web_scrap/lib/WPMenu.py

#!/usr/bin/python3

from bs4 import BeautifulSoup
from urllib.parse import urlparse
import requests, os, logging, re, json
from requests.adapters import HTTPAdapter
from requests.packages.urllib3.util.retry import Retry

class WPMenu:
    # Constructor
    def __init__(self, name="Thread-0", basic=None, canalblog="", wordpress="", logger=None, parser="html.parser", ssl_canalblog=True, ssl_wordpress=True):
        self._name = name
        self._basic = basic
        self._canalblog = canalblog
        self._wordpress = wordpress
        self._logger = logger
        self._parser = parser
        self._headers_json = {'Content-Type': 'application/json; charset=utf-8', 'Accept':'application/json'}
        self._protocol_wordpress = "https"
        self._protocol_canalblog = "https"
        self._directory = "backup"

        if ssl_wordpress is False:
            self._protocol_wordpress = "http"
        if ssl_canalblog is False:
            self._protocol_canalblog = "http"
        self._request_canalblog = requests.Session()
        self._request_wordpress = requests.Session()

        retries = Retry(connect=10, read=10, redirect=5,
                status_forcelist=[429, 500, 502, 503, 504], backoff_factor=2)

        self._request_canalblog.mount('{0}://'.format(self._protocol_canalblog), HTTPAdapter(max_retries=retries))
        self._request_wordpress.mount('{0}://'.format(self._protocol_wordpress), HTTPAdapter(max_retries=retries))

    
    # Destructor 
    def __del__(self):
        print("{0} : Import finished for {1}".format(self._name, self._wordpress))

    # Public method

    ## From file 

    def fromFile(self, files):
        if os.path.exists(files):
            with open(files, 'r') as f:
                self._logger.info("{0} : File is being processed : {1}".format(self._name, files))
                content = f.read()
                self._menu(content)
        else:
            self._logger.error("{0} : File isn't exist : {1}".format(self._name, files))



    ## Get from URL

    def fromUrl(self, canalblog):
        try:
            o = urlparse(canalblog)
            o = o._replace(scheme=self._protocol_canalblog)
            i = o.geturl().replace(":///", "://")
            page = self._request_canalblog.get(i)
            if page.status_code == 200:
                self._logger.info("{0} : Page web is being processed : {1}".format(self._name, i))
                self._menu(page.content)
            else:
                self._logger.error("{0} : index didn't get due status code : {1}".format(self._name, page.status_code))
                self._logger.debug("{0} : {1}".format(self._name, page.content))
        except ConnectionError as err:
            self._logger.error("{0} : Connection error for get url {1} : {2}".format(self._name, canalblog, err))
            exit(1)        
        except Exception as err:
            self._logger.error("{0} : Exception error for get url {1} : {2}".format(self._name, canalblog, err))

    
    def _menu(self, content):

        soup = BeautifulSoup(content, self._parser)
        ul = soup.find_all("ul", id="listsmooth")
        menu = list()
        parents = list()
        for anchor in ul[0].find_all("li"):
            parent = anchor.find("a").get_text()
            href = anchor.find("a").get("href")
            if href == "#":
                li = anchor.find_all("li")
                for child in li:
                    a = child.find("a")
                    self._logger.info("{0} {1} : {2}".format(self._name, anchor.find("a").get_text(), a.get_text()))
                    for i in ["categories", "tags"]:
                        try:
                            params = {"search":a.get_text(), "per_page":"100"}
                            page = self._request_wordpress.get("{2}://{0}/wp-json/wp/v2/{1}".format(self._wordpress, i, self._protocol_wordpress), auth=self._basic, params=params)
                            if page.status_code == 200:
                                result = page.json()
                                if len(result) > 0:
                                    menu.append({"id":result[0]["id"], "type":i, "title": a.get_text(), "parent":parent})
                            else:
                                self._logger.error("{0} : {2} didn't get due status code : {1}".format(self._name, page.status_code, i))
                                self._logger.debug("{0} : {1}".format(self._name, page.content))
                        except ConnectionError as err:
                            self._logger.error("{0} : Connection error for get url {1} : {2}".format(self._name, "{2}://{0}/wp-json/wp/v2/{1}".format(self._wordpress, i, self._protocol_wordpress), err))
                            exit(1)        
                        except Exception as err:
                            self._logger.error("{0} : Exception error for get url {1} : {2}".format(self._name, "{2}://{0}/wp-json/wp/v2/{1}".format(self._wordpress, i, self._protocol_wordpress), err))
            else:
                self._logger.info("{0} : Parent title : {1}".format(self._name, parent))
                parents.append(parent)
        for i in range(0, len(menu)-1):
            self._logger.info("{0} : Menu : {1} {2}".format(self._name, menu[i], len(parents)))
            for j in range(0, len(parents)-1):
                self._logger.info("{0} : Comparaison : {1} {2}".format(self._name, menu[i]["title"], parents[j]))

                if menu[i]["title"] == parents[j]:
                    self._logger.info("{0} del : {1}".format(self._name, parents[j]))
                    del parents[j]
        
        for i in parents:
            self._logger.info("{0} : Parent : {1}".format(self._name, i))