def test(self, input_dir):
"""Tests the model accuracy using source code files.
``input_dir`` -- source code files directory.
"""
report = {
'overall-accuracy': 0,
'per-language': {
lang: {
'nb-files': 0,
'accuracy': 0,
'predicted': {
predicted_lang: 0 for predicted_lang in self.languages
},
'predicted-files': {
predicted_lang: [] for predicted_lang in self.languages
}
} for lang in self.languages
}
}
# Test files found in input_dir
extensions = {
ext: lang for lang, exts in self.languages.items() for ext in exts}
for pos, path in enumerate(Path(input_dir).glob('**/*'), 1):
if not path.is_file():
continue
lang = extensions.get(path.suffix.lstrip('.'))
if lang is None:
continue
content = safe_read_file(path)
predicted_lang = self.language_name(content)
language_info = report['per-language'][lang]
language_info['nb-files'] += 1
language_info['predicted'][predicted_lang] += 1
language_info['predicted-files'][predicted_lang].append(str(path))
LOGGER.debug("[%d] files processed", pos)
# Fill the report accuracy data
total_success = 0
total_files = 0
for lang in self.languages:
nb_files = report['per-language'][lang]['nb-files']
if not nb_files:
continue
nb_success = report['per-language'][lang]['predicted'][lang]
report['per-language'][lang]['accuracy'] = nb_success / nb_files
total_success += nb_success
total_files += nb_files
report['overall-accuracy'] = (
total_success / total_files if total_files else 0.0)
return report
def _extract(path):
text = safe_read_file(path).lower()
tokens = [
token for token in split(text)
if not (len(token) > 1 and token.isdigit())
] # Drop numbers > 10
return Counter(tokens)
def test_safe_read_file():
for _ in range(10):
with tempfile.TemporaryDirectory() as dirname:
file_path = Path(dirname).joinpath('example_file')
# Write random bytes into the file
rand_value = random.getrandbits(8*NB_BYTES)
file_path.write_bytes(rand_value.to_bytes(NB_BYTES, 'little'))
# Retrieve text without raising errors
text = utils.safe_read_file(file_path)
assert text
def test_safe_read_file():
for _ in range(10):
with tempfile.NamedTemporaryFile('w+b') as tmp_file:
rand_value = random.getrandbits(8 * NB_BYTES)
# Write random bytes into the file
tmp_file.write(rand_value.to_bytes(NB_BYTES, 'little'))
tmp_file.seek(0)
text = utils.safe_read_file(Path(tmp_file.name))
assert text # Text retrieved without raising errors
