def test_parser():
'''test argument parsing for the command-line interface'''
testargs = [
'', '-fv', '--xmltei', '--notables', '-u', 'https://www.example.org'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert args.fast is True
assert args.verbose is True
assert args.notables is False
assert args.xmltei is True
assert args.URL == 'https://www.example.org'
args = cli.map_args(args)
assert args.output_format == 'xmltei'
testargs = ['', '-out', 'csv', '-u', 'https://www.example.org']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert args.fast is False
assert args.verbose is False
assert args.output_format == 'csv'
# test args mapping
testargs = ['', '--xml']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
args = cli.map_args(args)
assert args.output_format == 'xml'
args.xml, args.csv = False, True
args = cli.map_args(args)
assert args.output_format == 'csv'
args.csv, args.json = False, True
args = cli.map_args(args)
assert args.output_format == 'json'
# process_args
args.inputdir = '/dev/null'
args.verbose = True
args.blacklist = os.path.join(TEST_DIR, 'resources/list-discard.txt')
cli.process_args(args)
assert len(args.blacklist) == 2
# filter
testargs = [
'', '-i', 'resources/list-discard.txt', '--url-filter', 'test1',
'test2'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert args.inputfile == 'resources/list-discard.txt'
assert args.url_filter == ['test1', 'test2']
resources_dir = os.path.join(TEST_DIR, 'resources')
args.inputfile = os.path.join(resources_dir, 'list-discard.txt')
args.blacklist == os.path.join(resources_dir, 'list-discard.txt')
f = io.StringIO()
with redirect_stdout(f):
cli.process_args(args)
assert len(f.getvalue()) == 0
def test_parser():
'''test argument parsing for the command-line interface'''
testargs = [
'', '-fv', '--xmltei', '--notables', '-u', 'https://www.example.org'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert args.fast is True
assert args.verbose is True
assert args.notables is False
assert args.xmltei is True
assert args.URL == 'https://www.example.org'
args = cli.map_args(args)
assert args.output_format == 'xmltei'
testargs = ['', '-out', 'csv', '-u', 'https://www.example.org']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert args.fast is False
assert args.verbose is False
assert args.output_format == 'csv'
# test args mapping
testargs = ['', '--xml']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
args = cli.map_args(args)
assert args.output_format == 'xml'
args.xml, args.csv = False, True
args = cli.map_args(args)
assert args.output_format == 'csv'
args.csv, args.json = False, True
args = cli.map_args(args)
assert args.output_format == 'json'
# process_args
args.inputdir = '/dev/null'
args.verbose = True
args.blacklist = os.path.join(TEST_DIR, 'resources/list-discard.txt')
cli.process_args(args)
assert len(args.blacklist) == 4
def test_parser():
'''test argument parsing for the command-line interface'''
testargs = [
'', '-fvv', '--xmltei', '--notables', '-u', 'https://www.example.org'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert args.fast is True
assert args.verbose == 2
assert args.notables is False and args.no_tables is False
assert args.xmltei is True
assert args.URL == 'https://www.example.org'
args = cli.map_args(args)
assert args.output_format == 'xmltei'
testargs = [
'', '-out', 'csv', '--no-tables', '-u', 'https://www.example.org'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert args.fast is False
assert args.verbose == 0
assert args.output_format == 'csv'
assert args.no_tables is False
# test args mapping
testargs = ['', '--xml', '--nocomments', '--precision', '--recall']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
args = cli.map_args(args)
assert args.output_format == 'xml' and args.no_comments is False
# combination possible (?)
assert args.precision is True and args.recall is True
args.xml, args.csv = False, True
args = cli.map_args(args)
assert args.output_format == 'csv'
args.csv, args.json = False, True
args = cli.map_args(args)
assert args.output_format == 'json'
testargs = ['', '--with-metadata']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
args = cli.map_args(args)
assert args.only_with_metadata is True
# process_args
args.inputdir = '/dev/null'
args.verbose = 1
args.blacklist = os.path.join(RESOURCES_DIR, 'list-discard.txt')
cli.process_args(args)
assert len(args.blacklist) == 2
# filter
testargs = [
'', '-i', 'resources/list-discard.txt', '--url-filter', 'test1',
'test2'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert args.inputfile == 'resources/list-discard.txt'
assert args.url_filter == ['test1', 'test2']
args.inputfile = os.path.join(RESOURCES_DIR, 'list-discard.txt')
args.blacklist = os.path.join(RESOURCES_DIR, 'list-discard.txt')
f = io.StringIO()
with redirect_stdout(f):
cli.process_args(args)
assert len(f.getvalue()) == 0
# version
testargs = ['', '--version']
with pytest.raises(SystemExit) as e, redirect_stdout(f):
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert e.type == SystemExit
assert e.value.code == 0
assert re.match(
r'Trafilatura [0-9]\.[0-9]\.[0-9] - Python [0-9]\.[0-9]+\.[0-9]',
f.getvalue())
def test_cli_pipeline():
'''test command-line processing pipeline'''
# straight command-line input
#testargs = ['', '<html><body>Text</body></html>']
#with patch.object(sys, 'argv', testargs):
# args = cli.parse_args(testargs)
#f = io.StringIO()
#with redirect_stdout(f):
# cli.process_args(args)
#assert len(f.getvalue()) == 0
# test URL listing
# Force encoding to utf-8 for Windows in future processes spawned by multiprocessing.Pool
os.environ['PYTHONIOENCODING'] = "utf-8"
testargs = ['', '--list']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert cli_utils.url_processing_pipeline(args, dict()) is None
# test inputlist + blacklist
testargs = ['', '-i', os.path.join(RESOURCES_DIR, 'list-process.txt')]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
my_urls = cli_utils.load_input_urls(args)
assert my_urls is not None and len(my_urls) == 2
testargs = [
'', '-i',
os.path.join(RESOURCES_DIR, 'list-process.txt'), '--blacklist',
os.path.join(RESOURCES_DIR, 'list-discard.txt'), '--archived'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert args.blacklist is not None
# test backoff between domain requests
inputdict = add_to_compressed_dict(my_urls, args.blacklist, None, None)
reftime = datetime.now()
cli_utils.url_processing_pipeline(args, inputdict)
delta = (datetime.now() - reftime).total_seconds()
assert delta > 2
# test blacklist and empty dict
args.blacklist = cli_utils.load_blacklist(args.blacklist)
assert len(args.blacklist) == 2
inputdict = add_to_compressed_dict(my_urls, args.blacklist, None, None)
cli_utils.url_processing_pipeline(args, inputdict)
# test backup
testargs = ['', '--backup-dir', '/tmp/']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
cli_utils.archive_html('00Test', args)
# test date-based exclusion
testargs = ['', '-out', 'xml', '--with-metadata']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
with open(os.path.join(RESOURCES_DIR, 'httpbin_sample.html'), 'r') as f:
teststring = f.read()
assert cli.examine(teststring, args) is None
testargs = ['', '-out', 'xml', '--only-with-metadata', '--precision']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
with open(os.path.join(RESOURCES_DIR, 'httpbin_sample.html'), 'r') as f:
teststring = f.read()
assert cli.examine(teststring, args) is None
# test JSON output
testargs = ['', '-out', 'json', '--recall']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
with open(os.path.join(RESOURCES_DIR, 'httpbin_sample.html'), 'r') as f:
teststring = f.read()
assert cli.examine(teststring, args) is not None
# dry-run file processing pipeline
testargs = ['', '--parallel', '1', '--inputdir', '/dev/null']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
cli_utils.file_processing_pipeline(args)
# file processing pipeline on resources/
args.inputdir = RESOURCES_DIR
cli_utils.file_processing_pipeline(args)
# sitemaps
testargs = ['', '--sitemap', 'https://httpbin.org/', '--list']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
f = io.StringIO()
with redirect_stdout(f):
cli.process_args(args)
assert len(f.getvalue()) == 0
# config file
testargs = [
'', '--inputdir', '/dev/null', '--config-file', 'newsettings.cfg'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
with open(os.path.join(RESOURCES_DIR, 'httpbin_sample.html'), 'r') as f:
teststring = f.read()
args.config_file = os.path.join(RESOURCES_DIR, args.config_file)
# config = use_config(filename=args.config_file)
assert cli.examine(teststring, args) is None
# CLI options
testargs = ['', '--links', '--images']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
#with open(os.path.join(RESOURCES_DIR, 'http_sample.html'), 'r') as f:
# teststring = f.read()
#result = cli.examine(teststring, args)
#assert '[link](testlink.html)' in result # and 'test.jpg' in result
# Crawling
testargs = ['', '--crawl', 'https://httpbin.org/html']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
f = io.StringIO()
with redirect_stdout(f):
cli_utils.cli_crawler(args)
assert len(f.getvalue()) == 0
# links permitted
testargs = [
'', '--crawl', 'https://httpbin.org/links/1/1', '--list', '--parallel',
'1'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
f = io.StringIO()
with redirect_stdout(f):
cli_utils.cli_crawler(args)
assert f.getvalue() == 'https://httpbin.org/links/1/0\n'
# 0 links permitted
args.crawl = 'https://httpbin.org/links/4/4'
f = io.StringIO()
with redirect_stdout(f):
cli_utils.cli_crawler(args, n=0)
# print(f.getvalue())
assert len(f.getvalue().split('\n')) == 5
# Exploration (Sitemap + Crawl)
testargs = ['', '--explore', 'https://httpbin.org/html']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
f = io.StringIO()
with redirect_stdout(f):
cli.process_args(args)
assert len(f.getvalue()) == 0
def test_cli_pipeline():
'''test command-line processing pipeline'''
# straight command-line input
#testargs = ['', '<html><body>Text</body></html>']
#with patch.object(sys, 'argv', testargs):
# args = cli.parse_args(testargs)
#f = io.StringIO()
#with redirect_stdout(f):
# cli.process_args(args)
#assert len(f.getvalue()) == 0
# test URL listing
testargs = ['', '--list']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert cli_utils.url_processing_pipeline(args, dict()) is None
# test conversion and storage
inputdict = cli.convert_inputlist(None, ['ftps://www.example.org/'], None,
None)
assert inputdict == dict()
inputdict = cli.convert_inputlist(None, ['https://www.example.org/'], None,
None)
assert cli_utils.url_processing_pipeline(args, inputdict) is None
# test inputlist + blacklist
testargs = ['', '-i', os.path.join(RESOURCES_DIR, 'list-process.txt')]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
my_urls = cli_utils.load_input_urls(args.inputfile)
assert my_urls is not None and len(my_urls) == 2
testargs = [
'', '-i',
os.path.join(RESOURCES_DIR, 'list-process.txt'), '--blacklist',
os.path.join(RESOURCES_DIR, 'list-discard.txt'), '--archived'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert args.blacklist is not None
# test backoff between domain requests
inputdict = cli_utils.convert_inputlist(args.blacklist, my_urls, None,
None)
reftime = datetime.now()
cli_utils.url_processing_pipeline(args, inputdict)
delta = (datetime.now() - reftime).total_seconds()
assert delta > 2
# test blacklist and empty dict
args.blacklist = cli_utils.load_blacklist(args.blacklist)
assert len(args.blacklist) == 2
inputdict = cli_utils.convert_inputlist(args.blacklist, my_urls, None,
None)
cli_utils.url_processing_pipeline(args, inputdict)
# test backup
testargs = ['', '--backup-dir', '/tmp/']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
cli_utils.archive_html('00Test', args)
# test date-based exclusion
testargs = ['', '-out', 'xml', '--with-metadata']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
with open(os.path.join(RESOURCES_DIR, 'httpbin_sample.html'), 'r') as f:
teststring = f.read()
assert cli.examine(teststring, args) is None
# test JSON output
testargs = ['', '-out', 'json']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
with open(os.path.join(RESOURCES_DIR, 'httpbin_sample.html'), 'r') as f:
teststring = f.read()
assert cli.examine(teststring, args) is not None
# dry-run file processing pipeline
testargs = ['', '--parallel', '1', '--inputdir', '/dev/null']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
cli_utils.file_processing_pipeline(args)
# file processing pipeline on resources/
args.inputdir = RESOURCES_DIR
cli_utils.file_processing_pipeline(args)
# sitemaps
testargs = ['', '--sitemap', 'https://httpbin.org/', '--list']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
f = io.StringIO()
with redirect_stdout(f):
cli.process_args(args)
assert len(f.getvalue()) == 0
# config file
testargs = [
'', '--inputdir', '/dev/null', '--config-file', 'newsettings.cfg'
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
with open(os.path.join(RESOURCES_DIR, 'httpbin_sample.html'), 'r') as f:
teststring = f.read()
args.config_file = os.path.join(RESOURCES_DIR, args.config_file)
config = use_config(filename=args.config_file)
assert cli.examine(teststring, args) is None
# CLI options
testargs = ['', '--links', '--images']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
def test_cli_pipeline():
'''test command-line processing pipeline'''
# test URL listing
testargs = ['', '--list']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
assert cli_utils.url_processing_pipeline(args, [], 0) is None
assert cli_utils.url_processing_pipeline(
args, ['https://www.example.org/'], 0) is None
# test inputlist + blacklist
resources_dir = os.path.join(TEST_DIR, 'resources')
testargs = ['', '-i', os.path.join(resources_dir, 'list-process.txt')]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
my_urls = cli_utils.load_input_urls(args.inputfile)
assert my_urls is not None and len(my_urls) == 2
resources_dir = os.path.join(TEST_DIR, 'resources')
testargs = [
'', '-i',
os.path.join(resources_dir, 'list-process.txt'), '--blacklist',
os.path.join(resources_dir, 'list-discard.txt')
]
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
print(args.blacklist)
assert args.blacklist is not None
# test backoff between domain requests
reftime = datetime.now()
assert cli_utils.url_processing_pipeline(args, my_urls, 2) is None
delta = (datetime.now() - reftime).total_seconds()
print(delta)
assert delta > 2
# test backup
testargs = ['', '--backup-dir', '/tmp/']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
cli_utils.archive_html('00Test', args)
cli_utils.url_processing_pipeline(args, my_urls, 2)
# test date-based exclusion
testargs = ['', '-out', 'xml', '--with-metadata']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
with open(os.path.join(resources_dir, 'httpbin_sample.html'), 'r') as f:
teststring = f.read()
assert cli.examine(teststring, args) is None
# test timeout
testargs = ['', '-out', 'xml', '--timeout']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
with open(os.path.join(resources_dir, 'httpbin_sample.html'), 'r') as f:
teststring = f.read()
assert cli.examine(teststring, args) is not None
# test JSON output
testargs = ['', '-out', 'json']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
with open(os.path.join(resources_dir, 'httpbin_sample.html'), 'r') as f:
teststring = f.read()
assert cli.examine(teststring, args) is not None
# dry-run file processing pipeline
testargs = ['', '--parallel', '1', '--inputdir', '/dev/null']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
cli_utils.file_processing_pipeline(args)
# file processing pipeline on resources/
args.inputdir = resources_dir
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
cli_utils.file_processing_pipeline(args)
# sitemaps
testargs = ['', '--sitemap', 'https://httpbin.org/', '--list']
with patch.object(sys, 'argv', testargs):
args = cli.parse_args(testargs)
f = io.StringIO()
with redirect_stdout(f):
cli.process_args(args)
assert len(f.getvalue()) == 0
