Serialization and optionally escaped characters

[Sat-Ing]

Some characters like '>' are optionally escaped as of the W3 XML standard.
Some other tools are not as flexible as EMF on this part and cannot handle unescaped '>' characters.
I had a look at the export options, unable to find anything which would enable escaping more characters.

Have I overlooked any option? If not, I think that having this feature would be an enrichment to EMF.

[merks]

What is an example of such an inflexible tool that cannot process well-formed XML? (I've never heard of such a thing.)

Escaping is done by org.eclipse.emf.ecore.xmi.impl.XMLSaveImpl.Escape.convert(String)/convertText(String) so technically by specializing org.eclipse.emf.ecore.xmi.impl.XMLSaveImpl.init(XMLResource, Map) you could create a subclass that behaves differently.

[Sat-Ing]

In our case, it is a heritage software for space flight control. Sometimes, these can be quite hard to get changed.

Regarding the workaround: replacing the Escape instance in XMLSaveImpl is probably the easiest way to go? For example like so (note1: edited, because protected fields in Escape class seem to not be accessible in other osgi package
note2: copied some code from XMLSaveImpl.init to initialize Escape class):

package org.eclipse.emf.ecore.xmi.impl;

import java.util.Map;

import org.eclipse.emf.ecore.resource.Resource;
import org.eclipse.emf.ecore.xmi.XMLHelper;
import org.eclipse.emf.ecore.xmi.XMLResource;

public class CustomXMLSaveImpl extends XMLSaveImpl {

    public CustomXMLSaveImpl(final XMLHelper helper) {
        super(helper);
    }

    public CustomXMLSaveImpl(final Map<?, ?> options, final XMLHelper helper, final String encoding, final String xmlVersion) {
        super(options, helper, encoding, xmlVersion);
    }

    protected void init(XMLResource resource, Map<?, ?> options) {
        super.init(resource, options);
        overrideEscape(options);
    }

    /**
     * Replace the Escape instance with our custom version.
     */
    protected void overrideEscape(Map<?, ?> options) {
        if (this.escape == null) {
            return;
        }
        MyEscape myEscape = new MyEscape(options, encoding, xmlVersion);
        this.escape = myEscape;
    }

    protected static class MyEscape extends Escape {
        private static final int MAX_UTF_MAPPABLE_CODEPOINT = 0x10FFFF;
        private static final int MAX_LATIN1_MAPPABLE_CODEPOINT = 0xFF;
        private static final int MAX_ASCII_MAPPABLE_CODEPOINT = 0x7F;

        public MyEscape(Map<?, ?> options, String encoding, String xmlVersion) {
            String lineSeparator = (String) options.get(Resource.OPTION_LINE_DELIMITER);
            setLineFeed(lineSeparator);
            int maxSafeChar = MAX_UTF_MAPPABLE_CODEPOINT;
            if (encoding != null) {
                if (encoding.equalsIgnoreCase("ASCII") || encoding.equalsIgnoreCase("US-ASCII")) {
                    maxSafeChar = MAX_ASCII_MAPPABLE_CODEPOINT;
                } else if (encoding.equalsIgnoreCase("ISO-8859-1")) {
                    maxSafeChar = MAX_LATIN1_MAPPABLE_CODEPOINT;
                }
            }

            setMappingLimit(maxSafeChar);
            if (!"1.0".equals(xmlVersion)) {
                setAllowControlCharacters(true);
            }

            setUseCDATA(Boolean.TRUE.equals(options.get(XMLResource.OPTION_ESCAPE_USING_CDATA)));
        }

        @Override
        public String convertText(final String input) {
            String converted = super.convertText(input);
            return converted.replace(">", "&gt;");
        }

        @Override
        public String convert(final String input) {
            String converted = super.convertText(input);
            return converted.replace(">", "&gt;");
        }
    }
}

[merks]

Yes, that looks correct. Quite a bit of duplication though isn't it. 😢

You really only need this for convertText and not also for convert (as used to produce attribute values)?

[Sat-Ing]

Thanks for crosschecking! In our case, we do not need it, but for others, it may very well be important. I changed it in the example.

[merks]

Yes, that looks complete.