When I have unicode data to save, it seems that it does not save correctly, giving an encode error. I know this exists on 2.7 and from checking the code in xml/dom/minidom.py it looks like it does in 3.2 as well.

The method call that seem to be problematic is doc.writexml(open(filename, "wb"), "", "  ", "utf-8")

Currently I found this to work: doc.writexml(codecs.open(filename, "w", "utf-8"), "", "  ", "utf-8")

It seems like this should be handled by the writexml method since it already has the specified encoding.

On Python 3 you should not only open file in text mode with specified encoding, but also specify the "xmlcharrefreplace" error handler.

    doc.writexml(open(filename, "w", encoding="utf-8", errors="xmlcharrefreplace"), "", "  ", "utf-8")

I can suggest only one solution -- explicitly document this behavior.

Perhaps we also should add a special module level function for writing DOM tree to binary file. Low-level writexml() should not be used directly.

I am trying to resolve a issue for the first time. Can anybody please tell me or elaborate what is "module level function" specifically in this context.

It means a function defined in the module namespace, as opposed to as a method on a class, so that 'from xml.dom.minidom import <somefunction>' will get you that function.

This issue should be for documentation of the problem, since we won't add the function to 2.7.  A new issue should be opened for the enhancement request of adding a module level convenience function for writing a dom out to a binary file.

> On Python 3 you should not only open file in text mode with specified
> encoding, but also specify the "xmlcharrefreplace" error handler.

Isn't this only required in case there are non encodable characters?
If the encoding is utf-8, this shouldn't be necessary (unless there are lone surrogates).  Specifying xmlcharrefreplace might be useful while using ascii or latin1 though.

The docs of writexml don't seem to specify if the file should be opened in text or binary mode but istm that only text mode is supported.  The advice of using xmlcharrefreplace could be added in a note.

I added a PR for like this:

   .. note::

      You should specify the "xmlcharrefreplace" error handler when open a file with
      specified encoding::

         writer = open(
                filename, "w", encoding="utf-8",
                errors="xmlcharrefreplace")
         doc.writexml(writer, "", "  ", "utf-8")

Asking users unconditionally to use the "xmlcharrefreplace" replacement method seems wrong for UTF-8. It should not be necessary.

We should, however, document explicitly that the file will receive text and not bytes, i.e. that users are themselves responsible for opening the output file with the desired encoding. We should also make it clearer that the "encoding" argument to writexml() does not change that.

New changeset 5ac0b988fd5f1428efe35329c531c7b5c74d37f6 by Stefan Behnel (Windson yang) in branch 'master':
bpo-18911: clarify that the minidom XML writer receives texts but not bytes (GH-13352)
https://github.com/python/cpython/commit/5ac0b988fd5f1428efe35329c531c7b5c74d37f6

New changeset 18e23f227be59241cbb1eeb6d6669771dd7275fb by Stefan Behnel (Miss Islington (bot)) in branch '3.7':
bpo-18911: clarify that the minidom XML writer receives texts but not bytes (GH-13718)
https://github.com/python/cpython/commit/18e23f227be59241cbb1eeb6d6669771dd7275fb

messages: + msg257827
versions: - Python 3.4

components: + Documentation
nosy: + docs@python

messages: + msg196836
versions: + Python 3.3, Python 3.4