claude-code-jsonl-generator 0.0.0 → 0.1.0-beta.2

package/dist/emit-guards-DWi5YDuA.cjs

@@ -0,0 +1,312 @@
+//#region \0rolldown/runtime.js
+var __create = Object.create;
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __copyProps = (to, from, except, desc) => {
+	if (from && typeof from === "object" || typeof from === "function") for (var keys = __getOwnPropNames(from), i = 0, n = keys.length, key; i < n; i++) {
+		key = keys[i];
+		if (!__hasOwnProp.call(to, key) && key !== except) __defProp(to, key, {
+			get: ((k) => from[k]).bind(null, key),
+			enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable
+		});
+	}
+	return to;
+};
+var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", {
+	value: mod,
+	enumerable: true
+}) : target, mod));
+//#endregion
+let node_fs_promises = require("node:fs/promises");
+let node_fs = require("node:fs");
+let node_readline = require("node:readline");
+let node_path = require("node:path");
+node_path = __toESM(node_path, 1);
+//#region src/scanner.ts
+/**
+* `.jsonl` ファイルを再帰収集する。
+*
+* @param dir - 起点ディレクトリ
+* @returns 絶対パス配列
+*/
+async function collectJsonlFiles(dir) {
+	const out = [];
+	for (const entry of await (0, node_fs_promises.readdir)(dir)) {
+		const full = node_path.default.join(dir, entry);
+		if ((await (0, node_fs_promises.stat)(full)).isDirectory()) out.push(...await collectJsonlFiles(full));
+		else if (entry.endsWith(".jsonl")) out.push(full);
+	}
+	return out;
+}
+/**
+* コーパスを走査し type ごとに生サンプルを集約する。
+*
+* @param corpusDir - コーパスのルートディレクトリ
+* @param maxSamplesPerType - type ごとの保持上限 (メモリ保護、既定 5000)
+*/
+async function scanCorpus(corpusDir, maxSamplesPerType = 5e3) {
+	const samplesByType = /* @__PURE__ */ new Map();
+	let totalLines = 0;
+	let malformedLines = 0;
+	let typelessLines = 0;
+	for (const file of await collectJsonlFiles(corpusDir)) {
+		const rl = (0, node_readline.createInterface)({
+			input: (0, node_fs.createReadStream)(file, { encoding: "utf8" }),
+			crlfDelay: Infinity
+		});
+		for await (const raw of rl) {
+			if (raw.trim() === "") continue;
+			let obj;
+			try {
+				obj = JSON.parse(raw);
+			} catch {
+				malformedLines += 1;
+				continue;
+			}
+			totalLines += 1;
+			if (typeof obj !== "object" || obj === null || Array.isArray(obj) || typeof obj.type !== "string") {
+				typelessLines += 1;
+				continue;
+			}
+			const record = obj;
+			const type = record.type;
+			const bucket = samplesByType.get(type) ?? [];
+			if (bucket.length < maxSamplesPerType) {
+				bucket.push(record);
+				samplesByType.set(type, bucket);
+			}
+		}
+	}
+	return {
+		samplesByType,
+		totalLines,
+		malformedLines,
+		typelessLines
+	};
+}
+//#endregion
+//#region src/infer.ts
+/**
+* JSON 値の種別を返す。
+*
+* @param v - 対象値
+*/
+function jsonKind(v) {
+	if (v === null) return "null";
+	if (Array.isArray(v)) return "array";
+	const t = typeof v;
+	if ([
+		"string",
+		"number",
+		"boolean"
+	].includes(t)) return t;
+	return "object";
+}
+/** `InferOptions.maxLiterals` 省略時の既定値。 */
+const DEFAULT_MAX_LITERALS = 20;
+/**
+* 同一 type のサンプル群からフィールド形状を推論する。
+*
+* @param samples - 生 JSON オブジェクト配列
+* @param options - 推論オプション
+*/
+function inferShape(samples, options = {}) {
+	const maxLiterals = options.maxLiterals ?? DEFAULT_MAX_LITERALS;
+	const count = samples.length;
+	const presence = /* @__PURE__ */ new Map();
+	const kinds = /* @__PURE__ */ new Map();
+	const literals = /* @__PURE__ */ new Map();
+	for (const sample of samples) for (const [key, value] of Object.entries(sample)) {
+		presence.set(key, (presence.get(key) ?? 0) + 1);
+		const kset = kinds.get(key) ?? /* @__PURE__ */ new Set();
+		const kind = jsonKind(value);
+		kset.add(kind);
+		kinds.set(key, kset);
+		const cur = literals.get(key);
+		if (cur === null) {} else if (kind === "string") {
+			const set = cur ?? /* @__PURE__ */ new Set();
+			set.add(value);
+			if (set.size > maxLiterals) literals.set(key, null);
+			else literals.set(key, set);
+		}
+	}
+	const fields = /* @__PURE__ */ new Map();
+	for (const [key, seen] of presence) {
+		const lit = literals.get(key);
+		const kindSet = kinds.get(key) ?? /* @__PURE__ */ new Set();
+		const allString = kindSet.size === 1 && kindSet.has("string");
+		fields.set(key, {
+			required: seen === count,
+			kinds: kindSet,
+			literals: allString && lit ? lit : void 0
+		});
+	}
+	return {
+		count,
+		fields
+	};
+}
+//#endregion
+//#region src/emit-types.ts
+/** JS 識別子として安全な名前かどうかを判定する正規表現。 */
+const IDENTIFIER_RE$1 = /^[A-Za-z_$][\w$]*$/;
+/** シングルクォート文字列リテラルとして安全になるよう `'` をエスケープする。 */
+function escapeSingleQuote$1(s) {
+	return s.replaceAll("'", String.raw`\'`);
+}
+/**
+* 観測種別集合 (と任意のリテラル集合) を TypeScript 型文字列に写像する。
+*
+* @param kinds - 観測された JSON 種別
+* @param literals - 文字列リテラル集合 (あれば)
+*/
+function tsTypeOf(kinds, literals) {
+	const parts = [];
+	const mapped = {
+		string: "string",
+		number: "number",
+		boolean: "boolean",
+		null: "null",
+		array: "unknown[]",
+		object: "Record<string, unknown>"
+	};
+	if (literals && literals.size > 0 && kinds.has("string")) {
+		const lits = [...literals].toSorted().map((s) => `'${escapeSingleQuote$1(s)}'`);
+		parts.push(`${lits.join(" | ")} | (string & {})`);
+	} else if (kinds.has("string")) parts.push("string");
+	for (const k of [
+		"number",
+		"boolean",
+		"array",
+		"object",
+		"null"
+	]) if (kinds.has(k)) parts.push(mapped[k]);
+	return [...new Set(parts)].join(" | ");
+}
+/** JS 識別子として安全ならそのまま、そうでなければクォートする。 */
+function propKey(name) {
+	return IDENTIFIER_RE$1.test(name) ? name : `'${escapeSingleQuote$1(name)}'`;
+}
+/** interface 名として使うため type 文字列を PascalCase 化する (エミッタ内部用)。 */
+function pascalCase(type) {
+	return type.split(/[-_]/).map((p) => p.charAt(0).toUpperCase() + p.slice(1)).join("");
+}
+/**
+* 1 type の interface 文字列を生成する。
+*
+* @param interfaceName - 出力する interface 名 (例 LastPromptEntry)
+* @param type - 判別子文字列 (例 last-prompt)
+* @param shape - 推論結果
+*/
+function emitInterface(interfaceName, type, shape) {
+	const lines = [
+		`export interface ${interfaceName} {`,
+		`  _kind: 'known'`,
+		`  type: '${escapeSingleQuote$1(type)}'`
+	];
+	for (const [name, field] of shape.fields) {
+		if (name === "type") continue;
+		const opt = field.required ? "" : "?";
+		const tsType = tsTypeOf(field.kinds, field.literals);
+		lines.push(`  ${propKey(name)}${opt}: ${tsType}`);
+	}
+	lines.push("}");
+	return lines.join("\n");
+}
+//#endregion
+//#region src/emit-guards.ts
+/** JS 識別子として安全な名前かどうかを判定する正規表現。 */
+const IDENTIFIER_RE = /^[A-Za-z_$][\w$]*$/;
+/** シングルクォート文字列リテラルとして安全になるよう `'` をエスケープする。 */
+function escapeSingleQuote(s) {
+	return s.replaceAll("'", String.raw`\'`);
+}
+/** 単一種別の required フィールドに対する検査式を返す。 */
+function checkExpr(name, kinds) {
+	const access = `v.${IDENTIFIER_RE.test(name) ? name : `['${escapeSingleQuote(name)}']`}`;
+	const hasNull = kinds.has("null");
+	const nonNull = [...kinds].filter((k) => k !== "null");
+	if (nonNull.length !== 1) return `'${escapeSingleQuote(name)}' in v`;
+	const kind = nonNull[0];
+	const expr = {
+		string: `isString(${access})`,
+		number: `isNumber(${access})`,
+		boolean: `isBoolean(${access})`,
+		object: `isObject(${access})`,
+		array: `Array.isArray(${access})`
+	}[kind];
+	if (expr === void 0) return `'${escapeSingleQuote(name)}' in v`;
+	return hasNull ? `(${access} === null || ${expr})` : expr;
+}
+/**
+* 1 type のガード関数文字列を生成する。required フィールドのみ検査する。
+*
+* @param fnName - 関数名 (例 isMode)
+* @param type - 判別子
+* @param shape - 推論結果
+*/
+function emitGuardFunction(fnName, type, shape) {
+	const checks = [`isObject(v)`, `v.type === '${escapeSingleQuote(type)}'`];
+	for (const [name, field] of shape.fields) {
+		if (name === "type" || !field.required) continue;
+		const expr = checkExpr(name, field.kinds);
+		if (expr) checks.push(expr);
+	}
+	return [
+		`function ${fnName}(v: unknown): boolean {`,
+		`  return ${checks.join(" &&\n    ")}`,
+		`}`
+	].join("\n");
+}
+//#endregion
+Object.defineProperty(exports, "__toESM", {
+	enumerable: true,
+	get: function() {
+		return __toESM;
+	}
+});
+Object.defineProperty(exports, "emitGuardFunction", {
+	enumerable: true,
+	get: function() {
+		return emitGuardFunction;
+	}
+});
+Object.defineProperty(exports, "emitInterface", {
+	enumerable: true,
+	get: function() {
+		return emitInterface;
+	}
+});
+Object.defineProperty(exports, "inferShape", {
+	enumerable: true,
+	get: function() {
+		return inferShape;
+	}
+});
+Object.defineProperty(exports, "jsonKind", {
+	enumerable: true,
+	get: function() {
+		return jsonKind;
+	}
+});
+Object.defineProperty(exports, "pascalCase", {
+	enumerable: true,
+	get: function() {
+		return pascalCase;
+	}
+});
+Object.defineProperty(exports, "scanCorpus", {
+	enumerable: true,
+	get: function() {
+		return scanCorpus;
+	}
+});
+Object.defineProperty(exports, "tsTypeOf", {
+	enumerable: true,
+	get: function() {
+		return tsTypeOf;
+	}
+});

package/dist/index.cjs

@@ -0,0 +1,220 @@
+Object.defineProperty(exports, Symbol.toStringTag, { value: "Module" });
+const require_emit_guards = require("./emit-guards-DWi5YDuA.cjs");
+let zod = require("zod");
+//#region src/schemas.ts
+const entryBase = {
+	cwd: zod.z.string(),
+	entrypoint: zod.z.string(),
+	gitBranch: zod.z.string(),
+	isSidechain: zod.z.boolean(),
+	parentUuid: zod.z.string().nullable(),
+	sessionId: zod.z.string(),
+	timestamp: zod.z.string(),
+	userType: zod.z.string(),
+	uuid: zod.z.string(),
+	version: zod.z.string()
+};
+/** type 文字列 → 公開 zod スキーマ。 */
+const entrySchemas = {
+	assistant: zod.z.object({
+		...entryBase,
+		type: zod.z.literal("assistant"),
+		message: zod.z.object({}).loose(),
+		agentId: zod.z.string().optional(),
+		agentName: zod.z.string().optional(),
+		teamName: zod.z.string().optional(),
+		slug: zod.z.string().optional(),
+		sessionKind: zod.z.string().optional(),
+		session_id: zod.z.string().optional(),
+		requestId: zod.z.string().optional(),
+		isApiErrorMessage: zod.z.boolean().optional(),
+		apiErrorStatus: zod.z.number().optional(),
+		error: zod.z.unknown().optional(),
+		attributionAgent: zod.z.string().optional(),
+		attributionSkill: zod.z.string().optional(),
+		attributionPlugin: zod.z.string().optional(),
+		attributionMcpServer: zod.z.string().optional(),
+		attributionMcpTool: zod.z.string().optional()
+	}).loose(),
+	user: zod.z.object({
+		...entryBase,
+		type: zod.z.literal("user"),
+		message: zod.z.object({}).loose(),
+		agentId: zod.z.string().optional(),
+		agentName: zod.z.string().optional(),
+		teamName: zod.z.string().optional(),
+		slug: zod.z.string().optional(),
+		sessionKind: zod.z.string().optional(),
+		session_id: zod.z.string().optional(),
+		interruptedMessageId: zod.z.string().optional(),
+		isCompactSummary: zod.z.boolean().optional(),
+		isMeta: zod.z.boolean().optional(),
+		isVisibleInTranscriptOnly: zod.z.boolean().optional(),
+		mcpMeta: zod.z.unknown().optional(),
+		origin: zod.z.object({
+			kind: zod.z.string(),
+			from: zod.z.string().optional(),
+			senderTaskId: zod.z.string().optional()
+		}).loose().optional(),
+		permissionMode: zod.z.string().optional(),
+		planContent: zod.z.string().optional(),
+		promptId: zod.z.string().optional(),
+		promptSource: zod.z.string().optional(),
+		queuePriority: zod.z.string().optional(),
+		sourceToolAssistantUUID: zod.z.string().optional(),
+		sourceToolUseID: zod.z.string().optional(),
+		toolDenialKind: zod.z.string().optional(),
+		toolEndsTurn: zod.z.boolean().optional(),
+		toolUseResult: zod.z.unknown().optional()
+	}).loose(),
+	attachment: zod.z.object({
+		...entryBase,
+		type: zod.z.literal("attachment"),
+		attachment: zod.z.object({}).loose(),
+		agentId: zod.z.string().optional(),
+		agentName: zod.z.string().optional(),
+		teamName: zod.z.string().optional(),
+		slug: zod.z.string().optional(),
+		sessionKind: zod.z.string().optional(),
+		session_id: zod.z.string().optional()
+	}).loose(),
+	system: zod.z.object({
+		...entryBase,
+		type: zod.z.literal("system"),
+		subtype: zod.z.string(),
+		agentId: zod.z.string().optional(),
+		agentName: zod.z.string().optional(),
+		teamName: zod.z.string().optional(),
+		slug: zod.z.string().optional(),
+		sessionKind: zod.z.string().optional(),
+		session_id: zod.z.string().optional(),
+		compactMetadata: zod.z.unknown().optional(),
+		content: zod.z.unknown().optional(),
+		durationMs: zod.z.number().optional(),
+		error: zod.z.unknown().optional(),
+		hasOutput: zod.z.boolean().optional(),
+		hookAdditionalContext: zod.z.unknown().optional(),
+		hookCount: zod.z.number().optional(),
+		hookErrors: zod.z.unknown().optional(),
+		hookInfos: zod.z.unknown().optional(),
+		isMeta: zod.z.boolean().optional(),
+		level: zod.z.string().optional(),
+		logicalParentUuid: zod.z.string().optional(),
+		maxRetries: zod.z.number().optional(),
+		messageCount: zod.z.number().optional(),
+		pendingBackgroundAgentCount: zod.z.number().optional(),
+		pendingWorkflowCount: zod.z.number().optional(),
+		preventedContinuation: zod.z.boolean().optional(),
+		retryAttempt: zod.z.number().optional(),
+		retryInMs: zod.z.number().optional(),
+		stopReason: zod.z.string().optional(),
+		toolUseID: zod.z.string().optional()
+	}).loose(),
+	mode: zod.z.object({
+		type: zod.z.literal("mode"),
+		mode: zod.z.string(),
+		sessionId: zod.z.string()
+	}),
+	"permission-mode": zod.z.object({
+		type: zod.z.literal("permission-mode"),
+		permissionMode: zod.z.string(),
+		sessionId: zod.z.string()
+	}),
+	"ai-title": zod.z.object({
+		type: zod.z.literal("ai-title"),
+		aiTitle: zod.z.string(),
+		sessionId: zod.z.string()
+	}),
+	"agent-name": zod.z.object({
+		type: zod.z.literal("agent-name"),
+		agentName: zod.z.string(),
+		sessionId: zod.z.string()
+	}),
+	"agent-setting": zod.z.object({
+		type: zod.z.literal("agent-setting"),
+		agentSetting: zod.z.string(),
+		sessionId: zod.z.string()
+	}),
+	relocated: zod.z.object({
+		type: zod.z.literal("relocated"),
+		relocatedCwd: zod.z.string(),
+		sessionId: zod.z.string()
+	}),
+	"last-prompt": zod.z.object({
+		type: zod.z.literal("last-prompt"),
+		leafUuid: zod.z.string(),
+		sessionId: zod.z.string(),
+		lastPrompt: zod.z.string().optional()
+	}),
+	"worktree-state": zod.z.object({
+		type: zod.z.literal("worktree-state"),
+		worktreeSession: zod.z.object({}).loose().nullable(),
+		sessionId: zod.z.string()
+	}),
+	"bridge-session": zod.z.object({
+		type: zod.z.literal("bridge-session"),
+		bridgeSessionId: zod.z.string(),
+		lastSequenceNum: zod.z.number(),
+		sessionId: zod.z.string()
+	}),
+	"queue-operation": zod.z.object({
+		type: zod.z.literal("queue-operation"),
+		operation: zod.z.string(),
+		sessionId: zod.z.string(),
+		timestamp: zod.z.string(),
+		content: zod.z.unknown().optional()
+	}),
+	"pr-link": zod.z.object({
+		type: zod.z.literal("pr-link"),
+		prNumber: zod.z.number(),
+		prRepository: zod.z.string(),
+		prUrl: zod.z.string(),
+		sessionId: zod.z.string(),
+		timestamp: zod.z.string()
+	}),
+	"frame-link": zod.z.object({
+		type: zod.z.literal("frame-link"),
+		frameUrl: zod.z.string(),
+		path: zod.z.string(),
+		sessionId: zod.z.string(),
+		timestamp: zod.z.string()
+	}),
+	"file-history-snapshot": zod.z.object({
+		type: zod.z.literal("file-history-snapshot"),
+		isSnapshotUpdate: zod.z.boolean(),
+		messageId: zod.z.string(),
+		snapshot: zod.z.object({}).loose()
+	}),
+	started: zod.z.object({
+		type: zod.z.literal("started"),
+		key: zod.z.string(),
+		agentId: zod.z.string()
+	}),
+	result: zod.z.object({
+		type: zod.z.literal("result"),
+		key: zod.z.string(),
+		agentId: zod.z.string(),
+		result: zod.z.unknown().refine((v) => v !== void 0, { message: "result フィールドは必須です" })
+	})
+};
+/** 既知 type 名の配列。 */
+const KNOWN_TYPES = Object.keys(entrySchemas);
+/**
+* 指定 type のスキーマを返す。未知なら undefined。
+*
+* @param type - 判別子
+*/
+function knownTypeSchema(type) {
+	return entrySchemas[type];
+}
+//#endregion
+exports.KNOWN_TYPES = KNOWN_TYPES;
+exports.emitGuardFunction = require_emit_guards.emitGuardFunction;
+exports.emitInterface = require_emit_guards.emitInterface;
+exports.entrySchemas = entrySchemas;
+exports.inferShape = require_emit_guards.inferShape;
+exports.jsonKind = require_emit_guards.jsonKind;
+exports.knownTypeSchema = knownTypeSchema;
+exports.pascalCase = require_emit_guards.pascalCase;
+exports.scanCorpus = require_emit_guards.scanCorpus;
+exports.tsTypeOf = require_emit_guards.tsTypeOf;