"""
sphinx.builders.linkcheck
~~~~~~~~~~~~~~~~~~~~~~~~~
The CheckExternalLinksBuilder class.
:copyright: Copyright 2007-2021 by the Sphinx team, see AUTHORS.
:license: BSD, see LICENSE for details.
"""
import json
import queue
import re
import socket
import threading
import time
from datetime import datetime, timezone
from email.utils import parsedate_to_datetime
from html.parser import HTMLParser
from os import path
from typing import Any, Dict, List, NamedTuple, Optional, Set, Tuple
from urllib.parse import unquote, urlparse
from docutils import nodes
from docutils.nodes import Element, Node
from requests import Response
from requests.exceptions import HTTPError, TooManyRedirects
from sphinx.application import Sphinx
from sphinx.builders import Builder
from sphinx.locale import __
from sphinx.util import encode_uri, logging, requests
from sphinx.util.console import darkgray, darkgreen, purple, red, turquoise # type: ignore
from sphinx.util.nodes import get_node_line
logger = logging.getLogger(__name__)
uri_re = re.compile('([a-z]+:)?//') # matches to foo:// and // (a protocol relative URL)
RateLimit = NamedTuple('RateLimit', (('delay', float), ('next_check', float)))
DEFAULT_REQUEST_HEADERS = {
'Accept': 'text/html,application/xhtml+xml;q=0.9,*/*;q=0.8',
}
CHECK_IMMEDIATELY = 0
QUEUE_POLL_SECS = 1
DEFAULT_DELAY = 60.0
def node_line_or_0(node: Element) -> int:
"""
PriorityQueue items must be comparable. The line number is part of the
tuple used by the PriorityQueue, keep an homogeneous type for comparison.
"""
return get_node_line(node) or 0
class AnchorCheckParser(HTMLParser):
"""Specialized HTML parser that looks for a specific anchor."""
def __init__(self, search_anchor: str) -> None:
super().__init__()
self.search_anchor = search_anchor
self.found = False
def handle_starttag(self, tag: Any, attrs: Any) -> None:
for key, value in attrs:
if key in ('id', 'name') and value == self.search_anchor:
self.found = True
break
def check_anchor(response: requests.requests.Response, anchor: str) -> bool:
"""Reads HTML data from a response object `response` searching for `anchor`.
Returns True if anchor was found, False otherwise.
"""
parser = AnchorCheckParser(anchor)
# Read file in chunks. If we find a matching anchor, we break
# the loop early in hopes not to have to download the whole thing.
for chunk in response.iter_content(chunk_size=4096, decode_unicode=True):
if isinstance(chunk, bytes): # requests failed to decode
chunk = chunk.decode() # manually try to decode it
parser.feed(chunk)
if parser.found:
break
parser.close()
return parser.found
[ドキュメント]class CheckExternalLinksBuilder(Builder):
"""
Checks for broken external links.
"""
name = 'linkcheck'
epilog = __('Look for any errors in the above output or in '
'%(outdir)s/output.txt')
def init(self) -> None:
self.to_ignore = [re.compile(x) for x in self.app.config.linkcheck_ignore]
self.anchors_ignore = [re.compile(x)
for x in self.app.config.linkcheck_anchors_ignore]
self.auth = [(re.compile(pattern), auth_info) for pattern, auth_info
in self.app.config.linkcheck_auth]
self.good = set() # type: Set[str]
self.broken = {} # type: Dict[str, str]
self.redirected = {} # type: Dict[str, Tuple[str, int]]
# set a timeout for non-responding servers
socket.setdefaulttimeout(5.0)
# create output file
open(path.join(self.outdir, 'output.txt'), 'w').close()
# create JSON output file
open(path.join(self.outdir, 'output.json'), 'w').close()
# create queues and worker threads
self.rate_limits = {} # type: Dict[str, RateLimit]
self.wqueue = queue.PriorityQueue() # type: queue.PriorityQueue
self.rqueue = queue.Queue() # type: queue.Queue
self.workers = [] # type: List[threading.Thread]
for i in range(self.app.config.linkcheck_workers):
thread = threading.Thread(target=self.check_thread, daemon=True)
thread.start()
self.workers.append(thread)
def check_thread(self) -> None:
kwargs = {}
if self.app.config.linkcheck_timeout:
kwargs['timeout'] = self.app.config.linkcheck_timeout
def get_request_headers() -> Dict:
url = urlparse(uri)
candidates = ["%s://%s" % (url.scheme, url.netloc),
"%s://%s/" % (url.scheme, url.netloc),
uri,
"*"]
for u in candidates:
if u in self.config.linkcheck_request_headers:
headers = dict(DEFAULT_REQUEST_HEADERS)
headers.update(self.config.linkcheck_request_headers[u])
return headers
return {}
def check_uri() -> Tuple[str, str, int]:
# split off anchor
if '#' in uri:
req_url, anchor = uri.split('#', 1)
for rex in self.anchors_ignore:
if rex.match(anchor):
anchor = None
break
else:
req_url = uri
anchor = None
# handle non-ASCII URIs
try:
req_url.encode('ascii')
except UnicodeError:
req_url = encode_uri(req_url)
# Get auth info, if any
for pattern, auth_info in self.auth:
if pattern.match(uri):
break
else:
auth_info = None
# update request headers for the URL
kwargs['headers'] = get_request_headers()
try:
if anchor and self.app.config.linkcheck_anchors:
# Read the whole document and see if #anchor exists
response = requests.get(req_url, stream=True, config=self.app.config,
auth=auth_info, **kwargs)
response.raise_for_status()
found = check_anchor(response, unquote(anchor))
if not found:
raise Exception(__("Anchor '%s' not found") % anchor)
else:
try:
# try a HEAD request first, which should be easier on
# the server and the network
response = requests.head(req_url, allow_redirects=True,
config=self.app.config, auth=auth_info,
**kwargs)
response.raise_for_status()
except (HTTPError, TooManyRedirects) as err:
if isinstance(err, HTTPError) and err.response.status_code == 429:
raise
# retry with GET request if that fails, some servers
# don't like HEAD requests.
response = requests.get(req_url, stream=True,
config=self.app.config,
auth=auth_info, **kwargs)
response.raise_for_status()
except HTTPError as err:
if err.response.status_code == 401:
# We'll take "Unauthorized" as working.
return 'working', ' - unauthorized', 0
elif err.response.status_code == 429:
next_check = self.limit_rate(err.response)
if next_check is not None:
self.wqueue.put((next_check, uri, docname, lineno), False)
return 'rate-limited', '', 0
return 'broken', str(err), 0
elif err.response.status_code == 503:
# We'll take "Service Unavailable" as ignored.
return 'ignored', str(err), 0
else:
return 'broken', str(err), 0
except Exception as err:
return 'broken', str(err), 0
else:
netloc = urlparse(req_url).netloc
try:
del self.rate_limits[netloc]
except KeyError:
pass
if response.url.rstrip('/') == req_url.rstrip('/'):
return 'working', '', 0
else:
new_url = response.url
if anchor:
new_url += '#' + anchor
# history contains any redirects, get last
if response.history:
code = response.history[-1].status_code
return 'redirected', new_url, code
else:
return 'redirected', new_url, 0
def check(docname: str) -> Tuple[str, str, int]:
# check for various conditions without bothering the network
if len(uri) == 0 or uri.startswith(('#', 'mailto:', 'tel:')):
return 'unchecked', '', 0
elif not uri.startswith(('http:', 'https:')):
if uri_re.match(uri):
# non supported URI schemes (ex. ftp)
return 'unchecked', '', 0
else:
srcdir = path.dirname(self.env.doc2path(docname))
if path.exists(path.join(srcdir, uri)):
return 'working', '', 0
else:
for rex in self.to_ignore:
if rex.match(uri):
return 'ignored', '', 0
else:
self.broken[uri] = ''
return 'broken', '', 0
elif uri in self.good:
return 'working', 'old', 0
elif uri in self.broken:
return 'broken', self.broken[uri], 0
elif uri in self.redirected:
return 'redirected', self.redirected[uri][0], self.redirected[uri][1]
for rex in self.to_ignore:
if rex.match(uri):
return 'ignored', '', 0
# need to actually check the URI
for _ in range(self.app.config.linkcheck_retries):
status, info, code = check_uri()
if status != "broken":
break
if status == "working":
self.good.add(uri)
elif status == "broken":
self.broken[uri] = info
elif status == "redirected":
self.redirected[uri] = (info, code)
return (status, info, code)
while True:
next_check, uri, docname, lineno = self.wqueue.get()
if uri is None:
break
netloc = urlparse(uri).netloc
try:
# Refresh rate limit.
# When there are many links in the queue, workers are all stuck waiting
# for responses, but the builder keeps queuing. Links in the queue may
# have been queued before rate limits were discovered.
next_check = self.rate_limits[netloc].next_check
except KeyError:
pass
if next_check > time.time():
# Sleep before putting message back in the queue to avoid
# waking up other threads.
time.sleep(QUEUE_POLL_SECS)
self.wqueue.put((next_check, uri, docname, lineno), False)
self.wqueue.task_done()
continue
status, info, code = check(docname)
if status == 'rate-limited':
logger.info(darkgray('-rate limited- ') + uri + darkgray(' | sleeping...'))
else:
self.rqueue.put((uri, docname, lineno, status, info, code))
self.wqueue.task_done()
def limit_rate(self, response: Response) -> Optional[float]:
next_check = None
retry_after = response.headers.get("Retry-After")
if retry_after:
try:
# Integer: time to wait before next attempt.
delay = float(retry_after)
except ValueError:
try:
# An HTTP-date: time of next attempt.
until = parsedate_to_datetime(retry_after)
except (TypeError, ValueError):
# TypeError: Invalid date format.
# ValueError: Invalid date, e.g. Oct 52th.
pass
else:
next_check = datetime.timestamp(until)
delay = (until - datetime.now(timezone.utc)).total_seconds()
else:
next_check = time.time() + delay
netloc = urlparse(response.url).netloc
if next_check is None:
max_delay = self.app.config.linkcheck_rate_limit_timeout
try:
rate_limit = self.rate_limits[netloc]
except KeyError:
delay = DEFAULT_DELAY
else:
last_wait_time = rate_limit.delay
delay = 2.0 * last_wait_time
if delay > max_delay and last_wait_time < max_delay:
delay = max_delay
if delay > max_delay:
return None
next_check = time.time() + delay
self.rate_limits[netloc] = RateLimit(delay, next_check)
return next_check
def process_result(self, result: Tuple[str, str, int, str, str, int]) -> None:
uri, docname, lineno, status, info, code = result
filename = self.env.doc2path(docname, None)
linkstat = dict(filename=filename, lineno=lineno,
status=status, code=code, uri=uri,
info=info)
if status == 'unchecked':
self.write_linkstat(linkstat)
return
if status == 'working' and info == 'old':
self.write_linkstat(linkstat)
return
if lineno:
logger.info('(line %4d) ', lineno, nonl=True)
if status == 'ignored':
if info:
logger.info(darkgray('-ignored- ') + uri + ': ' + info)
else:
logger.info(darkgray('-ignored- ') + uri)
self.write_linkstat(linkstat)
elif status == 'local':
logger.info(darkgray('-local- ') + uri)
self.write_entry('local', docname, filename, lineno, uri)
self.write_linkstat(linkstat)
elif status == 'working':
logger.info(darkgreen('ok ') + uri + info)
self.write_linkstat(linkstat)
elif status == 'broken':
if self.app.quiet or self.app.warningiserror:
logger.warning(__('broken link: %s (%s)'), uri, info,
location=(filename, lineno))
else:
logger.info(red('broken ') + uri + red(' - ' + info))
self.write_entry('broken', docname, filename, lineno, uri + ': ' + info)
self.write_linkstat(linkstat)
elif status == 'redirected':
try:
text, color = {
301: ('permanently', purple),
302: ('with Found', purple),
303: ('with See Other', purple),
307: ('temporarily', turquoise),
308: ('permanently', purple),
}[code]
except KeyError:
text, color = ('with unknown code', purple)
linkstat['text'] = text
logger.info(color('redirect ') + uri + color(' - ' + text + ' to ' + info))
self.write_entry('redirected ' + text, docname, filename,
lineno, uri + ' to ' + info)
self.write_linkstat(linkstat)
def get_target_uri(self, docname: str, typ: str = None) -> str:
return ''
def get_outdated_docs(self) -> Set[str]:
return self.env.found_docs
def prepare_writing(self, docnames: Set[str]) -> None:
return
def write_doc(self, docname: str, doctree: Node) -> None:
logger.info('')
n = 0
# reference nodes
for refnode in doctree.traverse(nodes.reference):
if 'refuri' not in refnode:
continue
uri = refnode['refuri']
lineno = node_line_or_0(refnode)
uri_info = (CHECK_IMMEDIATELY, uri, docname, lineno)
self.wqueue.put(uri_info, False)
n += 1
# image nodes
for imgnode in doctree.traverse(nodes.image):
uri = imgnode['candidates'].get('?')
if uri and '://' in uri:
lineno = node_line_or_0(imgnode)
uri_info = (CHECK_IMMEDIATELY, uri, docname, lineno)
self.wqueue.put(uri_info, False)
n += 1
done = 0
while done < n:
self.process_result(self.rqueue.get())
done += 1
if self.broken:
self.app.statuscode = 1
def write_entry(self, what: str, docname: str, filename: str, line: int,
uri: str) -> None:
with open(path.join(self.outdir, 'output.txt'), 'a') as output:
output.write("%s:%s: [%s] %s\n" % (filename, line, what, uri))
def write_linkstat(self, data: dict) -> None:
with open(path.join(self.outdir, 'output.json'), 'a') as output:
output.write(json.dumps(data))
output.write('\n')
def finish(self) -> None:
self.wqueue.join()
# Shutdown threads.
for worker in self.workers:
self.wqueue.put((CHECK_IMMEDIATELY, None, None, None), False)
def setup(app: Sphinx) -> Dict[str, Any]:
app.add_builder(CheckExternalLinksBuilder)
app.add_config_value('linkcheck_ignore', [], None)
app.add_config_value('linkcheck_auth', [], None)
app.add_config_value('linkcheck_request_headers', {}, None)
app.add_config_value('linkcheck_retries', 1, None)
app.add_config_value('linkcheck_timeout', None, None, [int])
app.add_config_value('linkcheck_workers', 5, None)
app.add_config_value('linkcheck_anchors', True, None)
# Anchors starting with ! are ignored since they are
# commonly used for dynamic pages
app.add_config_value('linkcheck_anchors_ignore', ["^!"], None)
app.add_config_value('linkcheck_rate_limit_timeout', 300.0, None)
return {
'version': 'builtin',
'parallel_read_safe': True,
'parallel_write_safe': True,
}
