eyeling 1.28.0 → 1.28.2

package/lib/lexer.js

@@ -515,11 +515,114 @@ function normalizeRdfCompatibility(inputText) {
       return text.slice(at, j);
     }
 
+    function readBalancedTermAt(text, at) {
+      const open = text[at];
+      const matching = { '{': '}', '[': ']', '(': ')' };
+      if (!matching[open]) return null;
+
+      const stack = [matching[open]];
+      let j = at + 1;
+      while (j < text.length) {
+        const ch = text[j];
+        if (ch === '"' || ch === "'") {
+          const str = readStringAt(text, j);
+          j = str.end;
+          continue;
+        }
+        if (ch === '<' && !text.startsWith('<<', j)) {
+          const iri = readIriAt(text, j);
+          j = iri.end;
+          continue;
+        }
+        if (ch === '#') {
+          while (j < text.length && text[j] !== '\n' && text[j] !== '\r') j += 1;
+          continue;
+        }
+        if (matching[ch]) {
+          stack.push(matching[ch]);
+          j += 1;
+          continue;
+        }
+        if (ch === stack[stack.length - 1]) {
+          stack.pop();
+          j += 1;
+          if (stack.length === 0) return { text: text.slice(at, j), end: j };
+          continue;
+        }
+        j += 1;
+      }
+
+      throw new N3SyntaxError(`Unterminated term inside RDF 1.2 triple term, expected ${stack[stack.length - 1]}`);
+    }
+
+    function readRdfTripleTermComponent(text, at) {
+      const j = skipWsAndComments(text, at);
+      if (j >= text.length) return null;
+      const ch = text[j];
+
+      if (ch === '<') return readIriAt(text, j);
+
+      if (ch === '"' || ch === "'") {
+        const str = readStringAt(text, j);
+        let end = str.end;
+        let termText = str.text;
+        if (text.startsWith('^^', end)) {
+          const datatype = readRdfTripleTermComponent(text, end + 2);
+          if (datatype) {
+            termText += '^^' + datatype.text;
+            end = datatype.end;
+          }
+        } else if (text[end] === '@') {
+          let k = end + 1;
+          if (/[A-Za-z]/.test(text[k] || '')) {
+            while (k < text.length && /[A-Za-z0-9-]/.test(text[k])) k += 1;
+            termText += text.slice(end, k);
+            end = k;
+          }
+        }
+        return { text: termText, end };
+      }
+
+      if (ch === '{' || ch === '[' || ch === '(') return readBalancedTermAt(text, j);
+
+      let k = j;
+      while (k < text.length && !/\s/.test(text[k]) && !'{}[](),;'.includes(text[k])) k += 1;
+      if (k === j) return null;
+      const value = text.slice(j, k);
+      if (!value || value.startsWith('@')) return null;
+      return { text: value, end: k };
+    }
+
+    function validateSingleRdfTripleTerm(rawTriple) {
+      const triple = rawTriple.trim();
+      if (!triple) throw new N3SyntaxError('RDF 1.2 triple term must contain exactly one subject, predicate, and object');
+
+      let pos = 0;
+      for (const label of ['subject', 'predicate', 'object']) {
+        const term = readRdfTripleTermComponent(triple, pos);
+        if (!term) throw new N3SyntaxError(`RDF 1.2 triple term is missing a ${label}`);
+        pos = term.end;
+      }
+
+      const rest = skipWsAndComments(triple, pos);
+      if (rest >= triple.length) return;
+
+      const found = triple[rest];
+      if (found === ',') {
+        throw new N3SyntaxError("RDF 1.2 triple terms must contain exactly one object; object lists using ',' are not valid inside <<( ... )>>");
+      }
+      if (found === ';') {
+        throw new N3SyntaxError("RDF 1.2 triple terms must contain exactly one predicate-object pair; ';' is not valid inside <<( ... )>>");
+      }
+      throw new N3SyntaxError(`RDF 1.2 triple term must contain exactly one subject, predicate, and object; unexpected ${JSON.stringify(triple.slice(rest, rest + 20))}`);
+    }
+
     function graphTermFromTripleBody(rawBody, parenthesized) {
       let body = rawBody.trim();
       if (parenthesized && body.startsWith('(') && body.endsWith(')')) body = body.slice(1, -1).trim();
       const split = splitTopLevelReifier(body);
       const triple = split.triple;
+      validateSingleRdfTripleTerm(triple);
       const graph = '{ ' + triple + ' }';
       if (split.reifier) {
         const reifier = firstTerm(split.reifier);

package/notes/rdf-message-logs.md

@@ -0,0 +1,228 @@
+# RDF Message Logs internals
+
+Eyeling handles RDF Message Logs as an RDF-compatibility normalization step. A message log is not parsed as one flat graph first; it is split into message chunks and rewritten into ordinary N3 facts that describe the stream, its ordered message envelopes, and each message payload graph.
+
+There are two execution paths:
+
+- normal RDF mode, where a whole message log is normalized into one replay document;
+- `--stream-messages`, where the CLI reads one message at a time and runs the rules once per replayed message.
+
+Both paths expose the same basic `eymsg:` vocabulary and use N3 quoted formulas for payload graphs.
+
+## Input shape
+
+A log is recognized by a message-version directive:
+
+```trig
+VERSION "1.2-messages"
+PREFIX : <urn:example#>
+
+:a :value 1 .
+
+MESSAGE
+
+# Empty heartbeat.
+
+MESSAGE
+
+:b :value 2 .
+```
+
+The current code accepts message versions matching:
+
+```text
+VERSION "1.1-messages"
+VERSION "1.2-messages"
+VERSION "1.2-basic-messages"
+```
+
+Old-style `@version` and `@message` forms are also recognized.
+
+## Normal, whole-log RDF mode
+
+When `lex(input, { rdf: true })` sees a `*-messages` version directive, `normalizeRdfCompatibility()` dispatches to `normalizeRdfMessageLog()` in `lib/lexer.js`.
+
+That function:
+
+1. strips the version directive;
+2. splits the text at top-level `MESSAGE` / `@message` delimiters;
+3. creates deterministic stream, envelope, and payload IRIs from a hash of the whole source text;
+4. normalizes each message chunk through the RDF/TriG compatibility layer;
+5. rewrites blank-node labels so each message gets its own blank-node scope;
+6. emits ordinary N3 replay facts.
+
+Conceptually, the input above becomes something like:
+
+```n3
+<urn:eyeling:message-log:HASH#stream>
+  a eymsg:RDFMessageStream ;
+  eymsg:messageCount "3"^^xsd:integer ;
+  eymsg:orderedEnvelopes (
+    <urn:eyeling:message-log:HASH#m001>
+    <urn:eyeling:message-log:HASH#m002>
+    <urn:eyeling:message-log:HASH#m003>
+  ) ;
+  eymsg:firstEnvelope <urn:eyeling:message-log:HASH#m001> ;
+  eymsg:lastEnvelope <urn:eyeling:message-log:HASH#m003> ;
+  eymsg:envelope <urn:eyeling:message-log:HASH#m001>,
+                 <urn:eyeling:message-log:HASH#m002>,
+                 <urn:eyeling:message-log:HASH#m003> .
+
+<urn:eyeling:message-log:HASH#m001>
+  a eymsg:MessageEnvelope ;
+  eymsg:offset "1"^^xsd:integer ;
+  eymsg:payloadKind eymsg:nonEmpty ;
+  eymsg:nextEnvelope <urn:eyeling:message-log:HASH#m002> ;
+  eymsg:payloadGraph <urn:eyeling:message-log:HASH#m001/payload> .
+
+<urn:eyeling:message-log:HASH#m001/payload>
+  log:nameOf {
+    :a :value 1 .
+  } .
+
+<urn:eyeling:message-log:HASH#m002>
+  a eymsg:MessageEnvelope ;
+  eymsg:offset "2"^^xsd:integer ;
+  eymsg:payloadKind eymsg:empty ;
+  eymsg:nextEnvelope <urn:eyeling:message-log:HASH#m003> .
+
+<urn:eyeling:message-log:HASH#m003>
+  a eymsg:MessageEnvelope ;
+  eymsg:offset "3"^^xsd:integer ;
+  eymsg:payloadKind eymsg:nonEmpty ;
+  eymsg:payloadGraph <urn:eyeling:message-log:HASH#m003/payload> .
+
+<urn:eyeling:message-log:HASH#m003/payload>
+  log:nameOf {
+    :b :value 2 .
+  } .
+```
+
+After this rewrite, the ordinary N3 parser and reasoner handle the result. There is no separate RDF Message Log AST.
+
+## Payload graphs
+
+Payloads are represented exactly like other RDF/TriG named graphs in RDF compatibility mode:
+
+```n3
+?Payload log:nameOf { ...payload triples... } .
+```
+
+The object is an N3 quoted formula. Rules inspect it with formula-aware built-ins such as `log:includes`:
+
+```n3
+@prefix eymsg: <https://eyereasoner.github.io/eyeling/vocab/message#> .
+@prefix log: <http://www.w3.org/2000/10/swap/log#> .
+
+{
+  ?Envelope eymsg:payloadGraph ?Payload .
+  ?Payload log:nameOf ?Graph .
+  ?Graph log:includes { ?Subject :value ?Value . } .
+} => {
+  ?Envelope :sawValue ?Value .
+} .
+```
+
+This is the key design choice: message boundaries are preserved by putting each payload in its own quoted graph instead of merging all message triples into the top-level fact set.
+
+## Per-message normalization
+
+Each message body is normalized before it is embedded as a payload formula:
+
+- RDF 1.2 triple terms are converted to singleton N3 graph terms;
+- RDF 1.2 annotation syntax is expanded;
+- TriG named graphs are converted to `log:nameOf` triples;
+- blank-node labels are rewritten with a message-specific prefix.
+
+For example, the same blank-node label in two different messages does not denote the same internal blank node. Message 1 rewrites roughly to:
+
+```n3
+_:eyeling_m001_b :value 1 .
+```
+
+and message 3 rewrites roughly to:
+
+```n3
+_:eyeling_m003_b :value 2 .
+```
+
+If the message log appears as the second parsed source, the parser's usual source prefixing may add another source prefix around those labels, for example `_:src2_eyeling_m001_b`.
+
+## Empty heartbeat messages
+
+A message chunk is considered empty when, after directives and comments are ignored, it contains no RDF payload.
+
+Empty messages still get an envelope:
+
+```n3
+?Envelope a eymsg:MessageEnvelope ;
+  eymsg:offset "2"^^xsd:integer ;
+  eymsg:payloadKind eymsg:empty .
+```
+
+but they do not get an `eymsg:payloadGraph` triple. This lets rules distinguish heartbeats from non-empty data messages without accidentally reusing a previous payload.
+
+## `--stream-messages` mode
+
+The CLI streaming path lives in `lib/cli.js`.
+
+With:
+
+```bash
+eyeling --rdf --stream-messages rules.n3 messages.trig
+```
+
+the CLI separates ordinary sources from message-log sources. Ordinary sources are parsed once as the reusable rule/program documents. Message logs are then read chunk by chunk.
+
+For each chunk, `buildSingleMessageReplayDocument()` creates a small replay document containing one stream, one envelope, and optionally one payload graph:
+
+```n3
+<urn:eyeling:message-stream:HASH#stream>
+  a eymsg:RDFMessageStream ;
+  eymsg:envelope <urn:eyeling:message-stream:HASH#m000001> ;
+  eymsg:orderedEnvelopes (<urn:eyeling:message-stream:HASH#m000001>) ;
+  eymsg:firstEnvelope <urn:eyeling:message-stream:HASH#m000001> ;
+  eymsg:lastEnvelope <urn:eyeling:message-stream:HASH#m000001> .
+
+<urn:eyeling:message-stream:HASH#m000001>
+  a eymsg:MessageEnvelope ;
+  eymsg:offset "1"^^xsd:integer ;
+  eymsg:payloadKind eymsg:nonEmpty ;
+  eymsg:payloadGraph <urn:eyeling:message-stream:HASH#m000001/payload> .
+
+<urn:eyeling:message-stream:HASH#m000001/payload>
+  log:nameOf {
+    ...one message payload...
+  } .
+```
+
+That one-message replay document is merged with the already-parsed program sources and run immediately. The next message gets a fresh replay document and a fresh run.
+
+The streaming replay document intentionally does not expose the whole global message list. It only exposes the current message envelope, because the point of `--stream-messages` is one-message-at-a-time processing without materializing the whole log.
+
+## Remote logs
+
+For local files, `--stream-messages` reads line by line. For HTTP(S) sources, the CLI first checks the prefix to detect a message-version directive. When processing a remote text/plain RDF Message Log, it downloads the source into a temporary file and then streams that local copy line by line.
+
+## Output behavior
+
+RDF Message Logs do not introduce special output syntax. Once replay facts have been produced, output is the normal Eyeling output path:
+
+- `log:outputString` facts are collected and printed as text;
+- `log:query` output is printed from query conclusions;
+- in RDF mode, triples are printed with RDF-compatible output formatting where possible.
+
+## Practical mental model
+
+RDF Message Log support is best understood as:
+
+```text
+message-log syntax
+  -> split into message chunks
+  -> normalize each chunk as RDF/TriG/RDF 1.2
+  -> wrap each chunk in an eymsg: envelope
+  -> expose the payload as log:nameOf { ... }
+  -> run the ordinary N3 reasoner
+```
+
+So the feature is not a separate streaming RDF reasoner. It is a replay encoding that turns message boundaries into explicit N3 facts and quoted payload graphs, which ordinary Eyeling rules can inspect.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "eyeling",
-  "version": "1.28.0",
+  "version": "1.28.2",
   "description": "A minimal Notation3 (N3) reasoner in JavaScript.",
   "main": "./index.js",
   "keywords": [

package/test/api.test.js

@@ -2916,6 +2916,22 @@ _:b a ex:Person ; ex:name "B" .
     expect: [/:observation\s+:entails\s+<<\(\s+:sensor\s+:needs\s+:inspection\s*\)>>\s*\./m],
   },
 
+  {
+    name: 'RDF mode rejects object lists inside RDF 1.2 triple terms',
+    opt: { proofComments: false, rdf: true },
+    input: `VERSION "1.2"
+@prefix : <http://example.org/ns#> .
+@prefix coll: <http://example.org/collection#> .
+
+:climateDataset coll:contains
+  <<( :asiaDataset coll:contains
+      <<( :China :avgTemp :18C )>>,
+      <<( :Thailand :avgTemp :20C )>>
+  )>> .
+`,
+    expectError: true,
+  },
+
   {
     name: 'RDF mode accepts PREFIX and reified triple terms with reifiers',
     opt: { proofComments: false, rdf: true },

package/test/stream_messages.test.js

@@ -161,6 +161,88 @@ function startFileServer(file) {
   throw new Error('test HTTP server did not start');
 }
 
+
+function startOpenEndedMessageServer() {
+  const dir = mkTmpDir();
+  const script = path.join(dir, 'server.js');
+  const portFile = path.join(dir, 'server.port');
+  fs.writeFileSync(
+    script,
+    [
+      "const http = require('node:http');",
+      "const fs = require('node:fs');",
+      'const portFile = process.argv[2];',
+      'const server = http.createServer((req, res) => {',
+      "  res.writeHead(200, { 'content-type': 'text/plain' });",
+      "  res.write('VERSION \\\"1.2-messages\\\"\\nPREFIX : <urn:test#>\\n');",
+      "  res.write(':a :line \\\"one\\\\n\\\".\\nMESSAGE\\n');",
+      "  setInterval(() => res.write('# keepalive\\n'), 1000);",
+      '});',
+      "server.listen(0, '127.0.0.1', () => fs.writeFileSync(portFile, String(server.address().port)));",
+      '',
+    ].join('\n'),
+    'utf8',
+  );
+  const child = cp.spawn(process.execPath, [script, portFile], { cwd: root, stdio: ['ignore', 'ignore', 'pipe'] });
+  const deadline = Date.now() + 5000;
+  while (Date.now() < deadline) {
+    if (fs.existsSync(portFile)) {
+      const port = fs.readFileSync(portFile, 'utf8').trim();
+      return {
+        url: `http://127.0.0.1:${port}/messages.txt`,
+        stop: () => {
+          child.kill();
+          rmrf(dir);
+        },
+      };
+    }
+    Atomics.wait(new Int32Array(new SharedArrayBuffer(4)), 0, 0, 20);
+  }
+  child.kill();
+  rmrf(dir);
+  throw new Error('test open-ended HTTP server did not start');
+}
+
+function waitForEyelingOutput(args, pattern, opts = {}) {
+  const timeoutMs = opts.timeoutMs || 4000;
+  return new Promise((resolve, reject) => {
+    const child = cp.spawn(process.execPath, [eyelingJsPath, ...args], {
+      cwd: root,
+      stdio: ['ignore', 'pipe', 'pipe'],
+    });
+    let stdout = '';
+    let stderr = '';
+    let settled = false;
+    const timer = setTimeout(() => {
+      if (settled) return;
+      settled = true;
+      child.kill();
+      reject(new Error(`timed out waiting for ${pattern}; stdout=${JSON.stringify(stdout)} stderr=${JSON.stringify(stderr)}`));
+    }, timeoutMs);
+
+    function finish(err) {
+      if (settled) return;
+      settled = true;
+      clearTimeout(timer);
+      child.kill();
+      if (err) reject(err);
+      else resolve(stdout);
+    }
+
+    child.stdout.on('data', (chunk) => {
+      stdout += chunk.toString('utf8');
+      if (pattern.test(stdout)) finish();
+    });
+    child.stderr.on('data', (chunk) => {
+      stderr += chunk.toString('utf8');
+    });
+    child.on('error', finish);
+    child.on('exit', (code) => {
+      if (!settled && code !== 0) finish(new Error(`eyeling exited with ${code}; stdout=${stdout}; stderr=${stderr}`));
+    });
+  });
+}
+
 const cases = [
   {
     name: 'scoped payload rules run once per RDF Message',
@@ -220,6 +302,20 @@ const cases = [
       }
     },
   },
+  {
+    name: 'remote HTTP RDF Message Logs emit before the response ends',
+    async run(tmp) {
+      const rules = path.join(tmp, 'rules.n3');
+      writeScopedPayloadRules(rules);
+      const server = startOpenEndedMessageServer();
+      try {
+        const out = await waitForEyelingOutput(['-r', '--stream-messages', rules, server.url], /^one\n/);
+        assert.equal(out, 'one\n');
+      } finally {
+        server.stop();
+      }
+    },
+  },
   {
     name: 'MARC extraction rules fire over each streamed payload graph',
     run(tmp) {
@@ -251,7 +347,7 @@ const cases = [
   },
 ];
 
-(function main() {
+(async function main() {
   const suiteStart = msNow();
   info(`Running ${cases.length} stream-message tests`);
   let passed = 0;
@@ -261,7 +357,7 @@ const cases = [
     const testName = numberedName(index, tc.name);
     const start = msNow();
     try {
-      tc.run(tmp);
+      await tc.run(tmp);
       ok(`${testName} ${C.dim}(${msNow() - start} ms)${C.n}`);
       passed++;
     } catch (e) {
@@ -281,4 +377,7 @@ const cases = [
   }
   fail(`Some stream-message tests failed (${passed}/${cases.length})`);
   process.exit(1);
-})();
+})().catch((e) => {
+  fail(e && e.stack ? e.stack : String(e));
+  process.exit(1);
+});

package/tools/bundle.js

@@ -195,7 +195,13 @@ function buildBundleSource({ autoRunMain }) {
     out.push(
       '    if (__outerModule && __outerRequire && __outerRequire.main === __outerModule && typeof __entry.main === "function") {',
     );
-    out.push('      __entry.main();');
+    out.push('      const __mainResult = __entry.main();');
+    out.push('      if (__mainResult && typeof __mainResult.then === "function") {');
+    out.push('        __mainResult.catch((e) => {');
+    out.push('          try { if (typeof console !== "undefined" && console.error) console.error(e && e.stack ? e.stack : e && e.message ? e.message : String(e)); } catch (ignoredError) {}');
+    out.push('          try { if (typeof process !== "undefined" && process.exit) process.exit(1); } catch (ignoredError) {}');
+    out.push('        });');
+    out.push('      }');
     out.push('    }');
     out.push('  } catch (ignoredError) {}');
   }