Merge pull request #3 from TeamHG-Memex/cdr

Export in CDRv2 format

Author: kmike

README.md

@@ -51,3 +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_CRAWLER``, ``CDR_TEAM`` - CDR export metadata constants

undercrawler/items.py

@@ -1,20 +1,44 @@
 import scrapy
 
 
-class PageItem(scrapy.Item):
-    url = scrapy.Field()
-    text = scrapy.Field()
-    is_page = scrapy.Field()
-    depth = scrapy.Field()
+class CDRItem(scrapy.Item):
 
-    def __repr__(self):
-        return repr({
-            'url': self['url'],
-            'is_page': self['is_page'],
-            'depth': self['depth'],
-        })
+    # (url)-(crawl timestamp), SHA-256 hashed, UPPERCASE (string)
+    _id = scrapy.Field()
+
+    # MIME type (multi (strings))
+    content_type = scrapy.Field()
+
+    # Text label identifying the software used by the crawler (string)
+    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)
+    extracted_text = scrapy.Field()
 
+    # Original source text/html (string)
+    raw_content = scrapy.Field()
 
-class FormItem(scrapy.Item):
+    # Text label identifying the team responsible for the crawler (string)
+    team = scrapy.Field()
+
+    # Timestamp of COLLECTION of data from the web (datetime)
+    # https://www.elastic.co/guide/en/elasticsearch/reference/current/mapping-date-format.html#built-in-date-formats
+    timestamp = scrapy.Field()
+
+    # Full URL requested by the crawler (multi (strings))
     url = scrapy.Field()
-    form_type = scrapy.Field()
+
+    # Schema version. This document describes schema version 2.0. (float)
+    version = scrapy.Field()
+
+    def __repr__(self):
+        fields = ['_id', 'url', 'timestamp', 'extracted_metadata']
+        return '<CDRItem: {}>'.format(', '.join(
+            '{}: {}'.format(f, repr(self[f])) for f in fields))

undercrawler/settings.py

@@ -12,6 +12,9 @@
 AUTOLOGIN_URL = 'http://127.0.0.1:8089'
 AUTOLOGIN_ENABLED = True
 
+CDR_CRAWLER = 'scrapy undercrawler'
+CDR_TEAM = 'HG'
+
 PREFER_PAGINATION = True
 
 DOWNLOADER_MIDDLEWARES = {

undercrawler/spiders/base_spider.py

@@ -1,5 +1,7 @@
 import re
 import contextlib
+from datetime import datetime
+import hashlib
 
 import autopager
 import formasaurus
@@ -8,7 +10,7 @@
 from scrapy.utils.url import canonicalize_url
 from scrapy.utils.python import unique
 
-from ..items import PageItem, FormItem
+from ..items import CDRItem
 
 
 class BaseSpider(scrapy.Spider):
@@ -35,17 +37,16 @@ def splash_request(self, url, callback=None, **kwargs):
 
     def parse(self, response):
         url = response.url
-        self.logger.info(url)
-        yield PageItem(
-            url=url,
-            text=response.text,
+        if not self.link_extractor.matches(url):
+            return
+
+        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),
-            )
-        if response.text:
-            for _, meta in formasaurus.extract_forms(response.text):
-                yield FormItem(url=url, form_type=meta['form'])
-                self.logger.info('Found a %s form at %s', meta['form'], url)
+            forms=[meta for _, meta in forms],
+            ))
 
         if self.settings.getbool('PREFER_PAGINATION'):
             # Follow pagination links; pagination is not a subject of
@@ -62,6 +63,25 @@ def parse(self, response):
         for link in self.link_extractor.extract_links(response):
             yield self.splash_request(link.url)
 
+    def cdr_item(self, response, metadata):
+        url = response.url
+        timestamp = int(datetime.utcnow().timestamp() * 1000)
+        return CDRItem(
+            _id=hashlib.sha256('{}-{}'.format(url, timestamp).encode('utf-8'))\
+                .hexdigest().upper(),
+            content_type=response.headers['content-type']\
+                .decode('ascii', 'ignore'),
+            crawler=self.settings.get('CDR_CRAWLER'),
+            extracted_metadata=metadata,
+            extracted_text='\n'.join(
+                response.xpath('//body').xpath('string()').extract()),
+            raw_content=response.text,
+            team=self.settings.get('CDR_TEAM'),
+            timestamp=timestamp,
+            url=url,
+            version=2.0,
+        )
+
     def _normalize_url(self, url):
         if not url.startswith('http'):
             url = 'http://' + url