xml-adapters

relax-ng, xpath-eval and schematron-engine take documents through traits and parse none themselves. This crate implements those traits for roxmltree, so a host that has XML text validates it against a RELAX NG and a Schematron schema without writing the glue first. It sits in L1 because it knows the L0 traits; the L0 crates do not know it (rule “no host format in the core”), apart from a dev-dependency of xpath-eval’s tests.

Status

0.1.0, not yet published. The first version of a new crate has to be published by hand (see Releases). Only roxmltree; other trees (xmltree, quick-xml DOMs) are open under casoon/barrierlab#75.

Structure

Public surface

• RngElement (feature relax-ng): relax_ng::Element over a roxmltree element. Text is merged across comments and processing instructions, xmlns declarations are not attributes, bindings in scope (inherited too) serve QName values.
• XmlDoc, XmlNode (feature xpath): an arena copy of the tree in document order implementing xpath_eval::Document and Node. Namespace axis with the implicit xml node, attribute nodes, comments, processing instructions, empty namespace URI as “none”, xml:id as ID attribute. XmlDoc::parse(&str) or XmlDoc::from_roxmltree(&Document) for other parser options.
• Position: line and column from 1, byte offset from 0. It is the Location of RngElement (so a relax_ng::ValidationError carries it) and XmlNode::position() (so a Schematron Report::node has it).

Limits

• roxmltree exposes no DTD attribute types, so id() finds xml:id only.
• Positions are those of the start tag’s <; text, attribute and comment nodes have none.
• xpath-eval’s own tests use XmlDoc through a path-only dev-dependency that loops back to xpath-eval. Cargo allows this for dev-dependencies and drops the entry when packaging; the packaged tests of xpath-eval therefore do not build outside this repository.

Verification

crates/xml-adapters/tests/validate.rs: one document against an RNG and a Schematron schema (valid, invalid, positions), the namespace axis, comments, processing instructions, xml:id and document order, plus the README example as a doctest. The Jaxen and regression suites of xpath-eval run over this adapter.