Fix indenting CData

The reason of all CData indenting issues was replacing CDATA
tags before indent and then replacing it back. Due to this
CDATA wasn't hanlded by minidom parser correctly.

After this fix, CDATA will remain untouched. However,
indenting CDATA block with writexml() (which is called from
toprettyxml) works wierd - it starts cdata on the beginning
of the next line (without indent) and keeps parent closing
tag on the same line as well. E.g. formatted XML might look
like:

<a>
    <content>
    <![CDATA[hello & goodbye]]> </content>
</a>

To fix this, plugin now does monkey patching of
CDataSection.writexml(). This is just a hack and potentially
destructive action, so warning message is displayed when
patched version is used and this behaviour can be disabled
in settings.

This commit fixes #57, #93, #101

Author: alek-sys

indent_xml.Sublime-settings

@@ -0,0 +1,3 @@
+{
+  "disable_patch_minidom": false
+}

indentxml.py

@@ -98,8 +98,6 @@ def indent(self, s):
         xml_header = re.compile(b"<\?.*\?>").match(s)
         # convert to plain string without indents and spaces
         s = re.compile(b'>\s+([^\s])', re.DOTALL).sub(b'>\g<1>', s)
-        # replace tags to convince minidom process cdata as text
-        s = s.replace(b'<![CDATA[', b'%CDATAESTART%').replace(b']]>', b'%CDATAEEND%')
         try:
             s = parseString(s).toprettyxml()
         except ExpatError as err:
@@ -108,8 +106,6 @@ def indent(self, s):
             return
         # remove line breaks
         s = re.compile('>\n\s+([^<>\s].*?)\n\s+</', re.DOTALL).sub('>\g<1></', s)
-        # restore cdata
-        s = s.replace('%CDATAESTART%', '<![CDATA[').replace('%CDATAEEND%', ']]>')
         # remove xml header
         s = s.replace("<?xml version=\"1.0\" ?>", "").strip()
         if xml_header:

setup.py

@@ -0,0 +1,45 @@
+import sublime
+from xml.dom.minidom import CDATASection
+
+original_writexml = None
+warning_shown = False
+WARNING_MESSAGE = """
+Warning! You are using patched version of xml.dom.minidom.CDataSection.writexml from indent_xml plugin.
+For more details and how to disable please refer to https://github.com/alek-sys/sublimetext_indentxml
+"""
+
+
+def show_warning():
+    global warning_shown
+
+    if not warning_shown:
+        print(WARNING_MESSAGE)
+        warning_shown = True
+
+
+def patch_minidom_cdata():
+    global original_writexml
+
+    def patched_writexml(self, writer, indent="", addindent="", newl=""):
+        show_warning()
+        if self.data.find("]]>") >= 0:
+            raise ValueError("']]>' not allowed in a CDATA section")
+        writer.write("%s<![CDATA[%s]]>%s" % (indent, self.data, newl))
+
+    original_writexml = CDATASection.writexml
+    setattr(CDATASection, "writexml", patched_writexml)
+
+
+def is_minidom_patching_disabled():
+    settings = sublime.load_settings("indent_xml.Sublime-settings")
+    return settings.get('disable_patch_minidom', False)
+
+
+def plugin_loaded():
+    if not is_minidom_patching_disabled():
+        patch_minidom_cdata()
+
+
+def plugin_unloaded():
+    if not is_minidom_patching_disabled() and original_writexml:
+        setattr(CDATASection, "writexml", original_writexml)

tests/fixtures/03_input.xml

@@ -0,0 +1 @@
+<a><content><![CDATA[hello & goodbye]]></content></a>

tests/fixtures/03_output.xml

@@ -0,0 +1,5 @@
+<a>
+	<content>
+		<![CDATA[hello & goodbye]]>
+	</content>
+</a>