websucker-pip/mongo/mongocwarler.py

import pymongo
import trafilatura
import trafilatura.feeds
import trafilatura.sitemaps
import trafilatura.spider
import trafilatura.utils
import trafilatura.external
import sys
import courlan
import urllib
from datetime import datetime
import click
import logging as LOGGER
import os
import pprint

LANGUAGE= os.getenv("SUCKER_LANGUAGE","sk")
DOMAIN = os.getenv("SUCKER_DOMAIN","sk")
BATCHSIZE=os.getenv("SUCKER_BATCHSIZE",10)
CONNECTION=os.getenv("SUCKER_CONNECTION","mongodb://root:example@localhost:27017/")
DBNAME=os.getenv("SUCKER_DBNAME","crawler")
MINFILESIZE=300
MAXFILESIZE=10000000
MINTEXTSIZE=200

def put_queue(db,channel,message):
    queuecol = db["queue"]
    queuecol.insert_one({"channel":channel,"message":message,"created_at":datetime.utcnow(),"started_at":None})

def reserve_queue(db,channel,message):
    queuecol = db["queue"]
    r = queuecol.find_one_and_delete({"channel":channel},sort={"created_at":-1})

def delete_queue(db,channel):
    queuecol = db["queue"]
    pass

def calculate_checksums(text):
    """
    @return fingerprints of a paragraphs in text. Paragraphs are separated by a blank line
    """
    checksums = []
    sizes = []
    hval = 0
    hsz = 0
    sz = 0
    for c in text:
        cv = ord(c)
        sz += 1
        if cv > 64:
            hval += (hval << 3) + cv
            zv = hval >> 31
            hval &= 0x7fffffff
            hval += zv
            hsz += 1
        if c == "\n" and hsz > 0:
            if hsz > 100:
                checksums.append(hval)
                sizes.append(sz)
            sz = 0
            hsz = 0
    if hsz > 100:
        checksums.append(hval)
        sizes.append(sz)
    return checksums, sizes

def is_robot_good(link,rules):
    # check robots.txt rules
    if rules is not None and not rules.can_fetch("*", link):
        return False
    return True

def is_link_good(link):
    r = courlan.check_url(link,strict=True,language=LANGUAGE)
    if r is None:
        #print(link)
        return None
    llink,lhostname = r
    #print(llink,lhostname)
    # hostname rules
    if not lhostname.endswith(DOMAIN):
        LOGGER.debug("bad hostname")
        return None
    if courlan.is_not_crawlable(llink):
        LOGGER.debug("not crawlable")
        return None
    return llink

def get_link_doc(link,status="frontlink"):
    r  = courlan.check_url(link)
    assert r is not None
    link,host = r
    domain = courlan.extract_domain(link)
    return {"url":link,"host":host,"domain":domain,"status":status,"created_at":datetime.utcnow()}


def fetch_pages(link_batch):
    htmls  = []
    #print(link_batch)
    #print("zzzzzzzzzz")
    for link in link_batch:
        print("fetching:::::")
        print(link)
        final_link = link
        response = trafilatura.fetch_url(link,decode=False)
        html = None
        if response is not None :
            good = True
            if response.status != 200:
                good = False
                LOGGER.error('not a 200 response: %s for URL %s', response.status, url)
            elif response.data is None or len(response.data) < MINFILESIZE:
                LOGGER.error('too small/incorrect for URL %s', link)
                good = False
            # raise error instead?
            elif len(response.data) > MAXFILESIZE:
                good = False
                LOGGER.error('too large: length %s for URL %s', len(response.data), link)
            if good:
                html = trafilatura.utils.decode_response(response) 
                final_link = response.url
            if html is not None:
                html, final_link = trafilatura.spider.refresh_detection(html, final_link)
                # is there a meta-refresh on the page?
                if final_link is None:  # malformed or malicious content
                    html = None
        htmls.append((final_link,html))
    return htmls

def fetch_robot(base_url):
    rules = urllib.robotparser.RobotFileParser()
    rules.set_url("https://" + base_url + '/robots.txt')
    # exceptions happening here
    try:
        rules.read()
        print("GOT robot")
        print(rules)
        LOGGER.info('got robots')
    except Exception as exc:
        LOGGER.error('cannot read robots.txt: %s', exc)
        rules = None
    return rules


def extract_pages(link_batch,responses):
    out = []
    for original_link,(final_link,html) in zip(link_batch,responses):
        doc = None
        assert original_link is not None
        if html is not None:
            doc = trafilatura.bare_extraction(html,url=final_link,with_metadata=True,include_formatting=True,target_language=LANGUAGE)
            print("html2doc")
            print(text)
            if doc is not None:
                if not "text" in doc or len(doc["text"]) < MINTEXTSIZE:
                    # text too small
                    doc = None

        out.append((original_link,final_link,html,doc))
    return out


def index_pages(db,hostname,extracted_pages):
    linkcol = db["links"]
    htmlcol = db["html"]
    contentcol = db["content"]
    links = []
    for original_link,final_link,html,doc in extracted_pages:
        state = "good"
        link = original_link
        if original_link != final_link:
            linkcol.update_one({"url":original_link},{"$set":{"status":"redirect"}})
            link = final_link
        if html is None:
            state = "html_error"
        elif doc is None:
            state = "content_error"
        if doc is not None:
            text = doc["text"]
            checksums,sizes = calculate_checksums(text)
            doc["text_size"] = len(text)
            doc["paragraph_checksums"] = checksums
            doc["paragraph_sizes"] = sizes
            goodsz = sum(sizes)
            if len(text) < 200 or goodsz/len(text) < 0.4:
                stat = "trash"
        if state == "good":
            htdoc = get_link_doc(link,state)
            htdoc["html"] = html
            htdoc["html_size"] = len(html)
            # can be revisited - upsert
            del htdoc["url"]
            htmlcol.update_one({"url":link},{"$set":htdoc},upsert=True)
            doc.update(get_link_doc(link,"good"))
            # todo extract links
            print(doc)
            del doc["url"]
            contentcol.update_one({"url":link},{"$set":doc},upsert=True)
        linkcol.update_one({"url":original_link},{"$set":{"status":state}})


def extract_links(link_batch,responses,hostname,rules,default_status="frontlink"):
    links = {}
    for original_link,(final_link,html) in zip(link_batch,responses):
        status = default_status
        external_links = courlan.extract_links(html,final_link,external_bool=True,language=LANGUAGE)
        for link in external_links:
            links[link] = "frontlink"
        internal_links = courlan.extract_links(html,final_link,external_bool=False,language=LANGUAGE)
        #print(extracted_links)
        for link in internal_links:
            status = str(default_status)
            if courlan.is_navigation_page(link):
                status = "navigation"
            #print(link,status)
            links[link] = status
    outlinks = []
    badrobot = 0
    badlink = 0
    for link,status in links.items():
        if not is_robot_good(link,rules):
            badrobot += 1
            continue
        link = is_link_good(link)
        if link is None:
            badlink += 1
            continue
        outlinks.append((link,status))
    print(f"{len(links)} total links, {badrobot} badrobot {badlink} badlinks")
    return outlinks

def index_links(db,extracted_links):
    linkcol=db["links"]
    for link,status in extracted_links:
        doc = get_link_doc(link,status)
        try:
            linkcol.insert_one(doc)
        except pymongo.errors.DuplicateKeyError as ex:
            pass

def get_links(db,hostname,status,batch_size=BATCHSIZE):
    linkcol = db["links"]
    res  = linkcol.find({"status":status,"host":hostname},{"url":1},limit=batch_size)
    links = []
    for doc in res:
        #print(">>>>>" + status)
        #print(doc)
        links.append(doc["url"])
    return links



def process_links(db,hostname,status,links=[],rules=None,batch_size=BATCHSIZE):
    #print(links)
    responses = fetch_pages(links)
    #print(responses)
    extracted_pages = extract_pages(links,responses)
    #print(extracted_pages)
    extracted_links = extract_links(links,responses,hostname,rules,status)
    #print(extracted_links)
    index_links(db,extracted_links)
    index_pages(db,hostname,extracted_pages)


def link_summary(db,hostname):
    linkcol = db["links"]
    #res = linkcol.distinct("hostname",{"hostname":hostname})
    
    # count links
    res = linkcol.aggregate([
        {"$match":{"host":hostname}},
        {"$group":{"_id":"$status","count":{"$sum":1}}},
    ])
    for item in res:
        print(item)
    contentcol = db["content"]
    res = contentcol.aggregate([
        {"$match":{"hostname":hostname}},
        {"$group":{"_id":None,"text_size_sum":{"$sum":"text_size"}}},
    ])
    for item in res:
        print(item)


@click.group()
def cli():
    pass

@cli.command()
def createdb():
    myclient = pymongo.MongoClient(CONNECTION)
    db=myclient[DBNAME]
    linkcol = db["links"]
    linkcol.create_index("url",unique=True)
    linkcol.create_index("host")
    contentcol = db["content"]
    contentcol.create_index("url",unique=True)
    #contentcol.create_index({"paragraph_checksums":1})
    #contentcol.create_index({"hostname":1})
    htmlcol = db["html"]
    htmlcol.create_index("url",unique=True)

@cli.command()
@click.argument("link")
def parseurl(link):
    html = trafilatura.fetch_url(link,decode=True)
    doc = trafilatura.bare_extraction(html)
    import pprint
    pprint.pprint(doc)

@cli.command()
@click.argument("link")
def externaldomains(link):
    html = trafilatura.fetch_url(link,decode=True)
    external_links = courlan.extract_links(html,link,external_bool=True,language=LANGUAGE)
    domains = set()
    for l in external_links:
        r = courlan.check_url(l)
        if r is None:
            pass
        link,domain = r
        domains.add(domain)
    for d in domains:
        print(d)

@cli.command()
@click.argument("start_link")
def parseurl(start_link):
    link,hostname = courlan.check_url(start_link)
    links = [link]
    responses = fetch_pages(links)
    #pprint.pprint(responses)
    extracted_pages = extract_pages(links,responses)
    for ol,bl,html,doc in extracted_pages:
        pprint.pprint(doc)
    extracted_links = extract_links(links,responses,hostname,None,"backlink")
    pprint.pprint(extracted_links)


@cli.command()
@click.argument("start_link")
def visit(start_link):
    myclient = pymongo.MongoClient(CONNECTION)
    db=myclient[DBNAME]
    start_link,hostname = courlan.check_url(start_link)
    rules = fetch_robot(hostname)
    print(rules)
    batch_size = BATCHSIZE
    navigation_links = get_links(db,hostname,"navigation",batch_size)
    if start_link is not None:
        navigation_links.append(start_link)
    print(f"Navigation links {len(navigation_links)}")
    process_links(db,hostname,"frontlink",navigation_links,rules)
    links = get_links(db,hostname,"frontlink",batch_size)
    bl = len(links) - batch_size
    print(f"Got {len(links)} frontlinks")
    if bl > 0:
        print("Getting  backlinks")
        front_links = get_links(db,hostname,"backlink",bl)
        links += front_links
    print("Processing backlinks")
    process_links(db,hostname,"backlink",links,rules=rules)
    link_summary(db,hostname)

if __name__ == "__main__":
    cli()