tools/eu4_clausewitz/src/node.mjs

// The unified API over a position-preserving tree (ast.mjs). One class for
// ALL Clausewitz data — provinces, countries, cultures, events, whatever —
// because it's all the same shape: a positioned key-value tree where
// duplicate keys and duplicate date blocks are normal and must never be
// collapsed.
//
// Mutating methods (`set`/`add`/`remove`) never touch `this.items` and never
// synthesize a new tree: they compute a byte-offset splice and hand it to
// this file's EditSession (edit.mjs). Reading a Node again after mutating it
// through it will NOT reflect the edit — the source of truth after a
// mutation is the pending splice list, not the in-memory tree. This is
// deliberate: the whole point of this package is that text output is
// derived only from {original text + splices}, never from re-walking a
// tree, so the tree is not kept in sync with pending edits.

const DATE_KEY_RE = /^(\d{1,4})\.(\d{1,2})\.(\d{1,2})$/;

/**
 * Whether a string is a Clausewitz date key (e.g. "1444.11.11"). Exported so
 * callers/tests can reuse the exact same rule `at()` uses internally.
 *
 * @param {string} value
 * @returns {boolean}
 */
export function isDateKey(value) {
  return DATE_KEY_RE.test(value);
}

function parseDate(value) {
  const m = DATE_KEY_RE.exec(value);
  if (!m) return null;
  return { year: Number(m[1]), month: Number(m[2]), day: Number(m[3]) };
}

function compareDates(a, b) {
  return (a.year - b.year) || (a.month - b.month) || (a.day - b.day);
}

export class Node {
  /**
   * @param {Object} opts
   * @param {Array} [opts.items] - this scope's items (pair|bare), from ast.mjs. Absent in resolved (at()) mode.
   * @param {import('./lexer.mjs').Token[]} opts.tokens - full token array for the file
   * @param {string} opts.text - full original text for the file
   * @param {string} opts.path - absolute path of the file this node belongs to
   * @param {'vanilla'|'mod'} [opts.layer]
   * @param {import('./edit.mjs').EditSession|null} opts.session - null for read-only (at()) nodes
   * @param {boolean} [opts.readonly]
   * @param {import('./lexer.mjs').Token|null} [opts.lbraceToken] - null for the root node
   * @param {import('./lexer.mjs').Token|null} [opts.rbraceToken] - null for root, or an unclosed block
   * @param {Map|null} [opts.resolved] - resolved-view backing map, set only by at()
   * @param {{start:number,end:number}|null} [opts.baseSpan] - span to report as .raw when resolved
   */
  constructor(opts) {
    this.items = opts.items ?? null;
    this.tokens = opts.tokens;
    this.text = opts.text;
    this.path = opts.path;
    this.layer = opts.layer;
    this.session = opts.session ?? null;
    this.readonly = Boolean(opts.readonly);
    this.lbraceToken = opts.lbraceToken ?? null;
    this.rbraceToken = opts.rbraceToken ?? null;
    this._resolved = opts.resolved ?? null;

    if (this._resolved) {
      this._span = opts.baseSpan ?? { start: 0, end: this.text.length };
    } else if (this.lbraceToken) {
      const endTok = this.rbraceToken ?? this.tokens[this.tokens.length - 1];
      this._span = {
        start: this.lbraceToken.index,
        end: this.rbraceToken
          ? this.rbraceToken.index + this.rbraceToken.length
          : endTok.index + endTok.length,
      };
    } else {
      this._span = { start: 0, end: this.text.length };
    }
  }

  // ---- internal helpers -------------------------------------------------

  _matching(key) {
    return this.items.filter((it) => it.kind === 'pair' && it.keyToken.value === key);
  }

  _unwrap(valueNode) {
    if (valueNode.type === 'scalar') {
      return valueNode.token ? valueNode.token.value : undefined;
    }
    return this._wrapBlock(valueNode);
  }

  _wrapBlock(blockValue) {
    const lbraceToken = this.tokens[blockValue.startTokenIndex];
    const rbraceToken = blockValue.closed ? this.tokens[blockValue.endTokenIndex] : null;
    return new Node({
      items: blockValue.items,
      tokens: this.tokens,
      text: this.text,
      path: this.path,
      layer: this.layer,
      session: this.session,
      readonly: this.readonly,
      lbraceToken,
      rbraceToken,
    });
  }

  _toResolvedEntry(valueNode) {
    if (valueNode.type === 'scalar') {
      return { kind: 'scalar', value: valueNode.token ? valueNode.token.value : undefined };
    }
    return { kind: 'block', value: this._wrapBlock(valueNode) };
  }

  // For add_X/remove_X verbs, the operand is expected to be a plain scalar
  // (a tag, a culture id, etc.) in every real case measured. If some future
  // data ever puts a block there, fall back to its raw source slice rather
  // than throw — at() must never crash on unexpected-but-real shapes.
  _scalarOf(valueNode) {
    if (valueNode.type === 'scalar') {
      return valueNode.token ? valueNode.token.value : '';
    }
    const start = this.tokens[valueNode.startTokenIndex];
    const end = this.tokens[valueNode.endTokenIndex];
    return this.text.slice(start.index, end.index + end.length);
  }

  _assertWritable() {
    if (this.readonly) {
      throw new Error(
        'cannot mutate a node produced by at() — it is a read-only resolved view; '
        + 'edit the underlying block via node.blocks() instead',
      );
    }
  }

  _quoteIfNeeded(value, wasQuoted) {
    const needsQuote = wasQuoted || /\s/.test(value);
    if (!needsQuote) return value;
    if (value.includes('"')) {
      throw new Error(`cannot quote value containing a double-quote character: ${JSON.stringify(value)}`);
    }
    return `"${value}"`;
  }

  _locate(index) {
    let line = 1;
    let col = 1;
    for (let i = 0; i < index && i < this.text.length; i += 1) {
      if (this.text[i] === '\n') {
        line += 1;
        col = 1;
      } else {
        col += 1;
      }
    }
    return { line, col };
  }

  // ---- the 12-method API (+ bareValues, see below) -----------------------

  get(key) {
    if (this._resolved) {
      const entry = this._resolved.get(key);
      if (!entry) return undefined;
      if (entry.kind === 'accum') {
        throw new Error(`get(): "${key}" is a date-accumulated set with ${entry.set.size} value(s); use all()`);
      }
      return entry.value;
    }
    const matches = this._matching(key);
    if (matches.length > 1) {
      throw new Error(`get(): key "${key}" occurs ${matches.length} times; use all()`);
    }
    if (matches.length === 0) return undefined;
    return this._unwrap(matches[0].value);
  }

  all(key) {
    if (this._resolved) {
      const entry = this._resolved.get(key);
      if (!entry) return [];
      if (entry.kind === 'accum') return [...entry.set];
      return [entry.value];
    }
    return this._matching(key).map((m) => this._unwrap(m.value));
  }

  has(key) {
    if (this._resolved) {
      const entry = this._resolved.get(key);
      if (!entry) return false;
      if (entry.kind === 'accum') return entry.set.size > 0;
      return true;
    }
    return this._matching(key).length > 0;
  }

  keys() {
    if (this._resolved) {
      return [...this._resolved.keys()].filter((k) => this.has(k));
    }
    return this.items.filter((it) => it.kind === 'pair').map((it) => it.keyToken.value);
  }

  // Not one of the spec's headline 12 methods, but necessary: a block whose
  // contents are a bare-value list (`historical_idea_groups = { economic_ideas
  // offensive_ideas }`) or an RGB triple (`color = { 20 50 210 }`) has no
  // keys at all — `keys()` on it is correctly `[]`. This is the reader for
  // that shape. Not meaningful on an at() view (date blocks never hold bare
  // lists in real data), so it returns [] there rather than throwing.
  bareValues() {
    if (this._resolved) return [];
    return this.items.filter((it) => it.kind === 'bare').map((it) => it.token.value);
  }

  block(key) {
    if (this._resolved) {
      const entry = this._resolved.get(key);
      if (!entry) return undefined;
      if (entry.kind === 'accum') {
        throw new Error(`block(): "${key}" is a date-accumulated set, not a block`);
      }
      if (entry.kind !== 'block') {
        throw new TypeError(`block(): value for key "${key}" is not a block`);
      }
      return entry.value;
    }
    const matches = this._matching(key);
    if (matches.length > 1) {
      throw new Error(`block(): key "${key}" occurs ${matches.length} times; use blocks()`);
    }
    if (matches.length === 0) return undefined;
    const value = matches[0].value;
    if (value.type !== 'block') {
      throw new TypeError(`block(): value for key "${key}" is not a block`);
    }
    return this._wrapBlock(value);
  }

  blocks(key) {
    if (this._resolved) {
      if (key === undefined) {
        const out = [];
        for (const entry of this._resolved.values()) {
          if (entry.kind === 'block') out.push(entry.value);
        }
        return out;
      }
      const entry = this._resolved.get(key);
      if (!entry) return [];
      if (entry.kind !== 'block') {
        throw new TypeError(`blocks(): value for key "${key}" is not a block`);
      }
      return [entry.value];
    }
    if (key === undefined) {
      return this.items
        .filter((it) => it.kind === 'pair' && it.value.type === 'block')
        .map((it) => this._wrapBlock(it.value));
    }
    return this._matching(key).map((m) => {
      if (m.value.type !== 'block') {
        throw new TypeError(`blocks(): value for key "${key}" is not a block (found a scalar)`);
      }
      return this._wrapBlock(m.value);
    });
  }

  set(key, value) {
    this._assertWritable();
    if (typeof value !== 'string') throw new TypeError('set(): value must be a string');
    const matches = this._matching(key);
    if (matches.length !== 1) {
      throw new Error(
        `set(): requires exactly one existing entry for key "${key}"; found ${matches.length} `
        + '(use add() to create one)',
      );
    }
    const valueNode = matches[0].value;
    if (valueNode.type !== 'scalar') {
      throw new Error(`set(): value for key "${key}" is a block; edit its contents via block() instead`);
    }
    const token = valueNode.token;
    if (!token) throw new Error(`set(): key "${key}" has no value token to replace (malformed entry)`);
    const raw = this._quoteIfNeeded(value, token.kind === 'string');
    this.session.splice(token.index, token.length, raw);
  }

  add(key, value) {
    this._assertWritable();
    if (typeof value !== 'string') throw new TypeError('add(): value must be a string');
    const raw = this._quoteIfNeeded(value, false);
    const entryText = `${key} = ${raw}`;

    if (this.items.length > 0) {
      const lastItem = this.items[this.items.length - 1];
      const lastToken = this.tokens[lastItem.endTokenIndex];
      const insertAt = lastToken.index + lastToken.length;
      const beforeTok = this.tokens[lastItem.startTokenIndex - 1];
      const leading = (beforeTok && beforeTok.kind === 'whitespace') ? beforeTok.raw : '\n';
      this.session.splice(insertAt, 0, leading + entryText);
      return;
    }

    // Empty scope: no existing entry to copy indentation style from. This
    // path has no real-data precedent to match, so it's a documented
    // heuristic rather than a derived rule.
    if (this.lbraceToken) {
      const insertAt = this.lbraceToken.index + this.lbraceToken.length;
      this.session.splice(insertAt, 0, ` ${entryText} `);
    } else {
      this.session.splice(this.text.length, 0, `${entryText}\n`);
    }
  }

  remove(key, value) {
    this._assertWritable();
    const matches = this.items.filter((it) => it.kind === 'pair' && it.keyToken.value === key
      && (value === undefined || this._unwrap(it.value) === value));

    for (const m of matches) {
      const startTok = this.tokens[m.startTokenIndex];
      const endTok = this.tokens[m.endTokenIndex];
      let spanStart = startTok.index;
      let spanEnd = endTok.index + endTok.length;

      // Consume one adjacent whitespace run so deletion doesn't leave a
      // blank line: prefer the run right after (up to and including its
      // first newline, so an intentional blank line further down survives),
      // falling back to the run right before when this is the last entry
      // in the scope (nothing whitespace-shaped follows it).
      const afterTok = this.tokens[m.endTokenIndex + 1];
      if (afterTok && afterTok.kind === 'whitespace') {
        const nl = afterTok.raw.indexOf('\n');
        spanEnd += nl !== -1 ? nl + 1 : afterTok.length;
      } else {
        const beforeTok = this.tokens[m.startTokenIndex - 1];
        if (beforeTok && beforeTok.kind === 'whitespace') {
          spanStart -= beforeTok.length;
        }
      }

      this.session.splice(spanStart, spanEnd - spanStart, '');
    }

    return matches.length;
  }

  at(dateStr) {
    if (this._resolved) {
      throw new Error('at(): cannot call at() on a node that is already an at() view');
    }
    const target = parseDate(dateStr);
    if (!target) {
      throw new Error(`at(): "${dateStr}" is not a valid date key (expected e.g. "1444.11.11")`);
    }

    const resolved = new Map();

    // Applies one pair item's verb semantics (plain overwrite, or add_X /
    // remove_X accumulation into bucket X) into `resolved`. Used for BOTH
    // this node's own bare (pre-date) top-level fields and every date
    // block's inner items — `add_core = FRI` sitting bare at the top of a
    // province file is exactly as much an accumulation into bucket `core`
    // as a later `add_core` inside a date block is; there is nothing
    // date-block-specific about the verb, only about which entries qualify
    // by date. This is what makes duplicate top-level `add_core`/`add_core`
    // entries (confirmed normal in real data, e.g. Toledo's TLD+CAS cores)
    // resolve correctly through at() instead of colliding as if plain
    // overwrites of the same key.
    //
    // None of this affects `node.get('add_core')`/`node.get('owner')` on
    // THIS node (no at()) — those never call this method at all, so the
    // literal bare fields stay reachable through the normal read path
    // completely unaffected by at()'s bucket renaming or overwrite
    // resolution.
    const applyItem = (inner) => {
      if (inner.kind !== 'pair') return;
      const k = inner.keyToken.value;

      if (k.startsWith('add_')) {
        const bucket = k.slice(4);
        let entry = resolved.get(bucket);
        if (!entry || entry.kind !== 'accum') entry = { kind: 'accum', set: new Set() };
        entry.set.add(this._scalarOf(inner.value));
        resolved.set(bucket, entry);
      } else if (k.startsWith('remove_')) {
        const bucket = k.slice(7);
        const entry = resolved.get(bucket);
        if (entry && entry.kind === 'accum') entry.set.delete(this._scalarOf(inner.value));
        // Removing something never added: no-op, per spec.
      } else {
        resolved.set(k, this._toResolvedEntry(inner.value));
      }
    };

    for (const it of this.items) {
      if (it.kind !== 'pair' || isDateKey(it.keyToken.value)) continue;
      applyItem(it);
    }

    const dateBlocks = this.items
      .map((it, idx) => ({ it, idx }))
      .filter(({ it }) => it.kind === 'pair' && isDateKey(it.keyToken.value) && it.value.type === 'block');

    // Chronological order; original file order breaks ties (duplicate date
    // blocks are normal in real data — see history/countries/FRA - France.txt).
    dateBlocks.sort((a, b) => {
      const cmp = compareDates(parseDate(a.it.keyToken.value), parseDate(b.it.keyToken.value));
      return cmp !== 0 ? cmp : a.idx - b.idx;
    });

    for (const { it } of dateBlocks) {
      const blockDate = parseDate(it.keyToken.value);
      if (compareDates(blockDate, target) > 0) continue;

      for (const inner of it.value.items) applyItem(inner);
    }

    return new Node({
      tokens: this.tokens,
      text: this.text,
      path: this.path,
      layer: this.layer,
      session: null,
      readonly: true,
      resolved,
      baseSpan: this._span,
    });
  }

  get raw() {
    const { line, col } = this._locate(this._span.start);
    return {
      text: this.text.slice(this._span.start, this._span.end),
      index: this._span.start,
      line,
      col,
      length: this._span.end - this._span.start,
    };
  }

  get file() {
    return { path: this.path, layer: this.layer };
  }
}