> |
| 133 |
<tags-item tags="adj.*"/> |
| 134 |
<tags-item tags="adj"/> |
| 135 |
</def-label> |
| 136 |
</tagset> |
| 137 |
</tagger> |
| 138 |
""".strip() |
| 139 |
|
| 140 |
TRAIN_NO_PROBLEM_UNTAGGED = """ |
| 141 |
^The/the<det><def><sp>$ |
| 142 |
^cat/cat<n><sg>$ |
| 143 |
^books/book<n><pl>/book<vblex><pri><p3><sg>$ |
| 144 |
^the/the<det><def><sp>$ |
| 145 |
^room/room<n><sg>$ |
| 146 |
^./.<sent>$ |
| 147 |
|
| 148 |
^The/the<det><def><sp>$ |
| 149 |
^red/red<adj><sint>$ |
| 150 |
^cat/cat<n><sg>$ |
| 151 |
^books/book<n><pl>/book<vblex><pri><p3><sg>$ |
| 152 |
^the/the<det><def><sp>$ |
| 153 |
^red/red<adj><sint>$ |
| 154 |
^room/room<n><sg>$ |
| 155 |
^./.<sent>$ |
| 156 |
|
| 157 |
^The/the<det><def><sp>$ |
| 158 |
^red/red<adj><sint>$ |
| 159 |
^cat/cat<n><sg>$ |
| 160 |
^books/book<n><pl>/book<vblex><pri><p3><sg>$ |
| 161 |
^the/the<det><def><sp>$ |
| 162 |
^room/room<n><sg>$ |
| 163 |
^./.<sent>$ |
| 164 |
""".strip() |
| 165 |
|
| 166 |
TRAIN_NO_PROBLEM_TAGGED = """ |
| 167 |
^The/the<det><def><sp>$ |
| 168 |
^cat/cat<n><sg>$ |
| 169 |
^books/book<vblex><pri><p3><sg>$ |
| 170 |
^the/the<det><def><sp>$ |
| 171 |
^room/room<n><sg>$ |
| 172 |
^./.<sent>$ |
| 173 |
|
| 174 |
^The/the<det><def><sp>$ |
| 175 |
^red/red<adj><sint>$ |
| 176 |
^cat/cat<n><sg>$ |
| 177 |
^books/book<vblex><pri><p3><sg>$ |
| 178 |
^the/the<det><def><sp>$ |
| 179 |
^red/red<adj><sint>$ |
| 180 |
^room/room<n><sg>$ |
| 181 |
^./.<sent>$ |
| 182 |
|
| 183 |
^The/the<det><def><sp>$ |
| 184 |
^red/red<adj><sint>$ |
| 185 |
^cat/cat<n><sg>$ |
| 186 |
^books/book<vblex><pri><p3><sg>$ |
| 187 |
^the/the<det><def><sp>$ |
| 188 |
^room/room<n><sg>$ |
| 189 |
^./.<sent>$ |
| 190 |
""".strip() |
| 191 |
|
| 192 |
TRAIN_CAT_TO_BE_A_VERB_UNTAGGED = """ |
| 193 |
^The/The<det><def><sp>$ |
| 194 |
^falling/fall<vblex><pprs>/fall<vblex><ger>/fall<vblex><subs>$ |
| 195 |
^cat/cat<n><sg>$ |
| 196 |
^has/have<vbhaver><pres><p3><sg>$ |
| 197 |
^booked/book<vblex><pp>/book<vblex><past>$ |
| 198 |
^books/book<n><pl>/book<vblex><pres><p3><sg>$ |
| 199 |
^./.<sent>$ |
| 200 |
|
| 201 |
^Close/close<adj><sint>/close<n><sg>/close<vblex><inf>/close<vblex><pres>/close<vblex><imp>$ |
| 202 |
^the/the<det><def><sp>$ |
| 203 |
^books/book<n><pl>/book<vblex><pri><p3><sg>$ |
| 204 |
^./.<sent>$ |
| 205 |
|
| 206 |
^The/the<det><def><sp>$ |
| 207 |
^falling/fall<vblex><pprs>/fall<vblex><ger>/fall<vblex><subs>$ |
| 208 |
^cat/cat<n><sg>$ |
| 209 |
^has/have<vbhaver><pres><p3><sg>$ |
| 210 |
^books/book<n><pl>/book<vblex><pres><p3><sg>$ |
| 211 |
^./.<sent>$ |
| 212 |
""".strip() |
| 213 |
|
| 214 |
TRAIN_CAT_TO_BE_A_VERB_TAGGED = """ |
| 215 |
^The/The<det><def><sp>$ |
| 216 |
^falling/fall<vblex><pprs>$ |
| 217 |
^cat/cat<n><sg>$ |
| 218 |
^has/have<vbhaver><pres><p3><sg>$ |
| 219 |
^booked/book<vblex><pp>$ |
| 220 |
^books/book<n><pl>$ |
| 221 |
^./.<sent>$ |
| 222 |
|
| 223 |
^Close/close<vblex><imp>$ |
| 224 |
^the/the<det><def><sp>$ |
| 225 |
^books/book<n><pl>$ |
| 226 |
^./.<sent>$ |
| 227 |
|
| 228 |
^The/the<det><def><sp>$ |
| 229 |
^falling/fall<vblex><pprs>$ |
| 230 |
^cat/cat<n><sg>$ |
| 231 |
^has/have<vbhaver><pres><p3><sg>$ |
| 232 |
^books/book<n><pl>$ |
| 233 |
^./.<sent>$ |
| 234 |
""".strip() |
| 235 |
|
| 236 |
TEST_SUCCESS = """ |
| 237 |
^The/the<det><def><sp>$ |
| 238 |
^cat/cat<n><sg>$ |
| 239 |
^books/book<n><pl>/book<vblex><pri><p3><sg>$ |
| 240 |
^the/the<det><def><sp>$ |
| 241 |
^room/room<n><sg>$ |
| 242 |
^./.<sent>$ |
| 243 |
""".strip() |
| 244 |
|
| 245 |
TEST_NEW_AMBG_CLASS = """ |
| 246 |
^The/the<det><def><sp>$ |
| 247 |
^cat/cat<n><sg>/cat<adj>$ |
| 248 |
^books/book<n><pl>/book<vblex><pri><p3><sg>$ |
| 249 |
^the/the<det><def><sp>$ |
| 250 |
^room/room<n><sg>$ |
| 251 |
^./.<sent>$ |
| 252 |
""".strip() |
| 253 |
|
| 254 |
# Expected strings |
| 255 |
EXPECTED_SUBST = """ |
| 256 |
Error: A new ambiguity class was found. |
| 257 |
Retraining the tagger is necessary so as to take it into account. |
| 258 |
Word 'cat'. |
| 259 |
New ambiguity class: {NOUN,ADJ} |
| 260 |
""".strip().split("\n") |
| 261 |
|
| 262 |
|
| 263 |
# Tests |
| 264 |
class AmbiguityClassTest(unittest.TestCase): |
| 265 |
def setUp(self): |
| 266 |
self.tsx_fn = tmp(TSX) |
| 267 |
self.dic_fn = tmp(DIC) |
| 268 |
self.devnull = open(devnull, 'w') |
| 269 |
|
| 270 |
def changing_class_impl(self, flags, model_fn): |
| 271 |
test1 = tmp(TEST_SUCCESS) |
| 272 |
test2 = tmp(TEST_NEW_AMBG_CLASS) |
| 273 |
success_stderr = check_stderr( |
| 274 |
[APERTIUM_TAGGER, '-d'] + flags + |
| 275 |
['-g', model_fn, test1], |
| 276 |
stdout=self.devnull) |
| 277 |
self.assertEqual(success_stderr.strip(), "") |
| 278 |
subst_stderr = check_stderr( |
| 279 |
[APERTIUM_TAGGER, '-d'] + flags + |
| 280 |
['-g', model_fn, test2], |
| 281 |
stdout=self.devnull) |
| 282 |
subst_stderr = [line.strip() |
| 283 |
for line in subst_stderr.strip().split("\n")] |
| 284 |
self.assertEqual(subst_stderr, EXPECTED_SUBST) |
| 285 |
ambg_class = check_output( |
| 286 |
[rel('test-find-similar-ambiguity-class'), model_fn], |
| 287 |
input="NOUN ADJ\n") |
| 288 |
substituted_class = set(ambg_class.split(" ")) |
| 289 |
# Should get open class |
| 290 |
self.assertSetEqual(substituted_class, set(("VERB", "NOUN", "ADJ"))) |
| 291 |
|
| 292 |
def test_changing_class_hmm_sup(self): |
| 293 |
model_fn = tmp("") |
| 294 |
untagged = tmp(TRAIN_NO_PROBLEM_UNTAGGED) |
| 295 |
tagged = tmp(TRAIN_NO_PROBLEM_TAGGED) |
| 296 |
check_call( |
| 297 |
[APERTIUM_TAGGER, '-s', '0', self.dic_fn, untagged, self.tsx_fn, |
| 298 |
model_fn, tagged, untagged], |
| 299 |
stderr=self.devnull) |
| 300 |
self.changing_class_impl([], model_fn) |
| 301 |
|
| 302 |
def test_changing_class_hmm_unsup(self): |
| 303 |
model_fn = tmp("") |
| 304 |
untagged = tmp(TRAIN_NO_PROBLEM_UNTAGGED) |
| 305 |
check_call( |
| 306 |
[APERTIUM_TAGGER, '-t', '1', self.dic_fn, untagged, self.tsx_fn, |
| 307 |
model_fn], |
| 308 |
stderr=self.devnull) |
| 309 |
self.changing_class_impl([], model_fn) |
| 310 |
|
| 311 |
def test_changing_class_sliding_window(self): |
| 312 |
model_fn = tmp("") |
| 313 |
untagged = tmp(TRAIN_NO_PROBLEM_UNTAGGED) |
| 314 |
check_call( |
| 315 |
[APERTIUM_TAGGER, '--sliding-window', '-t', '1', self.dic_fn, |
| 316 |
untagged, self.tsx_fn, model_fn], |
| 317 |
stderr=self.devnull) |
| 318 |
self.changing_class_impl(['--sliding-window'], model_fn) |
| 319 |
|
| 320 |
def test_cat_is_a_verb(self): |
| 321 |
model_fn = tmp("") |
| 322 |
untagged = tmp(TRAIN_CAT_TO_BE_A_VERB_UNTAGGED) |
| 323 |
tagged = tmp(TRAIN_CAT_TO_BE_A_VERB_TAGGED) |
| 324 |
new_ambg_class = tmp(TEST_NEW_AMBG_CLASS) |
| 325 |
check_call( |
| 326 |
[APERTIUM_TAGGER, '-s', '0', self.dic_fn, untagged, self.tsx_fn, |
| 327 |
model_fn, tagged, untagged], |
| 328 |
stderr=self.devnull) |
| 329 |
subst_stdout = check_output( |
| 330 |
[APERTIUM_TAGGER, '-d', '-g', model_fn, new_ambg_class], |
| 331 |
stderr=self.devnull) |
| 332 |
acceptable = False |
| 333 |
for line in subst_stdout.split("\n"): |
| 334 |
if (line.startswith('^cat') and ('<adj>' in line or '<n>' in line)): |
| 335 |
acceptable = True |
| 336 |
self.assertTrue( |
| 337 |
acceptable, |
| 338 |
"'cat' must be output and tagged as an adjective or a noun.\n" + |
| 339 |
"Actual output:\n{}".format(subst_stdout)) |
| 340 |
|
| 341 |
@unittest.expectedFailure |
| 342 |
def test_g_vs_plus(self): |
| 343 |
inp = tmp(""" |
| 344 |
^TV-karriere/TV<np><al><cmp>+karriere<n><m><sg><ind>/Tv<n><m><sg><ind><cmp>+karriere<n><m><sg><ind>$ |
| 345 |
""".strip()) |
| 346 |
subst_stdout = check_output( |
| 347 |
[APERTIUM_TAGGER, '-g', "data/nob.prob", inp], |
| 348 |
stderr=self.devnull) |
| 349 |
out = subst_stdout.split("\n") |
| 350 |
wantanyof = {'^TV<np><al><cmp>+karriere<n><m><sg><ind>$', |
| 351 |
'^Tv<n><m><sg><ind><cmp>+karriere<n><m><sg><ind>$'} |
| 352 |
acceptable = set(out).intersection(wantanyof) |
| 353 |
msg = ("Couldn't find any of \n{}\nin output.\n" + |
| 354 |
"Actual output:\n {}\n").format(wantanyof, |
| 355 |
subst_stdout) |
| 356 |
self.assertTrue(bool(acceptable), msg) |
| 357 |
|
| 358 |
def test_undefined_mult_vs_plus(self): |
| 359 |
inp = tmp(""" |
| 360 |
^yz/y<n><f><cmp>+z<n><ut>$ |
| 361 |
""".strip()) |
| 362 |
subst_stdout = check_output( |
| 363 |
[APERTIUM_TAGGER, '-g', "data/nob.prob", inp], |
| 364 |
stderr=self.devnull) |
| 365 |
out = subst_stdout.split("\n") |
| 366 |
wantanyof = {'^y<n><f><cmp>+z<n><ut>$', |
| 367 |
} |
| 368 |
acceptable = set(out).intersection(wantanyof) |
| 369 |
msg = ("Couldn't find any of \n{}\nin output.\n" + |
| 370 |
"Actual output:\n {}\n").format(wantanyof, |
| 371 |
subst_stdout) |
| 372 |
self.assertTrue(bool(acceptable), msg) |
| 373 |
|
| 374 |
|
| 375 |
class FilterAmbiguityTest(unittest.TestCase): |
| 376 |
def test_unicode_tsx(self): |
| 377 |
stderrpath = tmp('') |
| 378 |
inputpath = tmp('^아/아<noun>$\n') |
| 379 |
output = check_output( |
| 380 |
[APERTIUM_FILTER_AMBIGUITY, rel('unicode.tsx'), inputpath], |
| 381 |
stderr=open(stderrpath, 'w')) |
| 382 |
self.assertEqual('^아/아<noun>$\n', output) |
| 383 |
stderr = open(stderrpath, 'r').read().strip() |
| 384 |
self.assertEqual('', stderr) |
| 385 |
|
| 386 |
class Alarm(Exception): |
| 387 |
pass |
| 388 |
|
| 389 |
class PerceptronNullFlushTest(unittest.TestCase): |
| 390 |
|
| 391 |
inputs = [ |
| 392 |
"^this/this<prn><dem><mf><sg>$ ^here/here<adv>$", |
| 393 |
"^is/be<vbser><pres><p3><sg>$ ^a/a<det><ind><sg>$", |
| 394 |
"^little/little<adj>/little<adv>$", |
| 395 |
"^flushing/*flushing$ ^test/test<n><sg>/test<vblex><inf>/test<vblex><pres>/test<vblex><imp>$" |
| 396 |
] |
| 397 |
|
| 398 |
outputs = [ |
| 399 |
"^this<prn><dem><mf><sg>$ ^here<adv>$", |
| 400 |
"^be<vbser><pres><p3><sg>$ ^a<det><ind><sg>$", |
| 401 |
"^little<adv>$", |
| 402 |
"^*flushing$ ^test<n><sg>$" |
| 403 |
] |
| 404 |
|
| 405 |
|
| 406 |
def alarmHandler(self, signum, frame): |
| 407 |
raise Alarm |
| 408 |
|
| 409 |
def withTimeout(self, seconds, cmd, *args, **kwds): |
| 410 |
signal.signal(signal.SIGALRM, self.alarmHandler) |
| 411 |
signal.alarm(seconds) |
| 412 |
ret = cmd(*args, **kwds) |
| 413 |
signal.alarm(0) # reset the alarm |
| 414 |
return ret |
| 415 |
|
| 416 |
def communicateFlush(self, string, process): |
| 417 |
if string: |
| 418 |
process.stdin.write(string.encode('utf-8')) |
| 419 |
process.stdin.write(b'\0') |
| 420 |
process.stdin.flush() |
| 421 |
|
| 422 |
output = [] |
| 423 |
char = None |
| 424 |
try: |
| 425 |
char = self.withTimeout(2, process.stdout.read, 1) |
| 426 |
except Alarm: |
| 427 |
print("Timeout before reading a single character!", file=stderr) |
| 428 |
while char and char != b'\0': |
| 429 |
output.append(char) |
| 430 |
try: |
| 431 |
char = self.withTimeout(2, process.stdout.read, 1) |
| 432 |
except Alarm: |
| 433 |
print("Timeout before reading %s chars" % len(output), |
| 434 |
file=stderr) |
| 435 |
break # send what we got up till now |
| 436 |
|
| 437 |
return b"".join(output).decode('utf-8').replace('\r\n', '\n') |
| 438 |
|
| 439 |
def test_null_flush(self): |
| 440 |
try: |
| 441 |
proc = Popen([APERTIUM_TAGGER, '-xg', '-z', "data/eng.prob"], |
| 442 |
stdin=PIPE, |
| 443 |
stdout=PIPE, |
| 444 |
stderr=PIPE) |
| 445 |
|
| 446 |
for inp, exp in zip(self.inputs, self.outputs): |
| 447 |
output = self.communicateFlush(inp+"[][\n]", proc) |
| 448 |
with self.subTest(input_line=inp): |
| 449 |
self.assertEqual(output, exp+"[][\n]") |
| 450 |
|
| 451 |
proc.communicate() # let it terminate |
| 452 |
proc.stdin.close() |
| 453 |
proc.stdout.close() |
| 454 |
proc.stderr.close() |
| 455 |
retCode = proc.poll() |
| 456 |
self.assertEqual(retCode, 0) |
| 457 |
|
| 458 |
finally: |
| 459 |
pass |