OxiXML

The Pure-Rust foundation for all XML and RDF processing in the COOLJAPAN ecosystem.

OxiXML is a single, foundation-pure workspace that natively implements the capabilities of Apache Xerces + Xalan + FOP, Saxon, and Oxigraph — from a byte-level XML scanner up through XSLT 3.0 / XQuery / XSD 1.1 / RELAX NG / Schematron / XMLDSig / XSL-FO, and the complete RDF 1.2 / Turtle / RDF-XML / JSON-LD / RDFa / canonicalization / SPARQL-results stack.

• Pure Rust, 0 external crates in default features. Optional features only (serialize, async-tokio, encoding, simd, crypto). Depends on nothing else in the ecosystem, so any project can rely on it with no dependency cycles.
• Drop-in migration. oxixml-xml mirrors the quick-xml 0.41 API (oxixml-quickxml-compat re-exports it under quick_xml::*); oxixml-model is oxrdf-shaped.
• Conformance-driven. Every parser, validator, and engine is gated on its W3C test suite.

Status: 0.1.2 (0.1.1 released 2026-08-06) — every implementation phase complete (XML core, trees/validation, query/transform, security/formatting, RDF/SPARQL, SVG); migrating the rest of the ecosystem onto OxiXML is a separate effort. 396,895 lines of Rust across 1,114 files in 46 crates; 9,775 tests and 900 doc-tests passing with --all-features (7,121 and 818 with default features), 0 failures. 0.1.2 breaks compatibility with 0.1.1 deliberately (oxrdf/oxttl API parity in oxixml-model) — see CHANGELOG.md. See TODO.md for phase-by-phase detail and the published gap list.

Installation

[dependencies]
oxixml = "0.1.2"

Every crate in the table below is also published independently; oxixml is a facade that re-exports them behind feature flags.

Workspace layout

Tier	Crates	Replaces / implements
Primitives	oxixml-support · -unicode · -encoding · -text · -regex	(internal foundation)
Atoms	oxixml-iri · -langtag · -xsd · -qname	oxiri · oxilangtag · oxsdatatypes
XML core	oxixml-xml · -quickxml-compat	quick-xml 0.41
Trees	oxixml-dom · -sax · -infoset · -html · -svg	Xerces DOM/SAX · XDM · HTML5 · SVG 1.1/2
Validation	oxixml-dtd · -schema · -relaxng · -schematron · -catalog · -xinclude	Xerces XSD · Jing · Schematron
Query/transform	oxixml-fn · -xpath · -xquery · -xslt	Xalan / Saxon
Security/format	oxixml-c14n · -dsig · -enc · -fo	XMLDSig · XMLEnc · Apache FOP
RDF	oxixml-model · -json · -turtle · -rdfxml · -jsonld · -rdfa · -trix · -hdt · -canon · -io · -vocab	the Oxigraph RDF stack
SPARQL	oxixml-sparql-results · -sparql-syntax	sparesults · spargebra/sparopt
Top	oxixml (facade) · oxixml-cli · oxixml-testsuite	—

Execution engines (SPARQL query execution, SHACL validation, FO→PDF rendering, OWL/RDFS reasoning) stay in their host projects (oxirs-arq, oxirs-shacl, fop, oxirs) and depend on OxiXML for parsing, the data model, and validation primitives.

Conformance

Every engine is gated on its W3C test suite. The five largest, measured on the 0.1.1 tree from one binary built once and driven repeatedly:

XSLT 3.0    xslt30-test        7340 / 7420 adjudicated passed,  80 failed,  7181 skipped
                               98.9% of adjudicated cases, over the whole 14601-case catalogue
XPath 3.1   QT3/FOTS          21731 / 21731 passed,   0 failed, 10090 skipped   100.0%
XQuery 3.1  QT3/FOTS          29744 / 29744 passed,   0 failed,  2077 skipped   100.0%
XSD 1.0     xsdtests          39271 / 39309 passed,  38 failed,     11 skipped   99.9%
XSD 1.1     xsdtests          41434 / 41463 passed,  29 failed,     11 skipped   99.9%

Every failure in that table is named in an expected-failure ledger inside the runner, and each run asserts that its failing set is exactly that ledger — in both directions. A case that starts failing is a regression; a ledgered case that stops failing is a stale claim; both turn the run red, which is what separates a ledger from an allowlist. An entry carries one of five classes and its evidence: suite-defect, the case or its expectation is wrong or unsatisfiable, with a specification citation; dependency, a capability this processor declares it does not have (checked against the engine's own system-property() answer) or a resource the run cannot reach; unicode-version, an expectation pinned to a Unicode version this workspace has moved past; performance, a case that does not finish inside the per-case budget — currently empty; and engine-gap, a genuine gap here, each with the mechanism that closes it, tracked as roadmap work in TODO.md. Of the 147 entries across the XSLT 3.0 and XML Schema suites, 111 are the suite's own defect or a declared dependency and 36 are this workspace's backlog. The QT3 drivers hold no entries: they are gated on zero wrong answers rather than zero failures, because a per-case timeout measures the machine while a wrong answer measures the processor.

The remaining suites, as measured for the 0.1.0 release and itemised in CHANGELOG.md, run at 100% of their cases: RDF 1.1/1.2 Turtle · TriG · N-Triples · N-Quads · RDF/XML · RDFC-1.0 · JSON-LD 1.1 · SPARQL 1.0/1.1/1.2 syntax · RELAX NG · html5lib-tests · RDFa 1.1 · Canonical XML 2.0 · XMLDSig interop · xmlconf DTD validation. Two sit just under 100% and say why: W3C XML conformance detects 984 of 988 not-well-formed cases (four documented allowlist entries), and ISO Schematron is 307/307 with 38 suite-defective fixtures documented and skipped.

Every suite is discovered from OXIXML_XML_SUITES_DIR / OXIXML_TESTSUITE_DIR and is skipped silently when unset, so the corpora are never vendored into this repository.

License

Licensed under the Apache License, Version 2.0 (LICENSE or http://www.apache.org/licenses/LICENSE-2.0).