@gregiteen/ssss-cli 0.11.0 → 0.11.1

package/src/frontmatter.mjs

@@ -10,8 +10,20 @@
  * operation patches without dropping nested frontmatter.
  */
 
+/**
+ * Repair a document a broken writer flattened onto one line with literal `\n`
+ * text in place of newlines (`---\ntype: "memory"\n...`). Real documents with
+ * any actual line break are returned unchanged.
+ */
+export function repairEscapedNewlines(content) {
+  if (typeof content !== 'string' || !content.startsWith('---\\n') || content.trim().includes('\n')) return content;
+  const fixed = content.replace(/\\n/g, '\n');
+  return fixed.endsWith('\n') ? fixed : `${fixed}\n`;
+}
+
 /** Split a document into its raw frontmatter block and body. */
 export function splitFrontmatter(content) {
+  content = repairEscapedNewlines(content);
   if (typeof content !== 'string' || !content.startsWith('---')) {
     return { fm: '', body: content || '', hasFm: false };
   }
@@ -25,36 +37,190 @@ export function splitFrontmatter(content) {
   return { fm: inner, body: content.slice(sliceEnd), hasFm: true };
 }
 
-function coerce(raw) {
-  const v = raw.trim();
-  if (v === '') return '';
-  if (v.length >= 2 && v.startsWith('"') && v.endsWith('"')) {
+function countIndent(line) {
+  const m = line.match(/^ */);
+  return m ? m[0].length : 0;
+}
+
+/** Index of the quote that closes the quoted scalar starting at `v[0]`, or -1. */
+function findClosingQuote(v) {
+  const q = v[0];
+  for (let i = 1; i < v.length; i++) {
+    if (q === '"') {
+      if (v[i] === '\\') i++;
+      else if (v[i] === '"') return i;
+    } else if (v[i] === "'") {
+      if (v[i + 1] === "'") i++;
+      else return i;
+    }
+  }
+  return -1;
+}
+
+const YAML_ESCAPES = {
+  0: '\0', a: '\x07', b: '\b', t: '\t', '\t': '\t', n: '\n', v: '\v', f: '\f', r: '\r', e: '\x1b',
+  ' ': ' ', '"': '"', '/': '/', '\\': '\\', N: '\x85', _: '\xa0', L: '\u2028', P: '\u2029'
+};
+
+/** Decode a YAML double-quoted scalar body (a superset of JSON string escapes). */
+function unescapeDouble(body) {
+  return body.replace(/\\(?:x([0-9a-fA-F]{2})|u([0-9a-fA-F]{4})|U([0-9a-fA-F]{8})|([\s\S]))/g, (m, x, u, U, c) => {
+    const hex = x ?? u ?? U;
+    if (hex) {
+      const cp = parseInt(hex, 16);
+      return cp <= 0x10ffff ? String.fromCodePoint(cp) : m;
+    }
+    return Object.hasOwn(YAML_ESCAPES, c) ? YAML_ESCAPES[c] : m;
+  });
+}
+
+function unquote(tok) {
+  if (tok[0] === '"') {
     // Double-quoted scalars carry escapes; the serializer writes them with JSON.stringify.
-    try { return JSON.parse(v); } catch { return v.slice(1, -1); }
+    try { return JSON.parse(tok); } catch { return unescapeDouble(tok.slice(1, -1)); }
   }
-  if (v.length >= 2 && v.startsWith("'") && v.endsWith("'")) return v.slice(1, -1).replace(/''/g, "'");
-  if (v === 'true') return true;
-  if (v === 'false') return false;
-  if (v === 'null' || v === '~') return null;
+  return tok.slice(1, -1).replace(/''/g, "'");
+}
+
+/** Drop a trailing ` # comment` from an unquoted value. */
+function stripComment(v) {
+  return v.replace(/(^|\s)#.*$/, '').trimEnd();
+}
+
+const INT_RE = /^[-+]?(?:0|[1-9]\d*)$/;
+const FLOAT_RE = /^[-+]?(?:(?:0|[1-9]\d*)?\.\d+(?:[eE][-+]?\d+)?|(?:0|[1-9]\d*)(?:\.\d*)?[eE][-+]?\d+)$/;
+
+function coercePlain(v) {
+  if (v === 'true' || v === 'True' || v === 'TRUE') return true;
+  if (v === 'false' || v === 'False' || v === 'FALSE') return false;
+  if (v === 'null' || v === 'Null' || v === 'NULL' || v === '~') return null;
   // A leading zero (zip 07003, license 00123) marks a code, not a number.
-  if (/^-?(0|[1-9]\d*)$/.test(v)) return parseInt(v, 10);
-  if (/^-?\d*\.\d+$/.test(v)) return parseFloat(v);
-  if (v.startsWith('[') && v.endsWith(']')) {
-    const inner = v.slice(1, -1).trim();
-    if (!inner) return [];
-    return inner.split(',').map((s) => coerce(s));
-  }
+  if (INT_RE.test(v)) return parseInt(v, 10);
+  if (/^0x[0-9a-fA-F]+$/.test(v)) return parseInt(v.slice(2), 16);
+  if (/^0o[0-7]+$/.test(v)) return parseInt(v.slice(2), 8);
+  if (FLOAT_RE.test(v)) return parseFloat(v);
   return v;
 }
 
-function countIndent(line) {
-  const m = line.match(/^ */);
-  return m ? m[0].length : 0;
+/**
+ * Parse an inline `[...]` / `{...}` value (JSON and YAML flow style). Quotes,
+ * nested brackets and commas inside quoted strings are respected. Throws on
+ * malformed input so the caller can fall back to the raw text.
+ */
+function parseFlow(src) {
+  let pos = 0;
+  const ws = () => { while (pos < src.length && /\s/.test(src[pos])) pos++; };
+  const quoted = () => {
+    const end = findClosingQuote(src.slice(pos));
+    if (end < 0) throw new Error('unterminated quote');
+    const tok = src.slice(pos, pos + end + 1);
+    pos += end + 1;
+    return unquote(tok);
+  };
+  const plain = (inKey) => {
+    const start = pos;
+    while (pos < src.length) {
+      const ch = src[pos];
+      if (ch === ',' || ch === ']' || ch === '}') break;
+      if (inKey && ch === ':' && (pos + 1 >= src.length || /[\s,\]}]/.test(src[pos + 1]))) break;
+      pos++;
+    }
+    return src.slice(start, pos).trim();
+  };
+  const node = () => {
+    ws();
+    const ch = src[pos];
+    if (ch === '[') return seq();
+    if (ch === '{') return map();
+    if (ch === '"' || ch === "'") return quoted();
+    const p = plain(false);
+    return p === '' ? null : coercePlain(p);
+  };
+  const seq = () => {
+    pos++;
+    const out = [];
+    for (;;) {
+      ws();
+      if (pos >= src.length) throw new Error('unterminated sequence');
+      if (src[pos] === ']') { pos++; return out; }
+      out.push(node());
+      ws();
+      if (src[pos] === ',') pos++;
+      else if (src[pos] !== ']') throw new Error('bad sequence');
+    }
+  };
+  const map = () => {
+    pos++;
+    const out = {};
+    for (;;) {
+      ws();
+      if (pos >= src.length) throw new Error('unterminated map');
+      if (src[pos] === '}') { pos++; return out; }
+      const key = src[pos] === '"' || src[pos] === "'" ? quoted() : plain(true);
+      ws();
+      if (src[pos] === ':') { pos++; out[key] = node(); } else out[key] = null;
+      ws();
+      if (src[pos] === ',') pos++;
+      else if (src[pos] !== '}') throw new Error('bad map');
+    }
+  };
+  const value = node();
+  ws();
+  if (pos < src.length && !src.slice(pos).startsWith('#')) throw new Error('trailing text');
+  return value;
+}
+
+function coerce(raw) {
+  const v = raw.trim();
+  if (v === '') return '';
+  if (v[0] === '"' || v[0] === "'") {
+    const end = findClosingQuote(v);
+    if (end > 0 && /^\s*(#.*)?$/.test(v.slice(end + 1))) return unquote(v.slice(0, end + 1));
+    return v;
+  }
+  if (v[0] === '[' || v[0] === '{') {
+    try { return parseFlow(v); } catch { return v; }
+  }
+  return coercePlain(stripComment(v));
 }
 
+// A key is a bare word (no whitespace, `:` or `#`; inner single spaces allowed)
+// or a quoted string, followed by `:` and then whitespace or end of line. The
+// whitespace requirement keeps tags like `context:universal` and URLs as values.
+const KEY_RE =
+  /^(?:([A-Za-z0-9_][^\s:#]*(?: [^\s:#]+)*)|("(?:[^"\\]|\\.)*")|('(?:[^']|'')*')):(?:[ \t]+(.*))?$/;
+
 function splitKeyValue(text) {
-  const m = text.match(/^([A-Za-z0-9_]+):\s*(.*)$/);
-  return m ? { key: m[1], rest: m[2] } : null;
+  const m = text.trimEnd().match(KEY_RE);
+  if (!m) return null;
+  const key = m[1] ?? unquote(m[2] ?? m[3]);
+  const rest = (m[4] ?? '').trim();
+  return { key, rest: rest.startsWith('#') ? '' : rest };
+}
+
+/**
+ * Read a scalar or flow value that starts on `lines[i]` (already split into
+ * `rest`) and may wrap onto deeper-indented continuation lines. Plain and
+ * quoted lines fold with a space; a blank line is one newline.
+ */
+function readScalar(lines, i, keyIndent, rest) {
+  const inline = /^["'[{]/.test(rest);
+  let text = rest;
+  let j = i + 1;
+  let blanks = 0;
+  let last = i;
+  while (j < lines.length) {
+    const line = lines[j];
+    if (!line.trim()) { blanks++; j++; continue; }
+    if (countIndent(line) <= keyIndent) break;
+    if (!inline && line.trim().startsWith('#')) break;
+    const piece = inline ? line.trim() : stripComment(line.trim());
+    text += blanks ? '\n'.repeat(blanks) + piece : ` ${piece}`;
+    blanks = 0;
+    last = j;
+    j++;
+  }
+  return { value: coerce(text), next: last + 1 };
 }
 
 function nextSignificant(lines, start) {
@@ -64,7 +230,7 @@ function nextSignificant(lines, start) {
   return -1;
 }
 
-const BLOCK_SCALAR_RE = /^([|>])([+-]?)$/;
+const BLOCK_SCALAR_RE = /^([|>])([+-]?)(?:\s+#.*)?$/;
 
 /** Read a `|` (literal) or `>` (folded) block scalar with clip/strip/keep chomping. */
 function parseBlockScalar(lines, start, parentIndent, style, chomp) {
@@ -84,23 +250,26 @@ function parseBlockScalar(lines, start, parentIndent, style, chomp) {
   while (end > 0 && raw[end - 1] === '') end--;
   const content = raw.slice(0, end);
   let text = content[0] ?? '';
+  let prevText = 0; // index of the last non-blank line already emitted
   for (let k = 1; k < content.length; k++) {
     const line = content[k];
-    const previous = content[k - 1];
-    if (style === '|') text += `\n${line}`;
-    // Folding: each blank line is one newline, and the break before it is
-    // consumed; more-indented lines keep their breaks; others join with a space.
-    else if (line === '') text += '\n';
-    else if (previous === '') text += line;
-    else if (/^\s/.test(line) || /^\s/.test(previous)) text += `\n${line}`;
-    else text += ` ${line}`;
+    if (line === '') continue;
+    const blanks = k - prevText - 1;
+    if (style === '|') text += `${'\n'.repeat(blanks + 1)}${line}`;
+    // Folding: a break between two ordinary lines becomes a space, and blank
+    // lines each add one newline; a more-indented line on either side keeps its break.
+    else if (/^\s/.test(line) || /^\s/.test(content[prevText])) text += `${'\n'.repeat(blanks + 1)}${line}`;
+    else text += blanks ? `${'\n'.repeat(blanks)}${line}` : ` ${line}`;
+    prevText = k;
   }
-  if (!content.length) text = '';
+  if (!content.length) text = chomp === '+' ? '\n'.repeat(raw.length) : '';
   else if (chomp === '+') text += '\n'.repeat(1 + raw.length - end);
   else if (chomp !== '-') text += '\n';
   return { value: text, next: i };
 }
 
+const isDash = (text) => text === '-' || text.startsWith('- ');
+
 function parseMap(lines, start, indent) {
   const data = {};
   let i = start;
@@ -120,19 +289,22 @@ function parseMap(lines, start, indent) {
       continue;
     }
     if (kv.rest !== '') {
-      data[kv.key] = coerce(kv.rest);
-      i++;
+      const scalar = readScalar(lines, i, indent, kv.rest);
+      data[kv.key] = scalar.value;
+      i = scalar.next;
       continue;
     }
 
     const next = nextSignificant(lines, i + 1);
-    if (next < 0 || countIndent(lines[next]) <= indent) {
+    // YAML allows a block list at the same indent as its key.
+    const sameIndentList = next >= 0 && countIndent(lines[next]) === indent && isDash(lines[next].slice(indent));
+    if (next < 0 || (countIndent(lines[next]) <= indent && !sameIndentList)) {
       data[kv.key] = {};
       i++;
       continue;
     }
     const childIndent = countIndent(lines[next]);
-    if (lines[next].slice(childIndent).startsWith('- ')) {
+    if (isDash(lines[next].slice(childIndent))) {
       const parsed = parseList(lines, next, childIndent);
       data[kv.key] = parsed.value;
       i = parsed.next;
@@ -154,21 +326,25 @@ function parseList(lines, start, indent) {
     const currentIndent = countIndent(line);
     if (currentIndent < indent) break;
     if (currentIndent > indent) { i++; continue; }
-    const item = line.slice(indent).match(/^-\s*(.*)$/);
+    const item = line.slice(indent).match(/^-(?:(\s+)(.*))?$/);
     if (!item) break;
-    const raw = item[1];
+    const raw = (item[2] ?? '').trim();
+    const keyIndent = indent + 1 + (item[1] ?? '').length;
     const kv = splitKeyValue(raw);
     if (kv) {
-      const obj = { [kv.key]: kv.rest === '' ? {} : coerce(kv.rest) };
-      i++;
-      const next = nextSignificant(lines, i);
-      if (next >= 0 && countIndent(lines[next]) > indent) {
-        const parsed = parseMap(lines, next, countIndent(lines[next]));
-        Object.assign(obj, parsed.value);
-        i = parsed.next;
-      }
-      items.push(obj);
-    } else if (raw === '') {
+      // Re-read the item as a map whose first key sits where the dash was, so
+      // nested children, block scalars and wrapped values of every key parse alike.
+      lines[i] = ' '.repeat(keyIndent) + raw;
+      const parsed = parseMap(lines, i, keyIndent);
+      items.push(parsed.value);
+      i = parsed.next;
+    } else if (isDash(raw)) {
+      // Nested sequence (`- - a`): re-read the inner dash as its own list item.
+      lines[i] = ' '.repeat(keyIndent) + raw;
+      const parsed = parseList(lines, i, keyIndent);
+      items.push(parsed.value);
+      i = parsed.next;
+    } else if (raw === '' || raw.startsWith('#')) {
       const next = nextSignificant(lines, i + 1);
       if (next >= 0 && countIndent(lines[next]) > indent) {
         const parsed = parseMap(lines, next, countIndent(lines[next]));
@@ -178,9 +354,15 @@ function parseList(lines, start, indent) {
         items.push('');
         i++;
       }
+    } else if (BLOCK_SCALAR_RE.test(raw)) {
+      const block = raw.match(BLOCK_SCALAR_RE);
+      const parsed = parseBlockScalar(lines, i + 1, indent, block[1], block[2]);
+      items.push(parsed.value);
+      i = parsed.next;
     } else {
-      items.push(coerce(raw));
-      i++;
+      const scalar = readScalar(lines, i, indent, raw);
+      items.push(scalar.value);
+      i = scalar.next;
     }
   }
   return { value: items, next: i };
@@ -208,15 +390,22 @@ function serializeValue(v) {
   if (v === null) return 'null';
   if (typeof v === 'boolean' || typeof v === 'number') return String(v);
   if (Array.isArray(v)) return `[${v.map((x) => serializeValue(x)).join(', ')}]`;
+  // Nested objects inside an inline array go out as JSON, which the flow parser reads back.
+  if (typeof v === 'object') return JSON.stringify(v);
   const s = String(v);
-  if (s === '' || /[:#\[\]{}"'\\\n\r\t|>]/.test(s) || /^\s|\s$/.test(s)) return JSON.stringify(s);
+  if (s === '' || /[:#,\[\]{}"'\\\n\r\t|>]/.test(s) || /^\s|\s$/.test(s) || /^[-?!&*%@`]/.test(s)) return JSON.stringify(s);
   // A string that would read back as a number, boolean or null stays quoted,
   // so "07003" and "true" survive a round trip (here and in YAML 1.1 readers).
   if (LOOKS_TYPED.test(s)) return JSON.stringify(s);
   return s;
 }
 
-function serializeEntry(lines, key, value, indent = 0) {
+function serializeKey(key) {
+  return /^[A-Za-z0-9_][A-Za-z0-9_.\-/]*$/.test(key) ? key : JSON.stringify(key);
+}
+
+function serializeEntry(lines, rawKey, value, indent = 0) {
+  const key = serializeKey(rawKey);
   const pad = ' '.repeat(indent);
   if (Array.isArray(value)) {
     if (!value.length) {
@@ -233,10 +422,10 @@ function serializeEntry(lines, key, value, indent = 0) {
         }
         const [first, ...rest] = entries;
         if (first[1] && typeof first[1] === 'object' && !Array.isArray(first[1])) {
-          lines.push(`${pad}  - ${first[0]}:`);
+          lines.push(`${pad}  - ${serializeKey(first[0])}:`);
           serializeObject(lines, first[1], indent + 6);
         } else {
-          lines.push(`${pad}  - ${first[0]}: ${serializeValue(first[1])}`);
+          lines.push(`${pad}  - ${serializeKey(first[0])}: ${serializeValue(first[1])}`);
         }
         for (const [childKey, childValue] of rest) serializeEntry(lines, childKey, childValue, indent + 4);
       } else {

package/src/kernel.mjs

@@ -184,6 +184,28 @@ export function createKernel(options = {}) {
       if (currentData?.type && currentData.type !== nextData.type) {
         return failure('validation_failed', envelope, operationId, ['Operation may not change an existing document type.'], [{ field: 'type', issue: 'Use an explicit migration.' }]);
       }
+      // Replaying an identical full operation is a no-op, not a replace. This must
+      // be checked before the append-only guard so that re-importing a bundle that
+      // is already present reports `unchanged` instead of failing (spec §17.1).
+      if (current && current.bytes.toString('utf8') === nextContent) {
+        const noopAuth = await authorize({
+          principal, workspaceId: envelope.workspace_id, definition,
+          action: envelope.type === 'operation' ? 'replace' : actionFor(envelope, true, isAppendType(definition)),
+          context,
+        });
+        if (!noopAuth.allowed) {
+          const code = validateVerifiedPrincipal(principal).valid ? 'forbidden' : 'unauthorized';
+          return failure(code, envelope, operationId, [`Access denied: ${noopAuth.reason}`], [{ field: 'principal', issue: noopAuth.reason }]);
+        }
+        const response = {
+          success: true, type: envelope.type, operation_id: operationId, path: envelope.path,
+          committed_at: current.version?.committed_at || null, dry_run: false,
+          unchanged: true,
+          validation: { valid: true, type: definition.qualified_type, errors: [], warnings: [] },
+        };
+        try { await idempotencyStore.put(envelope.workspace_id, envelope.idempotency_key, { request_hash: requestHash, response }); } catch {}
+        return response;
+      }
       // A full replace would rewrite recorded history (spec §5.3).
       if (current && isAppendType(definition)) {
         return failure('validation_failed', envelope, operationId, [`Append-only primitive '${declaredType}' cannot be replaced.`], [{ field: 'content', issue: 'Append with a patch carrying __body__.' }]);

package/src/registry.mjs

@@ -297,26 +297,68 @@ export function resolvePrimitiveDefinition(registrySet, type) {
   return local ? registrySet.types.get(local) || null : null;
 }
 
+/**
+ * Names of the extensions loaded when a caller does not scope the load.
+ *
+ * A registry directory accumulates one extension per product. Two unrelated
+ * products may each legitimately define a shared primitive name (e.g. both a
+ * CRM and a ticketing product needing `integration_connection`), and §5.6 makes
+ * that a load-time collision. `extensions/defaults.json` names the set that is
+ * safe to compose together; every other extension MUST be selected explicitly by
+ * the caller that owns it. Without the manifest, every extension is loaded and
+ * any cross-product collision stays a hard error rather than a silent preference.
+ */
+function loadExtensionDefaults(registryDir) {
+  const manifest = path.join(registryDir, 'extensions', 'defaults.json');
+  if (!fs.existsSync(manifest)) return null;
+  const parsed = JSON.parse(fs.readFileSync(manifest, 'utf8'));
+  const names = parsed?.default_extensions;
+  if (!Array.isArray(names) || names.some((n) => typeof n !== 'string')) {
+    throw new Error('extensions/defaults.json: default_extensions must be an array of registry names.');
+  }
+  return names;
+}
+
 /**
  * Load the core registry and merge extension registries.
+ *
+ * `only` scopes the load to a set of registry names (e.g. `['dabber']`). A
+ * registry directory may hold extensions for unrelated products that legitimately
+ * declare the same shared primitive name; loading them together is a load-time
+ * collision (§5.6), so a caller that knows which extension a vault belongs to
+ * MUST be able to load just that one. When `only` is omitted, the set named by
+ * `extensions/defaults.json` is loaded; failing that, every extension is loaded
+ * and any duplicate name remains a hard error rather than a silent preference.
+ *
  * @param {string} [registryDir] - directory containing core.json + extensions/
+ * @param {{ only?: string[]|Set<string> }} [options]
  * @returns {{ types: Map<string, object>, contractTypes: Set<string>, portabilityClasses: string[], core: object, extensions: Set<string> }}
  */
-export function loadRegistries(registryDir = DEFAULT_REGISTRY_DIR) {
+export function loadRegistries(registryDir = DEFAULT_REGISTRY_DIR, options = {}) {
   const corePath = path.join(registryDir, 'core.json');
   const core = JSON.parse(fs.readFileSync(corePath, 'utf8'));
+  const requested = options.only ? new Set(options.only) : null;
+  const defaults = requested ? null : loadExtensionDefaults(registryDir);
+  const only = requested || (defaults ? new Set(defaults) : null);
   const extensionInputs = [];
   const extDir = path.join(registryDir, 'extensions');
   if (fs.existsSync(extDir)) {
-    for (const file of fs.readdirSync(extDir).filter((f) => f.endsWith('.json')).sort()) {
+    for (const file of fs.readdirSync(extDir).filter((f) => f.endsWith('.json') && f !== 'defaults.json').sort()) {
       const extPath = path.join(extDir, file);
       if (fs.lstatSync(extPath).isSymbolicLink()) {
         throw new Error(`extensions/${file}: symlinked registries are not allowed.`);
       }
-      extensionInputs.push({
-        source: `extensions/${file}`,
-        registry: JSON.parse(fs.readFileSync(extPath, 'utf8')),
-      });
+      const registry = JSON.parse(fs.readFileSync(extPath, 'utf8'));
+      if (only && !only.has(registry.registry) && !only.has(file.replace(/\.json$/, ''))) continue;
+      extensionInputs.push({ source: `extensions/${file}`, registry });
+    }
+  }
+  if (only) {
+    const loaded = new Set(extensionInputs.map((e) => e.registry.registry));
+    for (const name of only) {
+      if (!loaded.has(name) && !extensionInputs.some((e) => e.source === `extensions/${name}.json`)) {
+        throw new Error(`registry '${name}' is not present in ${path.join(registryDir, 'extensions')}.`);
+      }
     }
   }
   return composeRegistries({ core, extensions: extensionInputs });