Fix whitespace around comments between block elements

markdownify/__init__.py

@@ -240,7 +240,7 @@ def process_tag(self, node, parent_tags=None):
         # adjacent to the inner/outer boundaries of block elements.
         should_remove_inside = should_remove_whitespace_inside(node)
 
-        def _can_ignore(el):
+        def _can_ignore(el, previous_content, next_content):
             if isinstance(el, Tag):
                 # Tags are always processed.
                 return False
@@ -252,10 +252,10 @@ def _can_ignore(el):
                 if six.text_type(el).strip() != '':
                     # Non-whitespace text nodes are always processed.
                     return False
-                elif should_remove_inside and (not el.previous_sibling or not el.next_sibling):
+                elif should_remove_inside and (previous_content is None or next_content is None):
                     # Inside block elements (excluding <pre>), ignore adjacent whitespace elements.
                     return True
-                elif should_remove_whitespace_outside(el.previous_sibling) or should_remove_whitespace_outside(el.next_sibling):
+                elif should_remove_whitespace_outside(previous_content) or should_remove_whitespace_outside(next_content):
                     # Outside block elements (including <pre>), ignore adjacent whitespace elements.
                     return True
                 else:
@@ -265,7 +265,18 @@ def _can_ignore(el):
             else:
                 raise ValueError('Unexpected element type: %s' % type(el))
 
-        children_to_convert = [el for el in node.children if not _can_ignore(el)]
+        # Ignore comments and whitespace when locating block boundaries. Advance
+        # through content siblings once, rather than rescanning long comment runs.
+        content_siblings = (el for el in node.children if _is_block_content_element(el))
+        previous_content = None
+        next_content = next(content_siblings, None)
+        children_to_convert = []
+        for el in node.children:
+            if el is next_content:
+                previous_content = el
+                next_content = next(content_siblings, None)
+            if not _can_ignore(el, previous_content, next_content):
+                children_to_convert.append(el)
 
         # Create a copy of this tag's parent context, then update it to include this tag
         # to propagate down into the children.

tests/test_advanced.py

@@ -1,3 +1,7 @@
+import pytest
+from bs4 import BeautifulSoup, Comment
+from markdownify import MarkdownConverter
+
 from .utils import md
 
 
@@ -27,6 +31,50 @@ def test_ignore_comments_with_other_tags():
     assert text == "[example link](http://example.com/)"
 
 
+@pytest.mark.parametrize('separator', ['\n', ' ', '\t'])
+@pytest.mark.parametrize('count', [1, 2, 4])
+def test_ignore_whitespace_between_comments_and_blocks(separator, count):
+    comments = separator + ('<!-- comment -->' + separator) * count
+    html = '<p>line 1</p>' + comments + '<p>line 2</p>'
+    assert MarkdownConverter().convert(html) == 'line 1\n\nline 2'
+
+
+def test_ignore_comment_whitespace_inside_block_boundaries():
+    html = '<div><!-- a --> \n<!-- b --> <p>text</p> <!-- c --> \n<!-- d --></div>'
+    assert md(html) == '\n\ntext\n\n'
+
+
+@pytest.mark.parametrize('html, expected', [
+    ('one <!-- comment --> <b>two</b>', 'one  **two**'),
+    ('one<code> a <!-- comment --> b </code>two', 'one `a  b` two'),
+])
+def test_comment_whitespace_between_inline_content(html, expected):
+    assert md(html) == expected
+
+
+@pytest.mark.parametrize('code', [False, True])
+def test_comment_whitespace_in_pre(code):
+    content = ' a\t\n<!-- comment -->\n\n b '
+    if code:
+        content = '<code>' + content + '</code>'
+    assert md('<pre>' + content + '</pre>', strip_pre=None) == '\n\n```\n a\t\n\n\n b \n```\n\n'
+
+
+def test_comment_whitespace_conversion_preserves_soup():
+    soup = BeautifulSoup('<p>one</p>\n<!-- a -->\n<!-- b -->\n<p>two</p>', 'html.parser')
+    original = str(soup)
+    comments = soup.find_all(string=lambda node: isinstance(node, Comment))
+    positions = [(node.parent, node.previous_sibling, node.next_sibling) for node in comments]
+    converter = MarkdownConverter()
+    assert converter.convert_soup(soup) == 'one\n\ntwo'
+    assert converter.convert_soup(soup) == 'one\n\ntwo'
+    assert str(soup) == original
+    for node, (parent, previous, following) in zip(comments, positions):
+        assert node.parent is parent
+        assert node.previous_sibling is previous
+        assert node.next_sibling is following
+
+
 def test_code_with_tricky_content():
     assert md('<code>></code>') == "`>`"
     assert md('<code>/home/</code><b>username</b>') == "`/home/`**username**"