Compare commits

..

2 Commits

Author SHA1 Message Date
1e7ac17d90 wip schema 2023-02-23 16:08:14 +01:00
56bf59e5e9 password and name 2023-02-23 16:08:02 +01:00
3 changed files with 105 additions and 4 deletions

View File

@ -12,7 +12,7 @@ import os
def create_database_from_context(ctx):
return Data(ctx.obj["cassandra_keyspace"],ctx.obj["cassandra_host"],ctx.obj["cassandra_port"])
return Data(ctx.obj["cassandra_keyspace"],ctx.obj["cassandra_host"],ctx.obj["cassandra_port"],ctx.obj["cassandra_username"],ctx.obj["cassandra_password"])
def create_queue_from_context(ctx):
return greenstalk.Client((ctx.obj["beanstalkd_host"],ctx.obj["beanstalkd_port"]),use=ctx.obj["beanstalkd_tube"],watch=ctx.obj["beanstalkd_tube"],encoding="utf8")
@ -23,13 +23,15 @@ def create_queue_from_context(ctx):
@click.option("--cassandra-keyspace",metavar="CASSANDRA_KEYSPACE",help="cassandra keyspace (if defined, value read from CASSANDRA_KEYSPACE env variable)",envvar="CASSANDRA_KEYSPACE",default="websucker",show_default=True)
@click.option("--cassandra-host",metavar="CASSANDRA_HOST",help="cassandra host (if defined, value read from CASSANDRA_HOST env variable)",envvar="CASSANDRA_HOST",default="127.0.0.1",show_default=True)
@click.option("--cassandra-port",metavar="CASSANDRA_PORT",help="cassandra port (if defined, value read from CASSANDRA_PORT env variable)",envvar="CASSANDRA_PORT",default=9042,show_default=True)
@click.option("--cassandra-username",metavar="CASSANDRA_USERNAME",help="cassandra username (if defined, value read from CASSANDRA_USERNAME env variable)",envvar="CASSANDRA_USERNAME")
@click.option("--cassandra-password",metavar="CASSANDRA_PASSWORD",help="cassandra password (if defined, value read from CASSANDRA_PASSWORD env variable)",envvar="CASSANDRA_PASSWORD")
@click.option("--beanstalkd-tube",metavar="BEANSTALKD_TUBE",help="beanstalkd keyspace (if defined, value read from BEANSTALKD_TUBE env variable)",envvar="BEANSTALKD_TUBE",default="websucker",show_default=True)
@click.option("--beanstalkd-host",metavar="BEANSTALKD_HOST",help="beanstalkd host (if defined, value read from beanstalkd_HOST env variable)",envvar="BEANSTALKD_HOST",default="127.0.0.1",show_default=True)
@click.option("--beanstalkd-port",metavar="BEANSTALKD_PORT",help="beanstalkd port (if defined, value read from BEANSTALKD_PORT env variable)",envvar="BEANSTALKD_PORT",default=11300,show_default=True)
@click.option("--parser",metavar="file_name",help="zzz")
@click.option("--visit",is_flag=True)
@click.option("--queue",is_flag=True)
def cli(ctx,cassandra_keyspace,cassandra_host,cassandra_port,beanstalkd_tube,beanstalkd_host,beanstalkd_port,parser,visit,queue):
def cli(ctx,cassandra_keyspace,cassandra_host,cassandra_port,cassandra_username,cassandra_password,beanstalkd_tube,beanstalkd_host,beanstalkd_port,parser,visit,queue):
ctx.ensure_object(dict)
p = BaseParser()
if parser is not None:
@ -44,6 +46,8 @@ def cli(ctx,cassandra_keyspace,cassandra_host,cassandra_port,beanstalkd_tube,bea
ctx.obj["parser"] = p
ctx.obj["cassandra_host"] = cassandra_host
ctx.obj["cassandra_port"] = cassandra_port
ctx.obj["cassandra_username"] = cassandra_username
ctx.obj["cassandra_password"] = cassandra_password
ctx.obj["cassandra_keyspace"] = cassandra_keyspace
ctx.obj["beanstalkd_host"] = beanstalkd_host
ctx.obj["beanstalkd_port"] = beanstalkd_port

View File

@ -1,5 +1,6 @@
import cassandra
import cassandra.cluster
from cassandra.auth import PlainTextAuthProvider
import random
import os
import pkg_resources
@ -20,12 +21,15 @@ class Data:
"""
Database of text documents
"""
def __init__(self,keyspace="websucker",cassandra_host="127.0.0.1",cassandra_port=9042):
def __init__(self,keyspace="websucker",cassandra_host="127.0.0.1",cassandra_port=9042,username=None,password=None):
print("Database {}@{}:{}".format(keyspace,cassandra_host, cassandra_port))
auth_provider = None
if username is not None and password is not None:
auth_provider = PlainTextAuthProvider(username=username, password=password)
# execution profile
ep = cassandra.cluster.ExecutionProfile(request_timeout=240.0)
profiles = {cassandra.cluster.EXEC_PROFILE_DEFAULT:ep}
self.cluster = cassandra.cluster.Cluster([cassandra_host],port=cassandra_port,execution_profiles=profiles)
self.cluster = cassandra.cluster.Cluster([cassandra_host],port=cassandra_port,execution_profiles=profiles,auth_provider=auth_provider)
self.session = self.cluster.connect(keyspace)
self.check_document_select_query = self.session.prepare("SELECT count(url_hash) FROM paragraph_checksums WHERE checksum=?" )

93
websucker/schema.py Normal file
View File

@ -0,0 +1,93 @@
from cassandra.cqlengine import columns
from cassandra.cqlengine.models import Model
class Links(Model):
domain_name = column.Text(primary_key=True)
url_path = column.Text(primary_key=True)
url_query = column.Text(primary_key=True)
url_schema = column.Text()
redirect_target = column.Text()
link_status = column.Text()
link_originality = column.Float()
body_size = column.Integer()
update_time = column.DateTime()
class DailyLinks(Model):
day = column.Integer(primary_key=True)
domain_name = column.Text(primary_key=True)
url_path = column.Text(primary_key=True)
url_query = column.Text(primary_key=True)
url_schema = column.Text()
redirect_target = column.Text()
link_status = column.Text()
link_originality = column.Float()
body_size = column.Integer()
update_time = column.DateTime()
CREATE TABLE domain_quality (
domain_name TEXT,
day DATE,
seen_count INT,
good_size INT,
good_count INT,
good_probability FLOAT,
good_originality FLOAT,
average_good_characters FLOAT,
content_size INT,
content_count INT,
content_probability FLOAT,
content_originality FLOAT,
average_content_characters FLOAT,
fetched_count INT,
average_fetched_good_characters FLOAT,
gain_ratio FLOAT,
update_time TIMESTAMP STATIC ,
PRIMARY KEY(domain_name,day)
) WITH CLUSTERING ORDER BY (day DESC);
CREATE TABLE content (
domain_name TEXT,
target_link TEXT,
agent_version TEXT,
title TEXT,
links SET<TEXT>,
authors SET<TEXT>,
tags SET<TEXT>,
description TEXT,
section TEXT,
article_published_time TEXT,
text_date TEXT,
body TEXT,
body_size INT,
update_time TIMESTAMP,
PRIMARY KEY(domain_name,target_link),
);
CREATE TABLE paragraph_checksums (
checksum BIGINT,
url_hash BIGINT,
PRIMARY KEY(checksum,url_hash),
);
CREATE TABLE html (
day DATE,
domain_name TEXT,
source_link TEXT,
target_link TEXT,
redirect_links LIST<TEXT>,
status INT,
content TEXT,
headers TEXT,
agent_version TEXT,
update_time TIMESTAMP,
PRIMARY KEY(day,domain_name,source_link)
);
CREATE TABLE domain_connections (
domain_name TEXT,
linked_domain TEXT,
PRIMARY KEY (domain_name,linked_domain)
);