More thorough cleanup of input whitespace

markdownify/__init__.py

@@ -7,7 +7,8 @@
 convert_heading_re = re.compile(r'convert_h(\d+)')
 line_beginning_re = re.compile(r'^', re.MULTILINE)
 whitespace_re = re.compile(r'[\t ]+')
-all_whitespace_re = re.compile(r'[\s]+')
+all_whitespace_re = re.compile(r'[\t \r\n]+')
+newline_whitespace_re = re.compile(r'[\t \r\n]*[\r\n][\t \r\n]*')
 html_heading_re = re.compile(r'h[1-6]')
 
 
@@ -66,6 +67,23 @@ def _todict(obj):
     return dict((k, getattr(obj, k)) for k in dir(obj) if not k.startswith('_'))
 
 
+def remove_whitespace_inside(el):
+    """Return to remove whitespace immediately inside a block-level element."""
+    if not el or not el.name:
+        return False
+    if html_heading_re.match(el.name) is not None:
+        return True
+    return el.name in ('p', 'blockquote',
+                       'ol', 'ul', 'li',
+                       'table', 'thead', 'tbody', 'tfoot',
+                       'tr', 'td', 'th')
+
+
+def remove_whitespace_outside(el):
+    """Return to remove whitespace immediately outside a block-level element."""
+    return remove_whitespace_inside(el) or (el and el.name == 'pre')
+
+
 class MarkdownConverter(object):
     class DefaultOptions:
         autolinks = True
@@ -119,27 +137,23 @@ def process_tag(self, node, convert_as_inline, children_only=False):
         if not children_only and (isHeading or isCell):
             convert_children_as_inline = True
 
-        # Remove whitespace-only textnodes in purely nested nodes
-        def is_nested_node(el):
-            return el and el.name in ['ol', 'ul', 'li',
-                                      'table', 'thead', 'tbody', 'tfoot',
-                                      'tr', 'td', 'th']
-
-        if is_nested_node(node):
-            for el in node.children:
-                # Only extract (remove) whitespace-only text node if any of the
-                # conditions is true:
-                # - el is the first element in its parent
-                # - el is the last element in its parent
-                # - el is adjacent to an nested node
-                can_extract = (not el.previous_sibling
-                               or not el.next_sibling
-                               or is_nested_node(el.previous_sibling)
-                               or is_nested_node(el.next_sibling))
-                if (isinstance(el, NavigableString)
-                        and six.text_type(el).strip() == ''
-                        and can_extract):
-                    el.extract()
+        # Remove whitespace-only textnodes just before, after or
+        # inside block-level elements.
+        remove_inside = remove_whitespace_inside(node)
+        for el in node.children:
+            # Only extract (remove) whitespace-only text node if any of the
+            # conditions is true:
+            # - el is the first element in its parent (block-level)
+            # - el is the last element in its parent (block-level)
+            # - el is adjacent to a block-level node
+            can_extract = (remove_inside and (not el.previous_sibling
+                                              or not el.next_sibling)
+                           or remove_whitespace_outside(el.previous_sibling)
+                           or remove_whitespace_outside(el.next_sibling))
+            if (isinstance(el, NavigableString)
+                    and six.text_type(el).strip() == ''
+                    and can_extract):
+                el.extract()
 
         # Convert the children first
         for el in node.children:
@@ -148,7 +162,13 @@ def is_nested_node(el):
             elif isinstance(el, NavigableString):
                 text += self.process_text(el)
             else:
-                text += self.process_tag(el, convert_children_as_inline)
+                text_strip = text.rstrip('\n')
+                newlines_left = len(text) - len(text_strip)
+                next_text = self.process_tag(el, convert_children_as_inline)
+                next_text_strip = next_text.lstrip('\n')
+                newlines_right = len(next_text) - len(next_text_strip)
+                newlines = '\n' * max(newlines_left, newlines_right)
+                text = text_strip + newlines + next_text_strip
 
         if not children_only:
             convert_fn = getattr(self, 'convert_%s' % node.name, None)
@@ -162,18 +182,26 @@ def process_text(self, el):
 
         # normalize whitespace if we're not inside a preformatted element
         if not el.find_parent('pre'):
-            text = whitespace_re.sub(' ', text)
+            if self.options['wrap']:
+                text = all_whitespace_re.sub(' ', text)
+            else:
+                text = newline_whitespace_re.sub('\n', text)
+                text = whitespace_re.sub(' ', text)
 
         # escape special characters if we're not inside a preformatted or code element
         if not el.find_parent(['pre', 'code', 'kbd', 'samp']):
             text = self.escape(text)
 
-        # remove trailing whitespaces if any of the following condition is true:
-        # - current text node is the last node in li
-        # - current text node is followed by an embedded list
-        if (el.parent.name == 'li'
-                and (not el.next_sibling
-                     or el.next_sibling.name in ['ul', 'ol'])):
+        # remove leading whitespace at the start or just after a
+        # block-level element; remove traliing whitespace at the end
+        # or just before a block-level element.
+        if (remove_whitespace_outside(el.previous_sibling)
+                or (remove_whitespace_inside(el.parent)
+                    and not el.previous_sibling)):
+            text = text.lstrip()
+        if (remove_whitespace_outside(el.next_sibling)
+                or (remove_whitespace_inside(el.parent)
+                    and not el.next_sibling)):
             text = text.rstrip()
 
         return text
@@ -221,7 +249,7 @@ def indent(self, text, level):
 
     def underline(self, text, pad_char):
         text = (text or '').rstrip()
-        return '%s\n%s\n\n' % (text, pad_char * len(text)) if text else ''
+        return '\n\n%s\n%s\n\n' % (text, pad_char * len(text)) if text else ''
 
     def convert_a(self, el, text, convert_as_inline):
         prefix, suffix, text = chomp(text)
@@ -246,7 +274,7 @@ def convert_a(self, el, text, convert_as_inline):
     def convert_blockquote(self, el, text, convert_as_inline):
 
         if convert_as_inline:
-            return text
+            return ' ' + text.strip() + ' '
 
         return '\n' + (line_beginning_re.sub('> ', text.strip()) + '\n\n') if text else ''
 
@@ -280,10 +308,11 @@ def convert_hn(self, n, el, text, convert_as_inline):
         if style == UNDERLINED and n <= 2:
             line = '=' if n == 1 else '-'
             return self.underline(text, line)
+        text = all_whitespace_re.sub(' ', text)
         hashes = '#' * n
         if style == ATX_CLOSED:
-            return '%s %s %s\n\n' % (hashes, text, hashes)
-        return '%s %s\n\n' % (hashes, text)
+            return '\n%s %s %s\n\n' % (hashes, text, hashes)
+        return '\n%s %s\n\n' % (hashes, text)
 
     def convert_hr(self, el, text, convert_as_inline):
         return '\n\n---\n\n'
@@ -318,7 +347,7 @@ def convert_list(self, el, text, convert_as_inline):
         if nested:
             # remove trailing newline if nested
             return '\n' + self.indent(text, 1).rstrip()
-        return text + ('\n' if before_paragraph else '')
+        return '\n\n' + text + ('\n' if before_paragraph else '')
 
     convert_ul = convert_list
     convert_ol = convert_list
@@ -343,13 +372,24 @@ def convert_li(self, el, text, convert_as_inline):
 
     def convert_p(self, el, text, convert_as_inline):
         if convert_as_inline:
-            return text
+            return ' ' + text.strip() + ' '
         if self.options['wrap']:
-            text = fill(text,
-                        width=self.options['wrap_width'],
-                        break_long_words=False,
-                        break_on_hyphens=False)
-        return '%s\n\n' % text if text else ''
+            # Preserve newlines (and preceding whitespace) resulting
+            # from <br> tags.  Newlines in the input have already been
+            # replaced by spaces.
+            lines = text.split('\n')
+            new_lines = []
+            for line in lines:
+                line = line.lstrip()
+                line_no_trailing = line.rstrip()
+                trailing = line[len(line_no_trailing):]
+                line = fill(line,
+                            width=self.options['wrap_width'],
+                            break_long_words=False,
+                            break_on_hyphens=False)
+                new_lines.append(line + trailing)
+            text = '\n'.join(new_lines)
+        return '\n\n%s\n\n' % text if text else ''
 
     def convert_pre(self, el, text, convert_as_inline):
         if not text:

tests/test_advanced.py

@@ -14,7 +14,7 @@ def test_chomp():
 
 def test_nested():
     text = md('<p>This is an <a href="http://example.com/">example link</a>.</p>')
-    assert text == 'This is an [example link](http://example.com/).\n\n'
+    assert text == '\n\nThis is an [example link](http://example.com/).\n\n'
 
 
 def test_ignore_comments():

tests/test_basic.py

@@ -11,3 +11,4 @@ def test_soup():
 
 def test_whitespace():
     assert md(' a  b \t\t c ') == ' a b c '
+    assert md(' a  b \n\n c ') == ' a b\nc '

tests/test_conversions.py

@@ -1,4 +1,4 @@
-from markdownify import markdownify as md, ATX, ATX_CLOSED, BACKSLASH, UNDERSCORE
+from markdownify import markdownify as md, ATX, ATX_CLOSED, BACKSLASH, SPACES, UNDERSCORE
 
 
 def inline_tests(tag, markup):
@@ -66,7 +66,7 @@ def test_blockquote_with_paragraph():
 
 def test_blockquote_nested():
     text = md('<blockquote>And she was like <blockquote>Hello</blockquote></blockquote>')
-    assert text == '\n> And she was like \n> > Hello\n\n'
+    assert text == '\n> And she was like\n> > Hello\n\n'
 
 
 def test_br():
@@ -112,36 +112,39 @@ def test_em():
 
 
 def test_header_with_space():
-    assert md('<h3>\n\nHello</h3>') == '### Hello\n\n'
-    assert md('<h4>\n\nHello</h4>') == '#### Hello\n\n'
-    assert md('<h5>\n\nHello</h5>') == '##### Hello\n\n'
-    assert md('<h5>\n\nHello\n\n</h5>') == '##### Hello\n\n'
-    assert md('<h5>\n\nHello   \n\n</h5>') == '##### Hello\n\n'
+    assert md('<h3>\n\nHello</h3>') == '\n### Hello\n\n'
+    assert md('<h3>Hello\n\n\nWorld</h3>') == '\n### Hello World\n\n'
+    assert md('<h4>\n\nHello</h4>') == '\n#### Hello\n\n'
+    assert md('<h5>\n\nHello</h5>') == '\n##### Hello\n\n'
+    assert md('<h5>\n\nHello\n\n</h5>') == '\n##### Hello\n\n'
+    assert md('<h5>\n\nHello   \n\n</h5>') == '\n##### Hello\n\n'
 
 
 def test_h1():
-    assert md('<h1>Hello</h1>') == 'Hello\n=====\n\n'
+    assert md('<h1>Hello</h1>') == '\n\nHello\n=====\n\n'
 
 
 def test_h2():
-    assert md('<h2>Hello</h2>') == 'Hello\n-----\n\n'
+    assert md('<h2>Hello</h2>') == '\n\nHello\n-----\n\n'
 
 
 def test_hn():
-    assert md('<h3>Hello</h3>') == '### Hello\n\n'
-    assert md('<h4>Hello</h4>') == '#### Hello\n\n'
-    assert md('<h5>Hello</h5>') == '##### Hello\n\n'
-    assert md('<h6>Hello</h6>') == '###### Hello\n\n'
+    assert md('<h3>Hello</h3>') == '\n### Hello\n\n'
+    assert md('<h4>Hello</h4>') == '\n#### Hello\n\n'
+    assert md('<h5>Hello</h5>') == '\n##### Hello\n\n'
+    assert md('<h6>Hello</h6>') == '\n###### Hello\n\n'
 
 
 def test_hn_chained():
-    assert md('<h1>First</h1>\n<h2>Second</h2>\n<h3>Third</h3>', heading_style=ATX) == '# First\n\n\n## Second\n\n\n### Third\n\n'
-    assert md('X<h1>First</h1>', heading_style=ATX) == 'X# First\n\n'
+    assert md('<h1>First</h1>\n<h2>Second</h2>\n<h3>Third</h3>', heading_style=ATX) == '\n# First\n\n## Second\n\n### Third\n\n'
+    assert md('X<h1>First</h1>', heading_style=ATX) == 'X\n# First\n\n'
+    assert md('X<h1>First</h1>', heading_style=ATX_CLOSED) == 'X\n# First #\n\n'
+    assert md('X<h1>First</h1>') == 'X\n\nFirst\n=====\n\n'
 
 
 def test_hn_nested_tag_heading_style():
-    assert md('<h1>A <p>P</p> C </h1>', heading_style=ATX_CLOSED) == '# A P C #\n\n'
-    assert md('<h1>A <p>P</p> C </h1>', heading_style=ATX) == '# A P C\n\n'
+    assert md('<h1>A <p>P</p> C </h1>', heading_style=ATX_CLOSED) == '\n# A P C #\n\n'
+    assert md('<h1>A <p>P</p> C </h1>', heading_style=ATX) == '\n# A P C\n\n'
 
 
 def test_hn_nested_simple_tag():
@@ -157,12 +160,12 @@ def test_hn_nested_simple_tag():
     ]
 
     for tag, markdown in tag_to_markdown:
-        assert md('<h3>A <' + tag + '>' + tag + '</' + tag + '> B</h3>') == '### A ' + markdown + ' B\n\n'
+        assert md('<h3>A <' + tag + '>' + tag + '</' + tag + '> B</h3>') == '\n### A ' + markdown + ' B\n\n'
 
-    assert md('<h3>A <br>B</h3>', heading_style=ATX) == '### A B\n\n'
+    assert md('<h3>A <br>B</h3>', heading_style=ATX) == '\n### A B\n\n'
 
     # Nested lists not supported
-    # assert md('<h3>A <ul><li>li1</i><li>l2</li></ul></h3>', heading_style=ATX) == '### A li1 li2 B\n\n'
+    # assert md('<h3>A <ul><li>li1</i><li>l2</li></ul></h3>', heading_style=ATX) == '\n### A li1 li2 B\n\n'
 
 
 def test_hn_nested_img():
@@ -172,18 +175,18 @@ def test_hn_nested_img():
         ("alt='Alt Text' title='Optional title'", "Alt Text", " \"Optional title\""),
     ]
     for image_attributes, markdown, title in image_attributes_to_markdown:
-        assert md('<h3>A <img src="/path/to/img.jpg" ' + image_attributes + '/> B</h3>') == '### A ' + markdown + ' B\n\n'
-        assert md('<h3>A <img src="/path/to/img.jpg" ' + image_attributes + '/> B</h3>', keep_inline_images_in=['h3']) == '### A ![' + markdown + '](/path/to/img.jpg' + title + ') B\n\n'
+        assert md('<h3>A <img src="/path/to/img.jpg" ' + image_attributes + '/> B</h3>') == '\n### A' + (' ' + markdown + ' ' if markdown else ' ') + 'B\n\n'
+        assert md('<h3>A <img src="/path/to/img.jpg" ' + image_attributes + '/> B</h3>', keep_inline_images_in=['h3']) == '\n### A ![' + markdown + '](/path/to/img.jpg' + title + ') B\n\n'
 
 
 def test_hn_atx_headings():
-    assert md('<h1>Hello</h1>', heading_style=ATX) == '# Hello\n\n'
-    assert md('<h2>Hello</h2>', heading_style=ATX) == '## Hello\n\n'
+    assert md('<h1>Hello</h1>', heading_style=ATX) == '\n# Hello\n\n'
+    assert md('<h2>Hello</h2>', heading_style=ATX) == '\n## Hello\n\n'
 
 
 def test_hn_atx_closed_headings():
-    assert md('<h1>Hello</h1>', heading_style=ATX_CLOSED) == '# Hello #\n\n'
-    assert md('<h2>Hello</h2>', heading_style=ATX_CLOSED) == '## Hello ##\n\n'
+    assert md('<h1>Hello</h1>', heading_style=ATX_CLOSED) == '\n# Hello #\n\n'
+    assert md('<h2>Hello</h2>', heading_style=ATX_CLOSED) == '\n## Hello ##\n\n'
 
 
 def test_head():
@@ -193,7 +196,7 @@ def test_head():
 def test_hr():
     assert md('Hello<hr>World') == 'Hello\n\n---\n\nWorld'
     assert md('Hello<hr />World') == 'Hello\n\n---\n\nWorld'
-    assert md('<p>Hello</p>\n<hr>\n<p>World</p>') == 'Hello\n\n\n\n\n---\n\n\nWorld\n\n'
+    assert md('<p>Hello</p>\n<hr>\n<p>World</p>') == '\n\nHello\n\n---\n\nWorld\n\n'
 
 
 def test_i():
@@ -210,12 +213,23 @@ def test_kbd():
 
 
 def test_p():
-    assert md('<p>hello</p>') == 'hello\n\n'
-    assert md('<p>123456789 123456789</p>') == '123456789 123456789\n\n'
-    assert md('<p>123456789 123456789</p>', wrap=True, wrap_width=10) == '123456789\n123456789\n\n'
-    assert md('<p><a href="https://example.com">Some long link</a></p>', wrap=True, wrap_width=10) == '[Some long\nlink](https://example.com)\n\n'
-    assert md('<p>12345<br />67890</p>', wrap=True, wrap_width=10, newline_style=BACKSLASH) == '12345\\\n67890\n\n'
-    assert md('<p>12345678901<br />12345</p>', wrap=True, wrap_width=10, newline_style=BACKSLASH) == '12345678901\\\n12345\n\n'
+    assert md('<p>hello</p>') == '\n\nhello\n\n'
+    assert md('<p>123456789 123456789</p>') == '\n\n123456789 123456789\n\n'
+    assert md('<p>123456789\n\n\n123456789</p>') == '\n\n123456789\n123456789\n\n'
+    assert md('<p>123456789\n\n\n123456789</p>', wrap=True, wrap_width=80) == '\n\n123456789 123456789\n\n'
+    assert md('<p>123456789 123456789</p>', wrap=True, wrap_width=10) == '\n\n123456789\n123456789\n\n'
+    assert md('<p><a href="https://example.com">Some long link</a></p>', wrap=True, wrap_width=10) == '\n\n[Some long\nlink](https://example.com)\n\n'
+    assert md('<p>12345<br />67890</p>', wrap=True, wrap_width=10, newline_style=BACKSLASH) == '\n\n12345\\\n67890\n\n'
+    assert md('<p>12345<br />67890</p>', wrap=True, wrap_width=50, newline_style=BACKSLASH) == '\n\n12345\\\n67890\n\n'
+    assert md('<p>12345<br />67890</p>', wrap=True, wrap_width=10, newline_style=SPACES) == '\n\n12345  \n67890\n\n'
+    assert md('<p>12345<br />67890</p>', wrap=True, wrap_width=50, newline_style=SPACES) == '\n\n12345  \n67890\n\n'
+    assert md('<p>12345678901<br />12345</p>', wrap=True, wrap_width=10, newline_style=BACKSLASH) == '\n\n12345678901\\\n12345\n\n'
+    assert md('<p>12345678901<br />12345</p>', wrap=True, wrap_width=50, newline_style=BACKSLASH) == '\n\n12345678901\\\n12345\n\n'
+    assert md('<p>12345678901<br />12345</p>', wrap=True, wrap_width=10, newline_style=SPACES) == '\n\n12345678901  \n12345\n\n'
+    assert md('<p>12345678901<br />12345</p>', wrap=True, wrap_width=50, newline_style=SPACES) == '\n\n12345678901  \n12345\n\n'
+    assert md('<p>1234 5678 9012<br />67890</p>', wrap=True, wrap_width=10, newline_style=BACKSLASH) == '\n\n1234 5678\n9012\\\n67890\n\n'
+    assert md('<p>1234 5678 9012<br />67890</p>', wrap=True, wrap_width=10, newline_style=SPACES) == '\n\n1234 5678\n9012  \n67890\n\n'
+    assert md('First<p>Second</p><p>Third</p>Fourth') == 'First\n\nSecond\n\nThird\n\nFourth'
 
 
 def test_pre():
@@ -289,3 +303,13 @@ def callback(el):
     assert md('<pre class="python">test\n    foo\nbar</pre>', code_language_callback=callback) == '\n```python\ntest\n    foo\nbar\n```\n'
     assert md('<pre class="javascript"><code>test\n    foo\nbar</code></pre>', code_language_callback=callback) == '\n```javascript\ntest\n    foo\nbar\n```\n'
     assert md('<pre class="javascript"><code class="javascript">test\n    foo\nbar</code></pre>', code_language_callback=callback) == '\n```javascript\ntest\n    foo\nbar\n```\n'
+
+
+def test_spaces():
+    assert md('<p> a b </p> <p> c d </p>') == '\n\na b\n\nc d\n\n'
+    assert md('<p> <i>a</i> </p>') == '\n\n*a*\n\n'
+    assert md('test <p> again </p>') == 'test\n\nagain\n\n'
+    assert md('test <blockquote> text </blockquote> after') == 'test\n> text\n\nafter'
+    assert md(' <ol> <li> x </li> <li> y </li> </ol> ') == '\n\n1. x\n2. y\n'
+    assert md(' <ul> <li> x </li> <li> y </li> </ol> ') == '\n\n* x\n* y\n'
+    assert md('test <pre> foo </pre> bar') == 'test\n```\n foo \n```\nbar'