/
gov2.py
393 lines (339 loc) · 16.8 KB
/
gov2.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
import re
import io
import os
import gzip
import codecs
from collections import Counter
from contextlib import contextmanager, ExitStack
from pathlib import Path
from typing import NamedTuple
from glob import glob
import ir_datasets
from ir_datasets.util import DownloadConfig, GzipExtract, TarExtract
from ir_datasets.formats import TrecQrels, TrecQueries, TrecColonQueries, BaseDocs, GenericQuery, BaseQrels, TrecPrels
from ir_datasets.datasets.base import Dataset, YamlDocumentation
from ir_datasets.indices import Docstore
_logger = ir_datasets.log.easy()
NAME = 'gov2'
QREL_DEFS = {
2: 'Highly Relevant',
1: 'Relevant',
0: 'Not Relevant',
}
NAMED_PAGE_QREL_DEFS = {
1: 'Relevant',
0: 'Not Relevant',
}
NAMED_PAGE_QTYPE_MAP = {
'<num> *(Number:)? *NP': 'query_id', # Remove NP prefix from QIDs
'<title> *(Topic:)?': 'text',
}
EFF_MAP_05 = {'751': '1192', '752': '1330', '753': '5956', '754': '6303', '755': '6939', '756': '7553', '757': '8784', '758': '9121', '759': '9266', '760': '10359', '761': '10406', '762': '11597', '763': '12750', '764': '15502', '765': '16895', '766': '17279', '767': '17615', '768': '18050', '769': '18678', '770': '19280', '771': '19963', '772': '20766', '773': '21329', '774': '21513', '775': '23212', '776': '24289', '777': '24781', '778': '24813', '779': '26593', '780': '27428', '781': '28120', '782': '28627', '783': '29561', '784': '33379', '785': '33820', '786': '34135', '787': '35192', '788': '36242', '789': '36530', '790': '36616', '791': '36738', '792': '37111', '793': '41088', '794': '41192', '795': '41506', '796': '44506', '797': '45081', '798': '47993', '799': '48890', '800': '49462'}
EFF_MAP_06 = {'801': '62937', '802': '63569', '803': '63582', '804': '63641', '805': '64227', '806': '64266', '807': '64310', '808': '64642', '809': '64687', '810': '64704', '811': '64723', '812': '64741', '813': '64752', '814': '64938', '815': '65024', '816': '65070', '817': '65222', '818': '65335', '819': '65486', '820': '65504', '821': '65599', '822': '65821', '823': '65826', '824': '65950', '825': '66084', '826': '66409', '827': '66725', '828': '67326', '829': '67531', '830': '67550', '831': '67782', '832': '67961', '833': '68322', '834': '68492', '835': '68967', '836': '69028', '837': '69127', '838': '69401', '839': '69552', '840': '69564', '841': '69935', '842': '70033', '843': '70041', '844': '70285', '845': '70579', '846': '70707', '847': '70751', '848': '70815', '849': '70935', '850': '71136'}
class Gov2Doc(NamedTuple):
doc_id: str
url: str
http_headers: str
body: bytes
body_content_type: str
class Gov2DocIter:
def __init__(self, gov2_docs, slice):
self.gov2_docs = gov2_docs
self.slice = slice
self.next_index = 0
self.file_iter = gov2_docs._docs_iter_source_files()
self.current_file = None
self.current_file_start_idx = 0
self.current_file_end_idx = 0
def __next__(self):
if self.slice.start >= self.slice.stop:
raise StopIteration
while self.next_index != self.slice.start or self.current_file is None or self.current_file_end_idx <= self.slice.start:
if self.current_file is None or self.current_file_end_idx <= self.slice.start:
# First iteration or no docs remaining in this file
if self.current_file is not None:
self.current_file.close()
self.current_file = None
# jump ahead to the file that contains the desired index
first = True
while first or self.current_file_end_idx < self.slice.start:
source_file = next(self.file_iter)
self.next_index = self.current_file_end_idx
self.current_file_start_idx = self.current_file_end_idx
self.current_file_end_idx = self.current_file_start_idx + self.gov2_docs._docs_file_counts()[source_file]
first = False
self.current_file = self.gov2_docs._docs_ctxt_iter_gov2(source_file)
else:
for _ in zip(range(self.slice.start - self.next_index), self.current_file):
# The zip here will stop at after either as many docs we must advance, or however
# many docs remain in the file. In the latter case, we'll just drop out into the
# next iteration of the while loop and pick up the next file.
self.next_index += 1
result = next(self.current_file)
self.next_index += 1
self.slice = slice(self.slice.start + (self.slice.step or 1), self.slice.stop, self.slice.step)
return result
def close(self):
self.file_iter = None
def __iter__(self):
return self
def __del__(self):
self.close()
def __getitem__(self, key):
if isinstance(key, slice):
# it[start:stop:step]
new_slice = ir_datasets.util.apply_sub_slice(self.slice, key)
return Gov2DocIter(self.gov2_docs, new_slice)
elif isinstance(key, int):
# it[index]
new_slice = ir_datasets.util.slice_idx(self.slice, key)
new_it = Gov2DocIter(self.gov2_docs, new_slice)
try:
return next(new_it)
except StopIteration as e:
raise IndexError((self.slice, slice(key, key+1), new_slice))
raise TypeError('key must be int or slice')
class Gov2Docs(BaseDocs):
def __init__(self, docs_dlc, doccount_dlc):
super().__init__()
self.docs_dlc = docs_dlc
self._doccount_dlc = doccount_dlc
self._docs_file_counts_cache = None
def docs_path(self, force=True):
return self.docs_dlc.path(force)
def _docs_iter_source_files(self):
dirs = sorted((Path(self.docs_dlc.path()) / 'GOV2_data').glob('GX???'))
for source_dir in dirs:
for source_file in sorted(source_dir.glob('*.gz')):
yield str(source_file)
def docs_iter(self):
return Gov2DocIter(self, slice(0, self.docs_count()))
def docs_cls(self):
return Gov2Doc
def _docs_ctxt_iter_gov2(self, gov2f):
with ExitStack() as stack:
if isinstance(gov2f, (str, Path)):
gov2f = stack.enter_context(gzip.open(gov2f, 'rb'))
inp = bytearray()
# incrementally read the input file with read1 -- this ends up being more than twice
# as fast as reading the input line-by-line and searching for <DOC> and </DOC> lines
inp.extend(gov2f.read1())
START, END = b'<DOC>\n', b'</DOC>\n'
while inp != b'':
inp, next_doc = self._extract_next_block(inp, START, END)
while next_doc is not None:
yield self._process_gov2_doc(next_doc)
inp, next_doc = self._extract_next_block(inp, START, END)
inp.extend(gov2f.read1())
def _process_gov2_doc(self, raw_doc):
# read the file by exploiting the sequence of blocks in the document -- this ends
# up being several times faster than reading line-by-line
raw_doc, doc_id = self._extract_next_block(raw_doc, b'<DOCNO>', b'</DOCNO>\n')
assert doc_id is not None
doc_id = doc_id.strip().decode()
doc_body, doc_hdr = self._extract_next_block(raw_doc, b'<DOCHDR>\n', b'</DOCHDR>\n')
assert doc_hdr is not None
for encoding in ['utf8', 'ascii', 'latin1']:
try:
doc_url, doc_hdr = doc_hdr.decode(encoding).split('\n', 1)
break
except UnicodeDecodeError:
continue
content_type_match = re.search('^content-type:(.*)$', doc_hdr, re.I|re.M)
content_type = 'text/html' # default to text/html
if content_type_match:
content_type = content_type_match.group(1)
if ';' in content_type:
content_type, _ = content_type.split(';', 1)
content_type = content_type.strip()
return Gov2Doc(doc_id, doc_url, doc_hdr, bytes(doc_body), content_type)
def _extract_next_block(self, inp, START, END):
# if START and END appear in inp, then return (everything after END in inp, the content between START and END),
# or if they don't appear, return (inp, None).
i_start = inp.find(START)
i_end = inp.find(END)
if i_start == -1 or i_end == -1:
return inp, None
return inp[i_end+len(END):], inp[i_start+len(START):i_end]
def _docs_id_to_source_file(self, doc_id):
parts = doc_id.split('-')
if len(parts) != 3:
return None
s_dir, file, doc = parts
source_file = os.path.join(self.docs_dlc.path(), 'GOV2_data', s_dir, f'{file}.gz')
return source_file
def _docs_file_counts(self):
if self._docs_file_counts_cache is None:
result = {}
with self._doccount_dlc.stream() as f:
f = codecs.getreader('utf8')(f)
for line in f:
path, count = line.strip().split()
file = os.path.join(self.docs_dlc.path(), 'GOV2_data', path)
result[file] = int(count)
self._docs_file_counts_cache = result
return self._docs_file_counts_cache
def docs_store(self):
docstore = Gov2Docstore(self)
return ir_datasets.indices.CacheDocstore(docstore, f'{self.docs_path(force=False)}.cache')
def docs_count(self):
return sum(self._docs_file_counts().values())
def docs_namespace(self):
return NAME
def docs_lang(self):
return 'en'
class Gov2Docstore(Docstore):
def __init__(self, gov2_docs):
super().__init__(gov2_docs.docs_cls(), 'doc_id')
self.gov2_docs = gov2_docs
def get_many_iter(self, doc_ids):
result = {}
files_to_search = {}
for doc_id in doc_ids:
source_file = self.gov2_docs._docs_id_to_source_file(doc_id)
if source_file is not None:
if source_file not in files_to_search:
files_to_search[source_file] = []
files_to_search[source_file].append(doc_id)
for source_file, doc_ids in files_to_search.items():
doc_ids = sorted(doc_ids)
for doc in self.gov2_docs._docs_ctxt_iter_gov2(source_file):
if doc_ids[0] == doc.doc_id:
yield doc
doc_ids = doc_ids[1:]
if not doc_ids:
break # file finished
class RewriteQids(BaseQrels):
def __init__(self, base_qrels, qid_map):
self._base_qrels = base_qrels
self._qid_map = qid_map
def qrels_iter(self):
cls = self.qrels_cls()
for qrel in self._base_qrels.qrels_iter():
if qrel.query_id in self._qid_map:
qrel = cls(self._qid_map[qrel.query_id], *qrel[1:])
yield qrel
def qrels_defs(self):
return self._base_qrels.qrels_defs()
def qrels_path(self):
return self._base_qrels.qrels_path()
def qrels_cls(self):
return self._base_qrels.qrels_cls()
class Gov2DocCountFile:
def __init__(self, path, docs_dlc):
self._path = path
self._docs_dlc = docs_dlc
def path(self, force=True):
if force and not os.path.exists(self._path):
docs_urls_path = os.path.join(self._docs_dlc.path(), 'GOV2_extras/url2id.gz')
result = Counter()
with _logger.pbar_raw(desc='building doccounts file', total=25205179, unit='doc') as pbar:
with gzip.open(docs_urls_path, 'rt') as fin:
for line in fin:
url, doc_id = line.rstrip().split()
d, f, i = doc_id.split('-') # formatted like: GX024-52-0546388
file = f'{d}/{f}.gz'
result[file] += 1
pbar.update()
with ir_datasets.util.finialized_file(self._path, 'wt') as fout:
for file in sorted(result):
fout.write(f'{file}\t{result[file]}\n')
return self._path
@contextmanager
def stream(self):
with open(self.path(), 'rb') as f:
yield f
def _init():
documentation = YamlDocumentation(f'docs/{NAME}.yaml')
base_path = ir_datasets.util.home_path()/NAME
dlc = DownloadConfig.context(NAME, base_path)
subsets = {}
docs_dlc = dlc['docs']
doccount_dlc = Gov2DocCountFile(os.path.join(base_path, 'corpus.doccounts'), docs_dlc)
collection = Gov2Docs(docs_dlc, doccount_dlc)
base = Dataset(collection, documentation('_'))
subsets['trec-tb-2004'] = Dataset(
collection,
TrecQueries(dlc['trec-tb-2004/queries'], namespace=NAME, lang='en'),
TrecQrels(dlc['trec-tb-2004/qrels'], QREL_DEFS),
documentation('trec-tb-2004')
)
subsets['trec-tb-2005'] = Dataset(
collection,
TrecQueries(dlc['trec-tb-2005/queries'], namespace=NAME, lang='en'),
TrecQrels(dlc['trec-tb-2005/qrels'], QREL_DEFS),
documentation('trec-tb-2005')
)
subsets['trec-tb-2005/named-page'] = Dataset(
collection,
TrecQueries(dlc['trec-tb-2005/named-page/queries'], qtype=GenericQuery, qtype_map=NAMED_PAGE_QTYPE_MAP, namespace=NAME, lang='en'),
TrecQrels(dlc['trec-tb-2005/named-page/qrels'], NAMED_PAGE_QREL_DEFS),
documentation('trec-tb-2005/named-page')
)
subsets['trec-tb-2005/efficiency'] = Dataset(
collection,
TrecColonQueries(GzipExtract(dlc['trec-tb-2005/efficiency/queries']), encoding='latin1', namespace=NAME, lang='en'),
RewriteQids(TrecQrels(dlc['trec-tb-2005/qrels'], QREL_DEFS), EFF_MAP_05),
documentation('trec-tb-2005/efficiency')
)
subsets['trec-tb-2006'] = Dataset(
collection,
TrecQueries(dlc['trec-tb-2006/queries'], namespace=NAME, lang='en'),
TrecQrels(dlc['trec-tb-2006/qrels'], QREL_DEFS),
documentation('trec-tb-2006')
)
subsets['trec-tb-2006/named-page'] = Dataset(
collection,
TrecQueries(dlc['trec-tb-2006/named-page/queries'], qtype=GenericQuery, qtype_map=NAMED_PAGE_QTYPE_MAP, namespace=NAME, lang='en'),
TrecQrels(dlc['trec-tb-2006/named-page/qrels'], NAMED_PAGE_QREL_DEFS),
documentation('trec-tb-2006/named-page')
)
subsets['trec-tb-2006/efficiency'] = Dataset(
collection,
TrecColonQueries(TarExtract(dlc['trec-tb-2006/efficiency/queries'], '06.efficiency_topics.all'), encoding='latin1', namespace=NAME, lang='en'),
RewriteQids(TrecQrels(dlc['trec-tb-2006/qrels'], QREL_DEFS), EFF_MAP_06),
documentation('trec-tb-2006/efficiency')
)
subsets['trec-tb-2006/efficiency/10k'] = Dataset(
collection,
TrecColonQueries(TarExtract(dlc['trec-tb-2006/efficiency/queries'], '06.efficiency_topics.10k'), encoding='latin1', namespace=NAME, lang='en'),
documentation('trec-tb-2006/efficiency/10k')
)
subsets['trec-tb-2006/efficiency/stream1'] = Dataset(
collection,
TrecColonQueries(TarExtract(dlc['trec-tb-2006/efficiency/queries'], '06.efficiency_topics.stream-1'), encoding='latin1', namespace=NAME, lang='en'),
documentation('trec-tb-2006/efficiency/stream1')
)
subsets['trec-tb-2006/efficiency/stream2'] = Dataset(
collection,
TrecColonQueries(TarExtract(dlc['trec-tb-2006/efficiency/queries'], '06.efficiency_topics.stream-2'), encoding='latin1', namespace=NAME, lang='en'),
documentation('trec-tb-2006/efficiency/stream2')
)
subsets['trec-tb-2006/efficiency/stream3'] = Dataset(
collection,
TrecColonQueries(TarExtract(dlc['trec-tb-2006/efficiency/queries'], '06.efficiency_topics.stream-3'), encoding='latin1', namespace=NAME, lang='en'),
RewriteQids(TrecQrels(dlc['trec-tb-2006/qrels'], QREL_DEFS), EFF_MAP_06),
documentation('trec-tb-2006/efficiency/stream3')
)
subsets['trec-tb-2006/efficiency/stream4'] = Dataset(
collection,
TrecColonQueries(TarExtract(dlc['trec-tb-2006/efficiency/queries'], '06.efficiency_topics.stream-4'), encoding='latin1', namespace=NAME, lang='en'),
documentation('trec-tb-2006/efficiency/stream4')
)
subsets['trec-mq-2007'] = Dataset(
collection,
TrecColonQueries(GzipExtract(dlc['trec-mq-2007/queries']), encoding='latin1'),
TrecPrels(dlc['trec-mq-2007/qrels'], QREL_DEFS),
documentation('trec-mq-2007')
)
subsets['trec-mq-2008'] = Dataset(
collection,
TrecColonQueries(GzipExtract(dlc['trec-mq-2008/queries']), encoding='latin1', namespace='trec-mq', lang='en'),
TrecPrels(TarExtract(dlc['trec-mq-2008/qrels'], '2008.RC1/prels'), QREL_DEFS),
documentation('trec-mq-2008')
)
ir_datasets.registry.register(NAME, base)
for s in sorted(subsets):
ir_datasets.registry.register(f'{NAME}/{s}', subsets[s])
return base, subsets
base, subsets = _init()