Re #3486 - Fixing and adding test for RSS/Atom not being converted to text when server sends "text/xml" instead of the "application/atom+xml" header

changedetectionio/html_tools.py

@@ -1,5 +1,4 @@
 from loguru import logger
-from lxml import etree
 from typing import List
 import html
 import json
@@ -58,13 +57,17 @@ def include_filters(include_filters, html_content, append_pretty_line_formatting
 
     return html_block
 
-def subtractive_css_selector(css_selector, html_content):
+def subtractive_css_selector(css_selector, content):
     from bs4 import BeautifulSoup
-    soup = BeautifulSoup(html_content, "html.parser")
+    soup = BeautifulSoup(content, "html.parser")
 
     # So that the elements dont shift their index, build a list of elements here which will be pointers to their place in the DOM
     elements_to_remove = soup.select(css_selector)
 
+    if not elements_to_remove:
+        # Better to return the original that rebuild with BeautifulSoup
+        return content
+
     # Then, remove them in a separate loop
     for item in elements_to_remove:
         item.decompose()
@@ -72,6 +75,7 @@ def subtractive_css_selector(css_selector, html_content):
     return str(soup)
 
 def subtractive_xpath_selector(selectors: List[str], html_content: str) -> str:
+    from lxml import etree
     # Parse the HTML content using lxml
     html_tree = etree.HTML(html_content)
 
@@ -83,6 +87,10 @@ def subtractive_xpath_selector(selectors: List[str], html_content: str) -> str:
         # Collect elements for each selector
         elements_to_remove.extend(html_tree.xpath(selector))
 
+    # If no elements were found, return the original HTML content
+    if not elements_to_remove:
+        return html_content
+
     # Then, remove them in a separate loop
     for element in elements_to_remove:
         if element.getparent() is not None:  # Ensure the element has a parent before removing

changedetectionio/processors/magic.py

@@ -94,24 +94,21 @@ class guess_stream_type():
             self.is_rss = True
         elif any(s in http_content_header for s in JSON_CONTENT_TYPES):
             self.is_json = True
+        elif 'pdf' in magic_content_header:
+            self.is_pdf = True
+        elif has_html_patterns or http_content_header == 'text/html':
+            self.is_html = True
+        elif any(s in magic_content_header for s in JSON_CONTENT_TYPES):
+            self.is_json = True
+        # magic will call a rss document 'xml'
+        # Rarely do endpoints give the right header, usually just text/xml, so we check also for <rss
+        # This also triggers the automatic CDATA text parser so the RSS goes back a nice content list
+        elif '<rss' in test_content_normalized or '<feed' in test_content_normalized or any(s in magic_content_header for s in RSS_XML_CONTENT_TYPES):
+            self.is_rss = True
         elif any(s in http_content_header for s in XML_CONTENT_TYPES):
             # Only mark as generic XML if not already detected as RSS
             if not self.is_rss:
                 self.is_xml = True
-        elif 'pdf' in magic_content_header:
-            self.is_pdf = True
-###
-        elif has_html_patterns or http_content_header == 'text/html':
-            self.is_html = True
-        # If magic says text/plain and we found no HTML patterns, trust it
-        elif magic_result == 'text/plain':
-            self.is_plaintext = True
-            logger.debug(f"Trusting magic's text/plain result (no HTML patterns detected)")
-        elif any(s in magic_content_header for s in JSON_CONTENT_TYPES):
-            self.is_json = True
-        # magic will call a rss document 'xml'
-        elif '<rss' in test_content_normalized or '<feed' in test_content_normalized or any(s in magic_content_header for s in RSS_XML_CONTENT_TYPES):
-            self.is_rss = True
         elif test_content_normalized.startswith('<?xml') or any(s in magic_content_header for s in XML_CONTENT_TYPES):
             # Generic XML that's not RSS/Atom (RSS/Atom checked above)
             self.is_xml = True
@@ -122,4 +119,8 @@ class guess_stream_type():
         # Only trust magic for 'text' if no other patterns matched
         elif 'text' in magic_content_header:
             self.is_plaintext = True
+        # If magic says text/plain and we found no HTML patterns, trust it
+        elif magic_result == 'text/plain':
+            self.is_plaintext = True
+            logger.debug(f"Trusting magic's text/plain result (no HTML patterns detected)")
 

changedetectionio/processors/text_json_diff/processor.py

@@ -404,6 +404,8 @@ class perform_site_check(difference_detection_processor):
         html_content = content
 
         # Apply include filters (CSS, XPath, JSON)
+        # Except for plaintext (incase they tried to confuse the system, it will HTML escape
+        #if not stream_content_type.is_plaintext:
         if filter_config.has_include_filters:
             html_content = content_processor.apply_include_filters(content, stream_content_type)
 

changedetectionio/tests/test_rss.py

@@ -110,8 +110,9 @@ def test_basic_cdata_rss_markup(client, live_server, measure_memory_usage):
     
 
     set_original_cdata_xml()
-
-    test_url = url_for('test_endpoint', content_type="application/atom+xml; charset=UTF-8", _external=True)
+    # Rarely do endpoints give the right header, usually just text/xml, so we check also for <rss
+    # This also triggers the automatic CDATA text parser so the RSS goes back a nice content list
+    test_url = url_for('test_endpoint', content_type="text/xml; charset=UTF-8", _external=True)
 
     # Add our URL to the import page
     uuid = client.application.config.get('DATASTORE').add_watch(url=test_url)