def initializeRawStartUrls():
print("Initializing all start URLs in the database")
with common.database.session_context() as sess:
for module in RawArchiver.RawActiveModules.ACTIVE_MODULES:
for starturl in module.get_start_urls():
have = sess.query(common.database.RawWebPages) \
.filter(common.database.RawWebPages.url == starturl) \
.count()
if not have:
netloc = urlFuncs.getNetLoc(starturl)
new = common.database.RawWebPages(
url = starturl,
starturl = starturl,
netloc = netloc,
priority = common.database.DB_IDLE_PRIORITY,
distance = common.database.DB_DEFAULT_DIST,
)
print("Missing start-url for address: '{}'".format(starturl))
sess.add(new)
try:
sess.commit()
except Exception:
print("Failure inserting start url for address: '{}'".format(starturl))
sess.rollback()
def initializeStartUrls(rules):
print("Initializing all start URLs in the database")
with common.database.session_context() as sess:
for ruleset in [
rset for rset in rules
if rset['starturls'] and rset['rewalk_disabled'] is False
]:
for starturl in ruleset['starturls']:
have = sess.query(db.WebPages) \
.filter(db.WebPages.url == starturl) \
.count()
if not have:
netloc = urlFuncs.getNetLoc(starturl)
new = db.WebPages(
url=starturl,
starturl=starturl,
netloc=netloc,
type=ruleset['type'],
priority=db.DB_IDLE_PRIORITY,
distance=db.DB_DEFAULT_DIST,
normal_fetch_mode=ruleset['normal_fetch_mode'],
epoch=0,
)
print(
"Missing start-url for address: '{}'".format(starturl))
sess.add(new)
try:
sess.commit()
except sqlalchemy.SQLAlchemyError:
print(
"Failure inserting start url for address: '{}'".format(
starturl))
sess.rollback()
def initializeStartUrls(rules):
print("Initializing all start URLs in the database")
sess = db.get_db_session()
for ruleset in [rset for rset in rules if rset['starturls']]:
for starturl in ruleset['starturls']:
have = sess.query(db.WebPages) \
.filter(db.WebPages.url == starturl) \
.count()
if not have:
netloc = urlFuncs.getNetLoc(starturl)
new = db.WebPages(
url = starturl,
starturl = starturl,
netloc = netloc,
type = ruleset['type'],
priority = db.DB_IDLE_PRIORITY,
distance = db.DB_DEFAULT_DIST,
normal_fetch_mode = ruleset['normal_fetch_mode'],
)
print("Missing start-url for address: '{}'".format(starturl))
sess.add(new)
try:
sess.commit()
except sqlalchemy.SQLAlchemyError:
print("Failure inserting start url for address: '{}'".format(starturl))
sess.rollback()
sess.close()
db.delete_db_session()
def initializeStartUrls(rules):
print("Initializing all start URLs in the database")
sess = db.get_db_session()
for ruleset in [rset for rset in rules if rset['starturls']]:
for starturl in ruleset['starturls']:
have = sess.query(db.WebPages) \
.filter(db.WebPages.url == starturl) \
.count()
if not have:
netloc = urlFuncs.getNetLoc(starturl)
new = db.WebPages(
url = starturl,
starturl = starturl,
netloc = netloc,
type = ruleset['type'],
priority = db.DB_IDLE_PRIORITY,
distance = db.DB_DEFAULT_DIST,
normal_fetch_mode = ruleset['normal_fetch_mode'],
)
print("Missing start-url for address: '{}'".format(starturl))
sess.add(new)
try:
sess.commit()
except sqlalchemy.SQLAlchemyError:
print("Failure inserting start url for address: '{}'".format(starturl))
sess.rollback()
sess.close()
db.delete_db_session()
def initializeRawStartUrls():
print("Initializing all start URLs in the database")
sess = common.database.get_db_session()
for module in RawArchiver.RawActiveModules.ACTIVE_MODULES:
for starturl in module.get_start_urls():
have = sess.query(common.database.RawWebPages) \
.filter(common.database.RawWebPages.url == starturl) \
.count()
if not have:
netloc = urlFuncs.getNetLoc(starturl)
new = common.database.RawWebPages(
url = starturl,
starturl = starturl,
netloc = netloc,
priority = common.database.DB_IDLE_PRIORITY,
distance = common.database.DB_DEFAULT_DIST,
)
print("Missing start-url for address: '{}'".format(starturl))
sess.add(new)
try:
sess.commit()
except Exception:
print("Failure inserting start url for address: '{}'".format(starturl))
sess.rollback()
sess.close()
common.database.delete_db_session()
