def _create_text_files(self):
"""
Creates test set text files from SGML or XML files.
If files are already in text format, do nothing.
"""
if self.file_format == TEXT_FILE:
return
elif self.file_format == SGML_FILE:
for sgml_file in (self.ref_file, self.src_file):
sgml_path = str(sgml_file.name)
text_path = sgml_path.replace('.sgm', '.txt')
if not Path(text_path).exists():
process_to_text(sgml_path, text_path)
elif self.file_format == XML_FILE:
# Extract source text
src_path = str(self.src_file.name)
txt_path = src_path.replace('.xml', '.txt')
if not Path(txt_path).exists():
# After validation it's guaranteed that src_langs has only one element
src_langs, _, _, _ = analyze_xml_file(src_path)
process_xml_to_text(src_path, txt_path, source=src_langs.pop())
# Extract reference texts; multiple references will be tab-separated
ref_path = str(self.ref_file.name)
txt_path = ref_path.replace('.xml', '.txt')
if not Path(txt_path).exists():
_, _, translators, _ = analyze_xml_file(ref_path)
# Sort to guarantee reproducibility
# Scores will be computed against the first reference only
translator = sorted(list(translators))[0]
process_xml_to_text(ref_path, txt_path, reference=translator)
def test_analyze_xml_file_with_hypothesis(self):
"""Checks if systems can be found in XML format."""
xml_path = TESTDATA_DIR + '/xml/sample-hyp.xml'
src_langs, _, _, systems = analyze_xml_file(xml_path)
self.assertSetEqual(src_langs, set(['en']))
self.assertSetEqual(systems, set(['test-team']))
def test_analyze_xml_file_with_multi_reference_testset(self):
"""Checks if multiple references can be found in XML format."""
xml_path = TESTDATA_DIR + '/xml/sample-src-multirefs.xml'
src_langs, ref_langs, translators, _ = analyze_xml_file(xml_path)
self.assertSetEqual(src_langs, set(['en']))
self.assertSetEqual(ref_langs, set(['ha']))
self.assertSetEqual(translators, set(['A', 'B']))
def test_analyze_xml_file_with_testset(self):
"""Checks if source and reference can be found in XML format."""
xml_path = TESTDATA_DIR + '/xml/sample-src-ref.xml'
src_langs, ref_langs, translators, _ = analyze_xml_file(xml_path)
self.assertSetEqual(src_langs, set(['en']))
self.assertSetEqual(ref_langs, set(['ha']))
self.assertSetEqual(translators, set(['A']))
def get_hyp_text(self, path_only=False):
"""Returns a list of hypothesis segments.
Args:
path_only (bool): Return a path to the hypothesis file instead of
a list of hypothesis segments
Returns:
list/str: A list of segments unless path_only and a file path
otherwise
"""
hyp_path = self.hyp_file.name
if self.file_format == SGML_FILE:
if self.test_set.file_format == SGML_FILE:
hyp_filtered_path = hyp_path.replace('.sgm', '.filtered.sgm')
if not Path(hyp_filtered_path).exists():
# Get docids from ref SGML path -- these are non "testsuite-"
ref_docids = Submission._get_docids_from_path(
self.test_set.ref_file.name)
# Filter hyp SGML in matching order, skipping testsuite-* docs
hyp_filtered_path = Submission._filter_sgml_by_docids(
self.hyp_file.name,
ref_docids,
)
else:
hyp_filtered_path = hyp_path
# Create text version of (possibly filtered) hyp SGML
hyp_text_path = hyp_filtered_path.replace('.sgm', '.txt')
if not Path(hyp_text_path).exists():
process_to_text(hyp_filtered_path, hyp_text_path)
elif self.file_format == XML_FILE:
hyp_text_path = hyp_path.replace('.xml', '.txt')
if not Path(hyp_text_path).exists():
_, _, _, sys_names = analyze_xml_file(hyp_path)
# It should never happen that there is no system translations
# thanks to validation, but better to check
if len(sys_names) > 0:
process_xml_to_text(hyp_path,
hyp_text_path,
system=sys_names.pop())
elif self.file_format == TEXT_FILE:
hyp_text_path = hyp_path
if path_only:
return hyp_text_path
else:
return (x for x in open(hyp_text_path, encoding='utf-8'))
def validate_xml_ref_testset(xml_file):
"""Validate reference texts in XML file."""
if not xml_file.name.endswith('.xml'):
return # Skip validation for other formats
_, ref_langs, translators, _ = analyze_xml_file(xml_file)
if len(ref_langs) == 0 or len(translators) == 0:
_msg = 'No reference found in the XML file {0}'.format(xml_file.name)
raise ValidationError(_msg)
if len(ref_langs) > 1:
_msg = (
'XML files with multiple reference languages are not supported')
raise ValidationError(_msg)
def validate_xml_src_testset(xml_file):
"""Validate source texts in XML file."""
if not xml_file.name.endswith('.xml'):
return # Skip validation for other formats
src_langs, _, _, _ = analyze_xml_file(xml_file)
if len(src_langs) == 0:
_msg = 'No source language found in the XML file {0}'.format(
xml_file.name)
raise ValidationError(_msg)
if len(src_langs) > 1:
_msg = 'XML files with multiple source languages are not supported'
raise ValidationError(_msg)
def validate_xml_submission(xml_file):
"""Validate submissions in XML format."""
if not xml_file.name.endswith('.xml'):
return # Skip validation for other formats
validate_xml_schema(xml_file)
xml_file.seek(0) # To be able to read() again
# Check if the submission has some translations from one system only
_, _, _, systems = analyze_xml_file(xml_file)
if len(systems) == 0:
_msg = 'No system found in the XML file {0}'.format(xml_file.name)
raise ValidationError(_msg)
if len(systems) > 1:
_msg = 'XML files with multiple systems are not supported'
raise ValidationError(_msg)