mdld-parse 0.2.7 → 0.2.9

package/README.md

@@ -14,8 +14,11 @@ MD-LD allows you to author RDF graphs directly in Markdown using explicit `{...}
 # Apollo 11 {=ex:apollo11 .SpaceMission}
 
 Launch: [1969-07-16] {startDate ^^xsd:date}
-Crew: [Neil Armstrong](ex:armstrong) {?crewMember}
+Crew: [Neil Armstrong] {=?ex:armstrong ?crewMember fullName}
 Description: [First crewed Moon landing] {description}
+
+[Section] {=?#overview ?hasPart}
+Overview: [Mission summary] {description}
 ```
 
 Generates valid RDF triples:
@@ -25,18 +28,20 @@ ex:apollo11 a schema:SpaceMission ;
   schema:startDate "1969-07-16"^^xsd:date ;
   schema:crewMember ex:armstrong ;
   schema:description "First crewed Moon landing" .
-```
 
-## Core Guarantees
+ex:armstrong schema:fullName "Neil Armstrong" .
+```
 
-MD-LD v0.2 provides strict semantic guarantees:
+## Core Features
 
-1. **CommonMark-preserving** — Removing `{...}` yields valid Markdown
-2. **Explicit semantics** — Every quad originates from explicit `{...}`
-3. **Single-pass parsing** — Streaming-friendly, deterministic
-4. **No blank nodes** — All subjects are stable IRIs
-5. **Complete traceability** — Every quad maps to source location
-6. **Round-trip capable** — Markdown ↔ RDF ↔ Markdown preserves structure
+- **Subject declarations**: `{=IRI}` and `{=#fragment}` for context setting
+- **Object IRIs**: `{=?IRI}` and `{=?#fragment}` for temporary object declarations  
+- **Four predicate forms**: `p` (S→L), `?p` (S→O), `^p` (L→S), `^?p` (O→S)
+- **Type declarations**: `.Class` for rdf:type triples
+- **Datatypes & language**: `^^xsd:date` and `@en` support
+- **Lists**: Explicit subject declarations for structured data
+- **Fragments**: Built-in document structuring with `{=#fragment}`
+- **Round-trip serialization**: Markdown ↔ RDF ↔ Markdown preserves structure
 
 ## Installation
 
@@ -246,6 +251,7 @@ ex:book schema:hasPart ex:part .
 ```markdown
 [ex] {: http://example.org/}
 [foaf] {: http://xmlns.com/foaf/0.1/}
+[@vocab] {: http://schema.org/}
 
 # Person {=ex:alice .foaf:Person}
 ```
@@ -385,6 +391,41 @@ MD-LD explicitly forbids to ensure deterministic parsing:
 - ❌ Predicate guessing from context
 - ❌ Multi-pass or backtracking parsers
 
+Below is a **tight, README-ready refinement** of the Algebra section.
+It keeps the math precise, examples exhaustive, and language compact.
+
+---
+
+## Algebra 
+
+> Every RDF triple `(s, p, o)` can be authored **explicitly, deterministically, and locally**, with no inference, guessing, or reordering.
+
+MD-LD models RDF authoring as a **closed edge algebra** over a small, explicit state. To be algebraically complete for RDF triple construction, a syntax must support:
+
+* Binding a **subject** `S`
+* Binding an **object** `O`
+* Emitting predicates in **both directions**
+* Distinguishing **IRI nodes** from **literal nodes**
+* Operating with **no implicit state or inference**
+
+MD-LD satisfies these requirements with four explicit operators.
+
+Each predicate is partitioned by **direction** and **node kind**:
+
+| Predicate form | Emitted triple |
+| -------------- | -------------- |
+| `p`            | `S ─p→ L`      |
+| `?p`           | `S ─p→ O`      |
+| `^p`           | `L ─p→ S`      |
+| `^?p`          | `O ─p→ S`      |
+
+This spans all **2 × 2** combinations of:
+
+* source ∈ {subject, object/literal}
+* target ∈ {subject, object/literal}
+
+Therefore, the algebra is **closed**.
+
 ## Use Cases
 
 ### Personal Knowledge Management
@@ -456,8 +497,10 @@ Contributions welcome! Please:
 
 ## Acknowledgments
 
+Developed by [Denis Starov](https://github.com/davay42).
+
 Inspired by:
-- Thomas Francart's [Semantic Markdown](https://blog.sparna.fr/2020/02/20/semantic-markdown/)
+- Thomas Francart's [Semantic Markdown](https://blog.sparna.fr/2020/02/20/semantic-markdown/) article
 - RDFa decades of structured data experience
 - CommonMark's rigorous parsing approach
 

package/index.js

@@ -24,7 +24,6 @@ export function hash(str) {
     return Math.abs(h).toString(16).slice(0, 12);
 }
 
-// IRI Utilities
 export function expandIRI(term, ctx) {
     if (term == null) return null;
     const raw = typeof term === 'string' ? term : (typeof term === 'object' && typeof term.value === 'string') ? term.value : String(term);
@@ -76,6 +75,13 @@ export function parseSemanticBlock(raw) {
                 continue;
             }
 
+            if (token.startsWith('=?#')) {
+                const fragment = token.substring(3);
+                result.object = `#${fragment}`;
+                result.entries.push({ kind: 'softFragment', fragment, relRange: { start: relStart, end: relEnd }, raw: token });
+                continue;
+            }
+
             if (token.startsWith('=?')) {
                 const iri = token.substring(2);
                 result.object = iri;
@@ -274,6 +280,41 @@ function extractInlineCarriers(text, baseOffset = 0) {
     let pos = 0;
 
     while (pos < text.length) {
+        // Try emphasis patterns first (before brackets)
+        const emphasisMatch = text.match(/^[*__`]+(.+?)[*__`]+\s*\{([^}]+)\}/, pos);
+        if (emphasisMatch) {
+            const carrierText = emphasisMatch[1];
+            const valueRange = [baseOffset + emphasisMatch[0].length, baseOffset + emphasisMatch[0].length + emphasisMatch[1].length];
+            carriers.push({
+                type: 'emphasis',
+                text: carrierText,
+                attrs: `{${emphasisMatch[2]}}`,
+                attrsRange: [baseOffset + emphasisMatch[0].length + emphasisMatch[1].length + 2, baseOffset + emphasisMatch[0].length + emphasisMatch[1].length + emphasisMatch[2].length],
+                valueRange,
+                range: [baseOffset + emphasisMatch[0].length, baseOffset + emphasisMatch[0].length + emphasisMatch[1].length]
+            });
+            pos = baseOffset + emphasisMatch[0].length + emphasisMatch[1].length + emphasisMatch[2].length;
+            continue;
+        }
+
+        // Try code spans
+        const codeMatch = text.match(/^``(.+?)``\s*\{([^}]+)\}/, pos);
+        if (codeMatch) {
+            const carrierText = codeMatch[1];
+            const valueRange = [baseOffset + 2, baseOffset + 2 + codeMatch[1].length];
+            carriers.push({
+                type: 'code',
+                text: carrierText,
+                attrs: `{${codeMatch[2]}}`,
+                attrsRange: [baseOffset + 2 + codeMatch[1].length + 2, baseOffset + 2 + codeMatch[1].length + 2],
+                valueRange,
+                range: [baseOffset + 2, baseOffset + 2 + codeMatch[1].length + 2]
+            });
+            pos = baseOffset + 2 + codeMatch[1].length + 2;
+            continue;
+        }
+
+        // Try bracket patterns (original logic)
         const bracketStart = text.indexOf('[', pos);
         if (bracketStart === -1) break;
 
@@ -365,7 +406,6 @@ function createBlock(subject, types, predicates, entries, range, attrsRange, val
     };
 }
 
-// Quad Utilities
 function quadIndexKey(subject, predicate, object) {
     const objKey = object.termType === 'Literal'
         ? JSON.stringify({ t: 'Literal', v: object.value, lang: object.language || '', dt: object.datatype?.value || '' })
@@ -406,7 +446,6 @@ function parseQuadIndexKey(key) {
     }
 }
 
-// Semantic Slot Utilities
 function createSemanticSlotId(subject, predicate) {
     return hash(`${subject.value}|${predicate.value}`);
 }
@@ -500,7 +539,17 @@ function processAnnotation(carrier, sem, state) {
 
     if (sem.object) {
         // Handle soft IRI object declaration - local to this annotation only
-        localObject = state.df.namedNode(expandIRI(sem.object, state.ctx));
+        if (sem.object.startsWith('#')) {
+            // Soft fragment - resolve against current subject base
+            const fragment = sem.object.substring(1);
+            if (state.currentSubject) {
+                const baseIRI = state.currentSubject.value.split('#')[0];
+                localObject = state.df.namedNode(`${baseIRI}#${fragment}`);
+            }
+        } else {
+            // Regular soft IRI
+            localObject = state.df.namedNode(expandIRI(sem.object, state.ctx));
+        }
     }
 
     if (newSubject) state.currentSubject = newSubject;
@@ -718,8 +767,6 @@ export function parse(text, options = {}) {
     return { quads: state.quads, origin: state.origin, context: state.ctx };
 }
 
-
-// Text Processing Utilities
 function readSpan(block, text, spanType = 'attrs') {
     const range = spanType === 'attrs' ? block?.attrsRange : block?.valueRange;
     if (!range) return null;
@@ -753,6 +800,16 @@ function removeObjectToken(tokens, iri) {
     return removeOneToken(tokens, t => t === objectToken);
 }
 
+function addSoftFragmentToken(tokens, fragment) {
+    const fragmentToken = `=?#${fragment}`;
+    return tokens.includes(fragmentToken) ? tokens : [...tokens, fragmentToken];
+}
+
+function removeSoftFragmentToken(tokens, fragment) {
+    const fragmentToken = `=?#${fragment}`;
+    return removeOneToken(tokens, t => t === fragmentToken);
+}
+
 function sanitizeCarrierValueForBlock(block, raw) {
     const s = String(raw ?? '');
     const t = block?.carrierType;
@@ -1057,6 +1114,17 @@ export function serialize({ text, diff, origin, options = {} }) {
                 return;
             }
 
+            // Handle soft fragment token removal
+            if (entry?.kind === 'softFragment') {
+                const fragment = entry.fragment;
+                const { tokens: updated, removed } = removeSoftFragmentToken(tokens, fragment);
+                if (!removed) return;
+
+                const newAttrs = updated.length === 0 ? '{}' : writeAttrsTokens(updated);
+                edits.push({ start: span.start, end: span.end, text: newAttrs });
+                return;
+            }
+
             const tokens = normalizeAttrsTokens(span.text);
             let updated = tokens;
             let removed = false;
@@ -1151,20 +1219,31 @@ export function serialize({ text, diff, origin, options = {} }) {
                     const objectShort = shortenIRI(full, ctx);
                     const predShort = shortenIRI(quad.predicate.value, ctx);
 
-                    // Check if this is a ?predicate form (should use object IRI)
-                    const span = readSpan(targetBlock, text, 'attrs');
-                    const tokens = blockTokensFromEntries(targetBlock) || normalizeAttrsTokens(span.text);
-                    const hasObjectToken = tokens.some(t => t.startsWith('=?'));
+                    // Check if this is a soft fragment
+                    const isSoftFragment = full.includes('#') && anchored?.entry?.kind === 'softFragment';
 
-                    if (hasObjectToken || anchored?.entry?.form === '?') {
-                        // Add object token if not present
-                        const updated = addObjectToken(tokens, objectShort);
-                        if (updated.length !== tokens.length) {
-                            edits.push({ start: span.start, end: span.end, text: writeAttrsTokens(updated) });
+                    if (isSoftFragment || anchored?.entry?.form === '?') {
+                        // Add soft fragment token if not present
+                        if (isSoftFragment) {
+                            const fragment = full.split('#')[1];
+                            const updated = addSoftFragmentToken(tokens, fragment);
+                            if (updated.length !== tokens.length) {
+                                edits.push({ start: span.start, end: span.end, text: writeAttrsTokens(updated) });
+                            }
+                        } else {
+                            const updated = addObjectToken(tokens, objectShort);
+                            if (updated.length !== tokens.length) {
+                                edits.push({ start: span.start, end: span.end, text: writeAttrsTokens(updated) });
+                            }
                         }
                     } else {
                         // Create new annotation with object token
-                        edits.push({ start: result.length, end: result.length, text: `\n[${objectShort}] {=?${objectShort} ?${predShort}}` });
+                        if (isSoftFragment) {
+                            const fragment = full.split('#')[1];
+                            edits.push({ start: result.length, end: result.length, text: `\n[${objectShort}] {=?#${fragment} ?${predShort}}` });
+                        } else {
+                            edits.push({ start: result.length, end: result.length, text: `\n[${objectShort}] {=?${objectShort} ?${predShort}}` });
+                        }
                     }
                     return;
                 }

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "mdld-parse",
-	"version": "0.2.7",
+	"version": "0.2.9",
 	"description": "A standards-compliant parser for **MD-LD (Markdown-Linked Data)** — a human-friendly RDF authoring format that extends Markdown with semantic annotations.",
 	"type": "module",
 	"main": "index.js",