def finish_volume(self):
if self.volume_id is None:
return
self.cur.execute(
"""update directory
set written=localtimestamp
where id=%s""", (self.volume_id, ))
if self.inst_id:
self.cur.execute(
"""delete from locality
where url_id in (
select url_id
from content
where volume_id=%s
)""", (self.volume_id, ))
self.cur.execute(
"""select url_id
from content
where volume_id=%s
order by url_id""", (self.volume_id, ))
rows = self.cur.fetchall()
for row in rows:
url_id = row[0]
self.cond_remove(get_loose_path(url_id))
self.cond_remove(get_loose_path(url_id, True))
if self.member_count > 0:
print("packed %d pages" % (self.member_count, ))
self.member_count = 0
self.volume_id = None
def open_repre(self, url_id, alt_repre):
f = None
loose_path = get_loose_path(url_id, alt_repre=alt_repre)
if os.path.exists(loose_path):
f = open(loose_path, "rb")
return f
def write_file(self, url_id, hdr, reader):
loose_path = get_loose_path(url_id, hdr)
writer = open(loose_path, "wb")
try:
shutil.copyfileobj(reader, writer)
finally:
writer.close()
def write(self, data):
if self.body_target is None:
if self.retrieve_body:
self.body_target = open(get_loose_path(self.url_id), 'wb')
else:
return -1
return self.body_target.write(data)
def delete_storage(self, url_id):
# for deletion, checking locality is required
if not self.check_locality:
return
for hdr in (True, False):
loose_path = get_loose_path(url_id, hdr)
if os.path.exists(loose_path):
os.remove(loose_path)
def run(self):
self.cur.execute("""select count(*)
from download_queue""")
row = self.cur.fetchone()
num_conn = row[0]
if not num_conn:
return
self.lazy_init()
batch_processed = 0
row = self.pop_work_item()
while row:
url_id = row[0]
url = self.get_url(url_id)
self.br.get(url)
error_code = None
try:
WebDriverWait(self.br, 10).until(
EC.presence_of_element_located((By.TAG_NAME, 'a')))
except exceptions.TimeoutException:
error_code = 500
eff_id = url_id
eff_url = self.br.current_url
msg = "got " + eff_url
if error_code:
msg += " with %d" % error_code
print(msg, file=sys.stderr)
if url != eff_url:
eff_id, known = self.add_redirect(url_id, eff_url)
body = self.br.page_source
with open(get_loose_path(url_id), 'w') as f:
f.write(body)
if error_code:
self.cur.execute(
"""insert into download_error(url_id, error_code, failed)
values(%s, %s, localtimestamp)
on conflict(url_id) do update
set error_code=%s, failed=localtimestamp""", (url_id, error_code, error_code))
self.br.close()
self.br = None
self.lazy_init()
self.finish_page(url_id, eff_id, not error_code)
batch_processed += 1
if batch_processed >= self.notification_threshold:
self.cond_notify()
batch_processed = 0
row = self.pop_work_item()
def close(self):
self.header_target.close()
self.header_target = None
if self.body_target:
self.body_target.close()
self.body_target = None
if self.url_id != self.eff_id:
os.rename(get_loose_path(self.url_id, True), get_loose_path(self.eff_id, True))
old_path = get_loose_path(self.url_id)
if os.path.exists(old_path):
if self.retrieve_body:
os.rename(old_path, get_loose_path(self.eff_id))
else:
os.remove(old_path)
self.owner.finish_page(self.url_id, self.eff_id, self.retrieve_body)
def open_page_ex(self, url_id, volume_id):
if not self.alt_repre:
return self.open_page(url_id, volume_id)
f = None
loose_path = get_loose_path(url_id, alt_repre=self.alt_repre)
if os.path.exists(loose_path):
f = open(loose_path, "rb")
return f
def __init__(self, owner, url, url_id):
self.owner = owner
self.url = url
self.url_id = url_id
self.eff_id = url_id
self.header_target = open(get_loose_path(url_id, True), 'wb')
self.body_target = None
self.retrieve_body = True
self.http_code = None
self.http_phrase = None
self.retry_after = None
def get_body_size_ex(self, url_id, volume_id):
if not self.alt_repre:
return self.get_body_size(url_id, volume_id)
sz = None
loose_path = get_loose_path(url_id, alt_repre=self.alt_repre)
if os.path.exists(loose_path):
statinfo = os.stat(loose_path)
sz = statinfo.st_size
return sz
def open_headers(self, url_id, volume_id=None):
f = None
if volume_id is None:
loose_path = get_loose_path(url_id, True)
if os.path.exists(loose_path):
f = open(loose_path, "rb")
else:
if volume_id != self.volume_id:
self.change_volume(volume_id)
try:
info = self.zp.getinfo(str(url_id) + 'h')
f = self.zp.open(info)
except KeyError:
pass
return f
def get_body_size(self, url_id, volume_id=None):
sz = None
if volume_id is None:
loose_path = get_loose_path(url_id)
if os.path.exists(loose_path):
statinfo = os.stat(loose_path)
sz = statinfo.st_size
else:
if volume_id != self.volume_id:
self.change_volume(volume_id)
try:
info = self.zp.getinfo(str(url_id))
sz = info.file_size
except KeyError:
pass
return sz
def purge_fast(self, url_id):
volume_id = self.get_volume_id(url_id)
if volume_id is None:
loose_path = get_loose_path(url_id)
self.ensure_removed(loose_path)
loose_path += 'h'
self.ensure_removed(loose_path)
else:
self.shrunk.add(volume_id)
self.doomed.add(url_id)
self.purge_from_set(url_id)
self.purge_to_set(url_id)
self.cur.execute("""delete from edges
where from_id=%s or to_id=%s""", (url_id, url_id))
self.cur.execute("""delete from nodes
where url_id=%s""", (url_id,))
self.cur.execute("""delete from extra
where url_id=%s""", (url_id,))
self.cur.execute("""delete from redirect
where from_id=%s or to_id=%s""", (url_id, url_id))
self.cur.execute("""delete from parse_queue
where url_id=%s""", (url_id,))
self.cur.execute("""delete from content
where url_id=%s""", (url_id,))
self.cur.execute("""delete from download_error
where url_id=%s""", (url_id,))
self.cur.execute("""delete from download_queue
where url_id=%s""", (url_id,))
self.cur.execute("""delete from field
where id=%s""", (url_id,))
def make_target(self):
return open(get_loose_path(self.url_id, True), 'wb')
def add_member_half(self, url_id, hdr):
path = get_loose_path(url_id, hdr)
if os.path.exists(path):
self.zip_front.write(path, os.path.basename(path))
