def output(self):
outputs = []
# Get HDFS client:
client = luigi.contrib.hdfs.WebHdfsClient()
# Find log files:
parent_path = "/heritrix/output/%s/%s/logs" % (self.job,
self.launch_id)
for listed_item in client.listdir(parent_path):
# Oddly, depending on the implementation, the listed_path may be absolute or basename-only, so fix here:
item = os.path.basename(listed_item)
item_path = os.path.join(parent_path, item)
if item.endswith(".lck"):
logger.error("Lock file should be be present on HDFS! %s" %
(item, item_path))
pass
elif item.startswith("crawl.log"):
outputs.append(
luigi.contrib.hdfs.HdfsTarget(path=item_path,
format=Plain))
#logger.debug("Including %s" % item)
else:
pass
#logger.debug("Skipping %s" % item)
# Return the logs to be processed:
return outputs
def find_watched_target_for(self, url, source, publishers):
'''
Given a URL and an array of publisher strings, determine which Watched Target to associate them with.
'''
# Find the list of Targets where a seed matches the given URL
surt = url_to_surt(url, host_only=True)
matches = []
for t in self.targets:
if t['watched']:
a_match = False
for seed in t['seeds']:
if surt.startswith(url_to_surt(seed, host_only=True)):
a_match = True
if a_match:
matches.append(t)
# No matches:
if len(matches) == 0:
logger.error("No match found for url %s" % url)
return None
# raise Exception("No matching target for url "+url)
# If one match, assume that is the right Target:
if len(matches) == 1:
return int(matches[0]['id'])
#
# Else multiple matches, so need to disambiguate.
#
# Attempt to disambiguate based on source ONLY:
if source is not None:
for t in matches:
for seed in t['seeds']:
logger.info("Looking for source match '%s' against '%s' " % (source, seed))
if seed == source:
# return int(t['id'])
logger.info("Found match source+seed but this is not enough to disambiguate longer crawls.")
break
# Then attempt to disambiguate based on publisher
# FIXME Make this a bit more forgiving of punctation/minor differences
title_matches = []
for t in matches:
for publisher in publishers:
logger.info("Looking for publisher match '%s' in title '%s' " % (publisher, t['title']))
if publisher and publisher.lower() in t['title'].lower():
title_matches.append(t)
break
if len(title_matches) == 0:
logger.warning("No matching title to associate with url %s " % url)
return None
# raise Exception("No matching title to associate with url %s " % url)
elif len(title_matches) == 1:
return int(title_matches[0]['id'])
else:
logger.warning("Too many matching titles for %s" % url)
for t in title_matches:
logger.warning("Candidate: %d %s " % (t['id'], t['title']))
logger.warning("Assuming first match is sufficient... (%s)" % title_matches[0]['title'])
return int(title_matches[0]['id'])
def run(self):
"""Zips up all log/config. files and copies said archive to HDFS; finds the
earliest timestamp in the logs."""
# Set up remote connection:
rf = luigi.contrib.ssh.RemoteFileSystem(self.host)
# Set up the output, first making sure the full path exists:
with self.output().open('w') as f:
f.write('')
self.output().remove()
# What to remove from the paths:
chop = len(str(h3().local_prefix))
with zipfile.ZipFile(self.output().path, 'w',
allowZip64=True) as zipout:
# Crawl log:
for crawl_log in remote_ls(
"%s/logs/%s/%s" %
(h3.local_output_folder, self.job, self.launch_id),
"/crawl.log%s" % get_stage_suffix(self.stage), rf):
logger.info("Found %s..." % os.path.basename(crawl_log))
self.add_remote_file(zipout, crawl_log[chop:], crawl_log, rf)
# Error log(s)
for log in remote_ls(
"%s/logs/%s/%s" %
(LOCAL_OUTPUT_FOLDER, self.job, self.launch_id),
"/*-errors.log%s" % get_stage_suffix(self.stage), rf):
logger.info("Found %s..." % os.path.basename(log))
self.add_remote_file(zipout, log[chop:], log, rf)
# Job text files
for txt in remote_ls(
"%s/%s/%s" % (LOCAL_JOB_FOLDER, self.job, self.launch_id),
"/*.txt", rf):
self.add_remote_file(zipout, txt[chop:], txt, rf)
# Job json files:
for txt in remote_ls(
"%s/%s/%s" % (LOCAL_JOB_FOLDER, self.job, self.launch_id),
"/*.json", rf):
logger.info("Found %s..." % os.path.basename(txt))
self.add_remote_file(zipout, txt[chop:], txt, rf)
# Job crawler definition:
cxml = "%s/%s/%s/crawler-beans.cxml" % (LOCAL_JOB_FOLDER, self.job,
self.launch_id)
if rf.exists(cxml):
logger.info("Found config...")
self.add_remote_file(zipout, cxml[chop:], cxml, rf)
else:
logger.error("Cannot find config.")
raise Exception("Cannot find config.")
def mdex(self):
'''
Metadata extraction and target association.
'''
# Pass the document through a different extractor based on how the URL starts.
try:
if (self.doc["document_url"].startswith("https://www.gov.uk/")):
self.mdex_gov_uk_publications()
elif (self.doc["document_url"].startswith("http://www.ifs.org.uk/")
):
self.mdex_ifs_reports()
else:
self.mdex_default()
except Exception as e:
logger.error(
"Ignoring error during extraction for document %s and landing page %s"
% (self.doc['document_url'], self.doc['landing_page_url']))
logger.exception(e)
if not 'title' in self.doc or not self.doc['title']:
logger.info("Falling back on default extraction logic...")
self.mdex_default()
logger.info("GOT %s" % self.doc)
# Look up which Target this URL should be associated with:
if self.targets and self.doc.has_key('landing_page_url'):
logger.info(
"Looking for match for %s source %s and publishers '%s'" %
(self.doc['landing_page_url'], self.source,
self.doc.get('publishers', [])))
self.doc['target_id'] = self.find_watched_target_for(
self.doc['landing_page_url'], self.source,
self.doc.get('publishers', []))
# If there is no association, drop it:
if not self.doc.get('target_id',
None) and self.null_if_no_target_found:
logger.critical(
"Failed to associated document with any target: %s" % self.doc)
return None
# If the publisher appears unambiguous, store it where it can be re-used
if len(self.doc.get('publishers', [])) is 1:
self.doc['publisher'] = self.doc['publishers'][0]
# Or return the modified version:
return self.doc
def __init__(self, url, email, password):
self.url = url.rstrip("/")
loginUrl = "%s/login" % self.url
logger.info("Logging into %s as %s " % (loginUrl, email))
response = requests.post(loginUrl, data={"email": email, "password": password})
if not response.history:
logger.error("W3ACT Login failed!")
raise Exception("W3ACT Login Failed!")
self.cookie = response.history[0].headers["set-cookie"]
self.get_headers = {
"Cookie": self.cookie,
}
self.up_headers = {
"Cookie": self.cookie,
"Content-Type": "application/json"
}
self.ld_cache = CachedDict()
def _file_exists(path, rf):
"""
Checks whether the given file exists and has content - allowed to be '.open' at this point.
Also checks on HDFS if there is no local file.
:type path: str
"""
logger.info("Looking for %s" % path)
if rf.exists(path) and not rf.isdir(path): # and rf.getsize(path) > 0:
return True
elif rf.exists("%s.open" % path) and not rf.isdir(
"%s.open" % path): # and rf.getsize("%s.open" % path) > 0:
return True
else:
try:
if get_hdfs_target(path).exists():
return True
except luigi.contrib.hdfs.error.HDFSCliError as e:
logger.error("Exception while checking HDFS.")
logger.exception(e)
return False
def mdex(self):
'''
Metadata extraction and target association.
'''
# Pass the document through a different extractor based on how the URL starts.
try:
if( self.doc["document_url"].startswith("https://www.gov.uk/")):
self.mdex_gov_uk_publications()
elif( self.doc["document_url"].startswith("http://www.ifs.org.uk/")):
self.mdex_ifs_reports()
else:
self.mdex_default()
except Exception as e:
logger.error("Ignoring error during extraction for document %s and landing page %s" % (self.doc['document_url'], self.doc['landing_page_url']))
logger.exception(e)
if not 'title' in self.doc or not self.doc['title']:
logger.info("Falling back on default extraction logic...")
self.mdex_default()
logger.info("GOT %s" % self.doc)
# Look up which Target this URL should be associated with:
if self.targets and self.doc.has_key('landing_page_url'):
logger.info("Looking for match for %s source %s and publishers '%s'" % (self.doc['landing_page_url'], self.source, self.doc.get('publishers',[])))
self.doc['target_id'] = self.find_watched_target_for(self.doc['landing_page_url'], self.source, self.doc.get('publishers', []))
# If there is no association, drop it:
if not self.doc.get('target_id', None) and self.null_if_no_target_found:
logger.critical("Failed to associated document with any target: %s" % self.doc)
return None
# If the publisher appears unambiguous, store it where it can be re-used
if len(self.doc.get('publishers',[])) is 1:
self.doc['publisher'] = self.doc['publishers'][0]
# Or return the modified version:
return self.doc
def find_watched_target_for(self, url, source, publishers):
'''
Given a URL and an array of publisher strings, determine which Watched Target to associate them with.
'''
# Find the list of Targets where a seed matches the given URL
surt = url_to_surt(url, host_only=True)
matches = []
for t in self.targets:
if t['watched']:
a_match = False
for seed in t['seeds']:
if surt.startswith(url_to_surt(seed, host_only=True)):
a_match = True
if a_match:
matches.append(t)
# No matches:
if len(matches) == 0:
logger.error("No match found for url %s" % url)
return None
# raise Exception("No matching target for url "+url)
# If one match, assume that is the right Target:
if len(matches) == 1:
return int(matches[0]['id'])
#
# Else multiple matches, so need to disambiguate.
#
# Attempt to disambiguate based on source ONLY:
if source is not None:
for t in matches:
for seed in t['seeds']:
logger.info("Looking for source match '%s' against '%s' " %
(source, seed))
if seed == source:
# return int(t['id'])
logger.info(
"Found match source+seed but this is not enough to disambiguate longer crawls."
)
break
# Then attempt to disambiguate based on publisher
# FIXME Make this a bit more forgiving of punctation/minor differences
title_matches = []
for t in matches:
for publisher in publishers:
logger.info("Looking for publisher match '%s' in title '%s' " %
(publisher, t['title']))
if publisher and publisher.lower() in t['title'].lower():
title_matches.append(t)
break
if len(title_matches) == 0:
logger.warning("No matching title to associate with url %s " % url)
return None
# raise Exception("No matching title to associate with url %s " % url)
elif len(title_matches) == 1:
return int(title_matches[0]['id'])
else:
logger.warning("Too many matching titles for %s" % url)
for t in title_matches:
logger.warning("Candidate: %d %s " % (t['id'], t['title']))
logger.warning("Assuming first match is sufficient... (%s)" %
title_matches[0]['title'])
return int(title_matches[0]['id'])
