web_scrap/lib/WPImport.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 WPimport:
    # Constructor
    def __init__(self, basic, wordpress, logger, parser):
        self._basic = basic
        self._wordpress = wordpress
        self._logger = logger
        self._parser = parser
        self._headers_json = {'Content-Type': 'application/json', 'Accept':'application/json'}

        self._request = requests.Session()

        retries = Retry(total=10,
                status_forcelist=[429, 500, 502, 503, 504], backoff_factor=2)

        self._request.mount('http://', HTTPAdapter(max_retries=retries))

    # Public method

    def setUrl(self, wordpress):
        self._wordpress = wordpress

    def fromUrl(self, webpage):
        for i in range(0, len(webpage)):
            try:
                r = self._request.get(webpage[i])
            except Exception as err:
                self._logger.error("Connection error : {0}".format(err))
                exit(1)
            if r.status_code == 200:
                self._logger.info("({0}/{1} : Page en cours d'import : {2}".format(i+1, len(webpage), webpage[i]))
                soup = BeautifulSoup(r.content, self._parser)
                articlebody = soup.find_all("div", class_="articlebody")
                if len(articlebody) > 0:
                    self._addOrUpdatePost(soup)
                else:
                    self._addOrUpdateFeaturedMedia(soup)
            else:
                self._logger.error("Connection error with status code : {0}".format(r.status_code))


    def fromDirectory(self, directory):
        directory = "{0}/archives".format(directory)
        directories = self._getDirectories([], "{0}".format(directory))
        files = self._getFiles(directories)
        self.fromFile(files)
        

    def fromFile(self, files):
        for file in files:
            if os.path.exists(file):
                self._logger.info("Fichier en cours de traitement : {0}".format(file))
                with open(file, 'r') as f:
                    content = f.read()
                    soup = BeautifulSoup(content, self._parser)
                    articlebody = soup.find_all("div", class_="articlebody")
                    if len(articlebody) > 0:
                        self._addOrUpdatePost(soup)
                    else:
                        self._addOrUpdateFeaturedMedia(soup)
    
    # Private method

    ## Get all files

    def _getFiles(self, item):
        files = []
        for i in item:
            for j in os.listdir(i):
                if os.path.isfile("{0}/{1}".format(i, j)):
                    files.append("{0}/{1}".format(i, j))
        return files

    ## Get directories

    def _getDirectories(self, subdirectory, item):
        sub = subdirectory
        for i in os.listdir(item):
            if os.path.isdir("{0}/{1}".format(item, i)):
                sub.append("{0}/{1}".format(item, i))
                subdirectory = self._getDirectories(sub, "{0}/{1}".format(item, i))
        return subdirectory

    ## Add or update featured media

    def _addOrUpdateFeaturedMedia(self, soup):
        item_div = soup.find_all("div", {"data-edittype": "post"})
        for i in item_div:
            h2 = i.find_all("h2")[0].text
            params = {"search":h2, "type":"post"}
            try:
                page = self._request.get("http://{0}/wp-json/wp/v2/search".format(self._wordpress), auth=self._basic, params=params)
            except Exception as err:
                self._logger.error("Connection error : {0}".format(err))
                exit(1)
            if page.status_code == 200:
                result = page.json()
                if len(result) > 0:
                    if h2 == result[0]["title"]:
                        img = i.find_all("img")
                        if len(img) > 0:
                            img_src = img[0].get("src")
                            try:
                                page = self._request.get(img_src)
                            except Exception as err:
                                self._logger.error("Connection error : {0}".format(err))
                                exit(1)
                            if page.status_code == 200:
                                name_img = img_src.replace("_q", "")
                                name_img = name_img.split("/")[len(name_img.split("/"))-1]
                                params = {"search": name_img}
                                try:
                                    page = self._request.get("http://{0}/wp-json/wp/v2/media".format(self._wordpress), auth=self._basic, params=params)
                                except Exception as err:
                                    self._logger.error("Connection error : {0}".format(err))
                                    exit(1)
                                if page.status_code == 200:
                                    res = page.json()
                                    if len(res) > 0:
                                        id_media = res[0]["id"]
                                        data = {"featured_media": id_media}
                                        try:
                                            r = self._request.post("http://{0}/wp-json/wp/v2/posts/{1}".format(self._wordpress, result[0]["id"]), auth=self._basic, headers=self._headers_json, data=json.dumps(data))
                                        except Exception as err:
                                            self._logger.error("Connection error : {0}".format(err))
                                            exit(1)
                                        if r.status_code == 200:
                                            self._logger.info("Ajout media featured : {0}".format(r.json()["title"]["raw"]))
                                        else:
                                            self._logger.error("Connection error with status code : {0}".format(r.status_code))
                                    else:
                                        self._logger.info("Aucun media trouvé pour {0}".format(h2))
                                else:
                                    self._logger.error("Connection error with status code : {0}".format(page.status_code))
                            else:
                                self._logger.error("Connection error with status code : {0}".format(page.status_code))
            else:
                self._logger.error("Connection error with status code : {0}".format(page.status_code))

    ## Association image to post

    def _linkImgPost(self, title, list_img, post_id):
        for i in list_img:
            data = {"post": post_id}
            try:
                r = self._request.post("http://{0}/wp-json/wp/v2/media/{1}".format(self._wordpress, i["id"]), auth=self._basic, data=data)
            except Exception as err:
                self._logger.error("Connection error : {0}".format(err))
                exit(1)
            if r.status_code == 200:
                self._logger.info("Association d'une image à l'article {0}".format(title))
            else:
                self._logger.error("Connection error with status code : {0}".format(r.status_code))
    
    ## Add or update img 

    def _addOrUpdateMedia(self, href_img, page):
        media = {"id":"", "rendered":""}
        split_fileimg = href_img.split("/")
        img_name = split_fileimg[len(split_fileimg)-1]
        params = { "search": img_name}
        try:
            r = self._request.get("http://{0}/wp-json/wp/v2/media".format(self._wordpress), auth=self._basic, params=params)
        except Exception as err:
            self._logger.error("Connection error : {0}".format(err))
            exit(1)
        if r.status_code == 200:
            res = r.json()
            if len(res) > 0:
                params = {"force":1}
                try:
                    r = self._request.delete("http://{0}/wp-json/wp/v2/media/{1}".format(self._wordpress, res[0]["id"]), auth=self._basic, params=params)
                except Exception as err:
                    self._logger.error("Connection error : {0}".format(err))
                    exit(1)
                if r.status_code == 200:
                    self._logger.info("Image removed {0}".format(img_name))
                else:
                    self._logger.error("Image not removed due status code : {0}".format(r.status_code))
            data = page.content
            img_type = "image/png"
            if img_name.split(".")[1] == "jpg" or img_name.split(".")[1] == "jpeg":
                img_type = "image/jpg"
            headers={ 'Content-Type': img_type,'Content-Disposition' : 'attachment; filename={0}'.format(img_name)}
            try:
                r = self._request.post("http://{0}/wp-json/wp/v2/media".format(self._wordpress), auth=self._basic, headers=headers, data=data)
            except Exception as err:
                self._logger.error("Connection error : {0}".format(err))
                exit(1)
            if r.status_code == 201:
                self._logger.info("Image added {0}".format(img_name))
                res = r.json()
                media["id"] = res["id"]
                media["rendered"] = res["guid"]["rendered"]
            else:
                self._logger.error("Image not added due status code : {0}".format(r.status_code))
        else:
            self._logger.error("Connection error with status code : {0}".format(r.status_code))
        return media

    ## Add or update comment

    def _addOrUpdateComment(self, post, comment, title):
        params = {"post": post}
        block = True
        try:
            page = self._request.get("http://{0}/wp-json/wp/v2/comments".format(self._wordpress), auth=self._basic, params=params)
        except Exception as err:
            self._logger.error("Connection error : {0}".format(err))
            exit(1)
        if page.status_code == 200:
            result = page.json()
            for i in comment:
                comment_exist = False
                for j in result:
                    if i["author"] == j["author_name"] and i["date"] == j["date"]:
                        comment_exist = True
                        id_comment = j["id"]
                data = {"post": post, "content": i["content"], "date": i["date"], "author_name": i["author"]}
                if comment_exist is True:
                    try:
                        page = page = self._request.post("http://{0}/wp-json/wp/v2/comments/{1}".format(self._wordpress, id_comment), auth=self._basic, data=data)
                    except Exception as err:
                        self._logger.error("Connection error : {0}".format(err))
                        exit(1)
                    if page.status_code == 200:
                        self._logger.info("Comment updated for {0}".format(title))
                    else:
                        self._logger.error("Comment not updated for {0} due status code : {1}".format(title, page.status_code))
                else:
                    if i["parent_id"] != -1:
                        parent_id = int(i["parent_id"])
                        try:
                            page = self._request.get("http://{0}/wp-json/wp/v2/comments".format(self._wordpress), auth=self._basic, params=params)
                        except Exception as err:
                            self._logger.error("Connection error : {0}".format(err))
                            exit(1)
                        if page.status_code == 200:
                            result = page.json()
                            for j in result:
                                if comment[parent_id]["author"] == j["author_name"] and comment[parent_id]["date"] == j["date"]:
                                    data["parent"]=j["id"]
                        else:
                            self._logger.error("Connection error with status code : {0}".format(page.status_code))
                    self._logger.info(data)
                    try:
                        page = self._request.post("http://{0}/wp-json/wp/v2/comments".format(self._wordpress), auth=self._basic, data=data)
                    except Exception as err:
                        self._logger.error("Connection error : {0}".format(err))
                        exit(1)
                    self._logger.info(page.status_code)
                    if page.status_code == 201:
                        self._logger.info("Comment added for {0}".format(title))
                    else:
                        self._logger.error("Comment not added for {0} due status code : {1}".format(title, page.status_code))
        else:
            self._logger.error("Connection error with status code : {0}".format(page.status_code))
    
    ## Check class name 

    def _hasClassName(self, tag, className):
        for i in tag["class"]:
            if i == className:
                return True
        return False
    
    ## Get class name 

    def _getClassName(self, tag, className):
        for i in tag["class"]:
            if re.match(className, i):
                return i
        return ""
    
    ## Get all comments

    def _getComment(self, comment):
        comment_post = []
        for i in range(0, len(comment)):
            comment_div = comment[i].find("div", class_="comment_item")
            comment_item = comment_div.text.split("\n")
            footer = comment_div.find_all("div", class_="itemfooter")
            comment_author = footer[0].text.split(",")[0].replace("Posté par ", "")
            comment_date = footer[0].find_all("abbr")[0].get("title")
            comment_content = "<p>"
            for j in range(0, len(comment_item)-2):
                if len(comment_item[j]) > 0:
                    comment_content = comment_content + comment_item[j] + "<br />"
            comment_content = comment_content + "</p>"
            parent = -1
            if self._hasClassName(comment[i], "level-1") is False:
                block = False
                className = self._getClassName(comment[i], "level-").split("-")
                level = 1
                if len(className) > 0:
                    level = int(className[1])
                for j in range(i-1, 0, -1):
                    if block is False:
                        levelName = "level-{0}".format(level - 1)
                        if self._hasClassName(comment[j], levelName) is True:
                            parent = j
                            block = True

            comment_post.append({"author": comment_author, "date": comment_date, "content": comment_content, "parent_id":parent})
        return comment_post
    ## Add or Update post 

    def _addOrUpdatePost(self, soup):
        tags = []
        month = {"janvier":"01", "février": "02", "mars": "03", "avril":"04", "mai": "05", "juin": "06", "juillet": "07", "août": "08", "septembre": "09", "octobre": "10", "novembre": "11", "décembre": "12"}
        liste = ["categories", "tags"]
        elements = {}
        element = {}
        listelement = {}

        for i in liste:
            try:
                page = self._request.get("http://{0}/wp-json/wp/v2/{1}".format(self._wordpress,i))
            except Exception as err:
                self._logger.error("Connection error : {0}".format(err))
                exit(1)
            if page.status_code == 200:
                elements[i] = page.json()
                element[i] = []
                listelement[i] = []
            else:
                self._logger.error("Connection error with status code : {0}".format(page.status_code))

        articletitle = soup.find_all("h2", class_="articletitle")
        articlebody = soup.find_all("div", class_="articlebody")
        articledate = soup.find_all("span", class_="articledate")
        articleacreator = soup.find_all("span", class_="articlecreator")
        dateheader = soup.find_all("div", class_="dateheader")
        itemfooter = soup.find_all("div", class_="itemfooter")
        comment = soup.find_all("li", class_="comment")
        img_a = articlebody[0].find_all("a", {"target": "_blank"})
        list_img = []
        for i in img_a:
            new_img = {}
            img = i.find_all("img")
            if len(img) > 0:
                href_a = i.get("href")
                href_img = img[0].get("src")
                new_img["old_src"]=href_img
                new_img["old_href"]=href_a
                try:
                    page_img = self._request.get(href_img)
                except Exception as err:
                    self._logger.error("Connection error : {0}".format(err))
                    exit(1)
                if page_img.status_code == 404:
                    href_img = href_a
                    try:
                        page_img = self._request.get(href_a)
                    except Exception as err:
                        self._logger.error("Connection error : {0}".format(err))
                        exit(1)
                
                if page_img.status_code == 200:
                    media=self._addOrUpdateMedia(href_img, page_img)
                    new_img["id"]=media["id"]
                    new_img["new_src"]=media["rendered"]
                    list_img.append(new_img)
                    if href_img != href_a:
                        media=self._addOrUpdateMedia(href_a, page_img)
                        new_img["id"]=media["id"]
                        new_img["new_src"]=media["rendered"]
                        list_img.append(new_img)
                if page_img.status_code not in [200, 404]:
                    self._logger.error("Connection error with status code : {0}".format(page_img.status_code))
                
        comment_post = self._getComment(comment)

        a = itemfooter[0].find_all("a", {"rel": True})
        for i in a:
            rel = i.get("rel")
            if rel[0] == 'tag':
                href = i.get("href")
                if re.search(r'/tag/', href):
                    element["tags"].append(i.text)
                if re.search(r'/archives/', href):
                    element["categories"].append(i.text)
        for i in liste:
            for j in element[i]:
                element_exist = False
                for k in elements[i]:
                    if k["name"] == j:
                        element_exist = True
                        listelement[i].append(k["id"])
                if element_exist is False:
                    data = {"name": j}
                    self._logger.debug("URL : {0} with data : {1} and with headers : {2}".format("http://{0}/wp-json/wp/v2/{1}".format(self._wordpress, i), data, self._headers_json))
                    try:
                        page = self._request.post("http://{0}/wp-json/wp/v2/{1}".format(self._wordpress, i), auth=self._basic, headers=self._headers_json, data=data)
                    except Exception as err:
                        self._logger.error("Connection error : {0}".format(err))
                        exit(1)
                    if page.status_code == 201:
                        result = page.json()
                        listelement[i].append(result["id"])
                    else:
                        self._logger.error("{0} not added due status code : {1}".format(i, page.status_code))

        title = articletitle[0].text
        author = articleacreator[0].text.lower()        
        body = articlebody[0].find_all("p")
        bodyhtml = "<p>"
        for i in body:
            if len(i.text) == 1:
                bodyhtml = bodyhtml + "<br />"
            else:
                bodyhtml = bodyhtml + str(i).replace("<p>", "").replace("</p>", "").replace("<br>", "<br />") + "<br />"
        bodyhtml = bodyhtml + "</p>"
        for i in list_img:
            o = urlparse(i["new_src"])
            bodyhtml = bodyhtml.replace(i["old_href"], o.path)
            bodyhtml = bodyhtml.replace(i["old_src"], o.path)
        hour = articledate[0].text
        time = dateheader[0].text.split(" ")
        data = {"title":title, "content":bodyhtml, "status":"publish", "date": "{0}-{1}-{2}T{3}:00".format(time[2],month[time[1]],time[0], hour), "tags": listelement["tags"], "categories": listelement["categories"]}
        params = {"search":author}
        try:
            page = self._request.get("http://{0}/wp-json/wp/v2/users".format(self._wordpress), auth=self._basic, params=params)
        except Exception as err:
            self._logger.error("Connection error : {0}".format(err))
            exit(1)
        if page.status_code == 200:
            result = page.json()
            data["author"] = result[0]["id"]
        else:
            self._logger.error("Connection error with status code : {0}".format(page.status_code))

        params = {"search":title}
        try:
            page = self._request.get("http://{0}/wp-json/wp/v2/posts".format(self._wordpress), auth=self._basic, params=params)
        except Exception as err:
            self._logger.error("Connection error : {0}".format(err))
            exit(1)
        page_exist = True
        headers = {'Content-Type': 'application/json', 'Accept':'application/json'}
        if page.status_code == 200:
            result = page.json()
            if len(result) == 0:
                page_exist = False
            else:
                self._logger.info("La page {0} existe deja et mis à jour".format(title))
                post_id = result[0]["id"]
                try:
                    page = self._request.post("http://{0}/wp-json/wp/v2/posts/{1}".format(self._wordpress, post_id), auth=self._basic, headers=headers, data=json.dumps(data))
                except Exception as err:
                    self._logger.error("Connection error : {0}".format(err))
                    exit(1)
                if page.status_code == 200:
                    result = page.json()
                    self._logger.info("Article mis à jour : {0}".format(result["title"]["raw"]))
                    self._addOrUpdateComment(result["id"], comment_post, result["title"]["raw"])
                    self._linkImgPost(result["title"]["raw"], list_img, result["id"])
                else:
                    self._logger.error("Post not updated due status code : {0}".format(page.status_code))
        else:
            self._logger.error("Connection error with status code : {0}".format(page.status_code))
                    
        if page_exist == False:
            try:
                page = self._request.post("http://{0}/wp-json/wp/v2/posts".format(self._wordpress), auth=self._basic, headers=headers, data=json.dumps(data))
            except Exception as err:
                self._logger.error("Connection error : {0}".format(err))
                exit(1)
            if page.status_code == 201:
                result = page.json()
                self._logger.info("Article ajoute : {0}".format(result["title"]["raw"]))
                self._addOrUpdateComment(result["id"], comment_post, result["title"]["raw"])
                self._linkImgPost(result["title"]["raw"], list_img, result["id"])
            else:
                self._logger.error("Post not added due status code : {0}".format(r.status_code))