XML writer

[ianprime0509]

There should be some way to write XML content. There are several potential APIs that could be used here: perhaps the most "elegant" API would just be a write function which accepts a reader Event, but this may come with its own issues and needs to be thought through a bit more.

Desired features:

• [ ] Basic functionality: must be able to produce an output for every well-formed XML document which is equivalent (under canonicalization) to the original document.
  • The "equivalent under canonicalization" condition here means that things like choosing the attribute value quote character, CDATA sections, etc. are not considered part of "basic functionality".
• [ ] Configurable "safety" validations
  • The writer should have a mode of operation (probably the default) which checks to make sure the user doesn't cause it to output non-well-formed XML. For example, outputting events out of sequence (e.g. attribute value in the middle of element content), writing characters which cannot be escaped in a given context (e.g. -- which cannot appear in the content of a comment, and there's no way to escape it), etc.
• [ ] Pretty printing
  • This is difficult in general, especially since a lot of the whitespace which many "XML formatters" mangle is actually considered significant in the XML spec.
• [ ] Custom encodings (not just UTF-8)

[ianprime0509]

The "elegant" API of a single writeEvent function turns out to be terrible to use in certain cases, such as when writing the start of an element with attributes which may or may not be included based on runtime conditions. It can still exist, perhaps, but there should be another API which is more flexible with cases like this.