v0.65.0: toRdf — drop literals with malformed language tags (+1 toRdf)

- A language-tagged literal whose @language is not a well-formed BCP47 tag
  (e.g. "a b", containing a space) is now dropped: the statement carries no
  valid RDF language, instead of emitting an invalid N-Quads literal (#twf05,
  §7.3). valueToLiteral returns null for the malformed case and objectToRdf
  propagates it so the quad is skipped.

W3C (toEqual gate): toRdf 453->454 (+1); expand 378, compact 246 (100%)
unchanged. Unit 282 green; phpstan max clean; pint clean; 0 regressions.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: skydudie

CHANGELOG.md

@@ -7,6 +7,25 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.65.0] - 2026-06-10
+
+toRdf: drop literals with malformed language tags.
+
+W3C JSON-LD 1.1 test suite (corrected `toEqual` gate):
+
+```
+            expand   compact   toRdf
+v0.64.0:      378      246      453
+v0.65.0:      378      246      454   (+1 toRdf)
+```
+
+### Fixed (toRdf, §7.3)
+
+- A language-tagged literal whose `@language` is not a well-formed BCP47
+  language tag (e.g. `"a b"`, which contains a space) is now dropped — the
+  statement carries no valid RDF language — instead of emitting an invalid
+  N-Quads literal (`#twf05`).
+
 ## [0.64.0] - 2026-06-10
 
 toRdf harness: sound blank-node isomorphism comparison.

README.md

@@ -31,7 +31,7 @@ A PHP implementation of the [JSON-LD 1.1](https://www.w3.org/TR/json-ld11/) spec
 - [x] Custom `DocumentLoader` interface
 - [~] Expansion (§5.5) — implemented; ~378/385 of the W3C expand suite (the remaining 7 are environment/spec-accommodation blockers)
 - [x] Compaction (§5.6) — **246/246 of the W3C compact suite (100%)**: container-maps (incl. property-valued `@index`, `@type`-map node refs), keyword recursion, `@nest`, nested `@list`, `@graph` maps, `@reverse` (incl. per-value term selection and containers), full inverse-context term scoring (list/direction-aware, `@json` literals), property-/type-scoped contexts with scoped `@base`/`@vocab`, IRI relativisation, expand-first normalisation, the `compactArrays` option, §5.7 prefix rules and error conditions
-- [~] Serialize JSON-LD to RDF (§7 / `toRdf`) — implemented; ~453/467 of the W3C toRdf suite (N-Quads output incl. `rdfDirection` and `produceGeneralizedRdf`)
+- [~] Serialize JSON-LD to RDF (§7 / `toRdf`) — implemented; ~454/467 of the W3C toRdf suite (N-Quads output incl. `rdfDirection` and `produceGeneralizedRdf`)
 
 > Suite numbers from v0.42.0 use the spec-accurate `toEqual` comparison
 > (object-key order insignificant, array order significant); earlier numbers

src/Algorithms/ToRdf.php

@@ -193,8 +193,10 @@ private function listToRdf(array $list, array &$listQuads, BlankNodeIssuer $issu
      * @param  list<RdfQuad>  $listQuads  Receives any auxiliary triples (used
      *                                    by the compound-literal rdfDirection
      *                                    mode).
+     * @return RdfTerm|null Null when the value carries a malformed language
+     *                      tag and the statement must be dropped (#twf05).
      */
-    private function valueToLiteral(array $item, array &$listQuads, BlankNodeIssuer $issuer): RdfTerm
+    private function valueToLiteral(array $item, array &$listQuads, BlankNodeIssuer $issuer): ?RdfTerm
     {
         $value = $item[Keyword::Value->value];
 
@@ -268,6 +270,13 @@ private function valueToLiteral(array $item, array &$listQuads, BlankNodeIssuer
         // String value.
         $stringValue = is_scalar($value) ? (string) $value : '';
 
+        // A language-tagged literal whose tag is not a well-formed BCP47
+        // language tag carries no valid RDF language: the statement is dropped
+        // (#twf05). A well-formed tag is ALPHA{1,8} (-(ALPHANUM){1,8})*.
+        if ($language !== null && preg_match('/^[a-zA-Z]{1,8}(-[a-zA-Z0-9]{1,8})*$/', $language) !== 1) {
+            return null;
+        }
+
         return RdfTerm::literal($stringValue, $datatype, $language);
     }
 

tests/W3c/README.md

@@ -50,9 +50,9 @@ composer test:all
 ## Current score
 
 ```
-Expansion:    378 passed /   7 failed (v0.64.0)
-Compaction:   246 passed /   0 failed (v0.64.0)  ← 100%
-toRdf:        453 passed /  14 failed (v0.64.0)
+Expansion:    378 passed /   7 failed (v0.65.0)
+Compaction:   246 passed /   0 failed (v0.65.0)  ← 100%
+toRdf:        454 passed /  13 failed (v0.65.0)
 ```
 
 > v0.42.0 corrected the expand/compact comparison from `toEqualCanonicalizing`
@@ -133,6 +133,7 @@ Each Phase 4 PR:
 | v0.62.0 |          378  |    +6  | expansion: expandContext option, @context:null full reset, @base:null, scoped @nest (+5 toRdf)|
 | v0.63.0 |          378  |    +0  | toRdf: produceGeneralizedRdf option — keep blank-node predicates (+2 toRdf)|
 | v0.64.0 |          378  |    +0  | toRdf harness: sound blank-node isomorphism compare — rescues 10 spec-correct false-failures (+10 toRdf)|
+| v0.65.0 |          378  |    +0  | toRdf: drop literals with malformed BCP47 language tags (+1 toRdf)|
 
 ### Notes on v0.4.0