@wcstack/state 3.5.4 → 4.0.0-rc.2

package/dist/parser.esm.js

@@ -1,878 +1,709 @@
-const DELIMITER = '.';
-const WILDCARD = '*';
-const MAX_WILDCARD_DEPTH = 128;
-/**
- * 1 本のパスが持てるセグメント数の上限（`getPathInfo` が初回 intern のときだけ検査する）。
- *
- * `PathInfo` は自分の**全ての接頭辞**を intern するので、深さ N のパス 1 本で N 個の
- * `PathInfo` ができ、それぞれが長さ k の配列と Set を持つ ＝ 時間もメモリも O(N²)。
- * 実測（Node 22・パス 1 本）で depth 400 → 145ms / 61MB、800 → 1162ms / 424MB、
- * 2000 → 既定 4GB ヒープで OOM。**約 4KB の `data-wcs` 属性値 1 つでタブを落とせた。**
- * ここで打ち切ると最悪でも 256² ≒ 65k 単位に収まる。
- *
- * 値は `MAX_WILDCARD_DEPTH`（128 = `$1..$128` の表の大きさ）を**使い切れる**ことから決めた:
- * `$recursion` の展開（`recursion/expand.ts`）が深さ 128 で作るパスは `a.*.b.*.…` の形で
- * 257 セグメントを超える。そこを割ると、ドキュメント済みの再帰深さが打ち切られてしまう。
- * 512 なら最悪でも 512² ≒ 262k 単位で、上の実測の 400 段（145ms / 61MB）の延長に収まる。
- * 実用のパスは 10 段に満たない。
- */
-const MAX_PATH_SEGMENTS = 512;
-// data-wcs バインディング構文 `[prop][#mod]: [path][|filter...]` の区切り文字（単一正本・`@state` は v2 で撤去）。
-// これらは「死守の壁（構文契約）」であり値は不変。manifest.syntax.delimiters で公開される。
-const BINDING_SEPARATOR = ';'; // 複数バインディングの区切り
-const PROP_VALUE_SEPARATOR = ':'; // 左辺(prop)と右辺(path)の区切り
-const MODIFIER_SEPARATOR = '#'; // prop と修飾子の区切り
-const FILTER_SEPARATOR = '|'; // フィルタパイプの区切り
-// bindingType 判別と左辺 namespace の語彙（単一正本）。manifest.syntax.bindingTypes で
-// 公開される。パーサ（parseBindTextsForElement）とイベント層はこの定数に分岐する。
-// apply 層のディスパッチマップ（apply/applyChange.ts の applyChangeByFirstSegment）の
-// キー集合との一致は __tests__/manifest.test.ts の drift テストが強制する —
-// manifest エントリ（DOM 非依存）から apply 層を import しないための分離。
-const ELSE_KEYWORD = 'else';
-const SPREAD_PROP = '...';
-const EVENT_PROP_PREFIX = 'on';
-const EVENT_TOKEN_NAMESPACE = 'eventToken';
-/**
- * `<wcs-state mount>` の左辺 `state.<key>: path` — ボリュームの注入口（要件 B14③・3.x 計画 D28）。
- * 束縛ではなくマウントの宣言なので、束縛の収集（getParseBindTextResults）は作らず、ボリュームが接ぎ木の時に読む
- */
-const VOLUME_INJECTION_PROP = 'state';
-const COMMAND_NAMESPACE = 'command';
-const CLASS_NAMESPACE = 'class';
-const ATTR_NAMESPACE = 'attr';
-const STYLE_NAMESPACE = 'style';
-// リストインデックス参照名（`$1`..`$N`）の接頭辞（単一正本）。
-// manifest.syntax.indexParam で公開される。
-const INDEX_PARAM_PREFIX = '$';
-/**
- * stackIndexByIndexName
- * インデックス名からスタックインデックスへのマッピング
- * $1 => 0
- * $2 => 1
- * :
- * ${i + 1} => i
- * i < MAX_WILDCARD_DEPTH
- */
-const tmpIndexByIndexName = {};
-for (let i = 0; i < MAX_WILDCARD_DEPTH; i++) {
-    tmpIndexByIndexName[`${INDEX_PARAM_PREFIX}${i + 1}`] = i;
-}
-Object.freeze(tmpIndexByIndexName);
-/**
- * 再帰ワイルドカード。オーサリング層（$recursion 宣言・getter キー・API 引数）にだけ
- * 現れ、PathInfo には決して降ろさない — wildcardCount が不定になると ListIndex 連鎖長・
- * $1..$n・$resolve の厳密一致・走査の段数が同時に壊れる
- * （docs/state-recursive-path-design.md §2-1）。
- */
-const RECURSION_WILDCARD = "**";
-
-function raiseError(message) {
-    throw new Error(`[@wcstack/state] ${message}`);
-}
+// src/hooks.ts
+var hooks = {};
 
-const _cache = new Map();
-/**
- * **tooling 専用**（`@wcstack/state/parser` の clearParserCaches からのみ呼ぶ）。
- * ランタイム文脈で呼んではならない — PathInfo のインスタンス同一性は正規化キー
- * （依存グラフ・アドレス比較）の前提であり、クリアすると同一パスの新旧インスタンスが
- * 併存して identity 比較が黙って壊れる。言語サーバー等の長時間プロセスが、編集中の
- * 中間パス（`user.n` 等）の恒久 intern によるメモリ単調増加を断つための出口。
- */
-function clearPathInfoCacheForTooling() {
-    _cache.clear();
-}
-let id = 0;
-function getPathInfo(path) {
-    let pathInfo = _cache.get(path);
-    if (typeof pathInfo !== "undefined") {
-        return pathInfo;
-    }
-    // 再帰ワイルドカードはオーサリング層の記号で、ここへ降りてきてはならない
-    // （降ろすと wildcardCount が不定になり ListIndex 連鎖長・$1..$n・$resolve の
-    //  厳密一致・走査の段数が同時に壊れる。設計書 D2）。到達したということは、
-    // `**` を解釈しない消費者に `**` パスが渡ったということ。通常のパスはこの検査を
-    // 初回 intern のときにしか払わない（`**` パスは intern されないので読むたびに落ちる）。
-    if (path.indexOf(RECURSION_WILDCARD) !== -1) {
-        raiseError(`[wcs/recursion-unsupported] "${path}" uses "${RECURSION_WILDCARD}", which is not accepted here. ` +
-            `It is only meaningful in a $recursion declaration, in a recursive getter key, and in the path ` +
-            `argument of $getAll / $setAll — and only when the state declares a $recursion anchor.`);
-    }
-    // 深さの上限（初回 intern のときだけ払う）。`PathInfo` は**全ての接頭辞**を intern するので
-    // 深さ N のパス 1 本で時間もメモリも O(N²) になり、上限が無いと約 4KB の属性値 1 つで
-    // タブが落ちた（実測値は `MAX_PATH_SEGMENTS` の注記）。
-    //
-    // **ワイルドカードの段数はここでは見ない。** 段数の上限（`MAX_WILDCARD_DEPTH` ＝ manifest の
-    // `indexParam.maxDepth`）は `$1..$N` の表が引けるかという別の話で、`recursion/expand.ts` の
-    // `[wcs/recursion-depth-exceeded]` と `proxy/traps/get.ts` の `$N` 範囲外診断が、原因を
-    // 名指しできる場所で受け持っている（番人: `integration.recursionPrerequisites.test.ts` が
-    // 「PathInfo 自体には段数の上限が無い」を固定している）。ここで先に落とすと、深さ超過か
-    // 循環かの切り分けが効かなくなる
-    const segmentCount = countSegments(path);
-    if (segmentCount > MAX_PATH_SEGMENTS) {
-        raiseError(`[wcs/binding-syntax] "${path}" has ${segmentCount} path segments — the limit is ${MAX_PATH_SEGMENTS}. ` +
-            `Every prefix of a path is interned, so the cost grows with the square of the depth.`);
+// src/diagnostics/guidance.ts
+function editDistance(a, b, max) {
+  if (Math.abs(a.length - b.length) > max) {
+    return max + 1;
+  }
+  const prev = new Array(b.length + 1);
+  const curr = new Array(b.length + 1);
+  for (let j = 0; j <= b.length; j++) {
+    prev[j] = j;
+  }
+  for (let i = 1; i <= a.length; i++) {
+    curr[0] = i;
+    for (let j = 1; j <= b.length; j++) {
+      const cost = a[i - 1] === b[j - 1] ? 0 : 1;
+      curr[j] = Math.min(prev[j] + 1, curr[j - 1] + 1, prev[j - 1] + cost);
     }
-    pathInfo = Object.freeze(new PathInfo(path));
-    _cache.set(path, pathInfo);
-    return pathInfo;
-}
-/** `.` の数 + 1。`split` の配列を作らずに数える（初回 intern のときだけ通る） */
-function countSegments(path) {
-    let count = 1;
-    for (let i = 0; i < path.length; i++) {
-        if (path[i] === DELIMITER) {
-            count++;
-        }
+    for (let j = 0; j <= b.length; j++) {
+      prev[j] = curr[j];
     }
-    return count;
+  }
+  return prev[b.length];
 }
-class PathInfo {
-    id = ++id;
-    path;
-    segments;
-    lastSegment;
-    cumulativePaths;
-    cumulativePathSet;
-    cumulativePathInfos;
-    cumulativePathInfoSet;
-    parentPath;
-    wildcardPaths;
-    wildcardPathSet;
-    indexByWildcardPath;
-    wildcardPathInfos;
-    wildcardPathInfoSet;
-    wildcardParentPaths;
-    wildcardParentPathSet;
-    wildcardParentPathInfos;
-    wildcardParentPathInfoSet;
-    wildcardPositions;
-    lastWildcardPath;
-    lastWildcardInfo;
-    wildcardCount;
-    parentPathInfo;
-    constructor(path) {
-        // Helper to get or create StructuredPathInfo instances, avoiding redundant creation for self-reference
-        const getPattern = (_path) => {
-            return (path === _path) ? this : getPathInfo(_path);
-        };
-        // Split the pattern into individual path segments (e.g., "items.*.name" → ["items", "*", "name"])
-        const segments = path.split(".");
-        // Arrays to track all cumulative paths from root to each segment
-        const cumulativePaths = [];
-        const cumulativePathInfos = [];
-        // Arrays to track wildcard-specific information
-        const wildcardPaths = [];
-        const indexByWildcardPath = {}; // Maps wildcard path to its index position
-        const wildcardPathInfos = [];
-        const wildcardParentPaths = []; // Paths of parent segments for each wildcard
-        const wildcardParentPathInfos = [];
-        const wildcardPositions = [];
-        let currentPatternPath = "", prevPatternPath = "";
-        let wildcardCount = 0;
-        // Iterate through each segment to build cumulative paths and identify wildcards
-        for (let i = 0; i < segments.length; i++) {
-            currentPatternPath += segments[i];
-            // If this segment is a wildcard, track it with all wildcard-specific metadata
-            if (segments[i] === WILDCARD) {
-                wildcardPaths.push(currentPatternPath);
-                indexByWildcardPath[currentPatternPath] = wildcardCount; // Store wildcard's ordinal position
-                wildcardPathInfos.push(getPattern(currentPatternPath));
-                wildcardParentPaths.push(prevPatternPath); // Parent path is the previous cumulative path
-                wildcardParentPathInfos.push(getPattern(prevPatternPath));
-                wildcardPositions.push(i);
-                wildcardCount++;
-            }
-            // Track all cumulative paths for hierarchical navigation (e.g., "items", "items.*", "items.*.name")
-            cumulativePaths.push(currentPatternPath);
-            cumulativePathInfos.push(getPattern(currentPatternPath));
-            // Save current path as previous for next iteration, then add separator
-            prevPatternPath = currentPatternPath;
-            currentPatternPath += ".";
-        }
-        // Determine the deepest (last) wildcard path and the parent path of the entire pattern
-        const lastWildcardPath = wildcardPaths.length > 0 ? wildcardPaths[wildcardPaths.length - 1] : null;
-        const parentPath = cumulativePaths.length > 1 ? cumulativePaths[cumulativePaths.length - 2] : null;
-        // Assign all analyzed data to readonly properties
-        this.path = path;
-        this.segments = segments;
-        this.lastSegment = segments[segments.length - 1];
-        this.cumulativePaths = cumulativePaths;
-        this.cumulativePathSet = new Set(cumulativePaths); // Set for fast lookup
-        this.cumulativePathInfos = cumulativePathInfos;
-        this.cumulativePathInfoSet = new Set(cumulativePathInfos);
-        this.wildcardPaths = wildcardPaths;
-        this.wildcardPathSet = new Set(wildcardPaths);
-        this.indexByWildcardPath = indexByWildcardPath;
-        this.wildcardPathInfos = wildcardPathInfos;
-        this.wildcardPathInfoSet = new Set(wildcardPathInfos);
-        this.wildcardParentPaths = wildcardParentPaths;
-        this.wildcardParentPathSet = new Set(wildcardParentPaths);
-        this.wildcardParentPathInfos = wildcardParentPathInfos;
-        this.wildcardParentPathInfoSet = new Set(wildcardParentPathInfos);
-        this.wildcardPositions = wildcardPositions;
-        this.lastWildcardPath = lastWildcardPath;
-        this.lastWildcardInfo = lastWildcardPath ? getPattern(lastWildcardPath) : null;
-        this.parentPath = parentPath;
-        this.parentPathInfo = parentPath ? getPattern(parentPath) : null;
-        this.wildcardCount = wildcardCount;
+function didYouMean(input, candidates) {
+  if (input.length === 0) {
+    return "";
+  }
+  const folded = input.toLowerCase();
+  let best = null;
+  let bestDistance = 3;
+  for (const candidate of candidates) {
+    const distance = editDistance(folded, candidate.toLowerCase(), 2);
+    if (distance < bestDistance) {
+      best = candidate;
+      bestDistance = distance;
     }
+  }
+  return best !== null ? ` Did you mean "${best}"?` : "";
 }
+var LINT_HINT = " Validate statically: npx @wcstack/lint <file>.";
 
-/**
- * errorGuidance.ts — エラーメッセージへの self-fix 誘導（GTM 2-5 /
- * docs/static-wiring-dx-design.md §3）。
- *
- * コンソールは「書き手（人間・AI とも）が誤った瞬間に必ず読む面」なので、
- * (a) did-you-mean 候補 (b) lint への誘導 をエラーメッセージ自体に埋め込む。
- * ここの関数は全て**エラーパスでのみ**呼ばれる — 正常系のコストはゼロ。
- * auto.min.js に同梱されるため文字列は最小限に保つ（エラーパス専用モジュールの
- * 遅延 import は `src/auto.ts` の SRI 自己完結制約で不可）。
- *
- * 診断 code の語彙はコンソール → lint → IDE の三面で共有する:
- * メッセージ先頭の `[wcs/...]` は wcstack-intellisense / @wcstack/lint の
- * 安定診断 code（packages/vscode-wcs/src/core/diagnostics.ts）と同一。
- */
-/** 挿入・削除・置換の編集距離。長さ差が max を超えたら早期に max+1 を返す。 */
-/**
- * lint への誘導（誘導付きメッセージ共通の一文）。
- * **lint が実際にそのケースを検出するサイトにだけ付ける** — 検出しないケースに
- * 付けると「エラー → lint 実行 → clean」の空振りで検証ループの信頼を毀損する。
- * 現在 lint 未検出のため付けないもの: DCC 宣言・watch の空キー / Object.prototype
- * 継承名 / ワイルドカード深度超過。
- * なお hint 付きサイト内でも被覆は部分的でありうる（例: `$watch: ident` の実体が
- * 非オブジェクトだった場合、ランタイムは評価後の値で raise するが lint は宣言 shape
- * から断定できず沈黙する）。サイト粒度の hint ではこの残余は構造的に避けられない。
- */
-const LINT_HINT = " Validate statically: npx @wcstack/lint <file>.";
+// src/parser/raiseError.ts
+function raiseError(message, subject, candidates) {
+  throw new Error(`[@wcstack/state] ${message}${hooks.explain?.(message, subject, candidates) ?? ""}`);
+}
 
-const STRUCTURAL_BINDING_TYPE_SET = new Set([
-    "if",
-    "elseif",
-    "else",
-    "for",
-]);
+// src/messages.ts
+var CODES = [
+  "",
+  "binding-syntax",
+  "template-syntax",
+  "binding-path-missing",
+  "binding-type-expectation",
+  "filter-unknown",
+  "filter-arity",
+  "getter-cycle",
+  "getter-depth-exceeded",
+  "index-arity",
+  "index-param-range",
+  "recursion-unsupported",
+  "token-misconfigured",
+  "token-undeclared",
+  "wildcard-rank",
+  "spread-no-bindable",
+  "declaration-alias",
+  "name-alias"
+];
+var codeOf = (id) => {
+  const c = CODES[id / 100 | 0];
+  return c ? `[wcs/${c}] ` : "";
+};
+function text(id, args = []) {
+  return hooks.render?.(id, args) ?? `${codeOf(id)}#${id}${args.map((a) => ` ${typeof a === "string" ? JSON.stringify(a) : String(a)}`).join("")}`;
+}
+function raise(id, args, subject, candidates) {
+  raiseError(text(id, args), subject, candidates);
+}
 
-/**
- * core/filterRegistry.ts — フィルタ実関数の登録簿（設計案 §4、要件 D16）。
- *
- * 文法（`path|filter(args)` の解析）は core に残り、**実関数は登録簿から束縛計画の段で引く**。
- * 解析の段は名前と引数しか作らない（`bindTextParser/parseFilters.ts`）ので、パーサだけを使う
- * tooling（`@wcstack/state/parser`）はフィルタの実装を 1 バイトも引き込まない。
- *
- * 書式フィルタ群（`uc` / `date` / `round` …）は `features/formats` が install で登録する。
- * core が自前で持つのは、エンジン自身が差し込む `not` だけ（`if` / `else` の反転 —
- * structural/notFilter.ts）。未知のフィルタは束縛計画の段で名指しで落ちる（従来は解析時）。
- */
-/** 名前 + 引数 + 入出力ごとに解決済みの実関数（解決は 1 回だけ） */
-const resolvedByKey = new Map();
-/** 解決済みの答えを捨てる（tooling: `@wcstack/state/parser` の clearParserCaches） */
-function clearFilterResolutionCache() {
-    resolvedByKey.clear();
+// src/filters/registry.ts
+var FORMATS_FILTER_NAMES = [
+  "toFixed",
+  "locale",
+  "upper",
+  "lower",
+  "capitalize",
+  "trim",
+  "slice",
+  "padStart",
+  "padEnd",
+  "repeat",
+  "reverse",
+  "truncate",
+  "join",
+  "round",
+  "floor",
+  "ceil",
+  "percent",
+  "unit",
+  "date",
+  "time",
+  "datetime",
+  "ymd",
+  "hms"
+];
+var definitions = /* @__PURE__ */ new Map();
+function hasFilter(name) {
+  return definitions.has(name);
 }
 
-/**
- * フィルタ引数リストのパース。`filter(a, b)` の `a, b` 部分を受け取る。
- *
- * トリムの規則は「**クォートの外側だけ**」。`fix( 2 )` のような書き癖を吸収するために
- * 素の引数は前後をトリムするが、クォートは「ここは literal」という宣言なので中身の
- * 空白は残す。両方まとめてトリムしていたため `pad(5, ' ')` が空文字パディング
- * （＝無変化）に化けており、空白区切りの `join(' / ')` も指定できなかった。
- */
-/** 引数 1 つを確定する。クォート由来の文字が入った範囲より外側だけをトリムする。 */
-function finalizeArg(text, firstQuoteStart, lastQuoteEnd) {
-    // 先頭側: 最初のクォート文字より前だけが削れる（クォートが無ければ全体が対象）
-    const startLimit = firstQuoteStart === -1 ? text.length : firstQuoteStart;
-    let start = 0;
-    while (start < startLimit && /\s/.test(text[start])) {
-        start++;
-    }
-    // 末尾側: 最後のクォート文字より後ろだけが削れる（クォートが無ければ全体が対象）
-    const endLimit = lastQuoteEnd === -1 ? 0 : lastQuoteEnd;
-    let end = text.length;
-    while (end > endLimit && /\s/.test(text[end - 1])) {
-        end--;
-    }
-    return text.slice(start, end);
+// src/pattern.ts
+var WILDCARD = "*";
+
+// src/parser/define.ts
+var DELIMITER = ".";
+var BINDING_SEPARATOR = ";";
+var PROP_VALUE_SEPARATOR = ":";
+var MODIFIER_SEPARATOR = "#";
+var FILTER_SEPARATOR = "|";
+var ELSE_KEYWORD = "else";
+var SPREAD_PROP = "...";
+var EVENT_PROP_PREFIX = "on";
+var EVENT_TOKEN_NAMESPACE = "eventToken";
+var VOLUME_INJECTION_PROP = "state";
+var COMMAND_NAMESPACE = "command";
+var CLASS_NAMESPACE = "class";
+var ATTR_NAMESPACE = "attr";
+var STYLE_NAMESPACE = "style";
+var MAX_PATH_SEGMENTS = 512;
+var MAX_INDEX_PARAM = 128;
+var RECURSION_WILDCARD = "**";
+
+// src/parser/parseFilterArgs.ts
+function finalizeArg(text2, firstQuoteStart, lastQuoteEnd) {
+  const startLimit = firstQuoteStart === -1 ? text2.length : firstQuoteStart;
+  const endLimit = lastQuoteEnd === -1 ? 0 : lastQuoteEnd;
+  return text2.slice(startLimit - text2.slice(0, startLimit).trimStart().length, endLimit + text2.slice(endLimit).trimEnd().length);
 }
-/** 引用符の無い引数の型（要件 B9）: true / false / null / 数値は型付き、それ以外は文字列 */
-const NUMBER_LITERAL = /^[+-]?(\d+\.?\d*|\.\d+)([eE][+-]?\d+)?$/;
-function toLiteral(text, quoted) {
-    if (quoted)
-        return text;
-    if (text === "true")
-        return true;
-    if (text === "false")
-        return false;
-    if (text === "null")
-        return null;
-    return NUMBER_LITERAL.test(text) ? Number(text) : text;
+var NUMBER_LITERAL = /^[+-]?(\d+\.?\d*|\.\d+)([eE][+-]?\d+)?$/;
+function toLiteral(text2, quoted) {
+  if (quoted) return text2;
+  if (text2 === "true") return true;
+  if (text2 === "false") return false;
+  if (text2 === "null") return null;
+  return NUMBER_LITERAL.test(text2) ? Number(text2) : text2;
 }
-/**
- * 引数の原文と、その型付きの値（要件 B9）を一緒に返す。原文は引用符を外したもの。
- *
- * 末尾の空引数だけが落ちる（`"a,"` → 1 個、`",a"` → 2 個、`","` → 1 個）。`filter()` を
- * 「引数 0 個」と読むための規則で、先頭・中間の空引数は位置を保つために残す。
- */
 function parseFilterArgsWithLiterals(argsText) {
-    const args = [];
-    const literals = [];
-    let current = '';
-    let inQuote = null;
-    let hasQuote = false;
-    let firstQuoteStart = -1;
-    let lastQuoteEnd = -1;
-    const flush = () => {
-        const arg = finalizeArg(current, firstQuoteStart, lastQuoteEnd);
-        args.push(arg);
-        literals.push(toLiteral(arg, hasQuote));
-        current = '';
-        hasQuote = false;
-        firstQuoteStart = -1;
-        lastQuoteEnd = -1;
-    };
-    for (let i = 0; i < argsText.length; i++) {
-        const char = argsText[i];
-        if (inQuote) {
-            if (char === inQuote) {
-                inQuote = null;
-            }
-            else {
-                if (firstQuoteStart === -1) {
-                    firstQuoteStart = current.length;
-                }
-                current += char;
-                lastQuoteEnd = current.length;
-            }
-        }
-        else if (char === '"' || char === "'") {
-            inQuote = char;
-            hasQuote = true;
-        }
-        else if (char === ',') {
-            flush();
-        }
-        else {
-            current += char;
-        }
-    }
-    if (inQuote !== null) {
-        // 閉じていない引用符は受理しない（要件 B2）。以前は黙って閉じたことにしていた
-        raiseError(`[wcs/binding-syntax] unterminated ${inQuote} quote in the filter arguments "(${argsText})". Close the quote.${LINT_HINT}`);
+  const args = [];
+  const literals = [];
+  let current = "";
+  let inQuote = null;
+  let hasQuote = false;
+  let firstQuoteStart = -1;
+  let lastQuoteEnd = -1;
+  const flush = (last) => {
+    const arg = finalizeArg(current, firstQuoteStart, lastQuoteEnd);
+    if (!last || arg || hasQuote) {
+      args.push(arg);
+      literals.push(toLiteral(arg, hasQuote));
     }
-    const last = finalizeArg(current, firstQuoteStart, lastQuoteEnd);
-    if (last || hasQuote) {
-        args.push(last);
-        literals.push(toLiteral(last, hasQuote));
+    current = "";
+    hasQuote = false;
+    firstQuoteStart = -1;
+    lastQuoteEnd = -1;
+  };
+  for (let i = 0; i < argsText.length; i++) {
+    const char = argsText[i];
+    if (inQuote) {
+      if (char === inQuote) {
+        inQuote = null;
+      } else {
+        if (firstQuoteStart === -1) {
+          firstQuoteStart = current.length;
+        }
+        current += char;
+        lastQuoteEnd = current.length;
+      }
+    } else if (char === '"' || char === "'") {
+      inQuote = char;
+      hasQuote = true;
+    } else if (char === ",") {
+      flush();
+    } else {
+      current += char;
     }
-    return { args, literals };
+  }
+  if (inQuote !== null) {
+    raise(107 /* UnterminatedQuote */, [inQuote, argsText]);
+  }
+  flush(true);
+  return { args, literals };
 }
 
-const trimFn = (s) => s.trim();
-const isQuote = (c) => c === "'" || c === '"';
-/**
- * `text` の中で、引用符（`'` / `"`）の外にある最初の `char` の位置。無ければ -1（要件 B1）。
- * 閉じていない引用符はそのまま末尾まで続く扱い — 不正な引用符はフィルタ引数の段で名指しで落ちる。
- */
-function indexOfOutsideQuotes(text, char) {
-    let quote = null;
-    for (let i = 0; i < text.length; i++) {
-        const c = text[i];
-        if (quote !== null) {
-            if (c === quote)
-                quote = null;
-        }
-        else if (isQuote(c)) {
-            quote = c;
-        }
-        else if (c === char) {
-            return i;
-        }
+// src/parser/utils.ts
+var trimFn = (s) => s.trim();
+function outsideQuotes(text2, char) {
+  const found = [];
+  let quote = null;
+  for (let i = 0; i < text2.length; i++) {
+    const c = text2[i];
+    if (quote !== null) {
+      if (c === quote) quote = null;
+    } else if (c === "'" || c === '"') {
+      quote = c;
+    } else if (c === char) {
+      found.push(i);
     }
-    return -1;
+  }
+  return found;
 }
-/**
- * `text` の中で、引用符の外にある**最後の** `char` の位置。無ければ -1（要件 B1）。
- * 引用符は前からしか追えないので、前向きに走査して最後の一致を控える。
- * フィルタの閉じ括弧を探す用（`a|foo(')')` の引用符内の `)` を終端と誤認しないため）。
- */
-function lastIndexOfOutsideQuotes(text, char) {
-    let quote = null;
-    let found = -1;
-    for (let i = 0; i < text.length; i++) {
-        const c = text[i];
-        if (quote !== null) {
-            if (c === quote)
-                quote = null;
-        }
-        else if (isQuote(c)) {
-            quote = c;
-        }
-        else if (c === char) {
-            found = i;
-        }
-    }
-    return found;
-}
-/**
- * `separator` で区切る。ただし引用符の中は区切らない（要件 B1）: `join(';')` や `join('|')` の
- * 区切り文字は引数であって、バインディングやフィルタの区切りではない。
- */
-function splitOutsideQuotes(text, separator) {
-    const parts = [];
-    let quote = null;
-    let start = 0;
-    for (let i = 0; i < text.length; i++) {
-        const c = text[i];
-        if (quote !== null) {
-            if (c === quote)
-                quote = null;
-        }
-        else if (isQuote(c)) {
-            quote = c;
-        }
-        else if (c === separator) {
-            parts.push(text.slice(start, i));
-            start = i + 1;
-        }
-    }
-    parts.push(text.slice(start));
-    return parts;
+var indexOfOutsideQuotes = (text2, char) => outsideQuotes(text2, char)[0] ?? -1;
+var lastIndexOfOutsideQuotes = (text2, char) => outsideQuotes(text2, char).pop() ?? -1;
+function splitOutsideQuotes(text2, separator) {
+  const parts = [];
+  let start = 0;
+  for (const i of outsideQuotes(text2, separator)) {
+    parts.push(text2.slice(start, i));
+    start = i + 1;
+  }
+  parts.push(text2.slice(start));
+  return parts;
 }
 
-/** tooling 専用（parser.ts の clearParserCaches からのみ呼ぶ）。 */
-function clearFilterFnCacheForTooling() {
-    clearFilterResolutionCache();
-}
-// format: filterName(arg1,arg2) or filterName
-/**
- * 文法の段（要件 D16）: 名前と引数だけを読む。**実関数は引かない** — 束縛計画の段で
- * 登録簿から解決する（`core/filterRegistry.ts`・`bindings/getBindingInfos.ts`）。
- * 未知のフィルタもここでは落とさない: パーサだけを使う tooling は実装を持たないので、
- * 「知らない名前」を解析の段で判定できない。
- */
-/**
- * @param filterTextList 個々のフィルタの原文（`|` で切った後）
- * @param filterIOType   入力（左辺）/ 出力（右辺）。実関数は引かないが、助言の出し分けに使う
- * @param sourceText     診断に埋める原文。呼び出し側は**その辺の全文**（`propPart` / `statePart`）を
- *                       渡すこと。`|` より後ろだけを渡すと、`textContent: a|` のように末尾が空の形で
- *                       空文字になり、つなぎ直し（`filterTextList.join("|")`）と同じで文脈が消える。
- *                       省略時のつなぎ直しは、原文を持たない直接の呼び出し用のフォールバック
- */
+// src/parser/parseFilters.ts
 function parseFilters(filterTextList, filterIOType, sourceText) {
-    const source = sourceText ?? filterTextList.join("|");
-    return filterTextList.map((filterText) => {
-        // 括弧も引用符の外だけで探す（要件 B1）。素の `indexOf` / `lastIndexOf` だと
-        // `foo(')')` の引用符内の `)` を終端に取り、「閉じ括弧が無い」ではなく
-        // 「引用符が閉じていない」という見当違いの診断になっていた
-        const openParenIndex = indexOfOutsideQuotes(filterText, '(');
-        let closeParenIndex = lastIndexOfOutsideQuotes(filterText, ')');
-        if (closeParenIndex === -1) {
-            // 引用符の外に閉じ括弧が無い ＝ 引用符が閉じていない形（`join('x)`）か、本当に無いか。
-            // 前者で「閉じ括弧が無い」と言うのは見当違いなので素の探索へ落とし、引数の段の
-            // `[wcs/binding-syntax] unterminated ' quote` に診断させる
-            closeParenIndex = filterText.lastIndexOf(')');
-        }
-        // check parentheses
-        if (openParenIndex !== -1 && closeParenIndex === -1) {
-            raiseError(`[wcs/binding-syntax] Invalid filter format: missing closing parenthesis in "${filterText}".${LINT_HINT}`);
-        }
-        if (closeParenIndex !== -1 && openParenIndex === -1) {
-            raiseError(`[wcs/binding-syntax] Invalid filter format: missing opening parenthesis in "${filterText}".${LINT_HINT}`);
-        }
-        if (closeParenIndex !== -1 && closeParenIndex < openParenIndex) {
-            // `foo)1(` — 以前は `substring` が start > end で引数を入れ替えるため、フィルタ名
-            // `foo)1` として受理されていた
-            raiseError(`[wcs/binding-syntax] Invalid filter format: ")" comes before "(" in "${filterText}".${LINT_HINT}`);
-        }
-        if (closeParenIndex !== -1 && filterText.slice(closeParenIndex + 1).trim().length > 0) {
-            // 閉じ括弧の後ろの残余を**黙って捨てていた**。`n|fix(2)uc` は `uc` が消えて診断ゼロで
-            // 通り、括弧の無い `n|ucuc` は `[wcs/filter-unknown]` で落ちる — 括弧の有無で非対称だった。
-            // 実害は「`|` の打ち忘れでフィルタが 1 本消えても無診断」
-            const trailing = filterText.slice(closeParenIndex + 1).trim();
-            raiseError(`[wcs/binding-syntax] "${filterText}": unexpected "${trailing}" after the filter's closing ")" — ` +
-                `separate filters with "|" (write "${filterText.slice(0, closeParenIndex + 1)}|${trailing}").${LINT_HINT}`);
-        }
-        const filterName = (openParenIndex === -1 ? filterText : filterText.substring(0, openParenIndex)).trim();
-        if (filterName.length === 0) {
-            // 空のフィルタ（`x|`・`x||y`・`x|(1)`）は文法の誤り。解析の段で名指しで落とす — 未知の
-            // フィルタとは別物で、実関数の解決（束縛計画の段）まで持ち越すと tooling の解析が素通りする
-            raiseError(`[wcs/binding-syntax] an empty filter in "${source}" — remove the extra "|" or name the filter.${LINT_HINT}`);
-        }
-        if (filterName.includes(MODIFIER_SEPARATOR)) {
-            // フィルタ名に `#` が飲まれた形。`[wcs/filter-unknown] filter not found: trim#ro` だと
-            // 診断が指す先が実際の誤りと違う（要件 B4 の並び）。
-            // 助言は辺で分ける — **修飾子は左辺にしか存在しない**ので、右辺で「修飾子をフィルタより
-            // 前に書け」と言うと成立しない直し方（`textContent#ro|trim: x`）を勧めることになる
-            const [name, modifiers] = filterName.split(MODIFIER_SEPARATOR);
-            raiseError(filterIOType === "input"
-                ? `[wcs/binding-syntax] "${filterName}" is not a filter name: a modifier list ` +
-                    `"${MODIFIER_SEPARATOR}${modifiers}" comes before the input filters, not inside one — write ` +
-                    `"<property>${MODIFIER_SEPARATOR}${modifiers}|${name}".${LINT_HINT}`
-                : `[wcs/binding-syntax] "${filterName}" is not a filter name: "${MODIFIER_SEPARATOR}" cannot appear ` +
-                    `in one. Modifiers belong on the left side of the binding, before the ":" — write "${name}" ` +
-                    `here.${LINT_HINT}`);
-        }
-        if (openParenIndex === -1) {
-            // no arguments
-            return { filterName, args: [], literals: [] };
-        }
-        const argsText = filterText.substring(openParenIndex + 1, closeParenIndex);
-        return { filterName, ...parseFilterArgsWithLiterals(argsText) };
-    });
-}
-
-const cacheFilterInfos$1 = new Map();
-/** tooling 専用（parser.ts の clearParserCaches からのみ呼ぶ）。 */
-function clearPropPartCacheForTooling() {
-    cacheFilterInfos$1.clear();
-}
-// format: propName#moodifier1,modifier2
-// propName-format: path.to.property (e.g., textContent, style.color, not include :)
-// special path: 
-//   'attr.attributeName' for attributes (e.g., attr.href, attr.data-id)
-//   'style.propertyName' for style properties (e.g., style.backgroundColor, style.fontSize)
-//   'class.className' for class names (e.g., class.active, class.hidden)
-//   'onclick', 'onchange' etc. for event listeners
-function parsePropPart(propPart) {
-    const pos = indexOfOutsideQuotes(propPart, FILTER_SEPARATOR);
-    let propText = '';
-    let filterTexts = [];
-    let filtersText = '';
-    let filters = [];
-    if (pos !== -1) {
-        propText = propPart.slice(0, pos).trim();
-        filtersText = propPart.slice(pos + 1).trim();
-        if (cacheFilterInfos$1.has(filtersText)) {
-            filters = cacheFilterInfos$1.get(filtersText);
-        }
-        else {
-            filterTexts = splitOutsideQuotes(filtersText, FILTER_SEPARATOR).map(trimFn);
-            // 診断に埋める原文は**左辺の全文**。`|` より後ろだけを渡すと `value|:` のように
-            // 末尾が空の形で空文字になる（解析結果のキャッシュ鍵は従来どおり `filtersText`。
-            // 落ちた解析はキャッシュに載らないので、原文を混ぜても鍵は汚れない）
-            filters = parseFilters(filterTexts, "input", propPart);
-            cacheFilterInfos$1.set(filtersText, filters);
-        }
+  const source = sourceText ?? filterTextList.join("|");
+  return filterTextList.map((filterText) => {
+    const openParenIndex = indexOfOutsideQuotes(filterText, "(");
+    let closeParenIndex = lastIndexOfOutsideQuotes(filterText, ")");
+    if (closeParenIndex === -1) {
+      closeParenIndex = filterText.lastIndexOf(")");
     }
-    else {
-        propText = propPart.trim();
+    if (openParenIndex !== -1 && closeParenIndex === -1) {
+      raise(108 /* FilterUnclosed */, [filterText]);
     }
-    // **不変条件**: ここから下の `split` は素で走らせてよい。`propText` は「引用符外の最初の `|`
-    // より前」のスライスであり、引用符を含みうるのは入力フィルタの引数（`|` の後ろ）だけなので、
-    // `#` も `,` も `.` も引用符の中に現れない。**修飾子の値に引用符を許す拡張（例
-    // `value#fmt('a,b'): x`）を入れるなら、この 3 つも `splitOutsideQuotes` に替えること** —
-    // 替え忘れると `structural/expandShorthandPaths.ts` で起きた「無言で 1 行も描画されない」
-    // と同じ欠陥クラスが再発する。
-    const modifierParts = propText.split(MODIFIER_SEPARATOR).map(trimFn);
-    if (modifierParts.length > 2) {
-        // 修飾子の並びは 1 つだけ（要件 B2）。`value#ro#wo` は以前 `ro` だけを残して黙って捨てていた
-        raiseError(`[wcs/binding-syntax] "${propText}": a binding takes one modifier list after a single "${MODIFIER_SEPARATOR}" — write "${modifierParts[0]}${MODIFIER_SEPARATOR}${modifierParts.slice(1).join(",")}".${LINT_HINT}`);
+    if (closeParenIndex !== -1 && openParenIndex === -1) {
+      raise(109 /* FilterUnopened */, [filterText]);
     }
-    const [propName, propModifiersText] = modifierParts;
-    const propSegments = propName.split(DELIMITER).map(trimFn);
-    // 明示のプロパティ形（`.name:` — 要件 B5 / D34）だけは先頭の空セグメントが正しい形。
-    // それ以外で空のセグメントが残るのは書き間違い: 左辺が空（`": x"` / `"#ro: x"` / `"|trim: x"`）だと
-    // `element[""] = value` の expando ができて完全に沈黙し、末尾が空（`"foo.: x"`）だと適用の段で
-    // 素の TypeError になる。どちらも解析の段で名指しで落とす（`.: x` 等は下の D34 の検査が受け持つ）
-    const isExplicitProperty = propSegments.length > 1 && propSegments[0] === '';
-    if (!isExplicitProperty && (propName.length === 0 || propSegments.some((segment) => segment.length === 0))) {
-        raiseError(`[wcs/binding-syntax] "${propPart}": the left side of a binding must name a property — ` +
-            `write "<property>: <path>" (modifiers and input filters come after the name).${LINT_HINT}`);
+    if (closeParenIndex !== -1 && closeParenIndex < openParenIndex) {
+      raise(110 /* FilterParenOrder */, [filterText]);
     }
-    const propModifiers = propModifiersText
-        ? propModifiersText.split(',').map(trimFn)
-        : [];
-    return {
-        propName,
-        propSegments,
-        propModifiers,
-        inFilters: filters,
-    };
+    if (closeParenIndex !== -1 && filterText.slice(closeParenIndex + 1).trim().length > 0) {
+      const trailing = filterText.slice(closeParenIndex + 1).trim();
+      raise(111 /* FilterTrailing */, [filterText, trailing]);
+    }
+    const filterName = (openParenIndex === -1 ? filterText : filterText.substring(0, openParenIndex)).trim();
+    if (filterName.length === 0) {
+      raise(112 /* FilterEmpty */, [source]);
+    }
+    if (filterName.includes(MODIFIER_SEPARATOR)) {
+      const [, modifiers] = filterName.split(MODIFIER_SEPARATOR);
+      raise(filterIOType === "input" ? 113 /* FilterNameHasModifiersInput */ : 114 /* FilterNameHasModifiersOutput */, [filterName, modifiers]);
+    }
+    if (openParenIndex === -1) {
+      return { filterName, args: [], literals: [] };
+    }
+    const argsText = filterText.substring(openParenIndex + 1, closeParenIndex);
+    return { filterName, ...parseFilterArgsWithLiterals(argsText) };
+  });
+}
+function splitFilters(part, filterIOType, cache2) {
+  const pos = indexOfOutsideQuotes(part, FILTER_SEPARATOR);
+  if (pos === -1) return [part.trim(), []];
+  const filtersText = part.slice(pos + 1).trim();
+  let filters = cache2.get(filtersText);
+  if (filters === void 0) {
+    filters = parseFilters(splitOutsideQuotes(filtersText, FILTER_SEPARATOR).map(trimFn), filterIOType, part);
+    cache2.set(filtersText, filters);
+  }
+  return [part.slice(0, pos).trim(), filters];
 }
 
-const cacheFilterInfos = new Map();
-/** tooling 専用（parser.ts の clearParserCaches からのみ呼ぶ）。 */
-function clearStatePartCacheForTooling() {
-    cacheFilterInfos.clear();
+// src/parser/parseStatePart.ts
+var cacheFilterInfos = /* @__PURE__ */ new Map();
+var clearStatePartCache = () => cacheFilterInfos.clear();
+function checkPathLikeGetPathInfo(path) {
+  if (path.includes(RECURSION_WILDCARD)) {
+    recursionUnsupported(path);
+  }
+  const segmentCount = path.split(DELIMITER).length;
+  if (segmentCount > MAX_PATH_SEGMENTS) {
+    raise(117 /* TooManySegments */, [path, segmentCount]);
+  }
+}
+function recursionUnsupported(path) {
+  raise(1101 /* RecursionUnsupported */, [path]);
 }
-// format: statePath|filter|filter
-// statePath-format: path.to.property (e.g., user.name.first, users.*.name, users.0.name, not include @)
-// filters-format: filterName or filterName(arg1,arg2)
 function parseStatePart(statePart) {
-    // 引用符の中の `|` はフィルタの区切りではない（要件 B1 — `join('|')`）
-    const pos = indexOfOutsideQuotes(statePart, FILTER_SEPARATOR);
-    let stateAndPath = '';
-    let filterTexts = [];
-    let filtersText = '';
-    let filters = [];
-    if (pos !== -1) {
-        stateAndPath = statePart.slice(0, pos).trim();
-        filtersText = statePart.slice(pos + 1).trim();
-        if (cacheFilterInfos.has(filtersText)) {
-            filters = cacheFilterInfos.get(filtersText);
-        }
-        else {
-            filterTexts = splitOutsideQuotes(filtersText, FILTER_SEPARATOR).map(trimFn);
-            // 診断に埋める原文は**右辺の全文**（`parsePropPart` と同じ理由 — `a|` が空文字になる）
-            filters = parseFilters(filterTexts, "output", statePart);
-            cacheFilterInfos.set(filtersText, filters);
-        }
-    }
-    else {
-        stateAndPath = statePart.trim();
-    }
-    if (stateAndPath.indexOf("@") !== -1) {
-        // 名前次元は v2 で撤去（docs/state-mount-design.md D16 / §9）。パスは 1 本のツリー。
-        raiseError(`"${stateAndPath}": the "@name" selector was removed in v2 — there is a single state tree. ` +
-            `Mount the named state onto the tree (<wcs-state mount="...">) and read it by its path prefix instead.`);
-    }
-    const statePathName = stateAndPath;
-    // 右辺も左辺（`parsePropPart`）と同じ規準で空セグメントを弾く（要件 B1）。放っておくと
-    // `a.` / `a..b` は診断ゼロで `["a","",""]` のまま intern され、`textContent:` /
-    // `textContent: |uc` は適用の段で「`Path ""` が無い」というパス名が空の診断になる。
-    //
-    // **判定の前にループ相対の短縮形を正規化する。** `.` で始まる右辺は `for` 行の相対パスで
-    // （`structural/expandShorthandPaths.ts` / `bindTextParser/expandSpread.ts` が生成し、正本の
-    // パーサは展開前の原文も解析する）、先頭の空セグメント 1 つは正当。とくに **`.` 単独は
-    // 「行そのもの」を指す正規の書き方**で、README と `examples/recursive-tree/index.html` の
-    // `state: .` がそれ — 「先頭は許すが末尾は拒否」と素朴に書くと `"."` は
-    // `["", ""]` ＝ 先頭かつ末尾なので落ちる。取り除いた残りが空でも通すこと
-    const isLoopRelative = statePathName.startsWith(DELIMITER);
-    const body = isLoopRelative ? statePathName.slice(DELIMITER.length) : statePathName;
-    const hasEmptySegment = body.length > 0 && body.split(DELIMITER).some((segment) => segment.length === 0);
-    if (hasEmptySegment || (!isLoopRelative && body.length === 0)) {
-        raiseError(`[wcs/binding-syntax] "${statePart}": the right side of a binding must name a state path — ` +
-            `write "<property>: <path>" (a path segment cannot be empty; "." alone and a leading "." are ` +
-            `the loop-relative shorthand).${LINT_HINT}`);
+  const [stateAndPath, filters] = splitFilters(statePart, "output", cacheFilterInfos);
+  if (stateAndPath.includes("@")) {
+    raise(32 /* SelectorRemoved */, [stateAndPath]);
+  }
+  const statePathName = stateAndPath;
+  const isLoopRelative = statePathName.startsWith(DELIMITER);
+  const body = isLoopRelative ? statePathName.slice(DELIMITER.length) : statePathName;
+  const hasEmptySegment = body.length > 0 && body.split(DELIMITER).some((segment) => segment.length === 0);
+  if (hasEmptySegment || !isLoopRelative && body.length === 0) {
+    raise(119 /* EmptySegment */, [statePart]);
+  }
+  checkPathLikeGetPathInfo(statePathName);
+  return {
+    statePathName,
+    outFilters: filters
+  };
+}
+
+// src/engine.ts
+var MAX_DRAIN_PASSES = 32;
+var MAX_RENDER_CHAIN = 100;
+
+// src/diagnostics/messages.ts
+var MOVED = ["enableMustache", "sameValueGuard", "enableDirectionalInitialSync"];
+var element = (tag, at) => `<${tag}${at.map((a, i) => i % 2 ? `="${a}"` : ` ${a}`).join("")}>`;
+var CSP_GUIDE = "https://github.com/wcstack/wcstack/blob/main/docs/csp.md";
+var SENTENCES = {
+  [1 /* ScanRemoved */]: () => "$scan was removed (use $watch or $on)",
+  [2 /* GetterWithoutSetter */]: (p) => `"${p}" is a getter without a setter`,
+  [3 /* NoRow */]: (p) => `no row for "${p}"`,
+  [4 /* ParentNotObject */]: (p, parent) => `cannot write "${p}": its parent is ${parent}`,
+  [5 /* NotAMethod */]: (name) => `"${name}" is not a method`,
+  [6 /* GetAllNoCommonLevel */]: (p) => `$getAll("${p}"): no loop level in common with the context`,
+  [7 /* EqIndexNoRow */]: (p) => `$eqIndex("${p}") needs a list row scope.`,
+  [8 /* Readonly */]: () => "This state is readonly.",
+  [9 /* SetAllNeedsIndexes */]: (p) => `$setAll("${p}") needs indexes ([] for every match)`,
+  [10 /* SetAllSpreadLength */]: (p, n) => `$setAll("${p}", \u2026, { spread: true }) needs an array of ${n} values`,
+  [11 /* DrainNotSettled */]: () => `updates did not settle after ${MAX_DRAIN_PASSES} passes`,
+  [41 /* RenderChain */]: () => `render chain depth limit exceeded (${MAX_RENDER_CHAIN} drains that rendering itself started); bindings for this batch were not applied.`,
+  [12 /* BindingFailed */]: (type, p) => `binding "${type}: ${p}" failed to apply.`,
+  [13 /* LoadFailed */]: (src, status) => `failed to load "${src}": ${status}`,
+  [14 /* ElementFailed */]: () => "this <wcs-state> failed to initialize; create a new one",
+  [15 /* NotInitialized */]: () => "state is not initialized",
+  [16 /* NoScript */]: (id) => `no <script> with id "${id}"`,
+  [42 /* InlineBlocked */]: () => `The inline <script> of <wcs-state> was blocked by Content-Security-Policy. Inline state is evaluated through a blob: URL: give the page's nonce to the <script> that loads @wcstack/state, or allow blob: in script-src. Moving the state into an external file (src="./state.js") needs neither. See ${CSP_GUIDE}`,
+  [43 /* InlineFailed */]: (detail) => `Failed to evaluate the inline <script> of <wcs-state>: ${detail}. If this page sets a Content-Security-Policy, see ${CSP_GUIDE}`,
+  [17 /* TokenSubscriberThrew */]: (name) => `a subscriber of token "${name}" threw.`,
+  [18 /* TokenListNotArray */]: (key) => `${key} must be an array of strings.`,
+  [19 /* TokenEntryEmpty */]: (key) => `${key} entries must be non-empty strings.`,
+  [20 /* TokenEntryReserved */]: (key, name, reserved) => `${key} entry "${name}" conflicts with the reserved namespace name "${reserved}".`,
+  [21 /* TokenEntryDuplicated */]: (key, name) => `${key} entry "${name}" is duplicated.`,
+  [22 /* OnNotObject */]: () => "$on must be an object of handlers.",
+  [23 /* OnEntryUndeclared */]: (name) => `$on entry "${name}" is not declared in $eventTokens.`,
+  [24 /* OnEntryNotFunction */]: (name) => `$on entry "${name}" must be a function.`,
+  [25 /* NamingLimit */]: (limit) => `view-transition naming-limit (${limit}) reached.`,
+  [26 /* FilterOptionsRequired */]: (fn) => `filter ${fn} requires at least one option`,
+  [27 /* FilterOptionNotNumber */]: (fn) => `filter ${fn} requires a number as option`,
+  [28 /* FilterValueNotNumber */]: (fn) => `filter ${fn} requires a number value`,
+  [29 /* FilterValueNotDate */]: (fn) => `filter ${fn} requires a date value`,
+  [30 /* FilterValueNotArray */]: (fn) => `filter ${fn} requires an array value`,
+  [31 /* DirectionalSyncDisabled */]: () => "init=/sync= modifiers require enableDirectionalInitialSync.",
+  [33 /* ModifierUnknown */]: (key, modifier) => `Unknown binding modifier "${key}" in "${modifier}".`,
+  [34 /* ModifierTwice */]: (key) => `Binding modifier "${key}" may only be specified once.`,
+  [35 /* ModifierValue */]: (key, value) => `Invalid ${key} modifier value "${value}".`,
+  [36 /* EventInitNone */]: () => "Event bindings only allow init=none.",
+  [37 /* InitUnsupported */]: (type, init) => `Binding type "${type}" does not support init=${init}.`,
+  [38 /* MemberUndeclared */]: (name) => `Property "${name}" is not declared by wcBindable.`,
+  [39 /* InitIncompatible */]: (init, name) => `init=${init} is incompatible with wcBindable member "${name}".`,
+  [40 /* SyncConnectNeedsOutput */]: (name) => `sync=connect requires observable property "${name}".`,
+  [44 /* OptionInvalid */]: (where, key) => `${where}: "${key}" is not one of its options, or not of the option's type.${where === "bootstrapState" && MOVED.includes(key) ? " 4.0 moved it to the state's $behavior." : ""}`,
+  [45 /* BehaviorChanged */]: () => "a re-set state may not change $behavior: create the element again.",
+  [47 /* SecondRoot */]: () => `a second <wcs-state> on the same root: there is one state tree per root \u2014 graft a subtree with <wcs-state mount="path"> (v1's name="\u2026" is gone: read the mounted state by its path).`,
+  [48 /* LocaleInvalid */]: (l) => `the locale "${l}" (<html lang> or bootstrapState's locale) is not a language tag Intl takes (en-US, not en_US): the locale filters use "en".`,
+  [46 /* FeaturesNotArray */]: () => '$features must be an array of add-on names (["temporal", "formats"]).',
+  [49 /* InitFailed */]: (tag, ...at) => `${element(tag, at)} failed to initialize.`,
+  [50 /* ConnectedFailed */]: (tag, ...at) => `${element(tag, at)} $connectedCallback failed.`,
+  [101 /* BindTextNoColon */]: (t) => `Invalid bindText: "${t}". Missing ':' separator between propPart and statePart.`,
+  [102 /* StructuralTakesNoModifiers */]: (t, keyword) => `"${t}": "${keyword}" takes no modifiers or filters on its left side \u2014 write "${keyword}:".`,
+  [103 /* ElseTakesNoValue */]: (t) => `"${t}": "else" takes no value \u2014 write "else:".`,
+  [104 /* SpreadNoPath */]: (t) => `Invalid spread binding "${t}": spread target path is required.`,
+  [105 /* SpreadNoFilters */]: (t) => `Invalid spread binding "${t}": filters are not allowed on spread targets.`,
+  [106 /* LeadingDotNamespace */]: (prop) => `"${prop}": a leading "." binds an element property by name \u2014 write a non-empty property that is not a namespace (class, attr, style, command, eventToken, state).`,
+  [107 /* UnterminatedQuote */]: (quote, args) => `unterminated ${quote} quote in the filter arguments "(${args})". Close the quote.`,
+  [108 /* FilterUnclosed */]: (f) => `Invalid filter format: missing closing parenthesis in "${f}".`,
+  [109 /* FilterUnopened */]: (f) => `Invalid filter format: missing opening parenthesis in "${f}".`,
+  [110 /* FilterParenOrder */]: (f) => `Invalid filter format: ")" comes before "(" in "${f}".`,
+  [111 /* FilterTrailing */]: (f, trailing) => `"${f}": unexpected "${trailing}" after the filter's closing ")" \u2014 separate filters with "|" (write "${f.slice(0, f.lastIndexOf(")") + 1)}|${trailing}").`,
+  [112 /* FilterEmpty */]: (source) => `an empty filter in "${source}" \u2014 remove the extra "|" or name the filter.`,
+  [113 /* FilterNameHasModifiersInput */]: (name, mods) => `"${name}" is not a filter name: a modifier list "${MODIFIER_SEPARATOR}${mods}" comes before the input filters, not inside one`,
+  [114 /* FilterNameHasModifiersOutput */]: (name) => `"${name}" is not a filter name: "${MODIFIER_SEPARATOR}" cannot appear in one.`,
+  [115 /* OneModifierList */]: (prop) => `"${prop}": a binding takes one modifier list after a single "#"`,
+  [116 /* NoPropertyName */]: (prop) => `"${prop}": the left side of a binding must name a property \u2014 write "<property>: <path>" (modifiers and input filters come after the name).`,
+  [117 /* TooManySegments */]: (p, n) => `"${p}" has ${n} path segments \u2014 the limit is ${MAX_PATH_SEGMENTS}.`,
+  [32 /* SelectorRemoved */]: (t) => `"${t}": the "@name" selector was removed in v2 \u2014 there is a single state tree. Mount the named state onto the tree (<wcs-state mount="...">) and read it by its path prefix instead.`,
+  [120 /* UnsafeSegment */]: (p) => `"${p}": a state path cannot go through "__proto__" or "prototype" (it would reach every object's prototype).`,
+  [121 /* ForNoFilters */]: (t) => `"${t}": "for:" takes no filters \u2014 a row is "<path>.<index>", so the rows of a filtered list would name other elements. Declare a getter that returns the filtered list and loop over it ("for: <getter>").`,
+  [119 /* EmptySegment */]: (t) => `"${t}": the right side of a binding must name a state path \u2014 write "<property>: <path>" (a path segment cannot be empty; "." alone and a leading "." are the loop-relative shorthand).`,
+  [201 /* StructuralNotSingle */]: (t) => `Invalid bindText: "${t}". 'if', 'elseif', 'else', and 'for' bindings must be single binding.`,
+  [202 /* ElseWithoutIf */]: (type) => `"${type}:" must follow an "if:" template`,
+  [204 /* TemplateHandedOver */]: (type) => `a "${type}:" template at the top of inserted content was not rendered, as it would render beside itself out of the inserter's reach: wrap it in an element.`,
+  [203 /* OuterInTemplate */]: (name) => `"${name}:" replaces its element, so it cannot be used inside a "for" / "if" template (a row or branch keeps its nodes by position): bind innerHTML: on a wrapper element instead.`,
+  [301 /* PathMissing */]: (p) => `Path "${p}" does not exist on the state tree.`,
+  [401 /* ClassNeedsBoolean */]: (name, type) => `class.${name} needs a boolean, got ${type}.`,
+  [501 /* FilterUnknown */]: (name) => `filter not found: ${name}.`,
+  [601 /* FilterTooFewArgs */]: (name, min, given) => `filter "${name}" requires at least ${min} argument(s) (${given} given).`,
+  [602 /* FilterTooManyArgs */]: (name, max, given) => `filter "${name}" accepts at most ${max} argument(s) (${given} given).`,
+  [701 /* GetterCycle */]: (p) => `"${p}" depends on itself`,
+  [801 /* GetterDepth */]: (p) => `"${p}"`,
+  [901 /* IndexArityExact */]: (api, p, depth, n) => `${api}("${p}") takes ${depth} index(es), got ${n}.`,
+  [902 /* IndexArityAtMost */]: (api, p, depth, n) => `${api}("${p}") takes at most ${depth} index(es), got ${n}.`,
+  [1001 /* IndexParamRange */]: (key) => `"${key}": list index parameters run from $1 to $${MAX_INDEX_PARAM}.`,
+  [1101 /* RecursionUnsupported */]: (p) => `"${p}" uses "${RECURSION_WILDCARD}", which is not accepted here.`,
+  [1201 /* CommandRightSide */]: (prop, p) => `"${prop}: ${p}": the right-hand side must be $command.<name>`,
+  [1202 /* NoBindable */]: (tag, what) => `<${tag}> declares no static wcBindable (${what}).`,
+  [1203 /* NoCommand */]: (tag, method) => `<${tag}> declares no command "${method}".`,
+  [1204 /* NoProperty */]: (tag, prop) => `<${tag}> declares no property "${prop}".`,
+  [1301 /* EventTokenUndeclared */]: (name) => `eventToken "${name}" is not declared in $eventTokens.`,
+  [1302 /* CommandTokenUndeclared */]: (name) => `"$command.${name}" is not declared in $commandTokens.`,
+  [1401 /* WildcardNoLoop */]: (p, depth, n = 0) => `"${p}" needs ${depth} enclosing loop level(s); the scope provides ${n}.`,
+  [1402 /* WildcardRelative */]: (p) => `"${p}" is relative: it needs an enclosing "for" template`,
+  [1403 /* WildcardOtherList */]: (p, over, loop) => `"${p}" ranges over the rows of "${over}", but the enclosing "for" template at that level renders "${loop}".`,
+  [1501 /* SpreadNoBindable */]: (tag, what) => `<${tag}> declares no static wcBindable (${what}).`,
+  [1601 /* DeclarationRemoved */]: (old, name) => `${old} was removed: write ${name}.`,
+  [1701 /* ApiRemoved */]: (old, name) => `${old} was removed: write ${name}.`
+};
+
+// src/diagnostics/explain.ts
+function render(id, args) {
+  const sentence = SENTENCES[id];
+  return sentence === void 0 ? `${codeOf(id)}#${id} ${args.join(" ")}` : codeOf(id) + sentence(...args);
+}
+var LINT_CODES = /* @__PURE__ */ new Set([
+  "binding-syntax",
+  "template-syntax",
+  "filter-unknown",
+  "filter-arity",
+  "index-arity",
+  "wildcard-rank",
+  "token-undeclared",
+  "binding-path-missing",
+  "index-param-range"
+]);
+var GUIDES = [
+  [/no loop level in common with the context/, "; pass indexes ([] for all)."],
+  [/"([^"#]+)#([^"]*)" is not a filter name: a modifier list .* comes before the input filters/, ' \u2014 write "<property>#$2|$1".'],
+  [/\[wcs\/recursion-unsupported\]/, " It is only meaningful in a $recursion declaration, in a recursive getter key, and in the path argument of $getAll / $setAll \u2014 and only when the state declares a $recursion anchor."],
+  [/must be single binding/, ' Put the structural binding alone in its own data-wcs (e.g. <template data-wcs="for: items">).'],
+  [/\[wcs\/wildcard-rank\] .* needs \d+ enclosing/, ' Wrap it in that many "for" templates, or use $resolve(path, indexes) to name the row explicitly.'],
+  [/\[wcs\/wildcard-rank\] .* ranges over the rows of/, ' A "*" in a binding is the row of the loop around it: read a row of another list in a getter, with $resolve(path, indexes).'],
+  [/\[wcs\/binding-type-expectation\] class\.([^ ]+)/, ' Write "class.$1: path|truthy" to toggle on truthiness.'],
+  [/path segments — the limit/, " Every prefix of a path is interned, so the cost grows with the square of the depth."],
+  [/\[wcs\/index-arity\] \$resolve/, ' $resolve takes one index per "*"; $getAll / $setAll take at most that many (fewer expands the rest).'],
+  [/"([^"#]+)#[^"]*" is not a filter name: "#" cannot appear in one/, ' Modifiers belong on the left side of the binding, before the ":" \u2014 write "$1" here.']
+];
+var RENAMED_FILTERS = {
+  inc: "add",
+  dec: "sub",
+  fix: "toFixed",
+  uc: "upper",
+  lc: "lower",
+  cap: "capitalize",
+  rep: "repeat",
+  rev: "reverse",
+  pad: "padStart",
+  null: "nullIfEmpty"
+};
+function removedFilter(name) {
+  if (name === "substr") return ' "substr" was removed in 4.0 \u2014 write slice(start, start + length): slice takes the end index, not a length.';
+  const to = Object.hasOwn(RENAMED_FILTERS, name) ? RENAMED_FILTERS[name] : null;
+  return to === null ? null : ` "${name}" was renamed "${to}" in 3.2 and removed in 4.0 \u2014 write "${to}".`;
+}
+function explain(message, subject, candidates) {
+  const removed = subject !== void 0 && message.includes("[wcs/filter-unknown]") ? removedFilter(subject) : null;
+  let out = removed ?? (subject !== void 0 && candidates !== void 0 ? didYouMean(subject, candidates) : "");
+  for (const [re, text2] of GUIDES) {
+    const m = re.exec(message);
+    if (m !== null) out += text2.replace(/\$(\d)/g, (_, i) => m[Number(i)]);
+  }
+  const mods = /"([^"#]+)#([^"]+)": a binding takes one modifier list/.exec(message);
+  if (mods !== null) out += ` \u2014 write "${mods[1]}#${mods[2].split("#").join(",")}".`;
+  if (message.includes("[wcs/filter-unknown]")) {
+    if (message.includes("formats add-on")) out += ` On a split auto page: features="formats" on the root, or "$features": ["formats"].`;
+    else if (!FORMATS_FILTER_NAMES.some(hasFilter)) {
+      out += ` No formatting filters are installed \u2014 add the formats add-on: installFeatures([formats]) from "@wcstack/state/features/formats" (a split auto page: features="formats", or "$features").`;
     }
-    const pathInfo = getPathInfo(statePathName);
-    return {
-        statePathName,
-        statePathInfo: pathInfo,
-        outFilters: filters,
-    };
+  }
+  const code = /\[wcs\/([\w-]+)\]/.exec(message);
+  if (code !== null && LINT_CODES.has(code[1]) && !/ path segments — the limit is |inserted content was not rendered|formats add-on/.test(message)) out += LINT_HINT;
+  return out;
 }
 
-// format: propPart:statePart; propPart:statePart; ...
-// special-propPart:
-//   if: statePart (single binding for conditional rendering)
-//   else: (single binding for conditional rendering, and statePart is ignored)
-//   elseif: statePart only (single binding for conditional rendering)
-//   for: statePart only (single binding for loop rendering)
-//   onclick: statePart, onchange: statePart etc. (event listeners)
-//   ...: statePart (spread — expand wcBindable properties+inputs of target object)
-/** 左辺に修飾子も入力フィルタも取らない束縛（構造ディレクティブと spread）— 付いていれば拒否する（要件 B4） */
-const KEYWORDS_WITHOUT_MODIFIERS = new Set([ELSE_KEYWORD, 'if', 'elseif', 'for', SPREAD_PROP]);
-/**
- * 明示のプロパティ形（`.name:`）の先頭に置けない語 — 名前空間として読まれる語（要件 B5）。
- * `state`（`VOLUME_INJECTION_PROP`）も含む: `<wcs-state mount>` 上の左辺 `state.<key>:` は
- * 注入の宣言（要件 B14③）なので、`.state.taxRate:` はプロパティか注入か曖昧になる。
- */
-const EXPLICIT_PROPERTY_REJECTED_HEADS = new Set([
-    CLASS_NAMESPACE, ATTR_NAMESPACE, STYLE_NAMESPACE, COMMAND_NAMESPACE, EVENT_TOKEN_NAMESPACE,
-    VOLUME_INJECTION_PROP,
+// src/parser/parsePropPart.ts
+var cacheFilterInfos2 = /* @__PURE__ */ new Map();
+var clearPropPartCache = () => cacheFilterInfos2.clear();
+function parsePropPart(propPart) {
+  const [propText, filters] = splitFilters(propPart, "input", cacheFilterInfos2);
+  const modifierParts = propText.split(MODIFIER_SEPARATOR).map(trimFn);
+  if (modifierParts.length > 2) {
+    raise(115 /* OneModifierList */, [propText]);
+  }
+  const [propName, propModifiersText] = modifierParts;
+  const propSegments = propName.split(DELIMITER).map(trimFn);
+  const isExplicitProperty = propSegments.length > 1 && propSegments[0] === "";
+  if (!isExplicitProperty && (propName.length === 0 || propSegments.some((segment) => segment.length === 0))) {
+    raise(116 /* NoPropertyName */, [propPart]);
+  }
+  const propModifiers = propModifiersText ? propModifiersText.split(",").map(trimFn) : [];
+  return {
+    propName,
+    propSegments,
+    propModifiers,
+    inFilters: filters
+  };
+}
+
+// src/parser/types.ts
+var STRUCTURAL_BINDING_TYPE_SET = /* @__PURE__ */ new Set([
+  "if",
+  "elseif",
+  "else",
+  "for"
+]);
+
+// src/parser/parseBindTextsForElement.ts
+var KEYWORDS_WITHOUT_MODIFIERS = /* @__PURE__ */ new Set([ELSE_KEYWORD, "if", "elseif", "for", SPREAD_PROP]);
+var EXPLICIT_PROPERTY_REJECTED_HEADS = /* @__PURE__ */ new Set([
+  CLASS_NAMESPACE,
+  ATTR_NAMESPACE,
+  STYLE_NAMESPACE,
+  COMMAND_NAMESPACE,
+  EVENT_TOKEN_NAMESPACE,
+  VOLUME_INJECTION_PROP
 ]);
-/**
- * `data-wcs` の値をバインディングごとに区切る（前後の空白は残す — tooling が位置を数えられるように）。
- * 引用符の中の `;` は区切りではない（要件 B1 — `join(';')`）。ランタイムと tooling（`@wcstack/state/parser`）で共有する
- */
 function splitBindTexts(bindText) {
-    return splitOutsideQuotes(bindText, BINDING_SEPARATOR);
+  return splitOutsideQuotes(bindText, BINDING_SEPARATOR);
 }
 function parseBindTextsForElement(bindText) {
-    const [...bindTexts] = splitBindTexts(bindText).map(trimFn).filter(s => s.length > 0);
-    const results = bindTexts.map((bindText) => {
-        // 左辺と右辺の区切りも引用符の外だけで探す（要件 B1）。`value|defaults(':'): path` の
-        // 引数の中の `:` を区切りとして拾っていた
-        const separatorIndex = indexOfOutsideQuotes(bindText, PROP_VALUE_SEPARATOR);
-        if (separatorIndex === -1) {
-            raiseError(`[wcs/binding-syntax] Invalid bindText: "${bindText}". Missing ':' separator between propPart and statePart.${LINT_HINT}`);
-        }
-        const propPart = bindText.slice(0, separatorIndex).trim();
-        const statePart = bindText.slice(separatorIndex + 1).trim();
-        // 種別は修飾子・入力フィルタより前の名前で決める（要件 B4）。以前は左辺全体との完全一致で
-        // 判定していたので、`radio#ro:` が汎用プロパティに落ちていた。
-        // **不変条件**: ここは素の `split` でよい — 取り出すのは最初の区切りより前の**先頭**片で、
-        // 引用符を含みうるのは入力フィルタの引数（引用符外の最初の `|` より後ろ）だけなので、
-        // `#` も `|` も引用符の中では出会わない。**修飾子の値に引用符を許す拡張を入れるなら
-        // `indexOfOutsideQuotes` に替えること**（`parsePropPart.ts` の同じ注記と対）。
-        const keyword = propPart.split(MODIFIER_SEPARATOR)[0].split(FILTER_SEPARATOR)[0].trim();
-        if (keyword !== propPart && KEYWORDS_WITHOUT_MODIFIERS.has(keyword)) {
-            raiseError(`[wcs/binding-syntax] "${bindText}": "${keyword}" takes no modifiers or filters on its left side — write "${keyword}:".${LINT_HINT}`);
-        }
-        if (propPart === ELSE_KEYWORD) {
-            if (statePart.length > 0) {
-                // else は値を取らない（要件 B2）。以前は右辺を黙って捨てていた
-                raiseError(`[wcs/binding-syntax] "${bindText}": "else" takes no value — write "else:".${LINT_HINT}`);
-            }
-            const pathInfo = getPathInfo('#else');
-            return {
-                propName: ELSE_KEYWORD,
-                propSegments: [ELSE_KEYWORD],
-                propModifiers: [],
-                statePathName: '#else',
-                statePathInfo: pathInfo,
-                inFilters: [],
-                outFilters: [],
-                bindingType: 'else',
-            };
-        }
-        else if (propPart === SPREAD_PROP) {
-            // 空の右辺は spread 専用の語彙で先に落とす。`parseStatePart` の一般の空パス診断
-            // （「the right side of a binding must name a state path」）より、ここでは
-            // 「spread target path is required」のほうが直し方を指している
-            if (statePart.length === 0) {
-                raiseError(`[wcs/binding-syntax] Invalid spread binding "${bindText}": spread target path is required.${LINT_HINT}`);
-            }
-            const stateResult = parseStatePart(statePart);
-            if (stateResult.outFilters.length > 0) {
-                raiseError(`[wcs/binding-syntax] Invalid spread binding "${bindText}": filters are not allowed on spread targets.${LINT_HINT}`);
-            }
-            return {
-                propName: SPREAD_PROP,
-                propSegments: [SPREAD_PROP],
-                propModifiers: [],
-                inFilters: [],
-                ...stateResult,
-                bindingType: 'spread',
-            };
-        }
-        else if (propPart === 'if'
-            || propPart === 'elseif'
-            || propPart === 'for') {
-            const stateResult = parseStatePart(statePart);
-            return {
-                propName: propPart,
-                propSegments: [propPart],
-                propModifiers: [],
-                inFilters: [],
-                ...stateResult,
-                bindingType: propPart,
-            };
-        }
-        else if (keyword === 'radio' || keyword === 'checkbox') {
-            // 修飾子（`#ro`・`#onchange` …）と入力フィルタは radio / checkbox のハンドラが読む（要件 B4）
-            const stateResult = parseStatePart(statePart);
-            const propResult = parsePropPart(propPart);
-            return {
-                ...propResult,
-                ...stateResult,
-                bindingType: keyword,
-            };
-        }
-        else {
-            const stateResult = parseStatePart(statePart);
-            const propResult = parsePropPart(propPart);
-            // 左辺の先頭ドット（`.online:`）は明示のプロパティ形（要件 B5・3.x 計画 D34）。ドットの無い形と
-            // 同じ束縛だが、`on` で始まってもイベントにはならない（`online:` は "line" イベントを待つ）。
-            // 名前空間の語（`.class.x` / `.command.x` …）はプロパティか名前空間か曖昧なので受けない
-            if (propResult.propSegments[0] === '' && propResult.propSegments.length > 1) {
-                const propSegments = propResult.propSegments.slice(1);
-                if (propSegments.includes('') || EXPLICIT_PROPERTY_REJECTED_HEADS.has(propSegments[0])) {
-                    raiseError(`[wcs/binding-syntax] "${propPart}": a leading "." binds an element property by name — ` +
-                        `write a non-empty property that is not a namespace (${[...EXPLICIT_PROPERTY_REJECTED_HEADS].join(", ")}).${LINT_HINT}`);
-                }
-                return {
-                    ...propResult,
-                    propName: propSegments.join(DELIMITER),
-                    propSegments,
-                    ...stateResult,
-                    bindingType: 'prop',
-                };
-            }
-            // eventToken.<prop>: <name> は要素 dispatch を state へ流す pub/sub 配線。
-            // 値適用ではないため bindingType 'event' として listener attach 経路に乗せる。
-            if (propResult.propSegments[0] === EVENT_TOKEN_NAMESPACE) {
-                return {
-                    ...propResult,
-                    ...stateResult,
-                    bindingType: 'event',
-                };
-            }
-            if (propResult.propSegments[0].startsWith(EVENT_PROP_PREFIX)) {
-                return {
-                    ...propResult,
-                    ...stateResult,
-                    bindingType: 'event',
-                };
-            }
-            else {
-                return {
-                    ...propResult,
-                    ...stateResult,
-                    bindingType: 'prop',
-                };
-            }
-        }
-    });
-    // check for sigle binding for 'if', 'elseif', 'else', 'for'
-    if (results.length > 1) {
-        const isIncludeSingleBinding = results.some(r => STRUCTURAL_BINDING_TYPE_SET.has(r.bindingType));
-        if (isIncludeSingleBinding) {
-            // lint 側の単独バインディング検査（bindingValidator の structuralMustBeSingle）が
-            // 同じケースを検出するため誘導を付ける（三面同語彙）。
-            raiseError(`[wcs/template-syntax] Invalid bindText: "${bindText}". 'if', 'elseif', 'else', and 'for' bindings must be single binding. Put the structural binding alone in its own data-wcs (e.g. <template data-wcs="for: items">).${LINT_HINT}`);
-        }
+  const bindTexts = splitBindTexts(bindText).map(trimFn).filter((s) => s.length > 0);
+  const results = bindTexts.map((bindText2) => {
+    const separatorIndex = indexOfOutsideQuotes(bindText2, PROP_VALUE_SEPARATOR);
+    if (separatorIndex === -1) {
+      raise(101 /* BindTextNoColon */, [bindText2]);
     }
-    return results;
-}
-
-function parseBindTextForEmbeddedNode(bindText) {
-    const stateResult = parseStatePart(bindText);
-    return {
-        propName: 'textContent',
-        propSegments: ['textContent'],
+    const propPart = bindText2.slice(0, separatorIndex).trim();
+    const statePart = bindText2.slice(separatorIndex + 1).trim();
+    const keyword = propPart.split(MODIFIER_SEPARATOR)[0].split(FILTER_SEPARATOR)[0].trim();
+    if (KEYWORDS_WITHOUT_MODIFIERS.has(keyword)) {
+      if (keyword !== propPart) {
+        raise(102 /* StructuralTakesNoModifiers */, [bindText2, keyword]);
+      }
+      let stateResult2;
+      if (keyword === ELSE_KEYWORD) {
+        if (statePart.length > 0) {
+          raise(103 /* ElseTakesNoValue */, [bindText2]);
+        }
+        stateResult2 = { statePathName: "#else", outFilters: [] };
+      } else if (keyword === SPREAD_PROP) {
+        if (statePart.length === 0) {
+          raise(104 /* SpreadNoPath */, [bindText2]);
+        }
+        stateResult2 = parseStatePart(statePart);
+        if (stateResult2.outFilters.length > 0) {
+          raise(105 /* SpreadNoFilters */, [bindText2]);
+        }
+      } else {
+        stateResult2 = parseStatePart(statePart);
+        if (keyword === "for" && stateResult2.outFilters.length > 0) {
+          raise(121 /* ForNoFilters */, [bindText2]);
+        }
+      }
+      return {
+        propName: keyword,
+        propSegments: [keyword],
         propModifiers: [],
         inFilters: [],
+        ...stateResult2,
+        bindingType: keyword === SPREAD_PROP ? "spread" : keyword
+      };
+    }
+    const stateResult = parseStatePart(statePart);
+    const propResult = parsePropPart(propPart);
+    const [head] = propResult.propSegments;
+    if (head === "" && propResult.propSegments.length > 1) {
+      const propSegments = propResult.propSegments.slice(1);
+      if (propSegments.includes("") || EXPLICIT_PROPERTY_REJECTED_HEADS.has(propSegments[0])) {
+        raise(106 /* LeadingDotNamespace */, [propPart]);
+      }
+      return {
+        ...propResult,
+        propName: propSegments.join(DELIMITER),
+        propSegments,
         ...stateResult,
-        bindingType: 'text',
+        bindingType: "prop"
+      };
+    }
+    return {
+      ...propResult,
+      ...stateResult,
+      // 修飾子（`#ro`・`#onchange` …）と入力フィルタは radio / checkbox のハンドラが読む（要件 B4）。
+      // eventToken.<prop>: <name> は要素 dispatch を state へ流す pub/sub 配線。
+      // 値適用ではないため bindingType 'event' として listener attach 経路に乗せる。
+      bindingType: keyword === "radio" || keyword === "checkbox" ? keyword : head === EVENT_TOKEN_NAMESPACE || head.startsWith(EVENT_PROP_PREFIX) ? "event" : "prop"
     };
+  });
+  if (results.length > 1) {
+    const isIncludeSingleBinding = results.some((r) => STRUCTURAL_BINDING_TYPE_SET.has(r.bindingType));
+    if (isIncludeSingleBinding) {
+      raise(201 /* StructuralNotSingle */, [bindText]);
+    }
+  }
+  return results;
 }
 
-/**
- * parser.ts — `data-wcs` バインディング構文の正本パーサを tooling 向けに公開する
- * サブパスエントリ（`@wcstack/state/parser`）。
- *
- * `./manifest` と同じ「実装が唯一の正本」パターン（docs/static-wiring-dx-design.md D2）。
- * vscode-wcs の正規表現パーサ・devtools の declaredScan 簡易パーサという複製実装を
- * 段階的にこの正本へ寄せるための土台。
- *
- * 契約:
- * - DOM 非依存・純関数（bindText 文字列 → ParseBindTextResult[]）。Node でそのまま動く
- *   （__tests__/parser.test.ts が node 環境で検証する）。
- * - **位置情報は持たず、不正構文は raiseError で throw する**。エラー耐性と診断 range の
- *   生成は消費側（vscode-wcs の positional ラッパー）の責務（同 D3）— ランタイムの
- *   サイズと責務をここで増やさない。
- * - `getPathInfo` はパス文字列の解析済みビュー（セグメント・ワイルドカード位置・親パス
- *   チェーン）を返す純関数。静的依存グラフの親チェーン展開はこの情報から機械的に再現できる。
- *   同一パス → 同一インスタンスの保証は**このエントリのモジュールインスタンス内**でのみ
- *   成立する（`.` エントリは別バンドル＝別キャッシュ。ランタイムの PathInfo と identity
- *   比較してはならない）。キャッシュは無制限（evict なし）— 言語サーバー等の長時間
- *   プロセスではメモリが増え続ける点に留意（断ち方は `clearPathInfoCacheForTooling`）。
- *   **増え方はパス種数への単調比例ではない**: `PathInfo` は自分の全ての接頭辞を intern
- *   するので、1 本のパスが持ち込む量はその深さの 2 乗に比例する。深さは
- *   `MAX_PATH_SEGMENTS` で頭打ちになる（超えたパスは `[wcs/binding-syntax]` で拒否）。
- * - `ParseBindTextResult.uuid` はランタイム内部（構造テンプレートのハイドレーション台帳）
- *   用のフィールドで、このパーサの戻り値では常に undefined。
- *
- * 公開面は意図的に最小（公開＝恒久契約）。`expandSpread` は live Element と
- * CustomElementRegistry を要するためここには含めない — ブラウザ内の消費者
- * （devtools の declared 正本化）は state 自身が pull API で答える。
- */
-/**
- * このエントリの内部キャッシュ（PathInfo intern・propPart/statePart のパース結果・フィルタ関数クロージャ）を全て捨てる。
- *
- * 言語サーバー等の**長時間プロセス専用**。編集中の中間パス（`user.n` 等）が
- * 無制限キャッシュに恒久 intern されてメモリが単調増加するため、ドキュメント
- * クローズ等の区切りで呼ぶ。クリア後の getPathInfo は同一パスに**新しい**
- * インスタンスを返す — 「同一パス → 同一参照」の保証はクリアを跨がない。
- * ランタイム（`.` エントリ）にはこの API は無く、呼ばれることもない。
- */
-function clearParserCaches() {
-    clearPathInfoCacheForTooling();
-    clearPropPartCacheForTooling();
-    clearStatePartCacheForTooling();
-    clearFilterFnCacheForTooling();
+// src/parser/parseBindTextForEmbeddedNode.ts
+function parseBindTextForEmbeddedNode(bindText) {
+  const stateResult = parseStatePart(bindText);
+  return {
+    propName: "textContent",
+    propSegments: ["textContent"],
+    propModifiers: [],
+    inFilters: [],
+    ...stateResult,
+    bindingType: "text"
+  };
 }
 
-export { clearParserCaches, getPathInfo, indexOfOutsideQuotes, parseBindTextForEmbeddedNode, parseBindTextsForElement, splitBindTexts, splitOutsideQuotes };
+// src/public/pathInfo.ts
+var cache = /* @__PURE__ */ new Map();
+var ids = 0;
+function clearPathInfoCache() {
+  cache.clear();
+}
+function getPathInfo(path) {
+  const known = cache.get(path);
+  if (known !== void 0) return known;
+  if (path.includes(RECURSION_WILDCARD)) raise(1101 /* RecursionUnsupported */, [path]);
+  const count = path.split(DELIMITER).length;
+  if (count > MAX_PATH_SEGMENTS) raise(117 /* TooManySegments */, [path, count]);
+  const info = Object.freeze(new PathInfo(path));
+  cache.set(path, info);
+  return info;
+}
+var PathInfo = class {
+  constructor(path) {
+    this.id = ++ids;
+    this.cumulativePaths = [];
+    this.cumulativePathInfos = [];
+    this.wildcardPaths = [];
+    this.indexByWildcardPath = {};
+    this.wildcardPathInfos = [];
+    this.wildcardParentPaths = [];
+    this.wildcardParentPathInfos = [];
+    this.wildcardPositions = [];
+    const info = (p) => p === path ? this : getPathInfo(p);
+    const segments = path.split(DELIMITER);
+    let current = "";
+    let prev = "";
+    for (let i = 0; i < segments.length; i++) {
+      current += segments[i];
+      if (segments[i] === WILDCARD) {
+        this.indexByWildcardPath[current] = this.wildcardPaths.length;
+        this.wildcardPaths.push(current);
+        this.wildcardPathInfos.push(info(current));
+        this.wildcardParentPaths.push(prev);
+        this.wildcardParentPathInfos.push(info(prev));
+        this.wildcardPositions.push(i);
+      }
+      this.cumulativePaths.push(current);
+      this.cumulativePathInfos.push(info(current));
+      prev = current;
+      current += DELIMITER;
+    }
+    const last = this.wildcardPaths.length > 0 ? this.wildcardPaths[this.wildcardPaths.length - 1] : null;
+    const parent = this.cumulativePaths.length > 1 ? this.cumulativePaths[this.cumulativePaths.length - 2] : null;
+    this.path = path;
+    this.segments = segments;
+    this.lastSegment = segments[segments.length - 1];
+    this.cumulativePathSet = new Set(this.cumulativePaths);
+    this.cumulativePathInfoSet = new Set(this.cumulativePathInfos);
+    this.wildcardPathSet = new Set(this.wildcardPaths);
+    this.wildcardPathInfoSet = new Set(this.wildcardPathInfos);
+    this.wildcardParentPathSet = new Set(this.wildcardParentPaths);
+    this.wildcardParentPathInfoSet = new Set(this.wildcardParentPathInfos);
+    this.lastWildcardPath = last;
+    this.lastWildcardInfo = last !== null ? info(last) : null;
+    this.parentPath = parent;
+    this.parentPathInfo = parent !== null ? info(parent) : null;
+    this.wildcardCount = this.wildcardPaths.length;
+  }
+};
+
+// src/public/parser.ts
+hooks.render = render;
+hooks.explain = explain;
+var withInfo = (b) => {
+  const path = b.statePathName;
+  const notPath = path.startsWith("$command.") || b.bindingType === "event" && (b.propSegments[0] === "eventToken" || !path.includes("."));
+  if (!notPath && path.split(".").some((s) => s === "__proto__" || s === "prototype")) raise(120 /* UnsafeSegment */, [path]);
+  return { ...b, statePathInfo: getPathInfo(path) };
+};
+function parseBindTextsForElement2(bindText) {
+  return parseBindTextsForElement(bindText).map(withInfo);
+}
+function parseBindTextForEmbeddedNode2(bindText) {
+  return withInfo(parseBindTextForEmbeddedNode(bindText));
+}
+function clearParserCaches() {
+  clearPathInfoCache();
+  clearPropPartCache();
+  clearStatePartCache();
+}
+export {
+  clearParserCaches,
+  getPathInfo,
+  indexOfOutsideQuotes,
+  parseBindTextForEmbeddedNode2 as parseBindTextForEmbeddedNode,
+  parseBindTextsForElement2 as parseBindTextsForElement,
+  splitBindTexts,
+  splitOutsideQuotes
+};