convert does not respect encoding

[DanielHabenicht]

Describe the bug
Convert should recognise "escaped" characters and parse them correctly.

To Reproduce

var xmlbuilder2 = require("xmlbuilder2")

const xmlBase = {
  xliff: {
    "@version": "1.2",
    file: {
      "@datatype": "plaintext",
      "@source-language": "en",
      body: {
      "@id": "test",
      source: "<this> test </this>"
      },
    },
  },
};

src = xmlbuilder2.convert(xmlBase, { prettyPrint: true })
console.log(src)
// "<?xml version=\"1.0\"?>\n<xliff version=\"1.2\">\n  <file … &lt;/this&gt;</source>\n    </body>\n  </file>\n</xliff>"

xmlbuilder2.convert(src, {format: "object"})

Expected behaviour
Convert does encode special characters like < or > as < and > But does not decode them if parsed again.

Version:

• node.js: 15.12.0
• xmlbuilder2 2.4.0

[DanielHabenicht]

Workaround: Use CDATA e.g.:

var xmlbuilder2 = require("xmlbuilder2")

const xmlBase = {
  xliff: {
    "@version": "1.2",
    file: {
      "@datatype": "plaintext",
      "@source-language": "en",
      body: {
      "@id": "test",
      $: "<this> test </this>"
      },
    },
  },
};

src = xmlbuilder2.convert(xmlBase, { prettyPrint: true })
console.log(src)
// "<?xml version=\"1.0\"?>\n<xliff version=\"1.2\">\n  <file … &lt;/this&gt;</source>\n    </body>\n  </file>\n</xliff>"

xmlbuilder2.convert(src, {format: "object"})

[kernwig]

I ran into this too. Being able to parse back its own output is a basic test. This isn't just failing to decode XML entities, it's actually encoding them again (thus escaping ' to ' instead of to ').

CDATA is not only unnecessary, it's impossible when you don't control the source.