createPricingEngine

首发版本：3.00.5

语法

createPricingEngine(name, instrument, [handler], [engineConfig])

详情

创建实时估值定价引擎。该引擎支持调用 DolphinDB 内置估值定价函数，也支持用户传入自定义函数。

定价引擎目前支持以下资产类型的内置定价函数：

资产类型 函数名 描述
固收 (Fixed Income) bondPricer 债券定价（支持贴现债、零息债、固定利率债）
bondFuturesPricer 国债期货定价
利率 (Interest Rate) irDepositPricer 存款定价
irFixedFloatingSwapPricer 标准利率互换定价
外汇 (FX) fxForwardPricer 外汇远期定价
fxSwapPricer 外汇掉期定价
fxEuropeanOptionPricer 外汇欧式期权定价

参数

name 字符串标量，表示引擎名称。该参数是引擎在一个数据节点上的唯一标识，可包含字母，数字和下划线，但必须以字母开头。

instrument INSTRUMENT 类型标量或向量，表示待定价的金融工具。定价过程中，市场数据匹配规则遵循 instrumentPricer 的规范。

handler 可选参数，一个自定义函数，用于处理定价完成后的结果输出。

  • 若 engineConfig 中设置 outputTime = true，函数签名为 def(eventTime, name, date, npv)；
  • 若 engineConfig 中设置 outputTime = false（默认值），函数签名为 def(name, date, npv)。

参数说明如下：

  • eventTime：NANOTIMESTAMP，触发事件的时间。
  • name：STRING，定价对象的 instrumentId。
  • date：DATE，定价日期。
  • npv：DOUBLE，定价结果的净值。

engineConfig 可选参数，一个字典，用于设置引擎运行配置。字典包含如下键值对：

  • numThreads：可选，整型标量，表示工作线程数，默认 8。设为 0 表示同步执行。

  • maxQueueDepth：可选，整型标量，表示最大队列深度，默认 10,000,000。

  • useSystemTime：可选，布尔标量，表示是否使用系统时间作为事件时间。默认为 true，使用系统时间作为事件时间。

  • timeColumn：可选，字符串标量，指定时间列作为事件时间。若指定该参数，必须将 useSystemTime 设置为 false。

  • outputTime：可选，布尔标量，表示是否输出事件时间列。默认值为 false。

返回值

返回一个定价引擎句柄。

例子

例1. 本例中定义待定价金融工具 240025.IB，根据 instrumentPricer 的默认规则，该工具将使用名为 CNY_TREASURY_BOND 的市场数据。通过调用 appendMktData 接口将相关市场数据写入定价引擎。

try{dropStreamEngine("PRICING_ENGINE")}catch(ex){}

bondDict  = {
    "productType": "Cash",
    "assetType": "Bond",
    "bondType": "FixedRateBond",
    "instrumentId": "240025.IB",
    "start": 2024.12.25,
    "maturity": 2031.12.25,
    "issuePrice": 100.0,
    "coupon": 0.0149,
    "frequency": "Annual",
    "dayCountConvention": "ActualActualISDA"
}
instrument = parseInstrument(bondDict)

pricingDate = 2025.08.18

curveDict = {
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": pricingDate,
    "currency": "CNY",
    "curveName": "CNY_TREASURY_BOND",
    "dayCountConvention": "ActualActualISDA",
    "compounding": "Compounded",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",
    "dates": [
        2025.09.18, 2025.11.18, 2026.02.18, 2026.08.18, 2027.08.18,
        2028.08.18, 2030.08.18, 2032.08.18, 2035.08.18, 2040.08.18,
        2045.08.18, 2055.08.18, 2065.08.18, 2075.08.18
    ],
    "values": [
        1.3000, 1.3700, 1.3898, 1.3865, 1.4299,
        1.4471, 1.6401, 1.7654, 1.7966, 1.9930,
        2.1834, 2.1397, 2.1987, 2.2225
    ] \ 100.0
}

discountCurve = parseMktData(curveDict)

share streamTable(1:0, `name`date`price, [STRING, DATE, DOUBLE]) as st

def myHandler(name, date, price) {
    tableInsert(st, name, date, price)
}

engine = createPricingEngine("PRICING_ENGINE", [instrument], myHandler)

appendMktData(engine, discountCurve)

select * from st
name date price
240025.IB 2025.08.18 99.60

例2. 本例以 10 年期国债期货 T2409 为例，先清理并创建 FuturesEngine，通过回调函数将引擎返回的 npv 写入结果表；随后分别构造可交割国债和国债期货合约，配置转换因子、结算日等关键参数，并插入国债收益率曲线作为贴现曲线，最终查看期货合约的估值结果。

// 清理历史同名引擎，避免重复创建时报错
try{dropStreamEngine("FuturesEngine")}catch(ex){}

// 创建结果表，用于接收并保存期货定价引擎回调返回的估值结果
share streamTable(1000:0, `engineName`instrumentId`npv`updateTime, [STRING, STRING, DOUBLE, TIMESTAMP]) as futuresResult

// 定义回调函数：每次引擎完成计算后，将合约名称、NPV 和更新时间写入结果表
def futuresHandler(name, date, npv){
    tableInsert(futuresResult, "FuturesEngine", name, npv, now())
}

// 构造可交割国债，作为国债期货 T2409 的标的券
underlyingBond = parseInstrument({
    "productType": "Cash",
    "assetType": "Bond",
    "bondType": "FixedRateBond",
    "instrumentId": "UNDERLYING_BOND",
    "start": 2020.09.15,
    "maturity": 2030.09.15,
    "issuePrice": 100.0,
    "coupon": 0.03,
    "frequency": "Annual",
    "dayCountConvention": "ActualActualISDA",
    "currency": "CNY",
    "subType": "TREASURY_BOND",
    "discountCurve": "CNY_TREASURY_BOND"
})

// 构造国债期货合约，重点设置标的券、名义票息、转换因子和交割结算日
futuresIns = parseInstrument({
    "productType": "Futures",
    "futuresType": "BondFutures",
    "instrumentId": "T2409",
    "nominal": 100.0,
    "maturity": 2024.09.13,
    "settlement": 2024.09.17,
    "underlying": underlyingBond,
    "nominalCouponRate": 0.03,
    "conversionFactor": 0.9812
})

// 创建期货定价引擎，并绑定待估值合约与回调函数
engine = createPricingEngine("FuturesEngine", futuresIns, futuresHandler)

// 构造国债收益率曲线；curveName 需与标的券中的 discountCurve 保持一致
discountCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": 2024.09.10,
    "currency": "CNY",
    "curveName": "CNY_TREASURY_BOND",
    "dayCountConvention": "ActualActualISDA",
    "compounding": "Compounded",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",
    "dates": [2024.12.10, 2025.12.10, 2026.12.10],
    "values": [0.000180, 0.000190, 0.000200]
})

// 向引擎追加市场数据，触发估值计算并通过 futuresHandler 写入结果表
appendMktData(engine, discountCurve)

// 查看 T2409 的估值结果
select * from futuresResult
engineName instrumentId npv updateTime
FuturesEngine T2409 117.84882688404052 2026.08.27 11:45:47.979

例3. 本例以 3 个月人民币定期存款为例，展示定价引擎如何根据本金、合同利率和贴现曲线，计算存款现值和应计利息。

// 清理历史同名引擎，避免重复创建时报错
try{dropStreamEngine("DepositEngine")}catch(ex){}

// 创建结果表，用于接收并保存存款定价引擎回调返回的估值结果
share streamTable(1000:0, `engineName`instrumentId`npv`updateTime, [STRING, STRING, DOUBLE, TIMESTAMP]) as depositResult

// 定义回调函数：每次引擎完成计算后，将合约名称、NPV 和更新时间写入结果表
def depositHandler(name, date, npv){
    tableInsert(depositResult, "DepositEngine", name, npv, now())
}

// 构造 3 个月人民币定期存款。Deposit 的本金币种字段必须使用 notionalCurrency；
// 同时本金金额使用 notionalAmount，与利率互换示例保持一致。
depositIns = parseInstrument({
    "productType": "Cash",
    "assetType": "Deposit",
    "instrumentId": "DEP_3M",
    "start": 2026.08.27,
    "maturity": 2026.11.27,
    "rate": 0.018,
    "dayCountConvention": "Actual360",
    "notionalCurrency": "CNY",
    "notionalAmount": 1000000.0,
    "payReceive": "Receive",
    "discountCurve": "CNY_DISC_CURVE"
})

// 创建存款定价引擎，并绑定待估值合约与回调函数
engine = createPricingEngine("DepositEngine", depositIns, depositHandler)

// 构造贴现曲线市场数据；curveName 与存款合约中的 discountCurve 对应
discountCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": 2026.08.27,
    "currency": "CNY",
    "curveName": "CNY_DISC_CURVE",
    "dayCountConvention": "Actual360",
    "compounding": "Compounded",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",
    "dates": [2026.09.27, 2026.11.27, 2027.02.27, 2027.08.27],
    "values": [0.0155, 0.0165, 0.0170, 0.0180]
})

// 向引擎追加市场数据，触发估值计算并通过 depositHandler 写入结果表
appendMktData(engine, discountCurve)

// 查看 DEP_3M 的估值结果
select * from depositResult
engineName instrumentId npv updateTime
DepositEngine DEP_3M 1,000,407.2780936392 2026.08.27 13:40:39.896

例4. 本例以一笔 5 年期固定兑浮动利率互换为例，帮助用户理解如何用贴现曲线和浮动端参考曲线估算互换价值，并查看 NPV、两端 PV 和 DV01。

// 清理历史同名引擎，避免重复创建时报错
try{dropStreamEngine("SwapEngine")}catch(ex){}

// 创建结果表，用于接收并保存利率互换定价引擎回调返回的估值结果
share streamTable(
    1000:0,
    `engineName`instrumentId`npv`updateTime,
    [STRING, STRING, DOUBLE, TIMESTAMP]
) as swapResult

// 定义回调函数：每次引擎完成计算后，将合约名称、NPV 和更新时间写入结果表
def swapHandler(name, date, npv){
    tableInsert(swapResult, "SwapEngine", name, npv, now())
}

// 构造 5 年期固定兑浮动利率互换；
// FR_007 互换除了 discountCurve / forwardCurve
// 还需要 assetPriceCurve，用于浮动端历史定盘数据
irsIns = parseInstrument({
    "productType": "Swap",
    "swapType": "IrSwap",
    "irSwapType": "IrFixedFloatingSwap",
    "instrumentId": "IRS_FR007_5Y",

    "start": 2026.08.27,
    "maturity": 2031.08.27,
    "frequency": "Quarterly",

    "fixedRate": 0.0385,
    "calendar": "CFET",
    "fixedDayCountConvention": "Actual365",
    "floatingDayCountConvention": "Actual365",

    "payReceive": "Pay",
    "iborIndex": "FR_007",
    "spread": 0.0,

    "notionalCurrency": "CNY",
    "notionalAmount": 10000000.0,

    // 三条曲线名称要和后面市场数据里的 curveName 保持一致
    "discountCurve": "CNY_DISC_CURVE",
    "forwardCurve": "FR007_FWD_CURVE",
    "assetPriceCurve": "PRICE_FR_007"
})

// 创建利率互换定价引擎，并绑定待估值合约与回调函数
engine = createPricingEngine(
    "SwapEngine",
    irsIns,
    swapHandler
)

// 构造人民币贴现曲线；curveName 与互换合约中的 discountCurve 对应
discountCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": 2026.08.27,
    "currency": "CNY",
    "curveName": "CNY_DISC_CURVE",

    "dayCountConvention": "Actual365",
    "compounding": "Compounded",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",

    "dates": [
        2026.11.27,
        2027.08.27,
        2028.08.27,
        2029.08.27,
        2031.08.27
    ],
    "values": [
        0.0180,
        0.0185,
        0.0190,
        0.0200,
        0.0220
    ]
})

// 构造 FR007 远期曲线；curveName 与互换合约中的 forwardCurve 对应
forwardCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": 2026.08.27,
    "currency": "CNY",
    "curveName": "FR007_FWD_CURVE",

    "dayCountConvention": "Actual365",
    "compounding": "Compounded",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",

    "dates": [
        2026.11.27,
        2027.02.27,
        2027.08.27,
        2028.08.27,
        2031.08.27
    ],
    "values": [
        0.0195,
        0.0200,
        0.0210,
        0.0220,
        0.0240
    ]
})

// 构造 FR007 历史定盘数据曲线。这是 IRS 定价非常关键的一条市场数据
// curveType 必须是 AssetPriceCurve
// curveName 要和 instrument 中的 assetPriceCurve 对应
fixingDates = 2026.05.19..2026.08.26
fixingRates = take(0.0190, size(fixingDates))

assetPriceCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "AssetPriceCurve",
    "referenceDate": 2026.08.27,
    "currency": "CNY",
    "curveName": "PRICE_FR_007",

    // 历史 FR007 定盘利率
    "dates": fixingDates,
    "values": fixingRates
})

// 一次性追加三条市场数据，触发估值
appendMktData(engine, [discountCurve, forwardCurve, assetPriceCurve])


// 查看估值结果
select * from swapResult
engineName instrumentId npv updateTime
SwapEngine IRS_FR007_5Y -701,240.0307065442 2026.08.27 14:35:55.568

例5. 本例以 6 个月 USD/CNY 外汇远期为例，说明远期汇率如何受即期汇率和两种货币利率曲线影响，并输出远期理论汇率和合约 NPV。

// 清理历史同名引擎
try{dropStreamEngine("FxFwdEngine")}catch(ex){}

// 结果表
share streamTable(1000:0, `engineName`instrumentId`npv`updateTime, [STRING, STRING, DOUBLE, TIMESTAMP]) as fxFwdResult

// 回调函数
def fxFwdHandler(name, date, npv){
    tableInsert(fxFwdResult, "FxFwdEngine", name, npv, now())
}

// 构造外汇远期合约：使用 domesticCurve / foreignCurve
fxFwdIns = parseInstrument({
    "productType": "Forward",
    "forwardType": "FxForward",
    "instrumentId": "USDCNY_6M_FWD",
    "expiry": 2027.02.27,
    "delivery": 2027.03.01,
    "currencyPair": "USDCNY",
    "direction": "Buy",
    "notionalCurrency": "USD",
    "notionalAmount": 1000000.0,
    "strike": 7.10,
    "domesticCurve": "CNY_DISC_CURVE",
    "foreignCurve": "USD_DISC_CURVE"
})

// 创建引擎
engine = createPricingEngine("FxFwdEngine", fxFwdIns, fxFwdHandler)

// FX 即期汇率市场数据
spot = parseMktData({
    "mktDataType": "Price",
    "priceType": "FxSpotRate",
    "referenceDate": 2026.08.27,
    "spotDate": 2026.08.29,
    "value": 7.1200,
    "unit": "USDCNY"
})

// 人民币贴现曲线
cnyCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": 2026.08.27,
    "currency": "CNY",
    "curveName": "CNY_DISC_CURVE",
    "dayCountConvention": "Actual365",
    "compounding": "Compounded",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",
    "dates": [2026.09.27, 2026.11.27, 2027.02.27],
    "values": [0.0155, 0.0165, 0.0170]
})

// 美元贴现曲线
usdCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": 2026.08.27,
    "currency": "USD",
    "curveName": "USD_DISC_CURVE",
    "dayCountConvention": "Actual365",
    "compounding": "Compounded",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",
    "dates": [2026.09.27, 2026.11.27, 2027.02.27],
    "values": [0.0440, 0.0430, 0.0415]
})

// 触发定价
appendMktData(engine, spot)
appendMktData(engine, cnyCurve)
appendMktData(engine, usdCurve)

// 查看结果
select * from fxFwdResult
engineName instrumentId npv updateTime
FxFwdEngine USDCNY_6M_FWD -65,285.58303391563 2026.08.27 14:16:54.106

例6. 本例以一笔 USD/CNY 外汇掉期为例，展示如何评估近端和远端两笔换汇现金流的价值，并得到掉期点、分段 PV 和整体 NPV。

// 清理历史同名引擎
try{dropStreamEngine("FxSwapEngine")}catch(ex){}

// 结果表
share streamTable(1000:0, `engineName`instrumentId`npv`updateTime, [STRING, STRING, DOUBLE, TIMESTAMP]) as fxSwapResult

// 回调函数
def fxSwapHandler(name, date, npv){
    tableInsert(fxSwapResult, "FxSwapEngine", name, npv, now())
}

// 构造 USD/CNY 外汇掉期；同时使用 domesticCurve / foreignCurve 指定曲线名
fxSwapIns = parseInstrument({
    "productType": "Swap",
    "swapType": "FxSwap",
    "instrumentId": "USDCNY_SWAP",
    "currencyPair": "USDCNY",
    "direction": "Buy",
    "notionalCurrency": "USD",
    "notionalAmount": 1000000.0,
    "nearStrike": 7.12,
    "nearExpiry": 2026.09.27,
    "nearDelivery": 2026.09.29,
    "farStrike": 7.15,
    "farExpiry": 2027.03.27,
    "farDelivery": 2027.03.29,
    "domesticCurve": "CNY_DISC_CURVE",
    "foreignCurve": "USD_DISC_CURVE"
})


// 创建引擎
engine = createPricingEngine("FxSwapEngine", fxSwapIns, fxSwapHandler)

// 即期汇率
spot = parseMktData({
    "mktDataType": "Price",
    "priceType": "FxSpotRate",
    "referenceDate": 2026.08.27,
    "spotDate": 2026.08.29,
    "value": 7.1200,
    "unit": "USDCNY"
})

// 人民币贴现曲线
cnyCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": 2026.08.27,
    "currency": "CNY",
    "curveName": "CNY_DISC_CURVE",
    "dayCountConvention": "Actual365",
    "compounding": "Compounded",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",
    "dates": [2026.09.27, 2027.03.27],
    "values": [0.0155, 0.0170]
})

// 美元贴现曲线
usdCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": 2026.08.27,
    "currency": "USD",
    "curveName": "USD_DISC_CURVE",
    "dayCountConvention": "Actual365",
    "compounding": "Compounded",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",
    "dates": [2026.09.27, 2027.03.27],
    "values": [0.0440, 0.0415]
})

// 触发定价
appendMktData(engine, spot)
appendMktData(engine, cnyCurve)
appendMktData(engine, usdCurve)

// 查看结果
select * from fxSwapResult
engineName instrumentId npv updateTime
FxSwapEngine USDCNY_SWAP 109,673.96002695739 2026.08.27 14:20:02.733

例7. 本例以 3 个月 USD/CNY 欧式看涨期权为例，帮助用户了解期权定价中即期汇率、利率曲线和隐含波动率的作用，并查看理论价值及 Delta、Gamma、Vega 等希腊值。

// 清理历史同名引擎，避免重复创建时报错
try{dropStreamEngine("FxOptEngine")}catch(ex){}

// 创建结果表
share streamTable(
    1000:0,
    `engineName`instrumentId`npv`updateTime,
    [STRING, STRING, DOUBLE, TIMESTAMP]
) as fxOptResult

// 定义回调函数
// createPricingEngine 默认 handler 签名为 def(name, date, npv)
def fxOptHandler(name, date, npv){
    tableInsert(fxOptResult, "FxOptEngine", name, npv, now())
}

// 构造 3 个月 USD/CNY 欧式看涨期权；
// 1）underlying 用货币对
// 2）domesticCurve / foreignCurve 指定曲线名
fxOptIns = parseInstrument({
    "productType": "Option",
    "optionType": "EuropeanOption",
    "assetType": "FxEuropeanOption",
    "instrumentId": "USDCNY_CALL_7.2",
    "notionalCurrency": "USD",
    "notionalAmount": 1000000.0,
    "strike": 7.20,
    "maturity": 2026.11.27,
    "delivery": 2026.11.29,
    "payoffType": "Call",
    "direction": "Buy",
    "dayCountConvention": "Actual365",
    "underlying": "USDCNY",
    // 自定义两条曲线名称
    // 后续 IrYieldCurve 的 curveName 必须与这里一致
    "domesticCurve": "CNY_DISC_CURVE",
    "foreignCurve": "USD_DISC_CURVE"
})

// 创建定价引擎
engine = createPricingEngine("FxOptEngine", fxOptIns, fxOptHandler)

// 构造 USD/CNY 即期汇率。
// 对定价引擎来说，FX spot 应使用 Price / FxSpotRate
// unit 使用货币对名称，和 underlying 对应
spot = parseMktData({
    "mktDataType": "Price",
    "priceType": "FxSpotRate",
    "referenceDate": 2026.08.27,
    "spotDate": 2026.08.29,
    "value": 7.1200,
    "unit": "USDCNY"
})

// 构造人民币贴现曲线
// curveName 与 instrument 中 domesticCurve 对应
cnyCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": 2026.08.27,
    "currency": "CNY",
    "curveName": "CNY_DISC_CURVE",

    "dayCountConvention": "Actual365",
    "compounding": "Continuous",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",

    "dates": [
        2026.09.27,
        2026.11.27,
        2027.02.27
    ],
    "values": [
        0.0155,
        0.0165,
        0.0170
    ]
})

// 构造美元贴现曲线
// curveName 与 instrument 中 foreignCurve 对应
usdCurve = parseMktData({
    "mktDataType": "Curve",
    "curveType": "IrYieldCurve",
    "referenceDate": 2026.08.27,
    "currency": "USD",
    "curveName": "USD_DISC_CURVE",

    "dayCountConvention": "Actual365",
    "compounding": "Continuous",
    "interpMethod": "Linear",
    "extrapMethod": "Flat",
    "frequency": "Annual",

    "dates": [
        2026.09.27,
        2026.11.27,
        2027.02.27
    ],
    "values": [
        0.0440,
        0.0430,
        0.0415
    ]
})

//  构造 USD/CNY 外汇波动率曲面
volSurf = parseMktData({
    "mktDataType": "Surface",
    "surfaceType": "FxVolatilitySurface",
    "referenceDate": 2026.08.27,
    "surfaceName": "USDCNY",
    "currencyPair": "USDCNY",

    "smileMethod": "Linear",

    // termDates 和 volSmiles 一一对应
    "termDates": [
        2026.11.27,
        2027.02.27
    ],

    "volSmiles": [
        {
            "strikes": [6.8, 7.2, 7.6],
            "vols": [0.0550, 0.0550, 0.0550]
        },
        {
            "strikes": [6.8, 7.2, 7.6],
            "vols": [0.0560, 0.0560, 0.0560]
        }
    ]
})

// 一次性追加全部市场数据，触发估值
appendMktData(engine, [spot, cnyCurve, usdCurve, volSurf])

// 查看估值结果
select * from fxOptResult
engineName instrumentId npv updateTime
FxOptEngine USDCNY_CALL_7.2 4,306.487653692539 2026.08.27 14:40:36.151

相关函数：appendMktData、createMktDataEngine