[Xmlunit-general] Validation API for XMLUnit 2

[Stefan B. <bo...@ap...>]

Hi,

Both current XMLUnit implementations support Validation of XML pieces
against XML Schema defitions or DTDs - in the .NET case this only
works if the instance document contains a Schema or DTD reference and
the reference can be resolved AFAICS.

The Java version has two different validator classes, one based on the
"old" validation model where the document is validated as a side
effect of parsing (similar to the way .NET works) and a different one
using javax.xml.validation - which doesn't do DTDs but in theory
supports RelaxNG and other schema languages.

All three of them return booleans to indicate whether an instance is
valid, the .NET code and the old Java validator provide a single
string message for the validation error(s) - the new Java validator
provides a list of exceptions.

The "new" Java validator can validate schema defitions in addition to
schema instances.  I like this even if I don't know how to implement
it for .NET or for DTDs in Java, yet.

The .NET and old Java code use a model of "one validator per instance
document", the new Java code uses one instance per schema definition.

All in all I tend to prefer the "new" Java API as a model for how
XMLUnit2 should do validation, but I'd rather use a ValidationError
class than a SAXParserException as a holder for the validation errors.

So a first sketch - leaving out things like resolvers for now - could
look like (no public/private qualifiers on attributes, just think .NET
properties or Java getters/setters):

class ValidationError {
    readonly int line; // with a negative constant for UNKNOWN
    readonly int column; // ditto

    readonly String message;
}

.NET knows warnings in addition to errors, I'm not sure how we should
handle them.

class ValidationResult {
    readonly bool valid;
    readonly Iterable/IEnumerable<ValidationError> errors;
}

class Validator {
    String schemaURI; // should that be an URI instead? probably wouldn't
                      // work for DTDs then
    Source[] schemaSources;

    ValidationResult validateSchema();
    ValidationResult validateInstance(Source);
}

Since the Java code will be prettier if it uses different
implementations for DTDs and W3C Schema I suggest to add a factory
method

class Validator {
    static Validator forLanguage(String);
}

using the strings defined in javax.xml.XmlConstants for DTD and W3C
schema.  We may also add a static isLanguageSupported method so people
can probe for e.g. RelaxNG support.  .NET would define a string for
XDR and support DTD, W3C schema and XDR, Java would support DTD (by
parsing) and anything SchemaFactory supports.

Am I on track here?

Stefan