@diagrammo/dgmo 0.47.0 → 0.48.0

package/src/sequence/parser.ts

@@ -1,1596 +0,0 @@
-// ============================================================
-// Sequence Diagram Parser (.dgmo format)
-// ============================================================
-
-import { inferParticipantType } from './participant-inference';
-import type { Brand, Writable } from '../utils/brand';
-import type { DgmoError } from '../diagnostics';
-import type { PaletteColors } from '../palettes/types';
-import {
-  formatDgmoError,
-  makeDgmoError,
-  makeFail,
-  NAME_DIAGNOSTIC_CODES,
-  nameMergedMessage,
-  suggest,
-} from '../diagnostics';
-import { normalizeName, displayName } from '../utils/name-normalize';
-import { parseArrow, parseInArrowLabel } from '../utils/arrows';
-import {
-  measureIndent,
-  extractColor,
-  splitNameAndMeta,
-  parseFirstLine,
-  OPTION_NOCOLON_RE,
-  warnUnknownMetaKeys,
-} from '../utils/parsing';
-import {
-  SEQUENCE_REGISTRY,
-  withTagAliases,
-} from '../utils/reserved-key-registry';
-import type { TagGroup } from '../utils/tag-groups';
-import {
-  matchTagBlockHeading,
-  validateTagValues,
-  validateTagGroupNames,
-  stripDefaultModifier,
-  finalizeAutoTagColors,
-  AUTO_TAG_COLOR_SENTINEL,
-} from '../utils/tag-groups';
-
-/** Known sequence-diagram options that take a value (space-separated). */
-const KNOWN_SEQ_OPTIONS = new Set(['active-tag']);
-
-/** Known sequence-diagram boolean options (bare keyword or `no-` prefix). */
-const KNOWN_SEQ_BOOLEANS = new Set(['solid-fill', 'no-title']);
-
-/** Boolean options that only support the `no-` prefix form (no bare or key-value). */
-const NO_PREFIX_ONLY_BOOLEANS = new Set(['activations']);
-
-/**
- * Participant types that can be declared via "Name is a type" syntax.
- *
- * The 0.16.0 trim retained only the types whose shapes carry semantic
- * weight at a glance: stick figure (actor), cylinder (database),
- * dashed cylinder (cache), horizontal pipe (queue), plus the default
- * rectangle. Any other type word falls back to `default`.
- */
-export type ParticipantType =
-  | 'default'
-  | 'database'
-  | 'actor'
-  | 'queue'
-  | 'cache';
-
-const VALID_PARTICIPANT_TYPES: ReadonlySet<string> = new Set([
-  'database',
-  'actor',
-  'queue',
-  'cache',
-]);
-
-/**
- * Branded participant identifier — a normalized name string that has
- * been minted through `addParticipant` and registered in the parser's
- * `participantMap`. Distinct from a raw display label or any other
- * `string`, so the type system catches "passed label where id expected"
- * at compile time.
- */
-export type ParticipantId = Brand<string, 'ParticipantId'>;
-
-/**
- * A declared or inferred participant in the sequence diagram.
- */
-export interface SequenceParticipant {
-  /** Internal identifier (e.g. "AuthService") */
-  readonly id: ParticipantId;
-  /** Display label — first-seen casing/spacing of the name */
-  readonly label: string;
-  /** Participant shape type */
-  readonly type: ParticipantType;
-  /** Source line number (1-based) */
-  readonly lineNumber: number;
-  /** Explicit layout position override (0-based from left, negative from right) */
-  readonly position?: number;
-  /** Pipe-delimited tag metadata (e.g. `| role: Gateway`) */
-  readonly metadata?: Readonly<Record<string, string>>;
-}
-
-/**
- * A message between two participants.
- *
- * `kind: 'message'` is the discriminator for the SequenceElement union.
- * Pre-1.0 type addition — Epic 105 Story 105.17.
- */
-export interface SequenceMessage {
-  readonly kind: 'message';
-  readonly from: ParticipantId;
-  readonly to: ParticipantId;
-  readonly label: string;
-  readonly lineNumber: number;
-  readonly async?: boolean;
-  /** Pipe-delimited tag metadata (e.g. `| c: Caching`) */
-  readonly metadata?: Readonly<Record<string, string>>;
-}
-
-/**
- * A conditional or loop block in the sequence diagram.
- */
-export interface ElseIfBranch {
-  readonly label: string;
-  readonly children: readonly SequenceElement[];
-  readonly lineNumber: number;
-}
-
-export interface SequenceBlock {
-  readonly kind: 'block';
-  readonly type: 'if' | 'loop' | 'parallel';
-  readonly label: string;
-  readonly children: readonly SequenceElement[];
-  readonly elseChildren: readonly SequenceElement[];
-  readonly elseIfBranches?: readonly ElseIfBranch[];
-  readonly elseLineNumber?: number;
-  readonly lineNumber: number;
-}
-
-/**
- * A labeled horizontal divider between message phases.
- */
-export interface SequenceSection {
-  readonly kind: 'section';
-  readonly label: string;
-  readonly lineNumber: number;
-}
-
-/**
- * An annotation attached to a message, rendered as a folded-corner box.
- */
-export interface SequenceNote {
-  readonly kind: 'note';
-  readonly text: string;
-  readonly position: 'right' | 'left';
-  readonly participantId: ParticipantId;
-  readonly lineNumber: number;
-  readonly endLineNumber: number;
-}
-
-export type SequenceElement =
-  | SequenceMessage
-  | SequenceBlock
-  | SequenceSection
-  | SequenceNote;
-
-export function isSequenceBlock(el: SequenceElement): el is SequenceBlock {
-  return el.kind === 'block';
-}
-
-export function isSequenceSection(el: SequenceElement): el is SequenceSection {
-  return el.kind === 'section';
-}
-
-export function isSequenceNote(el: SequenceElement): el is SequenceNote {
-  return el.kind === 'note';
-}
-
-/**
- * A named group of participants rendered as a labeled box.
- */
-export interface SequenceGroup {
-  readonly name: string;
-  readonly participantIds: readonly ParticipantId[];
-  readonly lineNumber: number;
-  /** Pipe-delimited tag metadata (e.g. `[Backend | t: Product]`) */
-  readonly metadata?: Readonly<Record<string, string>>;
-  /** Whether this group is collapsed by default */
-  readonly collapsed?: boolean;
-}
-
-/**
- * Parsed result from a .dgmo sequence diagram.
- */
-export interface ParsedSequenceDgmo {
-  readonly title: string | null;
-  readonly titleLineNumber: number | null;
-  readonly participants: readonly SequenceParticipant[];
-  readonly messages: readonly SequenceMessage[];
-  readonly elements: readonly SequenceElement[];
-  readonly groups: readonly SequenceGroup[];
-  readonly sections: readonly SequenceSection[];
-  readonly tagGroups: readonly TagGroup[];
-  readonly options: Readonly<Record<string, string>>;
-  readonly diagnostics: readonly DgmoError[];
-  readonly error: string | null;
-}
-
-// "Name is a type" pattern — e.g. "Auth Server is a database"
-// Participant names may contain spaces; [^:]+? stops at colons so that
-// note lines like "note right of A: this is an actor" are not falsely matched.
-// Remainder after type carries the optional `as <alias>` modifier.
-const IS_A_PATTERN = /^([^:]+?)\s+is\s+an?\s+(\w+)(?:\s+(.+))?$/i;
-
-// Colored participant declaration — e.g. "Tapin2(green)", "API(blue)"
-// Scoped to recognized 11-name palette colors only (§1.5) so legitimate
-// `funcCall(arg)` lines don't trigger the legacy-color diagnostic.
-const COLORED_PARTICIPANT_PATTERN =
-  /^(\S+?)\((red|orange|yellow|green|blue|purple|teal|cyan|gray|black|white)\)\s*$/;
-
-// Group heading pattern — "[Backend]", "[Backend] | t: Product"
-// Group 1: name (no ] or | inside brackets), Group 2: color in parens, Group 3: after-bracket text
-const GROUP_HEADING_PATTERN = /^\[([^\]|]+?)(?:\(([^)]+)\))?\]\s*(.*)$/;
-// Fallback: allows anything inside brackets (used to detect pipe-inside-brackets error)
-const GROUP_HEADING_FALLBACK = /^\[([^\]]+)\]\s*(.*)$/;
-
-// Section divider pattern — "== Label ==", "== Label(color) ==", or "== Label" (trailing == optional)
-const SECTION_PATTERN = /^==\s+(.+?)(?:\s*==)?\s*$/;
-
-// Arrow pattern for sequence inference — detects any arrow form
-const ARROW_PATTERN = /\S+\s*(?:<-\S+-|<~\S+~|-\S+->|~\S+~>|->|~>|<-|<~)\s*\S+/;
-
-// Note patterns — colon-free syntax only
-// Single-line: "note text", "note left text", "note right of X text", "note left X text"
-// Multi-line:  "note", "note right", "note right of X", "note left X" (body indented below)
-//
-// The colon-free positioned form requires participant resolution — the parser
-// already has participant collection infrastructure, so we match the general
-// structure here and resolve participant vs text in the parsing logic.
-const NOTE_BARE = /^note\s+(.+)$/i;
-const NOTE_MULTI = /^note(?:\s+(right|left)(?:\s+(?:of\s+)?(.+?))?)?\s*$/i;
-
-/** Result of parseNoteLine — indicates what the parser should do. */
-type NoteParseResult =
-  | {
-      kind: 'single';
-      position: 'right' | 'left';
-      participantId: ParticipantId;
-      text: string;
-    }
-  | {
-      kind: 'multi-head';
-      position: 'right' | 'left';
-      participantId: ParticipantId;
-    }
-  | { kind: 'skip' }
-  | null; // not a note line at all
-
-/**
- * Parse a note line, resolving participant names from the known participants list.
- *
- * Supports:
- * - `note text` — default position (right), last msg sender
- * - `note left of X text` / `note left X text`
- * - `note right` — multi-line head
- * - `note right of X` / `note left X` — multi-line head
- * - Quoted participant: `note left "Auth Service" text`
- */
-function parseNoteLine(
-  trimmed: string,
-  participants: SequenceParticipant[],
-  participantIds: Set<string>,
-  sortedParticipantsCache: SequenceParticipant[],
-  lastMsgFrom: string | null
-): NoteParseResult {
-  const lower = trimmed.toLowerCase();
-  if (!lower.startsWith('note')) return null;
-  // Must be exactly "note" or "note " — not "notebook" etc.
-  if (trimmed.length > 4 && trimmed[4] !== ' ') return null;
-
-  // 1. Try multi-line head (no text after note): `note`, `note right`, `note right of X`, `note left X`
-  // NOTE: NOTE_MULTI's (.+?) can greedily capture "participant text" as one group.
-  // Only trust this match if the captured participant actually exists. Otherwise,
-  // fall through to the bare-note handler which does proper participant-aware splitting.
-  const multiMatch = trimmed.match(NOTE_MULTI);
-  if (multiMatch) {
-    const position =
-      (multiMatch[1]?.toLowerCase() as 'right' | 'left') || 'right';
-    let participantId = multiMatch[2] || null;
-    if (!participantId) {
-      if (!lastMsgFrom) return { kind: 'skip' };
-      participantId = lastMsgFrom;
-    }
-    const lookupKey = normalizeName(participantId);
-    if (participantIds.has(lookupKey)) {
-      // Resolve to first-seen display id so the renderer can find it
-      const found = participants.find((p) => normalizeName(p.id) === lookupKey);
-      return {
-        kind: 'multi-head',
-        position,
-        // Verified participant — `participantIds.has(lookupKey)` confirmed.
-        participantId: found?.id ?? (participantId as ParticipantId),
-      };
-    }
-    // Participant not found — fall through to bare-note handler for proper resolution
-  }
-
-  // 2. Bare note: `note text` or `note left [of] X text`
-  const bareMatch = trimmed.match(NOTE_BARE);
-  if (bareMatch) {
-    // Capture group 1 guaranteed present after successful match.
-    const rest = bareMatch[1]!.trim();
-    const restLower = rest.toLowerCase();
-
-    // Check for positioned note: `note left/right ...`
-    if (restLower.startsWith('left') || restLower.startsWith('right')) {
-      const posWord = restLower.startsWith('left') ? 'left' : 'right';
-      const position = posWord as 'right' | 'left';
-      let afterPos = rest.substring(posWord.length).trim();
-
-      // Strip optional `of` keyword — track whether it was present
-      let hadOf = false;
-      if (afterPos.toLowerCase().startsWith('of ')) {
-        afterPos = afterPos.substring(3).trim();
-        hadOf = true;
-      }
-
-      if (!afterPos) {
-        // Just `note left` or `note right` — multi-line head
-        if (!lastMsgFrom) return { kind: 'skip' };
-        if (!participantIds.has(normalizeName(lastMsgFrom)))
-          return { kind: 'skip' };
-        // Verified — participantIds.has confirmed lastMsgFrom is registered.
-        return {
-          kind: 'multi-head',
-          position,
-          participantId: lastMsgFrom as ParticipantId,
-        };
-      }
-
-      // Try to match a known participant at the start of afterPos
-      const resolved = resolveParticipantAndText(
-        afterPos,
-        participants,
-        participantIds,
-        sortedParticipantsCache
-      );
-      if (resolved) {
-        if (resolved.text) {
-          return {
-            kind: 'single',
-            position,
-            participantId: resolved.participantId,
-            text: resolved.text,
-          };
-        } else {
-          // No text after participant — multi-line head
-          return {
-            kind: 'multi-head',
-            position,
-            participantId: resolved.participantId,
-          };
-        }
-      }
-
-      // No known participant matched.
-      // If `of` was explicit (`note right of Z ...`), the user intended a specific
-      // participant — skip when it doesn't exist rather than defaulting.
-      if (hadOf) return { kind: 'skip' };
-
-      // Without `of`, treat remaining text as note content on the last-msg sender
-      if (!lastMsgFrom) return { kind: 'skip' };
-      if (!participantIds.has(normalizeName(lastMsgFrom)))
-        return { kind: 'skip' };
-      return {
-        kind: 'single',
-        position,
-        // Verified by participantIds.has check above.
-        participantId: lastMsgFrom as ParticipantId,
-        text: afterPos,
-      };
-    }
-
-    // Plain `note text` — default position, last msg sender
-    if (!lastMsgFrom) return { kind: 'skip' };
-    if (!participantIds.has(normalizeName(lastMsgFrom)))
-      return { kind: 'skip' };
-    return {
-      kind: 'single',
-      position: 'right',
-      // Verified by participantIds.has check above.
-      participantId: lastMsgFrom as ParticipantId,
-      text: rest,
-    };
-  }
-
-  return null;
-}
-
-/**
- * Try to match a known participant name at the start of a string.
- * Returns the matched participant and remaining text, or null if no match.
- * Tries longest match first (multi-word participant names).
- */
-function resolveParticipantAndText(
-  input: string,
-  participants: SequenceParticipant[],
-  participantIds: Set<string>,
-  sortedParticipantsCache: SequenceParticipant[]
-): { participantId: ParticipantId; text: string } | null {
-  // Handle quoted participant: `"Auth Service" text`
-  if (input.startsWith('"') || input.startsWith("'")) {
-    // input[0] guaranteed by the startsWith check above.
-    const quote = input[0]!;
-    const endQuote = input.indexOf(quote, 1);
-    if (endQuote > 0) {
-      const name = input.substring(1, endQuote);
-      const key = normalizeName(name);
-      if (participantIds.has(key)) {
-        const text = input.substring(endQuote + 1).trim();
-        const found = participants.find((p) => normalizeName(p.id) === key);
-        // Verified by participantIds.has(key) above.
-        return { participantId: found?.id ?? (name as ParticipantId), text };
-      }
-    }
-    return null;
-  }
-
-  // Use pre-sorted participants (longest display id first) for greedy
-  // matching. Compare via the shared normalizer so 'auth service text' matches
-  // a participant declared as 'Auth Service'.
-  const sorted = sortedParticipantsCache;
-  for (const p of sorted) {
-    const idLen = p.id.length;
-    if (input.length < idLen) continue;
-    const candidate = input.substring(0, idLen);
-    if (normalizeName(candidate) !== normalizeName(p.id)) continue;
-    const remaining = input.substring(idLen);
-    // Must be followed by whitespace, end of string, or nothing
-    if (remaining === '' || remaining[0] === ' ' || remaining[0] === '\t') {
-      return { participantId: p.id, text: remaining.trim() };
-    }
-  }
-  return null;
-}
-
-/**
- * Parse a .dgmo file with `chart: sequence` into a structured representation.
- */
-export function parseSequenceDgmo(
-  content: string,
-  palette?: PaletteColors
-): ParsedSequenceDgmo {
-  // Diagram-level options accumulator (Readonly<Record<...>> on the public
-  // type; mutated locally during parse and assigned back via `result.options`).
-  const optionsAccumulator: Record<string, string> = {};
-  const result: Writable<ParsedSequenceDgmo> & {
-    options: Record<string, string>;
-  } = {
-    title: null,
-    titleLineNumber: null,
-    participants: [],
-    messages: [],
-    elements: [],
-    groups: [],
-    sections: [],
-    tagGroups: [],
-    options: optionsAccumulator,
-    diagnostics: [],
-    error: null,
-  };
-
-  // Per-parse entity alias literal → canonical participantId map
-  // (TD-18). Distinct from `aliasMap` further down, which is for
-  // tag-group aliases (per the A1 naming convention). Per C8:
-  // never persisted, fresh each parse.
-  const nameAliasMap = new Map<string, string>();
-  const resolveAlias = (token: string): string => {
-    const trimmed = token.trim();
-    return nameAliasMap.get(trimmed) ?? trimmed;
-  };
-
-  const fail = makeFail(result);
-
-  /** Push a recoverable error and continue parsing. */
-  const pushError = (line: number, message: string): void => {
-    const diag = makeDgmoError(line, message);
-    result.diagnostics.push(diag);
-    if (!result.error) result.error = formatDgmoError(diag);
-  };
-
-  /** Push a non-fatal warning (does not set result.error). */
-  const pushWarning = (line: number, message: string): void => {
-    result.diagnostics.push(makeDgmoError(line, message, 'warning'));
-  };
-
-  if (!content?.trim()) {
-    return fail(0, 'Empty content');
-  }
-
-  const lines = content.split('\n');
-  let hasExplicitChart = false;
-  let contentStarted = false;
-  // Whether the message body has begun (first message, section, block, or note).
-  // Unlike `contentStarted` — which any declaration trips to close the
-  // options/tag-group "headers first" window — `bodyStarted` stays false through
-  // the entire declaration preamble so bare and typed participant declarations
-  // can be freely interleaved. It gates bare-name declarations: once real body
-  // content appears, a bare word is treated as a stray line, not a participant.
-  let bodyStarted = false;
-  let firstLineIndex = -1; // line index of the `sequence [Title]` first line (to skip in main loop)
-
-  // Handle first non-empty, non-comment line for `sequence Title` syntax
-  for (let fi = 0; fi < lines.length; fi++) {
-    // In-bounds by loop guard.
-    const fl = lines[fi]!.trim();
-    if (!fl || fl.startsWith('//')) continue;
-    const parsed = parseFirstLine(fl);
-    if (parsed?.chartType === 'sequence') {
-      hasExplicitChart = true;
-      firstLineIndex = fi;
-      if (parsed.title) {
-        result.title = parsed.title;
-        result.titleLineNumber = fi + 1;
-      }
-    }
-    break;
-  }
-
-  // Group parsing state — tracks the active [Group] heading.
-  // Mutated during parse (participantIds.push); typed Writable so the
-  // local accumulator can grow before the readonly-typed value is exposed.
-  let activeGroup: Writable<SequenceGroup> | null = null;
-
-  // Fast lookup set for participant existence checks (mirrors result.participants).
-  // Holds NORMALIZED participant keys so 'Auth Service' and 'auth service' fold
-  // to the same entry. Display label survives in result.participants[i].label.
-  const participantIds = new Set<string>();
-
-  // Map<normalizedKey, first-seen participant.id> — the canonical
-  // identifier callers store on messages, group memberships, etc.
-  const idByKey = new Map<string, string>();
-
-  /**
-   * Get-or-create a participant by normalized key. Returns the FIRST-SEEN
-   * `participant.id` — which is the user's original casing/spacing.
-   * Callers should use the returned value on messages, group memberships,
-   * notes, etc. so all references resolve to the canonical id.
-   *
-   * Emits NAME_MERGED warning when an incoming source label normalizes to
-   * an existing key but its display form differs (case/whitespace).
-   */
-  const addParticipant = (
-    name: string,
-    lineNumber: number,
-    extras?: {
-      type?: ParticipantType;
-      position?: number;
-      label?: string;
-      metadata?: Record<string, string>;
-    }
-  ): ParticipantId => {
-    const key = normalizeName(name);
-    const trimmed = name.trim() as ParticipantId;
-    const incomingLabel = extras?.label ?? trimmed;
-    if (participantIds.has(key)) {
-      const existing = result.participants.find(
-        (p) => normalizeName(p.id) === key
-      );
-      if (existing) {
-        if (
-          displayName(existing.label) !== displayName(incomingLabel) ||
-          displayName(existing.id) !== displayName(trimmed)
-        ) {
-          result.diagnostics.push(
-            makeDgmoError(
-              lineNumber,
-              nameMergedMessage({
-                incomingDisplay: trimmed,
-                incomingLine: lineNumber,
-                existingDisplay: existing.id,
-                existingLine: existing.lineNumber,
-              }),
-              'warning',
-              NAME_DIAGNOSTIC_CODES.NAME_MERGED
-            )
-          );
-        }
-        return existing.id;
-      }
-      // Fallback: idByKey lookup hit but participant somehow missing.
-      return (idByKey.get(key) ?? trimmed) as ParticipantId;
-    }
-    participantIds.add(key);
-
-    idByKey.set(key, trimmed);
-    sortedCacheDirty = true;
-    result.participants.push({
-      id: trimmed,
-      label: incomingLabel,
-      type: extras?.type ?? inferParticipantType(name),
-      lineNumber,
-      ...(extras?.position !== undefined ? { position: extras.position } : {}),
-      // takePosition() may have emptied the metadata record (position was
-      // its only key) — don't attach an empty object.
-      ...(extras?.metadata && Object.keys(extras.metadata).length > 0
-        ? { metadata: extras.metadata }
-        : {}),
-    });
-    return trimmed;
-  };
-
-  // Cache sorted participants (longest ID first) for greedy name matching in notes.
-  // Invalidated whenever a new participant is added.
-  let sortedParticipantsCache: SequenceParticipant[] = [];
-  let sortedCacheDirty = true;
-
-  /** Get sorted participants (longest display label first), rebuilding only when dirty. */
-  const getSortedParticipants = (): SequenceParticipant[] => {
-    if (sortedCacheDirty) {
-      sortedParticipantsCache = [...result.participants].sort(
-        (a, b) => b.label.length - a.label.length
-      );
-      sortedCacheDirty = false;
-    }
-    return sortedParticipantsCache;
-  };
-
-  // Track participant → group name for duplicate membership detection
-  const participantGroupMap = new Map<string, string>();
-
-  // Tag group parsing state
-  let currentTagGroup: Writable<TagGroup> | null = null;
-  const aliasMap = new Map<string, string>();
-
-  /**
-   * Split same-line metadata from a line: "core k: v" (§1.4)
-   * → { core, meta }. Dispatched via splitNameAndMeta + SEQUENCE_REGISTRY.
-   */
-  const splitPipe = (
-    text: string,
-    ln?: number
-  ): { core: string; meta?: Record<string, string>; alias?: string } => {
-    // §1.4 unified metadata grammar — same-line cut, no pipe.
-    const registry = withTagAliases(
-      SEQUENCE_REGISTRY,
-      new Set(aliasMap.keys())
-    );
-    const split = splitNameAndMeta(text, registry, aliasMap);
-    if (ln != null) {
-      warnUnknownMetaKeys(
-        split.meta,
-        registry,
-        (msg) => pushWarning(ln, msg),
-        split.name
-      );
-    }
-    if (Object.keys(split.meta).length > 0) {
-      // splitNameAndMeta peels a trailing `as <alias>` off the name region
-      // when same-line metadata is present (the alias must precede the meta,
-      // which runs to EOL). Propagate it so callers can register the alias —
-      // the no-meta branch below leaves `as <alias>` in `core` for the
-      // caller's own peeling.
-      return {
-        core: split.name,
-        meta: split.meta,
-        ...(split.alias !== undefined && { alias: split.alias }),
-      };
-    }
-    return { core: text };
-  };
-
-  /**
-   * Pull the layout-order `position:` key out of a participant's
-   * parsed metadata (§2.2). Position is a layout directive, not display
-   * metadata, so it is *removed* from the record after extraction. A
-   * non-integer value raises an error and yields no position.
-   */
-  const takePosition = (
-    meta: Record<string, string> | undefined,
-    ln: number
-  ): number | undefined => {
-    if (meta?.['position'] === undefined) return undefined;
-    const raw = meta['position'];
-    delete meta['position'];
-    const n = parseInt(raw, 10);
-    if (!/^-?\d+$/.test(raw.trim()) || Number.isNaN(n)) {
-      pushError(ln, `'position: ${raw}' must be an integer slot index`);
-      return undefined;
-    }
-    return n;
-  };
-
-  // Block parsing state — blocks are built mutably during parse, then
-  // assigned back into the readonly-typed `ParsedSequenceDgmo`.
-  const blockStack: {
-    block: Writable<SequenceBlock>;
-    indent: number;
-    inElse: boolean;
-    activeElseIfBranch?: Writable<ElseIfBranch>;
-  }[] = [];
-  const currentContainer = (): SequenceElement[] => {
-    if (blockStack.length === 0) return result.elements;
-    // In-bounds by the length check above.
-    const top = blockStack[blockStack.length - 1]!;
-    if (top.activeElseIfBranch) return top.activeElseIfBranch.children;
-    return top.inElse ? top.block.elseChildren : top.block.children;
-  };
-
-  // Track last message sender for default note positioning
-  let lastMsgFrom: string | null = null;
-
-  for (let i = 0; i < lines.length; i++) {
-    // In-bounds by loop guard.
-    const raw = lines[i]!;
-    const trimmed = raw.trim();
-    const lineNumber = i + 1;
-
-    // Skip empty lines
-    if (!trimmed) {
-      activeGroup = null;
-      currentTagGroup = null;
-      continue;
-    }
-
-    // Skip first line already handled as `sequence [Title]`
-    if (i === firstLineIndex) continue;
-
-    // Parse group heading — [Group Name] or [Group Name] | k: v
-    const groupMatch = trimmed.match(GROUP_HEADING_PATTERN);
-    if (groupMatch) {
-      // Capture group 1 guaranteed present after successful match.
-      const groupName = groupMatch[1]!.trim();
-      const groupColor = groupMatch[2]?.trim();
-      let groupMeta: Record<string, string> | undefined;
-
-      // Parse metadata and collapse state AFTER the closing bracket
-      const afterBracket = groupMatch[3]?.trim() || '';
-      let isCollapsed = false;
-
-      // §1.4 same-line metadata on group header
-      if (afterBracket) {
-        const groupRegistry = withTagAliases(
-          SEQUENCE_REGISTRY,
-          new Set(aliasMap.keys())
-        );
-        const split = splitNameAndMeta(afterBracket, groupRegistry, aliasMap);
-        warnUnknownMetaKeys(
-          split.meta,
-          groupRegistry,
-          (msg) => pushWarning(lineNumber, msg),
-          split.name
-        );
-        if (Object.keys(split.meta).length > 0) groupMeta = split.meta;
-        if (groupMeta?.['collapsed']?.toLowerCase() === 'true') {
-          isCollapsed = true;
-          delete groupMeta['collapsed'];
-          if (Object.keys(groupMeta).length === 0) groupMeta = undefined;
-        }
-      }
-
-      if (groupColor) {
-        pushWarning(
-          lineNumber,
-          `'(${groupColor})' parens-color syntax removed from sequence diagrams — use 'tag:' groups for coloring`
-        );
-      }
-      contentStarted = true;
-      activeGroup = {
-        name: groupName,
-        participantIds: [],
-        lineNumber,
-        ...(groupMeta ? { metadata: groupMeta } : {}),
-        ...(isCollapsed ? { collapsed: true } : {}),
-      };
-      result.groups.push(activeGroup);
-      continue;
-    }
-
-    // Detect pipe-inside-brackets error: [Name | meta] → suggest [Name] | meta
-    if (trimmed.startsWith('[')) {
-      const fallbackMatch = trimmed.match(GROUP_HEADING_FALLBACK);
-      // Capture group 1 guaranteed present after successful match.
-      if (fallbackMatch && fallbackMatch[1]!.includes('|')) {
-        const rawInside = fallbackMatch[1]!;
-        const pipeIdx = rawInside.indexOf('|');
-        const cleanName = rawInside
-          .substring(0, pipeIdx)
-          .trim()
-          .replace(/\([^)]*\)$/, '')
-          .trim();
-        const metaPart = rawInside
-          .substring(pipeIdx)
-          .replace(/^\|\s*/, '')
-          .trim();
-        pushError(
-          lineNumber,
-          `Metadata goes after the bracket with no pipe — use '[${cleanName}] ${metaPart}' (e.g. '[Backend] t: Product'), not '[${rawInside.trim()}]'`
-        );
-        continue;
-      }
-    }
-
-    // Close active group on non-indented, non-group lines
-    if (activeGroup && measureIndent(raw) === 0) {
-      activeGroup = null;
-    }
-
-    // Skip comments — only // is supported
-    if (trimmed.startsWith('//')) continue;
-
-    // Reject # as comment syntax
-    if (trimmed.startsWith('#')) {
-      pushError(lineNumber, 'Use // for comments');
-      continue;
-    }
-
-    // ---- Tag group handling ----
-    // Tag block heading: "tag Name as <alias>"
-    const tagBlockMatch = matchTagBlockHeading(trimmed);
-    if (tagBlockMatch) {
-      if (contentStarted) {
-        pushError(lineNumber, 'Tag groups must appear before sequence content');
-        continue;
-      }
-      const newTagGroup: Writable<TagGroup> = {
-        name: tagBlockMatch.name,
-        ...(tagBlockMatch.alias !== undefined && {
-          alias: tagBlockMatch.alias,
-        }),
-        entries: [],
-        lineNumber,
-      };
-      currentTagGroup = newTagGroup;
-      if (tagBlockMatch.alias) {
-        aliasMap.set(
-          tagBlockMatch.alias.toLowerCase(),
-          tagBlockMatch.name.toLowerCase()
-        );
-      }
-      aliasMap.set(
-        normalizeName(tagBlockMatch.name),
-        tagBlockMatch.name.toLowerCase()
-      );
-      result.tagGroups.push(newTagGroup);
-      continue;
-    }
-
-    // Tag group entries (indented Value color under tag heading)
-    // First entry is the default unless another is marked `default`
-    if (currentTagGroup && !contentStarted && measureIndent(raw) > 0) {
-      const { text: cleanEntry, isDefault } = stripDefaultModifier(trimmed);
-      const { label, color } = extractColor(
-        cleanEntry,
-        palette,
-        result.diagnostics,
-        lineNumber
-      );
-      // Bare value (no explicit color) → keep it; finalized below.
-      if (isDefault) {
-        currentTagGroup.defaultValue = label;
-      } else if (currentTagGroup.entries.length === 0) {
-        currentTagGroup.defaultValue = label;
-      }
-      currentTagGroup.entries.push({
-        value: label,
-        color: color ?? AUTO_TAG_COLOR_SENTINEL,
-        lineNumber,
-      });
-      continue;
-    }
-
-    // Non-indented line after tag group — close it
-    if (currentTagGroup) {
-      currentTagGroup = null;
-    }
-
-    // Parse section dividers — "== Label ==" or "== Label(color) =="
-    // Close blocks first — sections at indent 0 should not nest inside blocks
-    const sectionMatch = trimmed.match(SECTION_PATTERN);
-    if (sectionMatch) {
-      const sectionIndent = measureIndent(raw);
-      while (blockStack.length > 0) {
-        // In-bounds by the length check above.
-        const top = blockStack[blockStack.length - 1]!;
-        if (sectionIndent > top.indent) break;
-        blockStack.pop();
-      }
-      // Capture group 1 guaranteed present after successful match.
-      const labelRaw = sectionMatch[1]!.trim();
-      // Scoped to recognized 11-name palette colors only (§1.5).
-      const colorMatch = labelRaw.match(
-        /^(.+?)\((red|orange|yellow|green|blue|purple|teal|cyan|gray|black|white)\)$/
-      );
-      if (colorMatch) {
-        pushWarning(
-          lineNumber,
-          `'(${colorMatch[2]})' parens-color syntax removed from sequence diagrams — use 'tag:' groups for coloring`
-        );
-      }
-      contentStarted = true;
-      bodyStarted = true;
-      const section: SequenceSection = {
-        kind: 'section',
-        // Capture group 1 guaranteed present after successful match.
-        label: colorMatch ? colorMatch[1]!.trim() : labelRaw,
-        lineNumber,
-      };
-      result.sections.push(section);
-      currentContainer().push(section);
-      continue;
-    }
-
-    // Parse header key: value lines (always top-level)
-    // Skip 'note' lines — parsed in the indent-aware section below
-    const colonIndex = trimmed.indexOf(':');
-    // §1.4: a real header key is a single identifier. Participant
-    // lines with same-line metadata (`CheckoutSvc o: Checkout`) have
-    // a space inside the pre-colon region — those are content lines,
-    // not header options.
-    const preColon =
-      colonIndex > 0 ? trimmed.substring(0, colonIndex).trim() : '';
-    const preColonIsSingleIdent = /^[A-Za-z][\w-]*$/.test(preColon);
-    if (
-      colonIndex > 0 &&
-      preColonIsSingleIdent &&
-      !trimmed.includes('->') &&
-      !trimmed.includes('~>') &&
-      !trimmed.includes('<-') &&
-      !trimmed.includes('<~') &&
-      !trimmed.includes('|')
-    ) {
-      const key = preColon.toLowerCase();
-      if (key === 'note' || key.startsWith('note ')) {
-        // Fall through to indent-aware note parsing below
-      } else {
-        const value = trimmed.substring(colonIndex + 1).trim();
-
-        // Enforce headers-before-content
-        if (contentStarted) {
-          pushError(
-            lineNumber,
-            `Options like '${key}: ${value}' must appear before the first message or declaration`
-          );
-          continue;
-        }
-
-        if (key === 'title') {
-          result.title = value;
-          result.titleLineNumber = lineNumber;
-          continue;
-        }
-
-        // Store other options
-        result.options[key] = value;
-        continue;
-      }
-    }
-
-    // Parse space-separated options (no colon): `no-activations`, `solid-fill`, `active-tag Priority`
-    {
-      const optLower = trimmed.toLowerCase();
-      // Negated boolean: `no-activations` → options.activations = 'off'
-      if (optLower.startsWith('no-')) {
-        const base = optLower.substring(3);
-        if (KNOWN_SEQ_BOOLEANS.has(base) || NO_PREFIX_ONLY_BOOLEANS.has(base)) {
-          if (contentStarted) {
-            pushError(
-              lineNumber,
-              `Options like '${trimmed}' must appear before the first message or declaration`
-            );
-            continue;
-          }
-          result.options[base] = 'off';
-          continue;
-        }
-      }
-      // Key-value option: `active-tag Priority`
-      const spaceMatch = trimmed.match(OPTION_NOCOLON_RE);
-      if (spaceMatch) {
-        // Capture groups 1 and 2 guaranteed present after successful match.
-        const optKey = spaceMatch[1]!.toLowerCase();
-        const optVal = spaceMatch[2]!.trim();
-        if (KNOWN_SEQ_OPTIONS.has(optKey) || KNOWN_SEQ_BOOLEANS.has(optKey)) {
-          if (contentStarted) {
-            pushError(
-              lineNumber,
-              `Options like '${trimmed}' must appear before the first message or declaration`
-            );
-            continue;
-          }
-          result.options[optKey] = optVal;
-          continue;
-        }
-      }
-      // Bare boolean keyword: `solid-fill` / `no-title`
-      if (
-        KNOWN_SEQ_BOOLEANS.has(optLower) &&
-        (optLower === 'solid-fill' || optLower === 'no-title')
-      ) {
-        if (contentStarted) {
-          pushError(
-            lineNumber,
-            `Options like '${trimmed}' must appear before the first message or declaration`
-          );
-          continue;
-        }
-        result.options[optLower] = 'on';
-        continue;
-      }
-    }
-
-    // Parse "Name is a type [position N]" declarations (always top-level).
-    // The legacy `aka Alias` modifier is no longer supported — Universal Name
-    // Handling makes aliasing unnecessary because forgiving normalization
-    // handles casing/whitespace differences automatically.
-    // Skip lines starting with 'note' — handled by note parsing below.
-    const {
-      core: isACore,
-      meta: isAMeta,
-      alias: isAAlias,
-    } = splitPipe(trimmed, lineNumber);
-    const isAMatch = !/^note(\s|$)/i.test(trimmed)
-      ? isACore.match(IS_A_PATTERN)
-      : null;
-    if (isAMatch) {
-      contentStarted = true;
-      // Capture groups 1 and 2 guaranteed present after successful match.
-      const id = isAMatch[1]!;
-      const typeStr = isAMatch[2]!.toLowerCase();
-      const remainder = isAMatch[3]?.trim() || '';
-
-      const participantType: ParticipantType = VALID_PARTICIPANT_TYPES.has(
-        typeStr
-      )
-        ? (typeStr as ParticipantType)
-        : 'default';
-
-      // TD-18: extract trailing `as <alias>` and register it. Order on the
-      // line is `Name is a TYPE as <alias> [position: N]`. When same-line
-      // metadata is present, splitPipe already peeled the alias (it must
-      // precede the metadata) and handed it back as `isAAlias`; otherwise
-      // the alias is still in `remainder` for us to peel here.
-      if (isAAlias !== undefined) {
-        nameAliasMap.set(isAAlias, id);
-      }
-      const asInRemainder = remainder.match(
-        /^(.*?)\s*\bas\s+([A-Za-z][A-Za-z0-9_]{0,11})\s*$/
-      );
-      if (asInRemainder) {
-        // Capture group 2 guaranteed present after successful match.
-        nameAliasMap.set(asInRemainder[2]!, id);
-      }
-
-      // Position is colon-keyed metadata (§2.2) — pull it from the parsed meta.
-      const position = takePosition(isAMeta, lineNumber);
-
-      // Avoid duplicate participant declarations
-      const key = addParticipant(id, lineNumber, {
-        type: participantType,
-        ...(position !== undefined && { position }),
-        ...(isAMeta !== undefined && { metadata: isAMeta }),
-      });
-      // Track group membership
-      if (activeGroup && !activeGroup.participantIds.includes(key)) {
-        const existingGroup = participantGroupMap.get(key);
-        if (existingGroup) {
-          pushError(
-            lineNumber,
-            `Participant '${id}' is already in group '${existingGroup}' — participants can only belong to one group`
-          );
-        } else {
-          activeGroup.participantIds.push(key);
-          // participantGroupMap is keyed by normalized participant key
-
-          participantGroupMap.set(key, activeGroup.name);
-        }
-      }
-      continue;
-    }
-
-    // Legacy `Name(color)` participant declaration at any level (§1.5 hard
-    // break). Scoped to the 11-name palette so `funcCall(arg)` doesn't trip.
-    const { core: colorCore, meta: colorMeta } = splitPipe(trimmed, lineNumber);
-    const coloredMatch = colorCore.match(COLORED_PARTICIPANT_PATTERN);
-    if (coloredMatch && !ARROW_PATTERN.test(colorCore)) {
-      // Capture groups 1 and 2 guaranteed present after successful match.
-      const id = coloredMatch[1]!;
-      const color = coloredMatch[2]!.trim();
-      pushError(
-        lineNumber,
-        `'${id}(${color})' parens-color syntax is no longer supported — use 'tag:' groups for coloring`
-      );
-      contentStarted = true;
-      const key = addParticipant(id, lineNumber, {
-        ...(colorMeta !== undefined && { metadata: colorMeta }),
-      });
-      if (activeGroup && !activeGroup.participantIds.includes(key)) {
-        const existingGroup = participantGroupMap.get(key);
-        if (existingGroup) {
-          pushError(
-            lineNumber,
-            `Participant '${id}' is already in group '${existingGroup}' — participants can only belong to one group`
-          );
-        } else {
-          activeGroup.participantIds.push(key);
-
-          participantGroupMap.set(key, activeGroup.name);
-        }
-      }
-      continue;
-    }
-
-    // Bare participant name — either inside an active group (indented) or top-level declaration
-    // Supports pipe metadata: "  API | c: Gateway" or "Tapin2 | l:Park"
-    {
-      const {
-        core: bareCore,
-        meta: bareMeta,
-        alias: bareAlias,
-      } = splitPipe(trimmed, lineNumber);
-      const inGroup = activeGroup && measureIndent(raw) > 0;
-      if (
-        /^\S+$/.test(bareCore) &&
-        !ARROW_PATTERN.test(bareCore) &&
-        (inGroup || !bodyStarted || bareMeta)
-      ) {
-        contentStarted = true;
-        const id = bareCore;
-        if (bareAlias !== undefined) nameAliasMap.set(bareAlias, id);
-        // Colon-keyed `position: N` (§2.2) arrives as metadata here.
-        const position = takePosition(bareMeta, lineNumber);
-        const key = addParticipant(id, lineNumber, {
-          ...(position !== undefined && { position }),
-          ...(bareMeta !== undefined && { metadata: bareMeta }),
-        });
-        if (activeGroup && !activeGroup.participantIds.includes(key)) {
-          const existingGroup = participantGroupMap.get(key);
-          if (existingGroup) {
-            pushError(
-              lineNumber,
-              `Participant '${id}' is already in group '${existingGroup}' — participants can only belong to one group`
-            );
-          } else {
-            activeGroup.participantIds.push(key);
-
-            participantGroupMap.set(key, activeGroup.name);
-          }
-        }
-        continue;
-      }
-    }
-
-    // ---- Indent-aware parsing for messages and block keywords ----
-    const indent = measureIndent(raw);
-
-    // Close blocks whose scope has ended (indent decreased)
-    while (blockStack.length > 0) {
-      // In-bounds by the length check above.
-      const top = blockStack[blockStack.length - 1]!;
-      if (indent > top.indent) break;
-      // Keep block on stack when 'else' or 'else if' matches current indent — handled below
-      if (
-        indent === top.indent &&
-        (top.block.type === 'if' || top.block.type === 'parallel')
-      ) {
-        const lower = trimmed.toLowerCase();
-        if (lower === 'else' || lower.startsWith('else if ')) break;
-      }
-      blockStack.pop();
-    }
-
-    // Split pipe metadata before arrow parsing (arrows use $ anchor)
-    const { core: arrowCore, meta: arrowMeta } = splitPipe(trimmed, lineNumber);
-
-    // Parse message lines first — arrows take priority over keywords
-    // Reject "async" keyword prefix — use ~> instead
-    const asyncPrefixMatch = arrowCore.match(/^async\s+(.+)$/i);
-    // Capture group 1 guaranteed present after successful match.
-    if (asyncPrefixMatch && ARROW_PATTERN.test(asyncPrefixMatch[1]!)) {
-      pushError(lineNumber, 'Use ~> for async messages: A ~> B: message');
-      continue;
-    }
-
-    // ---- Labeled arrows: -label->, ~label~> ----
-    // Must be checked BEFORE plain arrow patterns to avoid partial matches
-    const labeledArrow = parseArrow(arrowCore);
-    if (labeledArrow && 'error' in labeledArrow) {
-      pushError(lineNumber, labeledArrow.error);
-      continue;
-    }
-    if (labeledArrow) {
-      contentStarted = true;
-      bodyStarted = true;
-      const { from, to, label: rawLabel, async: isAsync } = labeledArrow;
-      const fromKey = addParticipant(resolveAlias(from), lineNumber);
-      const toKey = addParticipant(resolveAlias(to), lineNumber);
-      lastMsgFrom = fromKey;
-
-      // TD-13/TD-14: validate in-arrow label characters
-      const labelResult = parseInArrowLabel(rawLabel, lineNumber);
-      labelResult.diagnostics.forEach((d) => result.diagnostics.push(d));
-      const label = labelResult.label ?? rawLabel;
-
-      const msg: SequenceMessage = {
-        kind: 'message',
-        from: fromKey,
-        to: toKey,
-        label,
-        lineNumber,
-        ...(isAsync ? { async: true } : {}),
-        ...(arrowMeta ? { metadata: arrowMeta } : {}),
-      };
-      result.messages.push(msg);
-      currentContainer().push(msg);
-      continue;
-    }
-
-    // ---- Error: old colon-postfix syntax (A -> B: msg) ----
-    const colonPostfixSync = arrowCore.match(
-      /^(\S+)\s*->\s*([^\s:]+)\s*:\s*(.+)$/
-    );
-    const colonPostfixAsync = arrowCore.match(
-      /^(\S+)\s*~>\s*([^\s:]+)\s*:\s*(.+)$/
-    );
-    const colonPostfix = colonPostfixSync || colonPostfixAsync;
-    if (colonPostfix) {
-      // Capture groups 1-3 guaranteed present after successful match.
-      const a = colonPostfix[1]!;
-      const b = colonPostfix[2]!;
-      const msg = colonPostfix[3]!.trim();
-      const arrowChar = colonPostfixAsync ? '~' : '-';
-      const arrowEnd = colonPostfixAsync ? '~>' : '->';
-      pushError(
-        lineNumber,
-        `Colon syntax is no longer supported. Use '${a} ${arrowChar}${msg}${arrowEnd} ${b}' instead`
-      );
-      continue;
-    }
-
-    // ---- Error: plain bidirectional arrows (A <-> B, A <~> B) ----
-    const bidiPlainMatch = arrowCore.match(/^(.+?)\s*(?:<->|<~>)\s*(.+)/);
-    if (bidiPlainMatch) {
-      pushError(
-        lineNumber,
-        "Bidirectional arrows are no longer supported. Use two separate lines: 'A -msg-> B' and 'B -msg-> A'"
-      );
-      continue;
-    }
-
-    // ---- Deprecated bare return arrows: A <- B, A <~ B ----
-    const bareReturnSync = arrowCore.match(/^(.+?)\s*<-\s*(.+)$/);
-    const bareReturnAsync = arrowCore.match(/^(.+?)\s*<~\s*(.+)$/);
-    const bareReturn = bareReturnSync || bareReturnAsync;
-    if (bareReturn) {
-      const to = bareReturn[1];
-      const from = bareReturn[2];
-      pushError(
-        lineNumber,
-        `Left-pointing arrows are no longer supported. Write '${from} -> ${to}' instead`
-      );
-      continue;
-    }
-
-    // ---- Bare (unlabeled) call arrows: A -> B, A ~> B ----
-    const bareCallSync = arrowCore.match(/^(.+?)\s*->\s*(.+)$/);
-    const bareCallAsync = arrowCore.match(/^(.+?)\s*~>\s*(.+)$/);
-    const bareCall = bareCallSync || bareCallAsync;
-    if (bareCall) {
-      contentStarted = true;
-      bodyStarted = true;
-      // Capture groups 1 and 2 guaranteed present after successful match.
-      const from = bareCall[1]!;
-      const to = bareCall[2]!;
-      const fromKey = addParticipant(resolveAlias(from), lineNumber);
-      const toKey = addParticipant(resolveAlias(to), lineNumber);
-      lastMsgFrom = fromKey;
-
-      const msg: SequenceMessage = {
-        kind: 'message',
-        from: fromKey,
-        to: toKey,
-        label: '',
-        lineNumber,
-        ...(bareCallAsync ? { async: true } : {}),
-        ...(arrowMeta ? { metadata: arrowMeta } : {}),
-      };
-      result.messages.push(msg);
-      currentContainer().push(msg);
-      continue;
-    }
-
-    // Parse 'if <label>' block keyword
-    const ifMatch = trimmed.match(/^if\s+(.+)$/i);
-    if (ifMatch) {
-      contentStarted = true;
-      bodyStarted = true;
-      const block: Writable<SequenceBlock> = {
-        kind: 'block',
-        type: 'if',
-        // Capture group 1 guaranteed present after successful match.
-        label: ifMatch[1]!.trim(),
-        children: [],
-        elseChildren: [],
-        lineNumber,
-      };
-      currentContainer().push(block);
-      blockStack.push({ block, indent, inElse: false });
-      continue;
-    }
-
-    // Parse 'loop <label>' block keyword
-    const loopMatch = trimmed.match(/^loop\s+(.+)$/i);
-    if (loopMatch) {
-      contentStarted = true;
-      bodyStarted = true;
-      const block: Writable<SequenceBlock> = {
-        kind: 'block',
-        type: 'loop',
-        // Capture group 1 guaranteed present after successful match.
-        label: loopMatch[1]!.trim(),
-        children: [],
-        elseChildren: [],
-        lineNumber,
-      };
-      currentContainer().push(block);
-      blockStack.push({ block, indent, inElse: false });
-      continue;
-    }
-
-    // Parse 'parallel [label]' block keyword
-    const parallelMatch = trimmed.match(/^parallel(?:\s+(.+))?$/i);
-    if (parallelMatch) {
-      contentStarted = true;
-      bodyStarted = true;
-      const block: Writable<SequenceBlock> = {
-        kind: 'block',
-        type: 'parallel',
-        label: parallelMatch[1]?.trim() || '',
-        children: [],
-        elseChildren: [],
-        lineNumber,
-      };
-      currentContainer().push(block);
-      blockStack.push({ block, indent, inElse: false });
-      continue;
-    }
-
-    // Parse 'else if <label>' keyword (must come before bare 'else')
-    const elseIfMatch = trimmed.match(/^else\s+if\s+(.+)$/i);
-    if (elseIfMatch) {
-      if (
-        blockStack.length > 0 &&
-        // In-bounds by the length check above.
-        blockStack[blockStack.length - 1]!.indent === indent
-      ) {
-        // In-bounds by the length check above.
-        const top = blockStack[blockStack.length - 1]!;
-        if (top.block.type === 'parallel') {
-          pushError(
-            lineNumber,
-            "parallel blocks don't support else if — list all concurrent messages directly inside the block"
-          );
-          continue;
-        }
-        if (top.block.type === 'if') {
-          const branch: Writable<ElseIfBranch> = {
-            // Capture group 1 guaranteed present after successful match.
-            label: elseIfMatch[1]!.trim(),
-            children: [],
-            lineNumber,
-          };
-          // `elseIfBranches` is optional on SequenceBlock — Writable<>'s
-          // shallow strip doesn't unwrap optional readonly arrays, so seed
-          // and access through a typed-mutable local alias.
-          if (!top.block.elseIfBranches) top.block.elseIfBranches = [];
-          const branches = top.block.elseIfBranches as Writable<ElseIfBranch>[];
-          branches.push(branch);
-          top.activeElseIfBranch = branch;
-          top.inElse = false;
-        }
-      }
-      continue;
-    }
-
-    // Parse 'else' keyword (only applies to 'if' blocks)
-    if (trimmed.toLowerCase() === 'else') {
-      if (
-        blockStack.length > 0 &&
-        // In-bounds by the length check above.
-        blockStack[blockStack.length - 1]!.indent === indent
-      ) {
-        // In-bounds by the length check above.
-        const top = blockStack[blockStack.length - 1]!;
-        if (top.block.type === 'parallel') {
-          pushError(
-            lineNumber,
-            "parallel blocks don't support else — list all concurrent messages directly inside the block"
-          );
-          continue;
-        }
-        if (top.block.type === 'if') {
-          top.inElse = true;
-          delete top.activeElseIfBranch;
-          top.block.elseLineNumber = lineNumber;
-        }
-      }
-      continue;
-    }
-
-    // 'elif <label>' → suggest 'else if <label>'
-    const elifMatch = trimmed.match(/^elif\b\s*(.*)$/i);
-    if (elifMatch) {
-      // Capture group 1 guaranteed present after successful match.
-      const tailRaw = elifMatch[1]!.trim();
-      const tail = tailRaw ? ' ' + tailRaw : '';
-      pushError(
-        lineNumber,
-        `'elif' is not a keyword. Did you mean 'else if${tail}'?`
-      );
-      continue;
-    }
-
-    // 'else <text>' (where text isn't 'if …') → suggest 'else if <text>'
-    const elseLabelMatch = trimmed.match(/^else\s+(.+)$/i);
-    if (elseLabelMatch) {
-      // Capture group 1 guaranteed present after successful match.
-      const tail = elseLabelMatch[1]!.trim();
-      pushError(
-        lineNumber,
-        `'else' does not take a label. Did you mean 'else if ${tail}'?`
-      );
-      continue;
-    }
-
-    // ---- Note parsing (space-separated only) ----
-    // Strategy:
-    // 1. Try bare note: `note text` — position defaults, text is everything after `note`
-    // 2. For positioned: `note left [of] X text` — needs participant lookup to split name vs text
-    // 3. Multi-line: `note`, `note right`, `note right [of] X` (body indented below)
-    {
-      const noteParsed = parseNoteLine(
-        trimmed,
-        result.participants,
-        participantIds,
-        getSortedParticipants(),
-        lastMsgFrom
-      );
-      if (noteParsed) {
-        if (noteParsed.kind === 'single') {
-          const note: SequenceNote = {
-            kind: 'note',
-            text: noteParsed.text,
-            position: noteParsed.position,
-            participantId: noteParsed.participantId,
-            lineNumber,
-            endLineNumber: lineNumber,
-          };
-          bodyStarted = true;
-          currentContainer().push(note);
-          continue;
-        }
-        if (noteParsed.kind === 'multi-head') {
-          // Collect indented body lines
-          const noteLines: string[] = [];
-          while (i + 1 < lines.length) {
-            // In-bounds by the length check above.
-            const nextRaw = lines[i + 1]!;
-            const nextTrimmed = nextRaw.trim();
-            if (!nextTrimmed) break;
-            const nextIndent = measureIndent(nextRaw);
-            if (nextIndent <= indent) break;
-            noteLines.push(nextTrimmed);
-            i++;
-          }
-          if (noteLines.length === 0) continue; // no body yet — skip during live typing
-          const note: SequenceNote = {
-            kind: 'note',
-            text: noteLines.join('\n'),
-            position: noteParsed.position,
-            participantId: noteParsed.participantId,
-            lineNumber,
-            endLineNumber: i + 1, // i has advanced past the body lines (1-based)
-          };
-          bodyStarted = true;
-          currentContainer().push(note);
-          continue;
-        }
-        // 'skip' — note was incomplete (no preceding message, unknown participant)
-        continue;
-      }
-    }
-
-    // Catch-all: nothing matched this line
-    pushWarning(
-      lineNumber,
-      `Unexpected line: '${trimmed}'. Expected a message ('A -> B' or 'A -label-> B'), participant ('Name' or 'Name is a actor'), note ('note right of A Text'), section ('== Name =='), or block ('if/loop/parallel <label>').`
-    );
-  }
-
-  // Validate: if no explicit chart line, check for arrow-based inference
-  if (!hasExplicitChart && result.messages.length === 0) {
-    // Check if raw content has arrow patterns for inference
-    const hasArrows = lines.some((line) => ARROW_PATTERN.test(line.trim()));
-    if (!hasArrows) {
-      return fail(1, 'No "sequence" header and no sequence content detected');
-    }
-  }
-
-  // Warn about unused participants (only when the diagram has messages)
-  if (result.messages.length > 0) {
-    const usedIds = new Set<string>();
-    for (const msg of result.messages) {
-      // msg.from / msg.to / el.participantId are normalized participant keys
-      // (set via addParticipant during parsing)
-      // eslint-disable-next-line name-normalize/required-at-insertion
-      usedIds.add(msg.from);
-      // eslint-disable-next-line name-normalize/required-at-insertion
-      usedIds.add(msg.to);
-    }
-    // Walk elements recursively to find note participant references
-    const walkElements = (elements: readonly SequenceElement[]): void => {
-      for (const el of elements) {
-        if (isSequenceNote(el)) {
-          // eslint-disable-next-line name-normalize/required-at-insertion
-          usedIds.add(el.participantId);
-        } else if (isSequenceBlock(el)) {
-          walkElements(el.children);
-          walkElements(el.elseChildren);
-          if (el.elseIfBranches) {
-            for (const branch of el.elseIfBranches) {
-              walkElements(branch.children);
-            }
-          }
-        }
-      }
-    };
-    walkElements(result.elements);
-
-    for (const p of result.participants) {
-      if (!usedIds.has(p.id)) {
-        pushWarning(
-          p.lineNumber,
-          `Participant "${p.label}" is declared but never used in any message or note`
-        );
-      }
-    }
-  }
-
-  // Warn about empty groups
-  for (const group of result.groups) {
-    if (group.participantIds.length === 0) {
-      pushWarning(
-        group.lineNumber,
-        `Empty group '${group.name}' — did you mean '== ${group.name} ==' for a section divider?`
-      );
-    }
-  }
-
-  // Assign palette colors to bare (colorless) tag values.
-  finalizeAutoTagColors(result.tagGroups as Writable<TagGroup>[], palette);
-
-  // Validate tag group values on participants and messages
-  if (result.tagGroups.length > 0) {
-    const entities: Array<{
-      metadata: Record<string, string>;
-      lineNumber: number;
-    }> = [];
-    for (const p of result.participants) {
-      if (p.metadata)
-        entities.push({ metadata: p.metadata, lineNumber: p.lineNumber });
-    }
-    for (const m of result.messages) {
-      if (m.metadata)
-        entities.push({ metadata: m.metadata, lineNumber: m.lineNumber });
-    }
-    for (const g of result.groups) {
-      if (g.metadata)
-        entities.push({ metadata: g.metadata, lineNumber: g.lineNumber });
-    }
-    validateTagValues(entities, result.tagGroups, pushWarning, suggest);
-    validateTagGroupNames(result.tagGroups, pushWarning, pushError);
-  }
-
-  return result;
-}
-
-/**
- * Detect whether raw content looks like a sequence diagram.
- * Used by the chart type inference logic.
- */
-export function looksLikeSequence(content: string): boolean {
-  if (!content) return false;
-  const lines = content.split('\n');
-  return lines.some((line) => {
-    const trimmed = line.trim();
-    if (trimmed.startsWith('//')) return false;
-    return ARROW_PATTERN.test(trimmed);
-  });
-}