From 1ae22a218f384027ca9f94775894823014357ea0 Mon Sep 17 00:00:00 2001
From: dgw <dgw@technobabbl.es>
Date: Wed, 14 Jul 2021 00:29:24 -0500
Subject: [PATCH] wikipedia: skip messagebox template contents

---
 sopel/modules/wikipedia.py | 28 ++++++++++++++++++++++++++--
 1 file changed, 26 insertions(+), 2 deletions(-)

diff --git a/sopel/modules/wikipedia.py b/sopel/modules/wikipedia.py
index f9557eeed9..2e20001941 100644
--- a/sopel/modules/wikipedia.py
+++ b/sopel/modules/wikipedia.py
@@ -29,6 +29,7 @@ def __init__(self, section_name):
         self.section_name = section_name
 
         self.citations = False
+        self.messagebox = False
         self.span_depth = 0
         self.div_depth = 0
 
@@ -49,7 +50,9 @@ def handle_starttag(self, tag, attrs):
                     if attr[0] == 'class' and 'edit' in attr[1]:
                         self.span_depth += 1
 
-        elif tag == 'div':  # We want to skip thumbnail text and the inexplicable table of contents, and as such also need to track div depth
+        elif tag == 'div':
+            # We want to skip thumbnail text and the inexplicable table of contents,
+            # and as such also need to track div depth
             if self.div_depth:
                 self.div_depth += 1
             else:
@@ -57,6 +60,25 @@ def handle_starttag(self, tag, attrs):
                     if attr[0] == 'class' and ('thumb' in attr[1] or attr[1] == 'toc'):
                         self.div_depth += 1
 
+        elif tag == 'table':
+            # Message box templates are what we want to ignore here
+            for attr in attrs:
+                if (
+                    attr[0] == 'class'
+                    and any(classname in attr[1].lower() for classname in [
+                        # Most of list from https://en.wikipedia.org/wiki/Template:Mbox_templates_see_also
+                        'ambox',  # messageboxes on article pages
+                        'cmbox',  # messageboxes on category pages
+                        'imbox',  # messageboxes on file (image) pages
+                        'tmbox',  # messageboxes on talk pages
+                        'fmbox',  # header and footer messageboxes
+                        'ombox',  # messageboxes on other types of page
+                        'mbox',  # for messageboxes that are used in different namespaces and change their presentation accordingly
+                        'dmbox',  # for disambiguation messageboxes
+                    ])
+                ):
+                    self.messagebox = True
+
         elif tag == 'ol':
             for attr in attrs:
                 if attr[0] == 'class' and 'references' in attr[1]:
@@ -71,9 +93,11 @@ def handle_endtag(self, tag):
             self.span_depth -= 1
         if self.div_depth and tag == 'div':
             self.div_depth -= 1
+        if self.messagebox and tag == 'table':
+            self.messagebox = False
 
     def handle_data(self, data):
-        if self.consume and not any([self.citations, self.span_depth, self.div_depth]):
+        if self.consume and not any([self.citations, self.messagebox, self.span_depth, self.div_depth]):
             if not (self.is_header and data == self.section_name):  # Skip the initial header info only
                 self.result += data