fixed crawlqueue issue

onioningestor/__init__.py

@@ -61,7 +61,7 @@ class Ingestor:
 
             # Instantiate operator plugins.
             self.logger.debug("initializing operators")
-            self.operators = {name: operator(self.logger, **kwargs)
+            self.operators = {name: operator(self.logger, self.blacklist, **kwargs)
                               for name, operator, kwargs in self.config.operators()}
 
         except Exception as e:
@@ -99,6 +99,8 @@ class Ingestor:
                     ## Process onions with each operator.
                     for operator in self.operators:
                         self.logger.info(f"Processing found onions with operator '{operator}'")
+                        # Set CrawlQueue for every operator
+                        self.operators[operator].set_crawlQueue(self.queue)
                         # Process list of onions
                         self.operators[operator].process(onions)
                         done = True

onioningestor/databases/elasticsearch.py

@@ -61,7 +61,7 @@ class Plugin(PastieStorage):
                 'port':self.config['port']}])
             self.es.indices.create(
                     index=self.index,
-                    body=self.mapping,
+                    #body=self.mapping,
                     ignore=400)
         except Exception as e:
             self.logger.error(e)

onioningestor/operators/__init__.py

@@ -6,8 +6,7 @@ from itertools import islice
 from datetime import datetime as dt
 from concurrent.futures import ThreadPoolExecutor
 
-from collections import namedtuple
+from onioningestor.onion import Onion
-
 
 class Operator:
     """Base class for all Operator plugins.
@@ -47,12 +46,13 @@ class Operator:
         classes. Remember to do so *before* setting any default artifact_types.
         """
         self.logger = logger
-        self.processQueue = Queue()
+        self.onion = Onion
-        self.onions = {}
-        self.onion = namedtuple('onion',['url','source','type','index','monitor','denylist'])
         deny = allowed_sources or []
         self.blacklist = re.compile('|'.join([re.escape(word) for word in deny]), re.IGNORECASE)
 
+    def set_crawlQueue(self, queue):
+        self.queueCrawl = queue
+        
     def handle_onion(self, url):
         """Override with the same signature.
 
@@ -78,6 +78,7 @@ class Operator:
         blacklist = self.blacklist.findall(content)
         if blacklist:
             onion.denylist = blacklist
+            onion.status = 'blocked'
 
     def collect(self, onions):
         for onion in onions:

onioningestor/operators/html.py

@@ -23,8 +23,8 @@ class Plugin(Operator):
     This plugin collects HTML code from onion link
     """
 
-    def __init__(self, logger, **kwargs):
+    def __init__(self, logger, denylist, **kwargs):
-        super(Plugin, self).__init__(logger)
+        super(Plugin, self).__init__(logger, denylist)
         self.name = kwargs['name']
         self.logger.info(f"Initializing {self.name}")
 
@@ -90,7 +90,7 @@ class Plugin(Operator):
                                 "title": html.title.text,
                                 "language": detect(html.text),
                                 "status": "success",
-                                "interestingKeywords": Counter(self.interesting.findall(result)),
+                                "interestingKeywords": list(set(self.interesting.findall(result))),
                             }
                         else:
                             index = {
@@ -98,7 +98,7 @@ class Plugin(Operator):
                                 "title": None,
                                 "language": None,
                                 "status": "success",
-                                "interestingKeywords": Counter(self.interesting.findall(result)),
+                                "interestingKeywords": list(set(self.interesting.findall(result))),
                             }
                         return self.response("success", index)
 

onioningestor/operators/onionscan.py

@@ -17,7 +17,8 @@ class Plugin(Operator):
     Handles reading the config file, calling sources, maintaining state and
     sending artifacts to operators.
     """
-    def __init__(self, logger, **kwargs):
+    def __init__(self, logger, denylist, **kwargs):
+        super(Plugin, self).__init__(logger, denylist)
         self.name = kwargs['name']
         self.logger = logger
         self.logger.info(f'Initializing {self.name}')
@@ -27,16 +28,20 @@ class Plugin(Operator):
         self.torControl = "Zue5a29v4xE6FciWpPF93rR2M2T"
 
     def parseDoc(self, data):
-        data['onionscan'].pop('simpleReport', None)
+        data.pop('simpleReport', None)
-        crawls = data['onionscan'].pop('crawls', None)
+        crawls = data.pop('crawls', None)
-        hiddenService = data['onionscan'].pop('hiddenService', None)
+        hiddenService = data.pop('hiddenService', None)
-        data['onionscan']['crawls'] = [*crawls]
+        data['crawls'] = [*crawls]
-        data['hiddenService'] = hiddenService
+        crawl = set()
-        for onion in crawls.keys():
+        for onion in re.findall(r'\s?(\w+.onion)', str(crawls.keys())):
-            self.queueCrawl((
+            if onion != hiddenService:
+                crawl.add(onion)
+        for items in crawl:
+            print(f'crawling queue added: {item}')
+            self.queueCrawl.put((
                 3,
                 self.onion(
-                    url=onion,
+                    url=item,
                     source='crawled',
                     type='domain',
                     status='offline',
@@ -97,6 +102,7 @@ class Plugin(Operator):
                         self.parseDoc(stdout))
 
         self.logger.info("[!!!] Process timed out for %s", onion)
+        print(stdout)
         return self.response("failed",stdout)
 
     def handle_onion(self, onion):