Merge pull request 'https' (#13) from https into master

Reviewed-on: #13

backup_canalblog.sh

@@ -3,8 +3,8 @@
 TAR=/usr/bin/tar
 PYTHON=/usr/bin/python3
 GZIP=/usr/bin/gzip
-SCRIPTDIR=/home/valentin/script
+SCRIPTDIR=/home/valentin/script/webscrap
-WEBSCRAP=${SCRIPTDIR}/web_scrap.py
+WEBSCRAP=${SCRIPTDIR}/import_export_canalblog.py
 URL=www.clarissariviere.com
 DATE=$(date +%Y%m%d)
 DIRECTORY=/home/valentin/backup
@@ -24,7 +24,7 @@ else
 fi
 subject="${subject} ${URL} ${DATE}"
 echo > ${BACKUPDIR}/${LOGFILE}
-${PYTHON} ${WEBSCRAP} --url ${URL} --dir ${DIRECTORY}  --quiet --logfile ${BACKUPDIR}/${LOGFILE}
+${PYTHON} ${WEBSCRAP} --quiet --logfile ${BACKUPDIR}/${LOGFILE} --parallel 20 export --url ${URL} --directory ${DIRECTORY}  
 if [ ${?} -ne 0 ]; then 
 	subject="${subject} echoue : recuperation page"
 	echo ${subject} | mail -s "${subject}" -A ${BACKUPDIR}/${LOGFILE} ${SENDER}

import_export_canalblog.py

@@ -8,11 +8,33 @@ from concurrent.futures import as_completed, wait
 import argparse, logging, threading
 from lib.WPImport import WPimport
 from lib.WPExport import WPExport
+from lib.WPRemove import WPRemove
+
+def remove(args, basic, logger, ssl_wordpress):
+    removeWp = WPRemove(basic=basic, wordpress="", logger=logger, ssl_wordpress=ssl_wordpress)
+    if args.remove == True:
+        for i in args.wordpress.split(","):
+            removeWp.setUrl(i)
+            removeWp.cleanPosts()
+            removeWp.cleanTags()
+            removeWp.cleanCategories()
+            removeWp.cleanMedia()
+    else:
+        for i in args.wordpress.split(","):
+            removeWp.setUrl(i)
+            if args.posts == True:
+                removeWp.cleanPosts()
+            if args.categories == True:
+                removeWp.cleanCategories()
+            if args.tags == True:
+                removeWp.cleanTags()
+            if args.media == True:
+                removeWp.cleanMedia()
+    del removeWp
 
 
-
+def download(name_thread, max_thread, url, logger, parser, directory, html, img, ssl_canalblog):
-def download(name_thread, max_thread, url, logger, parser, directory, html, img):
+    exportWp = WPExport(name="Thread-{0}".format(int(name_thread) + 1), url=url, logger=logger, parser=parser, directory=directory, ssl_canalblog=ssl_canalblog)    
-    exportWp = WPExport(name="Thread-{0}".format(int(name_thread) + 1), url=url, logger=logger, parser=parser, directory=directory)    
     webpage = exportWp.getUrlPage(name_thread, max_thread)
     for i in ["article", "page"]:
         for j in ["publications", "principal"]:
@@ -25,25 +47,27 @@ def download(name_thread, max_thread, url, logger, parser, directory, html, img)
 
 
 
-def importUrl(name_thread, max_thread, canalblog, logger, parser, wordpress, basic, serial):
+def importUrl(name_thread, max_thread, canalblog, logger, parser, wordpress, basic, serial, ssl_wordpress, ssl_canalblog):
     canalblog = canalblog.split(",")
     wordpress = wordpress.split(",")
     name = "Thread-{0}".format(int(name_thread) + 1)
-
+    protocol = "https"
+    if ssl_canalblog is False:
+        protocol = "http"
     if serial is False:
         for canal in canalblog:
             try:
                 o = urlparse(canal)
-                o = o._replace(scheme="https")
+                o = o._replace(scheme=protocol)
                 url = o.geturl().replace(":///", "://")
             except Exception as err:
                 logger.error("{0} : parsing error : {1}".format(name, err))
                 exit(1)
-            exportWp = WPExport(name="Thread-{0}".format(int(name_thread) + 1), url=url, logger=logger, parser=parser)
+            exportWp = WPExport(name="Thread-{0}".format(int(name_thread) + 1), url=url, logger=logger, parser=parser, ssl_canalblog=ssl_canalblog)
             webpage = exportWp.getUrlPage(name_thread, max_thread)
             del exportWp
             for j in wordpress:
-                importWp = WPimport(name=name, basic=basic, wordpress=j, logger=logger, parser=parser)
+                importWp = WPimport(name=name, basic=basic, wordpress=j, logger=logger, parser=parser, ssl_wordpress=ssl_wordpress)
                 for k in ["article", "page"]:
                     for l in ["publications", "principal"]:
                         importWp.fromUrl(webpage[l][k])
@@ -56,15 +80,15 @@ def importUrl(name_thread, max_thread, canalblog, logger, parser, wordpress, bas
         for i in range(0, len(canalblog)-1):
             try:
                 o = urlparse(canalblog[i])
-                o = o._replace(scheme="https")
+                o = o._replace(scheme=protocol)
                 url = o.geturl().replace(":///", "://")
             except Exception as err:
                 logger.error("parsing error : {0}".format(err))
                 exit(1)
-            exportWp = WPExport(name=name, url=url, logger=logger, parser=parser)
+            exportWp = WPExport(name=name, url=url, logger=logger, parser=parser, ssl_canalblog=ssl_canalblog)
             webpage = exportWp.getUrlPage(name_thread, max_thread)
             del exportWp
-            importWp = WPimport(name=name, basic=basic, wordpress=wordpress[i], logger=logger, parser=parser)
+            importWp = WPimport(name=name, basic=basic, wordpress=wordpress[i], logger=logger, parser=parser, ssl_wordpress=ssl_wordpress)
 
             for k in ["article", "page"]:
                 for l in ["publications", "principal"]:
@@ -73,13 +97,13 @@ def importUrl(name_thread, max_thread, canalblog, logger, parser, wordpress, bas
             del importWp
     
 
-def importDirectory(name_thread, max_thread, directory, logger, parser, wordpress, basic, serial):
+def importDirectory(name_thread, max_thread, directory, logger, parser, wordpress, basic, serial, ssl_wordpress):
     name = "Thread-{0}".format(int(name_thread) + 1)
     directory = directory.split(",")
     wordpress = wordpress.split(",")
     if serial is False:
         for i in wordpress:
-            importWp = WPimport(name=name, basic=basic, wordpress=i, logger=logger, parser=parser)
+            importWp = WPimport(name=name, basic=basic, wordpress=i, logger=logger, parser=parser, ssl_wordpress=ssl_wordpress)
             for j in directory:
                 importWp.fromDirectory(j, name_thread, max_thread)
                 del importWp
@@ -89,7 +113,7 @@ def importDirectory(name_thread, max_thread, directory, logger, parser, wordpres
             logger.error("{0} : Error : Number directory is differant than wordpress".format(name))
             exit(1)
         for i in range(0, len(wordpress)-1):
-            importWp = WPimport(name=name, basic=basic, wordpress=wordpress[i], logger=logger, parser=parser)
+            importWp = WPimport(name=name, basic=basic, wordpress=wordpress[i], logger=logger, parser=parser, ssl_wordpress=ssl_wordpress)
             importWp.fromDirectory(directory[i])
             del importWp
 
@@ -102,6 +126,7 @@ if __name__ == '__main__':
     parser.add_argument("--quiet", help="No console output", action="store_true")
     parser.add_argument("--parser", help="Parser content", default="html.parser")
     parser.add_argument("--parallel", help="Define number thread (default : 1)", default=1)
+    parser.add_argument("--no-ssl", help="No ssl for canalblog and/or wordpress (example wordpress,canalblog)", dest="ssl", default="")
 
     subparsers = parser.add_subparsers(dest="command")
 
@@ -113,7 +138,22 @@ if __name__ == '__main__':
     import_parser.add_argument("--canalblog", help="URL Canalblog", default="")
     import_parser.add_argument("--wordpress", help="URL Wordpress", required=True)
     import_parser.add_argument("--serial", help="Serial execution", action="store_true")
-    import_parser.add_argument("--remove", help="Remove all articles", action="store_true")
+    import_parser.add_argument("--remove-all", dest="remove", help="Remove all", action="store_true")
+    import_parser.add_argument("--remove-posts", help="Remove all posts", dest="posts", action="store_true")
+    import_parser.add_argument("--remove-categories", help="Remove all categories", dest="categories", action="store_true")
+    import_parser.add_argument("--remove-tags", help="Remove all tags", dest="tags", action="store_true")
+    import_parser.add_argument("--remove-media", help="Remove all media", dest="media", action="store_true")
+
+
+    remove_parser = subparsers.add_parser("remove")
+    remove_parser.add_argument("--user", help="wordpress user", required=True)
+    remove_parser.add_argument("--password", help="password wordpress's user", default="")
+    remove_parser.add_argument("--wordpress", help="URL Wordpress", required=True)
+    remove_parser.add_argument("--all", dest="remove", help="Remove all (posts, media, tags, categories)", action="store_true")
+    remove_parser.add_argument("--posts", help="Remove all posts", action="store_true")
+    remove_parser.add_argument("--categories", help="Remove all categories", action="store_true")
+    remove_parser.add_argument("--tags", help="Remove all tags", action="store_true")
+    remove_parser.add_argument("--media", help="Remove all media", action="store_true")
 
 
 
@@ -134,6 +174,14 @@ if __name__ == '__main__':
 
     logger = logging.getLogger('import export canalblog')
     formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
+    ssl_canalblog = True
+    ssl_wordpress = True
+
+    for i in args.ssl.split(","):
+        if i == "canalblog":
+            ssl_canalblog = False
+        if i == "wordpress":
+            ssl_wordpress = False
 
     if args.quiet is False:
         ch = logging.StreamHandler()
@@ -156,7 +204,7 @@ if __name__ == '__main__':
         fileHandler.setFormatter(formatter)
         logger.addHandler(fileHandler)
 
-    if args.command == "import":
+    if args.command == "import" or args.command == "remove":
         password = args.password
         if len(args.password) == 0:
             password = getpass()
@@ -165,34 +213,29 @@ if __name__ == '__main__':
                 exit(1)
 
         basic = HTTPBasicAuth(args.user, password)
+    if args.command == "import":
         wordpress = args.wordpress.split(",")
-        importWp = WPimport(basic=basic, wordpress="", logger=logger, parser=args.parser)
+        importWp = WPimport(basic=basic, wordpress="", logger=logger, parser=args.parser, ssl_wordpress=ssl_wordpress)
         if len(args.file) > 0:
             for i in wordpress:
                 importWp.setUrl(i)
                 importWp.fromFile(files=args.file.split(","))
         if len(args.directory) > 0:
-            if args.remove:
+            remove(args, basic, logger, ssl_wordpress)
-                for i in args.wordpress.split(","):
-                    importWp.setUrl(i)
-                    importWp.removeAll()
             try:
                 with futures.ThreadPoolExecutor(max_workers=int(args.parallel)) as ex:
                     wait_for = [
-                    ex.submit(importDirectory, i, int(args.parallel), args.directory, logger, args.parser, args.wordpress, basic, args.serial)
+                    ex.submit(importDirectory, i, int(args.parallel), args.directory, logger, args.parser, args.wordpress, basic, args.serial, ssl_wordpress)
                     for i in range(0, int(args.parallel))
                     ]
             except Exception as err:
                 logger.error("Threading error : {0}".format(err))
         if len(args.canalblog) > 0:
-            if args.remove:
+            remove(args, basic, logger, ssl_wordpress)
-                for i in args.wordpress.split(","):
-                    importWp.setUrl(i)
-                    importWp.removeAll()
             try:
                 with futures.ThreadPoolExecutor(max_workers=int(args.parallel)) as ex:
                     wait_for = [
-                    ex.submit(importUrl, i, int(args.parallel), args.canalblog, logger, args.parser, args.wordpress, basic, args.serial)
+                    ex.submit(importUrl, i, int(args.parallel), args.canalblog, logger, args.parser, args.wordpress, basic, args.serial, ssl_wordpress, ssl_canalblog)
                     for i in range(0, int(args.parallel))
                     ]
             except Exception as err:
@@ -202,11 +245,14 @@ if __name__ == '__main__':
 
     if args.command == "export":
         canalblog = args.url.split(",")
-        exportWp = WPExport(logger=logger, parser=args.parser, directory=args.directory)
+        protocol = "https"
+        if ssl_canalblog is False:
+            protocol = "http"
+        exportWp = WPExport(logger=logger, parser=args.parser, directory=args.directory, ssl_canalblog=ssl_canalblog)
         for canal in canalblog:
             try:
                 o = urlparse(canal)
-                o = o._replace(scheme="https")
+                o = o._replace(scheme=protocol)
                 url = o.geturl().replace(":///", "://")
             except Exception as err:
                 logger.error("parsing error : {0}".format(err))
@@ -225,11 +271,14 @@ if __name__ == '__main__':
                 try:
                     with futures.ThreadPoolExecutor(max_workers=int(args.parallel)) as ex:
                         wait_for = [
-                        ex.submit(download, i, int(args.parallel), url, logger, args.parser, args.directory, args.html, args.img)
+                        ex.submit(download, i, int(args.parallel), url, logger, args.parser, args.directory, args.html, args.img, ssl_canalblog)
                         for i in range(0, int(args.parallel))
                         ]
                 except Exception as err:
                     logger.error("Threading error : {0}".format(err))
-
+        exit(0)
-                
+
+
+    if args.command == "remove":
+        remove(args, basic, logger, ssl_wordpress) 
         exit(0)

lib/WPExport.py

@@ -6,20 +6,22 @@ from requests.adapters import HTTPAdapter
 from requests.packages.urllib3.util.retry import Retry
 
 class WPExport:
-    def __init__(self, name = "Thread-0", url = "", logger = None, parser = "html.parser", directory = "backup"):
+    def __init__(self, name = "Thread-0", url = "", logger = None, parser = "html.parser", directory = "backup", ssl_canalblog=True):
         self._url = url
         self._logger = logger
         self._parser = parser
         self._dir = directory
         self._name = name
-
+        self._protocol = "https"
+        if ssl_canalblog is False:
+            self._protocol = "http"
         
         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))
+        self._request.mount('{0}://'.format(self._protocol), HTTPAdapter(max_retries=retries))
 
 
     # Destructor 
@@ -89,7 +91,7 @@ class WPExport:
         for i in page_url:
             section = "publications"
             o = urlparse(i)
-            o = o._replace(scheme="https")
+            o = o._replace(scheme=self._protocol)
             i = o.geturl().replace(":///", "://")
             if i == "{0}/".format(self._url):
                 section = "principal"
@@ -103,16 +105,20 @@ class WPExport:
                 if i not in webpage[section]["page"]:
                     webpage[section]["page"].append(i)
                 soup = BeautifulSoup(page.text, self._parser)
-                class_div = pagingfirstline = soup.find_all("div", class_="pagingfirstline")
+                class_div = soup.find_all("div", class_="pagingfirstline")
                 if len(class_div) > 0:
                     pagingfirstline = class_div[0].find_all("a")
                     if len(pagingfirstline) > 1:
                         lastpage = pagingfirstline[len(pagingfirstline)-1].get("href", "/")
+                        self._logger.debug("{0} : Last page {1}".format(self._name, lastpage))
+
                         element_lastpage = lastpage.split("/")[len(lastpage.split("/"))-1]
                         number_page = element_lastpage.split("-")[0].split("p")[1]
                         number_lastpage = int(number_page) / 10
 
                         setPageDivided = int(number_lastpage) / max_thread
+                        if setPageDivided > int(setPageDivided):
+                            setPageDivided = setPageDivided + 1
                         setPagePart = setPageDivided * (index_thread + 1)
                         firstPagePart = (setPagePart - setPageDivided)
 
@@ -120,7 +126,7 @@ class WPExport:
                         self._logger.debug("{0} : First range : {1}".format(self._name, int(firstPagePart)))
                         self._logger.debug("{0} : Last range : {1}".format(self._name, int(setPagePart)))
 
-                        for j in range(int(firstPagePart),int(setPagePart)):
+                        for j in range(int(firstPagePart),int(setPagePart)+1):
                             paging = j * 10
                             categorie = urlparse(i).path.split("/")
                             url_paging = "{0}/archives/p{1}-10.html".format(self._url, paging)
@@ -133,7 +139,9 @@ class WPExport:
                             if page.status_code == 200:
                                 soup = BeautifulSoup(page.text, self._parser)
                                 h2 = soup.find_all("h2")
+                                self._logger.debug("{0} : {1} H2 : {2}".format(self._name, url_paging, h2))
                                 for title in h2:
+                                    self._logger.debug("{0} : {1} a : {2}".format(self._name, url_paging, title.find_all("a")))
                                     href = title.find_all("a")[0].get("href", "/")
                                     if href not in webpage[section]["article"]:
                                         try:
@@ -262,9 +270,11 @@ class WPExport:
             self._mkdirPath("{0}/{1}/{2}".format(backup_dir, o.netloc, dir_page_web))
             try:
                 r = self._request.get(webpage[i])
-            except Exception as err:
+            except ConnectionError as err:
-                self._logger.error("Connection error : {0}".format(err))
+                self._logger.error("{0} : Connection error : {1}".format(self._name, err))
                 exit(1)
+            except Exception as err:
+                self._logger.error("{0} Exception error : {1}".format(self._name, err))
             if r.status_code == 200:
                 fileDownload = "{0}/{1}/index.html".format(backup_dir, o.netloc)
                 if len(dir_page_web) > 0 and len(filePageWeb) > 0:

lib/WPImport.py

@@ -8,20 +8,22 @@ from requests.packages.urllib3.util.retry import Retry
 
 class WPimport:
     # Constructor
-    def __init__(self, name="Thread-0", basic=None, wordpress="", logger=None, parser="html.parser"):
+    def __init__(self, name="Thread-0", basic=None, wordpress="", logger=None, parser="html.parser", ssl_wordpress=True):
         self._name = name
         self._basic = basic
         self._wordpress = wordpress
         self._logger = logger
         self._parser = parser
-        self._headers_json = {'Content-Type': 'application/json', 'Accept':'application/json'}
+        self._headers_json = {'Content-Type': 'application/json; charset=utf-8', 'Accept':'application/json'}
-
+        self._protocol = "https"
+        if ssl_wordpress is False:
+            self._protocol = "http"
         self._request = requests.Session()
 
         retries = Retry(connect=10, read=10, redirect=5,
                 status_forcelist=[429, 500, 502, 503, 504], backoff_factor=2)
 
-        self._request.mount('http://', HTTPAdapter(max_retries=retries))
+        self._request.mount('{0}://'.format(self._protocol), HTTPAdapter(max_retries=retries))
 
     
     # Destructor 
@@ -86,37 +88,26 @@ class WPimport:
                     else:
                         self._addOrUpdateFeaturedMedia(soup)
     
-    def removeAll(self):
-        params = {"per_page":100}
-        try:
-            self._logger.info("{0} : List posts to remove for url : {1}".format(self._name, self._wordpress))
-            
-            r = self._request.get("http://{0}/wp-json/wp/v2/posts".format(self._wordpress), auth=self._basic, params=params, headers=self._headers_json)
-        except Exception as err:
-            self._logger.error("{0} : Connection error for list post to remove : {1}".format(self._name, err))
-        if r.status_code == 200:
-            result = r.json()
-            if len(result) > 0:
-                for i in result:
-                    self._logger.info("{0} : Remove article for url {1} : {2}".format(self._name, self._wordpress, i["title"]["rendered"]))
-                    params = {"force":1}
-                    try:
-                        r = self._request.delete("http://{0}/wp-json/wp/v2/posts/{1}".format(self._wordpress, i["id"]), auth=self._basic, headers=self._headers_json , params=params)
-                        if r.status_code == 200:
-                            self._logger.info("{0} : Post removed for URL {1} : {2}".format(self._name, self._wordpress, i["title"]["rendered"]))
-                        else:
-                            self._logger.error("{0} : Connection error for post {1} with status code {2}".format(self._name, self._wordpress, i["title"]["rendered"]))
-                    except Exception as err:
-                        self._logger.error("{0} : Connection error for post remove : {1}".format(self._name, err))
-                        exit(1)
-                self.removeAll()
-        else:
-            self._logger.error("{0} : Error for list to remove due status code {1}".format(self._name, r.status_code))
-            self._logger.debug("{0} : Content error : {1}".format(self._name, r.content))
-
     
     # Private method
 
+    ## replace caracter 
+
+    def _replaceCaracter(self, title_rendered):
+        list_replace = {'’': "'", '–': '-', '…': '...', '« ': '"', ' »': '"', '« ': '"', ' »': '"', '’': "'", '"‘': "'"}
+        for old, new in list_replace.items():
+            title_rendered = title_rendered.replace(old, new)
+        return title_rendered
+
+    ## remove space 
+
+    def _removeSpace(self, title):
+        if title[len(title)-1] == " ":
+            title = title[:-1]
+        if title[0] == " ":
+            title = title[1:]
+        return title
+
     ## Get all files
 
     def _getFiles(self, item):
@@ -145,7 +136,7 @@ class WPimport:
             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)
+                page = self._request.get("{1}://{0}/wp-json/wp/v2/search".format(self._wordpress, self._protocol), auth=self._basic, params=params)
             except Exception as err:
                 self._logger.error("{0} : Connection error : {1}".format(self._name, err))
                 exit(1)
@@ -166,7 +157,7 @@ class WPimport:
                                 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)
+                                    page = self._request.get("{1}://{0}/wp-json/wp/v2/media".format(self._wordpress, self._protocol), auth=self._basic, params=params)
                                 except Exception as err:
                                     self._logger.error("{0} : Connection error search featured media : {1}".format(self._name, err))
                                     exit(1)
@@ -176,7 +167,7 @@ class WPimport:
                                         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))
+                                            r = self._request.post("{2}://{0}/wp-json/wp/v2/posts/{1}".format(self._wordpress, result[0]["id"], self._protocol), auth=self._basic, headers=self._headers_json, data=json.dumps(data))
                                         except Exception as err:
                                             self._logger.error("{0} : Connection error for post media featured : {1}".format(self._name, err))
                                             exit(1)
@@ -207,7 +198,7 @@ class WPimport:
         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)
+                r = self._request.post("{2}://{0}/wp-json/wp/v2/media/{1}".format(self._wordpress, i["id"], self._protocol), auth=self._basic, data=data)
             except Exception as err:
                 self._logger.error("{0} : Connection error for link image to post : {1}".format(self._name, err))
                 exit(1)
@@ -221,7 +212,7 @@ class WPimport:
     ## Add or update img 
 
     def _addOrUpdateMedia(self, href_img, page):
-        media_authorized = ["png", "jpg", "jpeg", "svg"]
+        media_authorized = ["png", "jpg", "jpeg", "svg", "gif"]
         media = {"id":"", "rendered":""}
         split_fileimg = href_img.split("/")
         img_name = split_fileimg[len(split_fileimg)-1]
@@ -231,10 +222,10 @@ class WPimport:
             self._logger.error("{0} : Element {1} is not image".format(self._name,img_name))
             is_img = False
         if is_img is True:
-            self._logger.debug("{0} : Search for image {1} with URL {2}".format(self._name, img_name, "http://{0}/wp-json/wp/v2/media".format(self._wordpress)))
+            self._logger.debug("{0} : Search for image {1} with URL {2}".format(self._name, img_name, "{1}://{0}/wp-json/wp/v2/media".format(self._wordpress, self._protocol)))
             params = { "search": img_name}
             try:
-                r = self._request.get("http://{0}/wp-json/wp/v2/media".format(self._wordpress), auth=self._basic, params=params)
+                r = self._request.get("{1}://{0}/wp-json/wp/v2/media".format(self._wordpress, self._protocol), auth=self._basic, params=params)
             except Exception as err:
                 self._logger.error("{0} : Connection error for search media : {1}".format(self._name, err))
                 exit(1)
@@ -245,7 +236,7 @@ class WPimport:
                 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)
+                        r = self._request.delete("{2}://{0}/wp-json/wp/v2/media/{1}".format(self._wordpress, res[0]["id"], self._protocol), auth=self._basic, params=params)
                     except Exception as err:
                         self._logger.error("{0} Connection error for delete image : {1}".format(self._name, err))
                         exit(1)
@@ -261,7 +252,7 @@ class WPimport:
                     img_type = "image/jpeg"
                 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)
+                    r = self._request.post("{1}://{0}/wp-json/wp/v2/media".format(self._wordpress, self._protocol), auth=self._basic, headers=headers, data=data)
                 except Exception as err:
                     self._logger.error("{0} : Connection error for add image : {1}".format(self._name, err))
                     exit(1)
@@ -287,7 +278,7 @@ class WPimport:
 
                 try:
                     params = {"post": post, "author_name":i["author"], "date":i["date"]}
-                    page = self._request.get("http://{0}/wp-json/wp/v2/comments".format(self._wordpress), auth=self._basic, params=params)
+                    page = self._request.get("{1}://{0}/wp-json/wp/v2/comments".format(self._wordpress, self._protocol), auth=self._basic, params=params)
                 except Exception as err:
                     self._logger.error("{0} : Connection error for search comment : {1}".format(self._name, err))
                     exit(1)
@@ -296,7 +287,7 @@ class WPimport:
                     for j in result:
                         try:
                             params = {"force":1}
-                            page = self._request.delete("http://{0}/wp-json/wp/v2/comments/{1}".format(self._wordpress, j["id"]), params=params, auth=self._basic)
+                            page = self._request.delete("{2}://{0}/wp-json/wp/v2/comments/{1}".format(self._wordpress, j["id"], self._protocol), params=params, auth=self._basic)
                         except Exception as err:
                             self._logger.error("{0} : Connection error for delete comment : {1}".format(self._name, err))
                             exit(1)
@@ -318,7 +309,7 @@ class WPimport:
                     parent_id = int(i["parent_id"])
                     params = {"post": post, "author_name":comment[parent_id]["author"], "date":comment[parent_id]["date"]}
                     try:
-                        page = self._request.get("http://{0}/wp-json/wp/v2/comments".format(self._wordpress), auth=self._basic, params=params)
+                        page = self._request.get("{1}://{0}/wp-json/wp/v2/comments".format(self._wordpress, self._protocol), auth=self._basic, params=params)
                     except Exception as err:
                         self._logger.error("{0} : Connection error for parent comment : {1}".format(self._name, err))
                         exit(1)
@@ -331,7 +322,7 @@ class WPimport:
                         self._logger.debug("{0} : {1}".format(self._name, page.content))
 
                 try:
-                    page = self._request.post("http://{0}/wp-json/wp/v2/comments".format(self._wordpress), auth=self._basic, data=data)
+                    page = self._request.post("{1}://{0}/wp-json/wp/v2/comments".format(self._wordpress, self._protocol), auth=self._basic, data=data)
                 except Exception as err:
                     self._logger.error("{0} : Connection error for add comment : {1}".format(self._name, err))
                     exit(1)
@@ -465,33 +456,56 @@ class WPimport:
         for i in liste:
             for j in element[i]:
                 element_exist = False
+                title_element = self._removeSpace(j)
+                for index in range(1,10):
+                    self._logger.info("{0} : search {1} with index {2} : {3}".format(self._name, i, index, title_element))
                     try:
-                    params = {"params":j}
+                        params = {"search":title_element, "per_page":"100", "page":index}
-                    page = self._request.get("http://{0}/wp-json/wp/v2/{1}".format(self._wordpress, i), auth=self._basic, params=params)
+                        page = self._request.get("{2}://{0}/wp-json/wp/v2/{1}".format(self._wordpress, i, self._protocol), auth=self._basic, params=params)
-                except Exception as err:
+                    except ConnectionError as err:
                         self._logger.error("{0} : Connection error for {1} : {2}".format(self._name, i, err))
                         exit(1)
+                    except Exception as err:
+                        self._logger.error("{0} : Exception error for {1} : {2}".format(self._name, i, err))
                     if page.status_code == 200:
-                    element_exist = True
                         result = page.json()
-                    listelement[i].append(result[0]["id"])
+                        self._logger.debug("{0} : content {3} {2} : {1}".format(self._name, result, title_element, i))
+                        if len(result) > 0:
+                            for k in result:
+                                title_rendered = k["name"]
+                                self._logger.debug("{0} : content {2} : {1}".format(self._name, title_rendered, i))
+                                self._logger.debug("{0} : size of content {3} : {2} - {1}".format(self._name, len(title_rendered), len(title_element), i))
+                                if len(title_element) != len(title_rendered):
+                                    title_rendered = self._replaceCaracter(title_rendered)
                         
+                                if title_element == title_rendered:    
+                                    self._logger.info("{0} : {1} found : {2}".format(self._name, i, title_rendered))
+                                    element_exist = True
+                                    listelement[i].append(k["id"])
+                        else:
+                            break
+                    if page.status_code == 400:
+                        self._logger.error("{0} : {1} not found due status code : {2}".format(self._name, i, page.status_code))
+                        self._logger.debug("{0} : {1}".format(self._name, page.content))
+                        break
                     else:
                         self._logger.error("{0} : {1} not found due status code : {2}".format(self._name, i, page.status_code))
                         self._logger.debug("{0} : {1}".format(self._name, page.content))
 
-
+                self._logger.debug("{0} : Element {3} {2} is {1}".format(self._name, element_exist, title_element, i))
-                if element_exist is False:
+                if element_exist == False:
-                    data = {"name": j}
+                    data = {"name": title_element}
-                    self._logger.debug("{0} : URL : {1} ".format("http://{1}/wp-json/wp/v2/{2}".format(self._name, self._wordpress, i)))
+                    self._logger.info("{0} : Create {1} : {2}".format(self._name, i, title_element))
-                    self._logger.debug("{0} : data : {1}".format(self._name, data))
+                    self._logger.debug("{0} : Data : {1}".format(self._name, data))
-                    self._logger.debug("{0} : headers : {1}".format(self._name, self._headers_form))
                     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)
+                        page = self._request.post("{2}://{0}/wp-json/wp/v2/{1}".format(self._wordpress, i, self._protocol), auth=self._basic, headers=self._headers_json, data=json.dumps(data))
-                    except Exception as err:
+                    except ConnectionError as err:
                         self._logger.error("{0} : Connection error for post {1} : {2}".format(self._name, i, err))
                         exit(1)
+                    except Exception as err:
+                        self._logger.error("{0} : Exception error for post {1} : {2}".format(self._name, i, err))
                     if page.status_code == 201:
+                        self._logger.info("{0} : {1} created : {2}".format(self._name, i, j))
                         result = page.json()
                         listelement[i].append(result["id"])
                     else:
@@ -514,44 +528,92 @@ class WPimport:
             bodyhtml = bodyhtml.replace(i["old_src"], o.path)
         hour = articledate[0].text
         time = dateheader[0].text.split(" ")
+        self._logger.debug("{0} : Title post : |{1}|".format(self._name, title))
+        title = self._removeSpace(title)
+        self._logger.debug("{0} : Rendered Title post : |{1}|".format(self._name, title))
+
         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}
+        self._logger.debug("{0} : Data for post : |{1}| : {2}" .format(self._name, title, data))
+
+        params = {"search":author, "per_page":100}
         try:
-            self._logger.info("{0} : Get author : {1}".format(self._name, author))
+            self._logger.info("{0} : Search author : {1}".format(self._name, author))
-            page = self._request.get("http://{0}/wp-json/wp/v2/users".format(self._wordpress), auth=self._basic, params=params)
+            page = self._request.get("{1}://{0}/wp-json/wp/v2/users".format(self._wordpress, self._protocol), auth=self._basic, headers=self._headers_json, params=params)
-        except Exception as err:
+            self._logger.debug("{0} : End Search author : {1}".format(self._name, author))
+            self._logger.debug("{0} : Debug requests : {1}".format(self._name, page.content))
+
+
+        except ConnectionError as err:
             self._logger.error("{0} : Connection error for get author : {1}".format(self._name, err))
             exit(1)
+        except Exception as err:
+            self._logger.error("{0} : Exception error for get author : {1}".format(self._name, err))
         if page.status_code == 200:
+            self._logger.info("{0} : Get author id : {1}".format(self._name, result))
             result = page.json()
-            data["author"] = result[0]["id"]
+            for a in result:
+                data["author"] = a["id"]
         else:
             self._logger.error("{0} : Connection error with status code for get author : {1}".format(self._name, page.status_code))
             self._logger.debug("{0} : {1}".format(page.content))
+        page_is_exist = False
 
-        params = {"search": title}
+        for index in range(1,10):
+            params = {"search": title, "per_page":100, "page": index}
             try:
-            page = self._request.get("http://{0}/wp-json/wp/v2/posts".format(self._wordpress), auth=self._basic, params=params)
+                self._logger.info("{0} : Search post with index {2} : {1}".format(self._name, title, index))
-        except Exception as err:
+                page = self._request.get("{1}://{0}/wp-json/wp/v2/posts".format(self._wordpress, self._protocol), auth=self._basic, params=params, headers=self._headers_json)
+            except ConnectionError as err:
                 self._logger.error("{0} : Connection error for search post : {1}".format(self._name, err))
                 exit(1)
-        page_is_exist = True
+            except Exception as err:
+                self._logger.error("{0} : Exception error for search post : {1}".format(self._name, err))
             if page.status_code == 200:
+                self._logger.debug("{0} : Encoding : {1}".format(self._name, page.encoding))
+                page.encoding = "utf-8"
                 result = page.json()
                 if len(result) == 0:
-                page_is_exist = False
+                    break
-            else:
+                self._logger.info("{0} : Number result posts : {1}".format(self._name, len(result)))
+                count = 0
                 for i in result:
-                    if i["title"]["rendered"] == title:
+                    title_rendered = i["title"]["rendered"]
+                    self._logger.info("{0} : Search title posts for |{2}| : |{1}|".format(self._name, title_rendered, title))
+                    if len(title_rendered) != len(title):
+                        title_rendered = self._replaceCaracter(title_rendered)
+                    self._logger.debug("{0} : Search title posts for |{2}| : |{1}|".format(self._name, title_rendered, title))
+                    self._logger.debug("{0} : SIze of title : {1} - {2}".format(self._name, len(title), len(title_rendered)))
+                    if title_rendered == title:
+                        page_is_exist = True
                         post_id = i["id"]
-                        self._logger.debug("{0} : Data for post to update : {1}".format(self._name, result[0]))
+                        count = count + 1
+                        if count > 1:
+                            self._logger.info("{0} : Page {1} is double and going to delete".format(self._name, title))
+                            try:
+                                params = {"force":1}
+                                page = self._request.delete("{2}://{0}/wp-json/wp/v2/posts/{1}".format(self._wordpress, post_id, self._protocol), auth=self._basic, headers=self._headers_json, params=params)
+                            except ConnectionError as err:
+                                self._logger.error("{0} : Connection error for deleted post : {1}".format(self._name, err))
+                                exit(1)
+                            except Exception as err:
+                                self._logger.error("{0} : Exception error for deleted post : {1}".format(self._name, err))
+                            if page.status_code == 200:
+                                self._logger.info("{0} : Post deleted : {1}".format(self._name, title))
+                            else:
+                                self._logger.error("{0} : Post not updated due status code : {1}".format(self._name, page.status_code))
+                                self._logger.debug("{0} : {1}".format(self._name, page.content))    
+                    
+                        else:
+                            self._logger.debug("{0} : Data for post to update : {1}".format(self._name, i))
                             self._logger.info("{0} : Page {1} already exist and going to update".format(self._name, title))
                 
                             try:
-                            page = self._request.post("http://{0}/wp-json/wp/v2/posts/{1}".format(self._wordpress, post_id), auth=self._basic, headers=self._headers_json, data=json.dumps(data))
+                                page = self._request.post("{2}://{0}/wp-json/wp/v2/posts/{1}".format(self._wordpress, post_id, self._protocol), auth=self._basic, headers=self._headers_json, data=json.dumps(data))
-                        except Exception as err:
+                            except ConnectionError as err:
                                 self._logger.error("{0} : Connection error for update post : {1}".format(self._name, err))
                                 exit(1)
+                            except Exception as err:
+                                self._logger.error("{0} : Exception error for update post : {1}".format(self._name, err))
                             if page.status_code == 200:
                                 result = page.json()
                                 self._logger.info("{0} : Post updated : {1}".format(self._name, title))
@@ -560,16 +622,23 @@ class WPimport:
                             else:
                                 self._logger.error("{0} : Post not updated due status code : {1}".format(self._name, page.status_code))
                                 self._logger.debug("{0} : {1}".format(self._name, page.content))
+            if page.status_code == 400:
+                self._logger.error("{0} : Connection for update post unauthorized : {1}".format(self._name, page.status_code))
+                self._logger.debug("{0} : {1}".format(self._name, page.content))
+                break
             else:
                 self._logger.error("{0} : Connection for update post error with status code : {1}".format(self._name, page.status_code))
                 self._logger.debug("{0} : {1}".format(self._name, page.content))
 
-        if page_is_exist == False:
+        if page_is_exist is False:
             try:
-                page = self._request.post("http://{0}/wp-json/wp/v2/posts".format(self._wordpress), auth=self._basic, headers=self._headers_json, data=json.dumps(data))
+                self._logger.info("{0} : Creating posts : {1}".format(self._name, data["title"]))
-            except Exception as err:
+                page = self._request.post("{1}://{0}/wp-json/wp/v2/posts".format(self._wordpress, self._protocol), auth=self._basic, headers=self._headers_json, data=json.dumps(data))
+            except ConnectionError as err:
                 self._logger.error("{0} : Connection error for create post : {1}".format(self._name, err))
                 exit(1)
+            except Exception as err:
+                self._logger.error("{0} : Exception error for create post : {1}".format(self._name, err))
             if page.status_code == 201:
                 result = page.json()
                 self._logger.info("{0} : Post added : {1}".format(self._name, result["title"]["raw"]))

lib/WPRemove.py

@@ -0,0 +1,78 @@
+#!/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 WPRemove:
+    # Constructor
+    def __init__(self, name="Thread-0", basic=None, wordpress="", logger=None, ssl_wordpress=True):
+        self._name = name
+        self._basic = basic
+        self._wordpress = wordpress
+        self._logger = logger
+        self._headers_json = {'Content-Type': 'application/json', 'Accept':'application/json'}
+        
+        self._protocol = "https"
+        if ssl_wordpress is False:
+            self._protocol = "http"
+        self._request = requests.Session()
+
+        retries = Retry(connect=10, read=10, redirect=5,
+                status_forcelist=[429, 500, 502, 503, 504], backoff_factor=2)
+
+        self._request.mount('{0}://'.format(self._protocol), HTTPAdapter(max_retries=retries))
+
+    
+    # Destructor 
+    def __del__(self):
+        print("{0} : Import finished for {1}".format(self._name, self._wordpress))
+
+    # Public method
+
+    def setUrl(self, wordpress):
+        self._wordpress = wordpress
+    
+    def cleanPosts(self):
+        self._removeAll("posts")
+
+    def cleanTags(self):
+        self._removeAll("tags")
+
+    def cleanCategories(self):
+        self._removeAll("categories")
+    
+    def cleanMedia(self):
+        self._removeAll("media")
+
+    # Private method
+    
+    def _removeAll(self, composant):
+        params = {"per_page":100}
+        try:
+            self._logger.info("{0} : List {2} to remove for url : {1}".format(self._name, self._wordpress, composant))
+            
+            r = self._request.get("{2}://{0}/wp-json/wp/v2/{1}".format(self._wordpress, composant, self._protocol), auth=self._basic, params=params, headers=self._headers_json)
+        except Exception as err:
+            self._logger.error("{0} : Connection error for list {1} to remove : {2}".format(self._name, composant, err))
+        if r.status_code == 200:
+            result = r.json()
+            if len(result) > 0:
+                for i in result:
+                    self._logger.info("{0} : Remove {2} for url {1} : {3}".format(self._name, self._wordpress, composant, i["title"]["rendered"]))
+                    params = {"force":1}
+                    try:
+                        r = self._request.delete("{3}://{0}/wp-json/wp/v2/{1}/{2}".format(self._wordpress, composant, i["id"], self._protocol), auth=self._basic, headers=self._headers_json , params=params)
+                        if r.status_code == 200:
+                            self._logger.info("{0} : Post removed for URL {1} {2} : {3}".format(self._name, self._wordpress, composant, i["title"]["rendered"]))
+                        else:
+                            self._logger.error("{0} : Connection error for post {1} {2} {3} with status code {4}".format(self._name, self._wordpress, composant, i["title"]["rendered"], r.status_code))
+                    except Exception as err:
+                        self._logger.error("{0} : Connection error for {1} remove : {2}".format(self._name, composant, err))
+                        exit(1)
+                self._removeAll(composant)
+        else:
+            self._logger.error("{0} : Error for list to remove {1} due status code {2}".format(self._name, composant, r.status_code))
+            self._logger.debug("{0} : Content error for {1} : {2}".format(self._name, composant, r.content))