Always use CDR format, add extracted_metadata

What was previously stored in PageItem and FormItem
is now stored in extracted_metadata: is_page, depth, forms.

Author: lopuhin

README.md

@@ -51,5 +51,4 @@ Useful options to tweak (add to the above command via ``-s NAME=value``):
 - ``DOWNLOAD_DELAY`` - set to 0 when crawling local test server
 - ``RUN_HH`` - set to 0 to skip running full headless-horesman scripts
 - ``PREFER_PAGINATION`` - set to 0 to disable pagination handling
-- ``CDR_EXPORT`` - set to 0 to disable export in CDR format
-- ``CDR_*`` - CDR export constants
+- ``CDR_CRAWLER``, ``CDR_TEAM`` - CDR export metadata constants

undercrawler/items.py

@@ -1,20 +1,6 @@
 import scrapy
 
 
-class PageItem(scrapy.Item):
-    url = scrapy.Field()
-    text = scrapy.Field()
-    is_page = scrapy.Field()
-    depth = scrapy.Field()
-
-    def __repr__(self):
-        return repr({
-            'url': self['url'],
-            'is_page': self['is_page'],
-            'depth': self['depth'],
-        })
-
-
 class CDRItem(scrapy.Item):
 
     # (url)-(crawl timestamp), SHA-256 hashed, UPPERCASE (string)
@@ -27,6 +13,10 @@ class CDRItem(scrapy.Item):
     crawler = scrapy.Field()
 
     # Tika/other extraction output (object)
+    # Our suff here:
+    #  forms: forms metadata as extracted by formasaurus
+    #  depth: page depth
+    #  is_page: this is a page reached by pagination
     extracted_metadata = scrapy.Field()
 
     # Tika/other extraction output (string)
@@ -49,5 +39,6 @@ class CDRItem(scrapy.Item):
     version = scrapy.Field()
 
     def __repr__(self):
-        fields = ['_id', 'url', 'timestamp']
-        return repr({f: self[f] for f in fields})
+        fields = ['_id', 'url', 'timestamp', 'extracted_metadata']
+        return '<CDRItem: {}>'.format(', '.join(
+            '{}: {}'.format(f, repr(self[f])) for f in fields))

undercrawler/settings.py

@@ -12,7 +12,6 @@
 AUTOLOGIN_URL = 'http://127.0.0.1:8089'
 AUTOLOGIN_ENABLED = True
 
-CDR_EXPORT = True
 CDR_CRAWLER = 'scrapy undercrawler'
 CDR_TEAM = 'HG'
 

undercrawler/spiders/base_spider.py

@@ -4,12 +4,13 @@
 import hashlib
 
 import autopager
+import formasaurus
 import scrapy
 from scrapy.linkextractors import LinkExtractor
 from scrapy.utils.url import canonicalize_url
 from scrapy.utils.python import unique
 
-from ..items import PageItem, CDRItem
+from ..items import CDRItem
 
 
 class BaseSpider(scrapy.Spider):
@@ -39,15 +40,13 @@ def parse(self, response):
         if not self.link_extractor.matches(url):
             return
 
-        if self.settings.getbool('CDR_EXPORT'):
-            yield self.cdr_item(response)
-        else:
-            yield PageItem(
-                url=url,
-                text=response.text,
-                is_page=response.meta.get('is_page', False),
-                depth=response.meta.get('depth', None),
-                )
+        forms = formasaurus.extract_forms(response.text) if response.text \
+                else []
+        yield self.cdr_item(response, dict(
+            is_page=response.meta.get('is_page', False),
+            depth=response.meta.get('depth', None),
+            forms=[meta for _, meta in forms],
+            ))
 
         if self.settings.getbool('PREFER_PAGINATION'):
             # Follow pagination links; pagination is not a subject of
@@ -64,7 +63,7 @@ def parse(self, response):
         for link in self.link_extractor.extract_links(response):
             yield self.splash_request(link.url)
 
-    def cdr_item(self, response):
+    def cdr_item(self, response, metadata):
         url = response.url
         timestamp = int(datetime.utcnow().timestamp() * 1000)
         return CDRItem(
@@ -73,7 +72,7 @@ def cdr_item(self, response):
             content_type=response.headers['content-type']\
                 .decode('ascii', 'ignore'),
             crawler=self.settings.get('CDR_CRAWLER'),
-            extracted_metadata={},
+            extracted_metadata=metadata,
             extracted_text='\n'.join(
                 response.xpath('//body').xpath('string()').extract()),
             raw_content=response.text,