Restore XML conformant logic for escaping < and > characters

[kwokcb]

Changes

Update #2956

From a user point of all run-time strings will still show < and >. The only difference is this enforces writing conformant escaped characters to file. This does not disallow users from using unescaped characters by hand editing.

Implementation

Replace MaterialX specific code for writing unescaped characters < and > with original PugiXML logic.

Tests

• Add unit test for writing escaped characters
• Add test to read back escaped output to make sure that unescaping works properly.

Example

Here is an original UDIM test file:

<?xml version="1.0"?>
<materialx version="1.38">
  <image name="image_color" type="color3">
    <input name="file" type="filename" value="resources/Images/grid_udim/grid.<UDIM>.png" colorspace="srgb_texture" />
  </image>
  ...
</materialx>

Here is the escaped version written out:

<?xml version="1.0"?>
<materialx version="1.39">
  <image name="image_color" type="color3" xpos="1.724638" ypos="0.000000">
    <input name="file" type="filename" value="resources/Images/grid_udim/grid.&lt;UDIM&gt;.png" colorspace="srgb_texture" />
  </image>
  ...
</materialx>

Loading either one in results in the same unescaped user string be stored at run-time and shown the UI. This is using the GraphEditor:

[kwokcb]

Pinging @ashwinbhat, @ld-kerley, and @jstone-lucasfilm for feedback / review.

I'm not sure why < and > was ever modified to not be escaped as from a user perspective the only time you'd notice is if you hand edit a file -- but even here you can write either escaped or non-escaped, and run-time is always unescaped. A standard XML parser will always handle this as does PugiXML before it was modified to have this non-conformant behaviour.

[kwokcb]

This was always the logic as read always unescapes.

[ashwinbhat]

HI @kwokcb having the option is helpful but it does create a situtation where mtlx generated by MaterialX will be non-conformant. As per

MaterialX/documents/Specification/MaterialX.Specification.md

Line 221 in 74d314d

filevalue = "cmsscheme:myassetdiffuse.<UDIM>.tif?ver=current"

UDIM tags are have < and >, should this be updated to mention that these should be encoded for XML?

[kwokcb]

I have put this up to allow maintaining "previous" (undesired) behaviour for discussion.

I would advocate to remove this option and always require encoding for any SGML derivatives (XML, HTML etc).
( HTML5 I believe has looser constraints.)

Hence the spec would need the clarification you note.

[kwokcb]

After discussing with @ashwinbhat offline, I have made this restore PugiXML behaviour for < and > to write conformant XML all the time.