def histogram(training_set, ruleset, trainee, training_cache, delay, tabs,
show_browser, buckets, rules):
"""Show a histogram of rule scores.
We also break down what proportion of each bucket comprised positive or
negative samples. Altogether, this gives you an idea whether a rule is
broadly applicable, discriminatory, and spitting out what you expect.
"""
training_set = Path(training_set)
if training_set.is_dir():
if not ruleset:
raise BadOptionUsage(
'ruleset',
'A --ruleset file must be specified when TRAINING_SET_FOLDER is passed a directory.'
)
if not trainee:
raise BadOptionUsage(
'trainee',
'A --trainee ID must be specified when TRAINING_SET_FOLDER is passed a directory.'
)
training_data = make_or_find_vectors(ruleset, trainee, training_set,
training_cache, show_browser,
'training', delay, tabs)
training_pages = training_data['pages']
x, y, num_yes, _ = tensors_from(training_pages)
feature_names = training_data['header']['featureNames']
print_feature_report(
feature_metrics(feature_names, x, y, buckets, rules or feature_names))
def get_tortoise_config(ctx: Context, tortoise_orm: str) -> dict:
"""
get tortoise config from module
:param ctx:
:param tortoise_orm:
:return:
"""
splits = tortoise_orm.split(".")
config_path = ".".join(splits[:-1])
tortoise_config = splits[-1]
try:
config_module = importlib.import_module(config_path)
except (ModuleNotFoundError, AttributeError):
raise BadOptionUsage(ctx=ctx,
message=f'No config named "{config_path}"',
option_name="--config")
config = getattr(config_module, tortoise_config, None)
if not config:
raise BadOptionUsage(
option_name="--config",
message=
f'Can\'t get "{tortoise_config}" from module "{config_module}"',
ctx=ctx,
)
return config
def main(testing_set, weights, confidence_threshold, ruleset, trainee,
testing_cache, delay, show_browser, verbose):
"""Compute the accuracy of the given coefficients and biases on a folder of
testing samples.
TESTING_SET_FOLDER is a directory of labeled testing pages. It can also be,
for backward compatibility, a JSON file of vectors from FathomFox's
Vectorizer.
WEIGHTS should be a JSON-formatted object like this. You can paste it
directly from the output of fathom-train.
{"coeffs": [["nextAnchorIsJavaScript", 1.1627885103225708],
["nextButtonTypeSubmit", 4.613410949707031],
["nextInputTypeSubmit", 4.374269008636475]],
"bias": -8.645608901977539}
"""
testing_set = Path(testing_set)
if testing_set.is_dir():
if not ruleset:
raise BadOptionUsage(
'ruleset',
'A --ruleset file must be specified when TESTING_SET_FOLDER is passed a directory.'
)
if not trainee:
raise BadOptionUsage(
'trainee',
'A --trainee ID must be specified when TESTING_SET_FOLDER is passed a directory.'
)
with make_or_find_vectors(ruleset, trainee, testing_set, testing_cache,
show_browser, 'testing',
delay).open(encoding='utf-8') as testing_file:
testing_data = load(testing_file)
testing_pages = testing_data['pages']
x, y, num_yes = tensors_from(testing_pages)
model = model_from_json(weights, len(y[0]),
testing_data['header']['featureNames'])
accuracy, false_positives, false_negatives = accuracy_per_tag(
y, model(x), confidence_threshold)
print(
pretty_accuracy('Testing', accuracy, len(x), false_positives,
false_negatives, num_yes))
if testing_pages and 'time' in testing_pages[0]:
print(speed_readout(testing_pages))
if verbose:
print('\nTesting per-tag results:')
print_per_tag_report([
per_tag_metrics(page, model, confidence_threshold)
for page in testing_pages
])
def check_settings(self):
if not self.settings.telegram_bot_token:
raise BadOptionUsage(
'channel',
'Telegram bot token is required for the Telegram channel'
)
if not self.settings.telegram_chat_id:
raise BadOptionUsage(
'channel',
'Telegram Chat ID is required for the Telegram channel'
)
def get_tortoise_config(ctx: Context, tortoise_orm: str) -> dict:
"""
get tortoise config from module
:param ctx:
:param tortoise_orm:
:return:
"""
splits = tortoise_orm.split(".")
config_path = ".".join(splits[:-1])
tortoise_config = splits[-1]
try:
config_module = importlib.import_module(config_path)
except ModuleNotFoundError as e:
raise ClickException(
f"Error while importing configuration module: {e}") from None
config = getattr(config_module, tortoise_config, None)
if not config:
raise BadOptionUsage(
option_name="--config",
message=
f'Can\'t get "{tortoise_config}" from module "{config_module}"',
ctx=ctx,
)
return config
def hunt(urls, threads, exclude_flags, include_flags, interesting_extensions, interesting_files, stdout_flags,
progress_enabled, timeout, max_depth, not_follow_subdomains, exclude_sources, proxies, delay,
not_allow_redirects):
"""Find web directories without bruteforce
"""
if exclude_flags and include_flags:
raise BadOptionUsage('--exclude-flags and --include-flags are mutually exclusive.')
welcome()
if not urls:
click.echo('•_•) OOPS! Add urls to analyze.\nFor example: dirhunt http://domain/path\n\n'
'Need help? Then use dirhunt --help', err=True)
return
exclude_flags, include_flags = flags_range(exclude_flags), flags_range(include_flags)
progress_enabled = (sys.stdout.isatty() or sys.stderr.isatty()) if progress_enabled is None else progress_enabled
crawler = Crawler(max_workers=threads, interesting_extensions=interesting_extensions,
interesting_files=interesting_files, std=sys.stdout if sys.stdout.isatty() else sys.stderr,
progress_enabled=progress_enabled, timeout=timeout, depth=max_depth,
not_follow_subdomains=not_follow_subdomains, exclude_sources=exclude_sources,
not_allow_redirects=not_allow_redirects, proxies=proxies, delay=delay)
crawler.add_init_urls(*urls)
try:
catch_keyboard_interrupt(crawler.print_results, crawler.restart)(set(exclude_flags), set(include_flags))
except SystemExit:
crawler.close()
crawler.print_urls_info()
if not sys.stdout.isatty():
output_urls(crawler, stdout_flags)
def main(database: str, output: str) -> None:
"""Export rdial data for use with timew.
Writes timew compatible data to ‘output’.
"""
if exists(output):
raise BadOptionUsage('output', 'Output path must not exist')
files = process_events(database)
write_events(output, files)
def main(input: TextIO, output: str) -> None:
"""Export timew data for use with rdial.
Reads the output of ‘timew export’, and writes rdial compatible data to
‘output’.
"""
if path.exists(output):
raise BadOptionUsage('output', 'Output path must not exist')
files = process_records(input)
write_events(output, files)
def hunt(urls, threads, exclude_flags, include_flags, interesting_extensions, interesting_files, stdout_flags,
progress_enabled, timeout, max_depth, not_follow_subdomains, exclude_sources, proxies, delay,
not_allow_redirects, limit, to_file):
"""Find web directories without bruteforce
"""
if exclude_flags and include_flags:
raise BadOptionUsage('--exclude-flags and --include-flags are mutually exclusive.')
welcome()
urls = flat_list(urls)
proxies = multiplier_args(proxies)
if not urls:
click.echo('•_•) OOPS! Add urls to analyze.\nFor example: dirhunt http://domain/path\n\n'
'Need help? Then use dirhunt --help', err=True)
return
exclude_flags, include_flags = flags_range(exclude_flags), flags_range(include_flags)
progress_enabled = (sys.stdout.isatty() or sys.stderr.isatty()) if progress_enabled is None else progress_enabled
crawler = Crawler(max_workers=threads, interesting_extensions=interesting_extensions,
interesting_files=interesting_files, std=sys.stdout if sys.stdout.isatty() else sys.stderr,
progress_enabled=progress_enabled, timeout=timeout, depth=max_depth,
not_follow_subdomains=not_follow_subdomains, exclude_sources=exclude_sources,
not_allow_redirects=not_allow_redirects, proxies=proxies, delay=delay, limit=limit,
to_file=to_file)
if os.path.exists(crawler.get_resume_file()):
click.echo('Resuming the previous program execution...')
try:
crawler.resume(crawler.get_resume_file())
except IncompatibleVersionError as e:
click.echo(e)
crawler.add_init_urls(*urls)
while True:
choice = catch_keyboard_interrupt_choices(crawler.print_results, ['abort', 'continue', 'results'], 'a')\
(set(exclude_flags), set(include_flags))
if choice == 'a':
crawler.close(True)
click.echo('Created resume file "{}". Run again using the same parameters to resume.'.format(
crawler.get_resume_file())
)
return
elif choice == 'c':
crawler.restart()
continue
else:
break
crawler.print_urls_info()
if not sys.stdout.isatty():
output_urls(crawler, stdout_flags)
if to_file:
crawler.create_report(to_file)
if not to_file and os.path.exists(crawler.get_resume_file()):
# The resume file exists. Deleting...
os.remove(crawler.get_resume_file())
def get_app_connection_name(config, app_name: str) -> str:
"""
get connection name
:param config:
:param app_name:
:return:
"""
app = config.get("apps").get(app_name)
if app:
return app.get("default_connection", "default")
raise BadOptionUsage(
option_name="--app",
message=f'Can\'t get app named "{app_name}"',
)
def handle_cli_args(
self,
url=None,
platform=None,
model_name=None,
scenario_name=None,
version=None,
local_data=None,
_store_as=("platform_info", "scenario_info"),
):
"""Handle command-line arguments.
May update the :attr:`data_path`, :attr:`platform_info`, :attr:`scenario_info`,
and/or :attr:`url` settings.
"""
# Store the path to command-specific data and metadata
if local_data:
self.local_data = local_data
# References to the Context settings to be updated
platform_info = self.setdefault(_store_as[0], dict())
scenario_info = self.setdefault(_store_as[1], dict())
# Store information for the target Platform
if url:
if platform or model_name or scenario_name or version:
raise BadOptionUsage(
"--platform --model --scenario and/or --version",
" redundant with --url",
)
self.url = url
urlinfo = ixmp.utils.parse_url(url)
platform_info.update(urlinfo[0])
scenario_info.update(urlinfo[1])
elif platform:
platform_info["name"] = platform
# Store information about the target Scenario
if model_name:
scenario_info["model"] = model_name
if scenario_name:
scenario_info["scenario"] = scenario_name
if version:
scenario_info["version"] = version
def __init__(self, settings: Settings):
self.settings = settings
self.mailbox = MailBox(self.settings.imap_hostname)
self.mailbox.login(self.settings.imap_username,
self.settings.imap_password)
self.spam_filters = [
re.compile(spam_filter) for spam_filter in settings.spam_filters
]
self.channels: typing.Dict[str, BaseChannel] = {}
for channel_name in settings.channels:
channel_class = self.CHANNELS.get(channel_name)
if channel_class:
self.channels[channel_name] = channel_class(settings)
else:
raise BadOptionUsage('channel',
f'Channel {channel_name} is not defined')
def main(training_set, validation_set, ruleset, trainee, training_cache,
validation_cache, delay, show_browser, stop_early, learning_rate,
iterations, pos_weight, comment, quiet, confidence_threshold, layers,
exclude):
"""Compute optimal numerical parameters for a Fathom ruleset.
The usual invocation is something like this::
fathom-train samples/training --validation-set samples/validation --ruleset rulesets.js --trainee new
The first argument is a directory of labeled training pages. It can also
be, for backward compatibility, a JSON file of vectors from FathomFox's
Vectorizer.
To see graphs of loss functions, install TensorBoard, then run
``tensorboard --logdir runs/``. These will tell you whether you need to
adjust the ``--learning-rate``.
Definitions of terms used in output:
\b
pruned
Said of a node that was prematurely eliminated from consideration
because it did not match the selector of any ``dom()`` call in the
ruleset
\b
target
A "right answer": a labeled, positive DOM node, one that should be
recognized.
"""
training_set = Path(training_set)
# If they pass in a dir for either the training or validation sets, we need
# a ruleset and a trainee for vectorizing:
if (validation_set and validation_set.is_dir()) or training_set.is_dir():
if not ruleset:
raise BadOptionUsage(
'ruleset',
'A --ruleset file must be specified when TRAINING_SET_FOLDER or --validation-set are passed a directory.'
)
if not trainee:
raise BadOptionUsage(
'trainee',
'A --trainee ID must be specified when TRAINING_SET_FOLDER or --validation-set are passed a directory.'
)
training_data = exclude_features(
exclude,
make_or_find_vectors(ruleset, trainee, training_set, training_cache,
show_browser, 'training', delay))
training_pages = training_data['pages']
x, y, num_yes, num_prunes = tensors_from(training_pages, shuffle=True)
num_samples = len(x) + num_prunes
if validation_set:
validation_pages = exclude_features(
exclude,
make_or_find_vectors(ruleset, trainee, validation_set,
validation_cache, show_browser, 'validation',
delay))['pages']
validation_ins, validation_outs, validation_yes, validation_prunes = tensors_from(
validation_pages)
validation_arg = validation_ins, validation_outs
else:
validation_arg = None
layers = list(layers) # Comes in as tuple
full_comment = '.LR={l},i={i}{c}'.format(l=learning_rate,
i=iterations,
c=(',' +
comment) if comment else '')
model = learn(learning_rate,
iterations,
x,
y,
confidence_threshold,
num_prunes,
validation=validation_arg,
stop_early=stop_early,
run_comment=full_comment,
pos_weight=pos_weight,
layers=layers)
print(pretty_coeffs(model, training_data['header']['featureNames']))
accuracy, false_positives, false_negatives = accuracy_per_tag(
y, model(x), confidence_threshold, num_prunes)
print(
pretty_accuracy('Training', accuracy, num_samples, false_positives,
false_negatives, num_yes + num_prunes))
if validation_set:
accuracy, false_positives, false_negatives = accuracy_per_tag(
validation_outs, model(validation_ins), confidence_threshold,
validation_prunes)
print(
pretty_accuracy('Validation', accuracy, len(validation_ins),
false_positives, false_negatives,
validation_yes + validation_prunes))
# Print timing information:
if training_pages and 'time' in training_pages[0]:
if validation_set and validation_pages and 'time' in validation_pages[
0]:
print(speed_readout(training_pages + validation_pages))
else:
print(speed_readout(training_pages))
if not quiet:
print('\nTraining per-tag results:')
print_per_tag_report([
per_tag_metrics(page, model, confidence_threshold)
for page in training_pages
])
if validation_set:
print('\nValidation per-tag results:')
print_per_tag_report([
per_tag_metrics(page, model, confidence_threshold)
for page in validation_pages
])
def option_error():
raise BadOptionUsage("Incorrect option", "What a pity")
def fetcher(config, show_list, fetchers, check, pool_size,
https_only, http_check_https, no_socks, save):
from .proxyfetcher import ProxyFetcher
from .proxychecker import ProxyChecker
if show_list:
for fetcher in ProxyFetcher.registry.values():
echo(fetcher.name +' '+ fetcher.__module__ + ':' + fetcher.__name__)
return
proxies = OrderedDict()
checker = None
if check:
conf = config.get('proxyfetcher', {})
if http_check_https:
conf['https_force_check'] = True
if pool_size:
conf['pool_size'] = pool_size
blacklist = conf.pop('blacklist', None)
if not blacklist:
conf['blacklist'] = proxies
else:
# Do not check already checked proxies
conf['blacklist'] = CompositeContains(blacklist, proxies)
checker = ProxyChecker(**conf)
json_encoder = JSONEncoder(**config.get('json', {}))
def proxy(proxy):
if proxy.addr in proxies:
proxies[proxy.addr].merge_meta(proxy)
else:
proxies[proxy.addr] = proxy
conf = config.get('proxyfetcher', {})
fetchers_ = conf.pop('fetchers', None)
if fetchers == '*':
fetchers_ = ProxyFetcher.registry
elif fetchers:
fetchers_ = fetchers.split(',')
if not fetchers:
raise BadOptionUsage('You should specify fetchers with option or in config.')
types = set(t.upper() for t in
conf.pop('types', ['HTTP', 'HTTPS', 'SOCKS4', 'SOCKS5']))
if https_only and not http_check_https:
types = set(('HTTPS', 'SOCKS4', 'SOCKS5'))
if no_socks:
types = types.difference(['SOCKS4', 'SOCKS5'])
if not types:
raise BadOptionUsage('Proxy types appears to be empty. '
'Check config and options compability.')
if pool_size:
conf['pool_size'] = pool_size
fetcher = ProxyFetcher(fetchers_, checker=checker, proxy=proxy, types=types, **conf)
fetcher(join=True)
http_count, socks_count, sources = 0, 0, {}
for p in proxies.values():
if tuple(p.types)[0].name.startswith('HTTP'):
http_count += 1
else:
socks_count += 1
for source in p.fetch_sources:
sources.setdefault(source, {'total': 0, 'uniq': 0})
sources[source]['total'] += 1
if len(p.fetch_sources) == 1:
sources[source]['uniq'] += 1
sources = ', '.join(['{0}:total={1[total]} uniq={1[uniq]}'.format(k, v)
for k, v in sources.items()])
logging.info('Fetched %s proxies (http(s)=%s, socks=%s %s)',
len(proxies), http_count, socks_count, sources)
json_encoder.dump(proxies.values(), save or sys.stdout)
def main(training_set, validation_set, ruleset, trainee, training_cache, validation_cache, delay, show_browser, stop_early, learning_rate, iterations, pos_weight, comment, quiet, confidence_threshold, layers, exclude):
"""Compute optimal numerical parameters for a Fathom ruleset.
There are a lot of options, but the usual invocation is something like...
fathom-train samples/training --validation-set samples/validation --stop-early --ruleset rulesets.js --trainee new
TRAINING_SET_FOLDER is a directory of labeled training pages. It can also
be, for backward compatibility, a JSON file of vectors from FathomFox's
Vectorizer.
To see graphs of the results, install TensorBoard, then run this:
tensorboard --logdir runs/. These will tell you whether you need to adjust
the --learning-rate.
Some vocab used in the output messages:
target -- A "right answer" DOM node, one that should be recognized
candidate -- Any node (target or not) brought into the ruleset, by a
dom() call, for consideration
negative sample -- A sample with no intended target nodes, used to bait
the recognizer into a false-positive choice
"""
training_set = Path(training_set)
# If they pass in a dir for either the training or validation sets, we need
# a ruleset and a trainee for vectorizing:
if (validation_set and validation_set.is_dir()) or training_set.is_dir():
if not ruleset:
raise BadOptionUsage('ruleset', 'A --ruleset file must be specified when TRAINING_SET_FOLDER or --validation-set are passed a directory.')
if not trainee:
raise BadOptionUsage('trainee', 'A --trainee ID must be specified when TRAINING_SET_FOLDER or --validation-set are passed a directory.')
with open(make_or_find_vectors(ruleset,
trainee,
training_set,
training_cache,
show_browser,
'training',
delay),
encoding='utf-8') as training_file:
training_data = exclude_features(exclude, load(training_file))
training_pages = training_data['pages']
x, y, num_yes = tensors_from(training_pages, shuffle=True)
if validation_set:
with open(make_or_find_vectors(ruleset,
trainee,
validation_set,
validation_cache,
show_browser,
'validation',
delay),
encoding='utf-8') as validation_file:
validation_pages = exclude_features(exclude, load(validation_file))['pages']
validation_ins, validation_outs, validation_yes = tensors_from(validation_pages)
validation_arg = validation_ins, validation_outs
else:
validation_arg = None
layers = list(layers) # Comes in as tuple
full_comment = '.LR={l},i={i}{c}'.format(
l=learning_rate,
i=iterations,
c=(',' + comment) if comment else '')
model = learn(learning_rate,
iterations,
x,
y,
confidence_threshold,
validation=validation_arg,
stop_early=stop_early,
run_comment=full_comment,
pos_weight=pos_weight,
layers=layers)
print(pretty_coeffs(model, training_data['header']['featureNames']))
accuracy, false_positives, false_negatives = accuracy_per_tag(y, model(x), confidence_threshold)
print(pretty_accuracy('Training',
accuracy,
len(x),
false_positives,
false_negatives,
num_yes))
if validation_set:
accuracy, false_positives, false_negatives = accuracy_per_tag(validation_outs, model(validation_ins), confidence_threshold)
print(pretty_accuracy('Validation',
accuracy,
len(validation_ins),
false_positives,
false_negatives,
validation_yes))
# Print timing information:
if training_pages and 'time' in training_pages[0]:
if validation_set and validation_pages and 'time' in validation_pages[0]:
print(speed_readout(training_pages + validation_pages))
else:
print(speed_readout(training_pages))
if not quiet:
print('\nTraining per-tag results:')
print_per_tag_report([per_tag_metrics(page, model, confidence_threshold) for page in training_pages])
if validation_set:
print('\nValidation per-tag results:')
print_per_tag_report([per_tag_metrics(page, model, confidence_threshold) for page in validation_pages])
