gh-81623: Do not add whitespace to significant content in toprettyxml()

Whitespace is no longer added inside an element which is marked with
xml:space="preserve", which is declared in the DTD as not having element
content, or, in absence of such declaration, which contains text.  Previously
such indentation changed the content of the element.

Author: serhiy-storchaka

Doc/library/xml.dom.minidom.rst

@@ -187,13 +187,23 @@ module documentation.  This section lists the differences between the API and
 
    The *standalone* argument behaves exactly as in :meth:`writexml`.
 
+   No indentation is added inside an element
+   which is marked with ``xml:space="preserve"``,
+   which is declared in the DTD as not having element content,
+   or, in absence of such declaration, which contains text,
+   because this would change its content.
+
    .. versionchanged:: 3.8
       The :meth:`toprettyxml` method now preserves the attribute order specified
       by the user.
 
    .. versionchanged:: 3.9
       The *standalone* parameter was added.
 
+   .. versionchanged:: next
+      Whitespace is no longer added inside an element with mixed content
+      or marked with ``xml:space="preserve"``.
+
 .. _dom-example:
 
 DOM Example

Doc/whatsnew/3.16.rst

@@ -642,6 +642,13 @@ xml
   and :meth:`!Document.createEntityReference`.
   (Contributed by Jason Orendorff and Serhiy Storchaka in :gh:`44871`.)
 
+* :meth:`~xml.dom.minidom.Node.toprettyxml` in :mod:`xml.dom.minidom`
+  no longer adds whitespace inside an element
+  which is marked with ``xml:space="preserve"``,
+  which is declared in the DTD as not having element content,
+  or, in absence of such declaration, which contains text.
+  (Contributed by Serhiy Storchaka in :gh:`81623`.)
+
 * Add :meth:`!GetSpecifiedAttributeCount` method
   to the :mod:`XML parser <xml.parsers.expat>` objects.
   It tells how many of the reported attributes were given in the start tag
@@ -881,6 +888,14 @@ that may require changes to your code.
   Attributes defaulted in the DTD are no longer omitted when parsing.
   (Contributed by Jason Orendorff and Serhiy Storchaka in :gh:`44871`.)
 
+* :meth:`~xml.dom.minidom.Node.toprettyxml` in :mod:`xml.dom.minidom`
+  no longer adds whitespace inside an element
+  which is marked with ``xml:space="preserve"``,
+  which is declared in the DTD as not having element content,
+  or, in absence of such declaration, which contains text,
+  because this changed the content of the element.
+  (Contributed by Serhiy Storchaka in :gh:`81623`.)
+
 * On Windows, seeking a pipe now fails instead of silently appearing to
   succeed: :func:`os.lseek` and :meth:`~io.IOBase.seek` raise :exc:`OSError`,
   and :meth:`~io.IOBase.seekable` returns ``False``.  As a consequence,

Lib/test/test_minidom.py

@@ -609,32 +609,76 @@ def testAltNewline(self):
         self.assertEqual(domstr, str.replace("\n", "\r\n"))
 
     def test_toprettyxml_with_text_nodes(self):
-        # see issue #4147, text nodes are not indented
+        # see gh-48397 and gh-81623,
+        # the content of an element with text is not changed
         decl = '<?xml version="1.0" ?>\n'
         self.assertEqual(parseString('<B>A</B>').toprettyxml(),
                          decl + '<B>A</B>\n')
         self.assertEqual(parseString('<C>A<B>A</B></C>').toprettyxml(),
-                         decl + '<C>\n\tA\n\t<B>A</B>\n</C>\n')
+                         decl + '<C>A<B>A</B></C>\n')
         self.assertEqual(parseString('<C><B>A</B>A</C>').toprettyxml(),
-                         decl + '<C>\n\t<B>A</B>\n\tA\n</C>\n')
+                         decl + '<C><B>A</B>A</C>\n')
         self.assertEqual(parseString('<C><B>A</B><B>A</B></C>').toprettyxml(),
                          decl + '<C>\n\t<B>A</B>\n\t<B>A</B>\n</C>\n')
         self.assertEqual(parseString('<C><B>A</B>A<B>A</B></C>').toprettyxml(),
-                         decl + '<C>\n\t<B>A</B>\n\tA\n\t<B>A</B>\n</C>\n')
+                         decl + '<C><B>A</B>A<B>A</B></C>\n')
+        # toprettyxml treats whitespace between elements as insignificant
+        self.assertEqual(parseString('<C> <B>A</B> </C>').toprettyxml(),
+                         decl + '<C>\n\t \n\t<B>A</B>\n\t \n</C>\n')
 
     def test_toprettyxml_with_adjacent_text_nodes(self):
-        # see issue #4147, adjacent text nodes are indented normally
+        # see gh-81623, adjacent text nodes are not separated
         dom = Document()
         elem = dom.createElement('elem')
         elem.appendChild(dom.createTextNode('TEXT'))
         elem.appendChild(dom.createTextNode('TEXT'))
         dom.appendChild(elem)
         decl = '<?xml version="1.0" ?>\n'
-        self.assertEqual(dom.toprettyxml(),
-                         decl + '<elem>\n\tTEXT\n\tTEXT\n</elem>\n')
+        self.assertEqual(dom.toprettyxml(), decl + '<elem>TEXTTEXT</elem>\n')
+
+    def test_toprettyxml_preserve(self):
+        decl = '<?xml version="1.0" ?>\n'
+        # xml:space="preserve" applies to the whole subtree
+        self.assertEqual(
+            parseString('<C xml:space="preserve"><B>A</B><B>A</B></C>'
+                        ).toprettyxml(),
+            decl + '<C xml:space="preserve"><B>A</B><B>A</B></C>\n')
+        self.assertEqual(
+            parseString('<C xml:space="preserve"><B><D/></B></C>'
+                        ).toprettyxml(),
+            decl + '<C xml:space="preserve"><B><D/></B></C>\n')
+        # other values do not preserve whitespace
+        self.assertEqual(
+            parseString('<C xml:space="default"><B>A</B></C>').toprettyxml(),
+            decl + '<C xml:space="default">\n\t<B>A</B>\n</C>\n')
+
+    def test_toprettyxml_with_non_xml_whitespace(self):
+        # only " \t\r\n" are whitespace in XML (see XML 1.0, 2.3)
+        decl = '<?xml version="1.0" ?>\n'
+        self.assertEqual(parseString('<C>\xa0<B>A</B></C>').toprettyxml(),
+                         decl + '<C>\xa0<B>A</B></C>\n')
+
+    def test_toprettyxml_with_dtd(self):
+        decl = '<?xml version="1.0" ?>\n'
+        # only whitespace in element content is ignorable
+        doctype = ('<!DOCTYPE C [<!ELEMENT C (#PCDATA|B)*>'
+                   '<!ELEMENT B (#PCDATA)>]>')
+        self.assertEqual(
+            parseString(doctype + '<C><B>A</B><B>A</B></C>').toprettyxml(),
+            decl + doctype + '\n<C><B>A</B><B>A</B></C>\n')
+        doctype = '<!DOCTYPE C [<!ELEMENT C (B)*><!ELEMENT B (#PCDATA)>]>'
+        self.assertEqual(
+            parseString(doctype + '<C><B>A</B><B>A</B></C>').toprettyxml(),
+            decl + doctype + '\n<C>\n\t<B>A</B>\n\t<B>A</B>\n</C>\n')
+
+    def test_toprettyxml_with_cdata_section(self):
+        decl = '<?xml version="1.0" ?>\n'
+        self.assertEqual(
+            parseString('<C><![CDATA[A]]><B>A</B></C>').toprettyxml(),
+            decl + '<C><![CDATA[A]]><B>A</B></C>\n')
 
     def test_toprettyxml_preserves_content_of_text_node(self):
-        # see issue #4147
+        # see gh-48397
         for str in ('<B>A</B>', '<A><B>C</B></A>'):
             dom = parseString(str)
             dom2 = parseString(dom.toprettyxml())

Lib/xml/dom/minidom.py

@@ -940,6 +940,10 @@ def writexml(self, writer, indent="", addindent="", newl=""):
                 self.childNodes[0].nodeType in (
                         Node.TEXT_NODE, Node.CDATA_SECTION_NODE)):
                 self.childNodes[0].writexml(writer, '', '', '')
+            elif self._preserves_whitespace():
+                # Adding whitespace here would change the content.
+                for node in self.childNodes:
+                    node.writexml(writer, '', '', '')
             else:
                 writer.write(newl)
                 for node in self.childNodes:
@@ -949,6 +953,25 @@ def writexml(self, writer, indent="", addindent="", newl=""):
         else:
             writer.write("/>%s"%(newl))
 
+    def _preserves_whitespace(self):
+        """Returns true iff whitespace in the content is significant.
+
+        This is the case if the element is marked with xml:space="preserve",
+        if the DTD declares that its content model is not element content,
+        or, in absence of such declaration, if it contains text.
+        """
+        if self.getAttribute("xml:space") == "preserve":
+            return True
+        doc = self.ownerDocument
+        info = doc and doc._get_elem_info(self)
+        if info is not None:
+            # Only whitespace in element content is ignorable
+            # (see XML 1.0, 3.2.1).
+            return not info.isElementContent()
+        return any(node.nodeType in (Node.TEXT_NODE, Node.CDATA_SECTION_NODE)
+                   and node.data.strip(_XML_WHITESPACE)
+                   for node in self.childNodes)
+
     def _get_attributes(self):
         self._ensure_attributes()
         return NamedNodeMap(self._attrs, self._attrsNS, self)

Misc/NEWS.d/next/Library/2026-08-30-18-00-00.gh-issue-81623.Vh2Kt6.rst

@@ -0,0 +1,5 @@
+:meth:`~xml.dom.minidom.Node.toprettyxml` in :mod:`xml.dom.minidom` no longer
+adds whitespace inside an element which is marked with
+``xml:space="preserve"``, which is declared in the DTD as not having element
+content, or, in absence of such declaration, which contains text.  Previously
+such indentation changed the content of the element.