guilhermelmello commited on
Commit
816a97f
1 Parent(s): 789cc32

Update loading script to v1.2

Browse files
Files changed (1) hide show
  1. corpus-carolina.py +13 -20
corpus-carolina.py CHANGED
@@ -18,6 +18,7 @@ from collections import defaultdict
18
  from lxml import etree
19
  import os
20
  import datasets
 
21
 
22
 
23
  _HOMEPAGE = "https://sites.usp.br/corpuscarolina/"
@@ -158,32 +159,23 @@ class Carolina(datasets.GeneratorBasedBuilder):
158
  # download checksum files
159
  checksum_urls = {t: _CHECKSUM_FNAME.format(tax=t) for t in taxonomies}
160
  checksum_paths = dl_manager.download(checksum_urls)
161
-
162
  # prepare xml file name and zip urls
163
- zip_urls = dict() # xml
164
  for tax, cpath in checksum_paths.items():
165
  tax_path = _CORPUS_URL.format(tax=tax)
166
  with open(cpath, encoding="utf-8") as cfile:
167
  for line in cfile:
168
  xml_tax_path = line.split()[1] # xml file inside taxonomy
169
- zip_fname = xml_tax_path + ".zip" # zip file inside taxonomy
170
- xml_fname = xml_tax_path.split('/')[-1] # xml file name only
171
  zip_fpath = os.path.join(tax_path, zip_fname) # path inside corpus
172
- zip_urls[xml_fname] = zip_fpath # xml file -> zip path
173
-
174
- # extractions are made in cache folders and
175
- # the path returned is the folder path, not the
176
- # extracted file (or files). It is necessary to
177
- # build the xml file path. It is made using the
178
- # zip_urls dict structure.
179
- extracted = dl_manager.download_and_extract(zip_urls)
180
- xml_files = [os.path.join(v, k) for k, v in extracted.items()]
181
- xml_files = sorted(xml_files)
182
 
 
183
  return [
184
  datasets.SplitGenerator(
185
  name="corpus",
186
- gen_kwargs={"filepaths": xml_files}
187
  )
188
  ]
189
 
@@ -197,10 +189,10 @@ class Carolina(datasets.GeneratorBasedBuilder):
197
 
198
  _key = 0
199
  for path in filepaths:
200
- # parse xml file
201
- for _, tei in etree.iterparse(path, **parser_params):
202
  header = tei.find(f"{TEI_NS}teiHeader")
203
-
204
  meta = etree.tostring(
205
  header, encoding="utf-8").decode("utf-8")
206
  text = ' '.join([e.text
@@ -208,9 +200,10 @@ class Carolina(datasets.GeneratorBasedBuilder):
208
  if e.text is not None
209
  ])
210
 
211
- example = {
212
  "meta": meta,
213
  "text": text
214
  }
215
- yield _key, example
216
  _key += 1
 
 
 
18
  from lxml import etree
19
  import os
20
  import datasets
21
+ import gzip
22
 
23
 
24
  _HOMEPAGE = "https://sites.usp.br/corpuscarolina/"
 
159
  # download checksum files
160
  checksum_urls = {t: _CHECKSUM_FNAME.format(tax=t) for t in taxonomies}
161
  checksum_paths = dl_manager.download(checksum_urls)
162
+
163
  # prepare xml file name and zip urls
164
+ gzip_urls = list()
165
  for tax, cpath in checksum_paths.items():
166
  tax_path = _CORPUS_URL.format(tax=tax)
167
  with open(cpath, encoding="utf-8") as cfile:
168
  for line in cfile:
169
  xml_tax_path = line.split()[1] # xml file inside taxonomy
170
+ zip_fname = xml_tax_path + ".gz" # zip file inside taxonomy
 
171
  zip_fpath = os.path.join(tax_path, zip_fname) # path inside corpus
172
+ gzip_urls.append(zip_fpath)
 
 
 
 
 
 
 
 
 
173
 
174
+ gzip_files = dl_manager.download(gzip_urls)
175
  return [
176
  datasets.SplitGenerator(
177
  name="corpus",
178
+ gen_kwargs={"filepaths": gzip_files}
179
  )
180
  ]
181
 
 
189
 
190
  _key = 0
191
  for path in filepaths:
192
+ gzip_file = gzip.GzipFile(path, "rb")
193
+ for _, tei in etree.iterparse(gzip_file, **parser_params):
194
  header = tei.find(f"{TEI_NS}teiHeader")
195
+
196
  meta = etree.tostring(
197
  header, encoding="utf-8").decode("utf-8")
198
  text = ' '.join([e.text
 
200
  if e.text is not None
201
  ])
202
 
203
+ yield _key, {
204
  "meta": meta,
205
  "text": text
206
  }
 
207
  _key += 1
208
+
209
+ gzip_file.close()