def clean_element(xml: str, selector=':root',
progress_iterator=None, num_processes=1, **kwargs) -> str:
"""Return xml with the selected elements cleaned up. kwargs are passed to
text_cleanup.raw.cleanup()"""
# Use html.parser so that it doesn't try to fix the structure
soup = BeautifulSoup(xml, 'html.parser')
# Build entire list first to avoid modifying a live iterator
nodes = set(
node
for element in soup.select(selector, **kwargs)
for node in element.strings
if not node.isspace())
if progress_iterator is None:
progress_iterator = lambda x: x # noqa: E731
text_iterator = map(str, nodes)
if num_processes > 1:
with multiprocessing.Pool(num_processes) as pool:
fixed = []
futures = [
pool.apply_async(cleanup, (text,), kwargs)
for text in text_iterator]
fixed = [future.get() for future in progress_iterator(futures)]
else:
fixed = progress_iterator([cleanup(t, **kwargs)
for t in text_iterator])
for node, new in zip(nodes, fixed):
node.replace_with(new)
# Maybe show small diff here?
return str(soup)
def test_complicated_sample(self):
sample = """
In the context of 1960, Stranger in a Strange Land was a book that his
publishers feared-itwas too far off the beaten path. So, in order to
mini- mize possible losses, Robert was asked to cutthe monuscript down
to 150,000 words-a loss of about 70,000 words. Other changes were
alsorequested, before the editor was willing to take a chance on
publication.
"""
expected = """
In the context of 1960, Stranger in a Strange Land was a book that his
publishers feared-it was too far off the beaten path. So, in order to
minimize possible losses, Robert was asked to cut the manuscript down
to 150,000 words-a loss of about 70,000 words. Other changes were
also requested, before the editor was willing to take a chance on
publication.
"""
result = raw.cleanup(sample)
self.assertEqual(result, expected)
def main(argv=None):
"""Entry point for text-cleanup cli."""
parser = argparse.ArgumentParser("Clean up text.")
parser.add_argument('input',
nargs='?',
type=argparse.FileType(encoding='utf-8'),
help="The input file to clean up.",
default=sys.stdin)
parser.add_argument('--output',
type=argparse.FileType(mode='w', encoding='utf-8'),
help="Write results to this filename.",
default=sys.stdout)
parser.add_argument(
'--selector',
'-s',
help="Only clean elements mathching this CSS selector. Implies --xml.")
parser.add_argument('--xml', action='store_true', help="Assume XML input.")
parser.add_argument('--num_processes',
'-n',
metavar='N',
help="Utilize N processes.",
type=int,
default=1)
parser.add_argument('--disallow_substitution',
action='store_false',
help='Allow the correction to substitute letters.')
parser.add_argument('--disallow_deletion',
action='store_false',
help='Allow the correction to delete letters.')
parser.add_argument('--disallow_insertion',
action='store_false',
help='Allow the correction to insert letters.')
parser.add_argument('--avoid_capitalized_words',
action='store_true',
help=("Ignore words starting with a capital letter"
"unless we're *really* sure."))
group = parser.add_mutually_exclusive_group()
group.add_argument(
'--reformat-only',
action='store_true',
help="Prettify XML input without changing any of the text.")
args = parser.parse_args(argv or sys.argv[1:])
# Fix dependencies between arguments (e.g. x implies y)
if args.selector or args.reformat_only:
args.xml = True
if args.selector is None:
args.selector = ':root'
if args.xml:
xml = args.input.read()
if args.reformat_only:
output = XML.reformat(xml)
else:
def make_bar(items):
return progressbar.progressbar(items)
output = XML.clean_element(
xml,
args.selector,
progress_iterator=make_bar,
num_processes=args.num_processes,
insertion=not args.disallow_insertion,
deletion=not args.disallow_deletion,
substitution=not args.disallow_substitution,
avoid_capitalized_words=args.avoid_capitalized_words,
)
else:
text = args.input.read()
output = raw.cleanup(
text,
insertion=not args.disallow_insertion,
deletion=not args.disallow_deletion,
substitution=not args.disallow_substitution,
avoid_capitalized_words=args.avoid_capitalized_words,
)
args.output.write(output)
def test_complicated_punctuation(self):
sample = """"Wait! I con't!" he said agaon-twice thot day now."""
expected = """"Wait! I can't!" he said again-twice that day now."""
result = raw.cleanup(sample)
self.assertEqual(result, expected)
def test_missing_spaces_with_errors(self):
sample = "This texthqs missingspaces, but also someerrors."
expected = "This text has missing spaces, but also some errors."
result = raw.cleanup(sample)
self.assertEqual(result, expected)
def test_missing_spaces(self):
sample = "This texthas a few missingspaces."
expected = "This text has a few missing spaces."
result = raw.cleanup(sample)
self.assertEqual(result, expected)
def test_simple_misspelling(self):
sample = "This tixt has one error."
expected = "This text has one error."
result = raw.cleanup(sample)
self.assertEqual(result, expected)
def test_noop(self):
expected = "This text has no errors."
result = raw.cleanup(expected)
self.assertEqual(result, expected)
