一键在 Manus 中运行任何 Skill

powerlanguage-python-conversion

星标8

分支4

更新时间2026年6月12日 01:52

Use when converting or porting code between MultiCharts PowerLanguage and Python in either direction — Strategy ABC scaffold, concept mapping tables, semantic differences, pandas-ta/TA-Lib indicators, conversion checklists.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

KasperChenGH

KasperChenGH/multicharts-powerlanguage

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

PowerLanguage ↔ Python Conversion

How to use

This skill covers the structural and semantic differences between MultiCharts PowerLanguage and Python for algorithmic trading. It does not duplicate the syntax rules of PowerLanguage — for declarations, bar references, control flow, and built-in function signatures use the powerlanguage-syntax skill. When performing a conversion in either direction, work through Part 0 (structural template) to set up the boilerplate, then Part 1 (concept mapping) for line-by-line translation, then review Part 2 (semantic differences and gotchas), and finally run through the relevant checklist in Part 3 before calling the output complete.

Part 0: Structural Template

PowerLanguage runs the entire script top-to-bottom on each bar close automatically. Python has no implicit bar loop, no built-in OHLCV type, and no order submission mechanism. Every PL-to-Python conversion targets the framework-agnostic scaffold below.

Bar dataclass

from dataclasses import dataclass

@dataclass
class Bar:
    open: float
    high: float
    low: float
    close: float
    volume: float
    time: int          # Unix epoch seconds
    bar_number: int    # 1-based to match PL convention

Order types

from enum import Enum, auto
from dataclasses import dataclass

class Side(Enum):
    LONG = auto()
    SHORT = auto()

class OrderType(Enum):
    MARKET = auto()
    LIMIT = auto()
    STOP = auto()

class Action(Enum):
    ENTRY = auto()
    EXIT = auto()
    STOP_LOSS = auto()
    PROFIT_TARGET = auto()
    BREAK_EVEN = auto()

@dataclass
class Order:
    label: str
    action: Action
    side: Side
    order_type: OrderType
    price: float = 0.0   # 0.0 for Market
    qty: float = 1.0

Strategy ABC

from abc import ABC, abstractmethod

class Strategy(ABC):
    @abstractmethod
    def on_bar(self, bars: list[Bar], orders: list[Order]) -> None:
        ...

Strategy subclass skeleton

import pandas as pd
import pandas_ta as ta

class MyStrategy(Strategy):
    def __init__(self, length: int = 20, threshold: float = 0.5):
        # Inputs (PL: Inputs: Length(20), Threshold(0.5))
        self.length = length
        self.threshold = threshold

        # Variables (PL: Variables: my_var(0))
        self.my_var = 0.0

        # Position state (PL: MarketPosition, EntryPrice, etc.)
        self.position = 0           # -1, 0, +1
        self.entry_price = 0.0
        self.bars_since_entry = 0
        self.current_contracts = 0.0

    def on_bar(self, bars: list[Bar], orders: list[Order]) -> None:
        if len(bars) < self.length:
            return

        closes = pd.Series([b.close for b in bars])

        # pandas-ta: compute indicator on the full series
        sma = ta.sma(closes, length=self.length)
        current_sma = sma.iloc[-1]

        # ... strategy logic, push to orders list ...

Alternative: TA-Lib indicator pattern

import talib
import numpy as np

class MyStrategy(Strategy):
    def on_bar(self, bars: list[Bar], orders: list[Order]) -> None:
        closes = np.array([b.close for b in bars])

        # TA-Lib: batch array API (NumPy in, NumPy out)
        sma = talib.SMA(closes, timeperiod=self.length)
        current_sma = sma[-1]

pandas-ta is recommended as the primary library (pure Python, zero install friction, pip-installable). TA-Lib is the performance alternative (C-based, requires compiling the C library). The concept mapping tables in Part 1 show both.

Main loop

def main():
    bars: list[Bar] = load_bars("data.csv")  # user-provided
    strategy = MyStrategy(length=20, threshold=0.5)

    for i in range(len(bars)):
        orders: list[Order] = []
        strategy.on_bar(bars[: i + 1], orders)
        # execute orders against simulated book...

The slice bars[:i+1] gives the strategy access to full history; bars[-1] is the current bar. This mirrors PL's implicit bar-by-bar execution.

Part 1: Concept Mapping

Table 1 — Declarations

PowerLanguage	Python	Notes
`Inputs: Length(20)`	`__init__` parameter with default: `def __init__(self, length: int = 20)`	Stored as `self.length`; treat as read-only after construction
`Variables: myVar(0)`	`self.my_var = 0.0` in `__init__`	Mutable instance attributes
`Value1` .. `Value99`	Named attribute (e.g., `self.rsi_val`)	Must rename to meaningful identifiers; Python has no pre-declared variables
`Condition1` .. `Condition99`	Named attribute (e.g., `self.is_oversold`)	Same: rename to typed `bool` attributes
`Arrays: buf[10](0)`	`self.buf = [0.0] * 11`	PL `[10]` means last index is 10 (11 elements); use a Python `list`
`IntraBarPersist myVar(0)`	`self.my_var = 0.0` (same as Variables)	Python has no bar-close vs intra-bar distinction; all instance attributes persist

Table 2 — Data Access

PowerLanguage	Python	Notes
`Open`, `High`, `Low`, `Close`, `Volume`	`bars[-1].open`, `.high`, `.low`, `.close`, `.volume`	`bars[-1]` for current bar
`Close[1]`	`bars[-2].close`	Lookback N: `bars[-1 - n].close`. Guard with `len(bars) > n`.
`Close of Data2`	separate `bars2: list[Bar]` parameter	PL Data2 is a second chart feed; pass a second list to `on_bar()`
`Date`	`datetime.fromtimestamp(bars[-1].time, tz=timezone.utc)`	PL `Date` is YYYMMDD integer; Python uses `datetime` module
`Time`	`dt.hour * 100 + dt.minute` (from datetime object)	PL `Time` is HHMM integer
`BarNumber` / `CurrentBar`	`bars[-1].bar_number` or `len(bars)`	PL is 1-based

Table 3 — Technical Indicators (dual-library)

PowerLanguage	pandas-ta (primary)	TA-Lib (alternative)	Notes
`Average(Close, N)`	`ta.sma(closes, length=N)`	`talib.SMA(closes, timeperiod=N)`	Current value: `.iloc[-1]` / `[-1]`
`XAverage(Close, N)`	`ta.ema(closes, length=N)`	`talib.EMA(closes, timeperiod=N)`
`RSI(Close, N)`	`ta.rsi(closes, length=N)`	`talib.RSI(closes, timeperiod=N)`	Returns 0–100
`BollingerBand(Close, N, 2)`	`ta.bbands(closes, length=N, std=2)`	`talib.BBANDS(closes, timeperiod=N, nbdevup=2, nbdevdn=2)`	Returns upper/mid/lower
`MACD(Close, 12, 26)`	`ta.macd(closes, fast=12, slow=26, signal=9)`	`talib.MACD(closes, fastperiod=12, slowperiod=26, signalperiod=9)`	Returns MACD/signal/histogram
`Stochastic(...)`	`ta.stoch(highs, lows, closes, k=14, d=3)`	`talib.STOCH(highs, lows, closes, fastk_period=14, slowk_period=3, slowd_period=3)`	Returns slowK/slowD
`ADX(N)`	`ta.adx(highs, lows, closes, length=N)`	`talib.ADX(highs, lows, closes, timeperiod=N)`
`CCI(N)`	`ta.cci(highs, lows, closes, length=N)`	`talib.CCI(highs, lows, closes, timeperiod=N)`
`AvgTrueRange(N)`	`ta.atr(highs, lows, closes, length=N, mamode='sma')` or rolling mean of `ta.true_range(...)`	`talib.SMA(talib.TRANGE(highs, lows, closes), timeperiod=N)`	PL `AvgTrueRange` is a SIMPLE average of TrueRange; pandas-ta/TA-Lib ATR default to Wilder smoothing — the default Wilder ATR matches PL `SmoothedAverage(TrueRange, Len)` instead
`Momentum(Close, N)`	`ta.mom(closes, length=N)`	`talib.MOM(closes, timeperiod=N)`
`Highest(High, N)`	`highs.rolling(N).max()`	`talib.MAX(highs, timeperiod=N)`
`Lowest(Low, N)`	`lows.rolling(N).min()`	`talib.MIN(lows, timeperiod=N)`
`DMIPlus(N)` / `DMIMinus(N)`	`ta.dm(highs, lows, length=N)`	`talib.PLUS_DI(...)` / `talib.MINUS_DI(...)`
`KeltnerChannel(Close, N, mult)`	`ta.kc(highs, lows, closes, length=N, scalar=mult)`	Manual: `talib.EMA(closes, N) ± mult * talib.ATR(highs, lows, closes, N)`
`Parabolic(step)`	`ta.psar(highs, lows, af0=step, max_af=0.2)`	`talib.SAR(highs, lows, acceleration=step, maximum=0.2)`
`PercentR(N)`	`ta.willr(highs, lows, closes, length=N) + 100`	`talib.WILLR(highs, lows, closes, timeperiod=N) + 100`	PL `PercentR` is POSITIVE 0..100 (= Williams %R + 100); library WILLR returns −100..0 — add 100, and thresholds become 80/20 on the positive scale
`RateOfChange(Close, N)`	`ta.roc(closes, length=N)`	`talib.ROC(closes, timeperiod=N)`
`Volatility(N)`	`ta.true_range(highs, lows, closes).ewm(span=N).mean()` (or `ta.atr(..., mamode='ema')`)	`talib.EMA(talib.TRANGE(highs, lows, closes), timeperiod=N)`	PL `Volatility` is a smoothed average of TrueRange weighted toward the most recent bar — NOT a stdev-based measure (`StandardDev` is stdev of closes; `VolatilityStdDev` is annualized stdev of log returns)
`StandardDev(Close, N, 1)`	`ta.stdev(closes, length=N)`	`talib.STDDEV(closes, timeperiod=N, nbdev=1)`
`TrueRange`	`ta.true_range(highs, lows, closes)`	`talib.TRANGE(highs, lows, closes)`	Per-bar, no period
`MoneyFlow(N)`	`ta.mfi(highs, lows, closes, volumes, length=N)`	`talib.MFI(highs, lows, closes, volumes, timeperiod=N)`	Requires HLCV
`AccumDist`	`ta.ad(highs, lows, closes, volumes)`	`talib.AD(highs, lows, closes, volumes)`	Cumulative, no period
`LinearRegValue(Close, N, 0)`	`ta.linreg(closes, length=N)`	`talib.LINEARREG(closes, timeperiod=N)`
`LinearRegSlope(Close, N)`	`ta.linreg(closes, length=N, slope=True)`	`talib.LINEARREG_SLOPE(closes, timeperiod=N)`
`Summation(Close, N)`	`closes.rolling(N).sum()`	`talib.SUM(closes, timeperiod=N)`
`Cum(Volume)`	`volumes.cumsum()`	`np.cumsum(volumes)`	Running total
`WAverage(Close, N)`	`ta.wma(closes, length=N)`	`talib.WMA(closes, timeperiod=N)`
`MidPoint(Close, N)`	`ta.midpoint(closes, length=N)`	`talib.MIDPOINT(closes, timeperiod=N)`
`TSI(Close, LongLen, ShortLen)`	`ta.tsi(closes, fast=ShortLen, slow=LongLen)`	Manual: double-EMA of momentum / double-EMA of abs(momentum)	pandas-ta `fast`=short, `slow`=long (reversed naming vs PL)
`SwingHigh(Occur, Price, Strength, Length)`	Manual: scan for pivot high	Manual: scan for pivot high	Occurrence is the FIRST PL arg; returns −1 when no swing found; `Length` must exceed `Strength`
`SwingLow(Occur, Price, Strength, Length)`	Manual: scan for pivot low	Manual: scan for pivot low	Same: Occurrence first, −1 when none found, `Length` > `Strength`
`HighestBar(High, N)`	`highs[-N:].idxmax()` → compute bars ago	`np.argmax(highs[-N:])`	Returns bars ago
`LowestBar(Low, N)`	`lows[-N:].idxmin()` → compute bars ago	`np.argmin(lows[-N:])`	Returns bars ago
`CountIF(cond, N)`	`cond_series.rolling(N).sum()`	Manual: `sum(1 for ...)`	Count True values
`Crosses Over`	`prev_a <= prev_b and a > b`	Same	No built-in; store previous values
`Crosses Under`	`prev_a >= prev_b and a < b`	Same	Same
`AverageFC(Close, N)`	`ta.sma(closes, length=N)`	`talib.SMA(closes, timeperiod=N)`	FC = "fast calculation": `AverageFC = SummationFC/Len` (running-sum SMA), numerically identical to `Average` — use SMA
`AdaptiveMovAvg(Close, N)`	`ta.kama(closes, length=N)`	`talib.KAMA(closes, timeperiod=N)`	Kaufman Adaptive Moving Average
`UltimateOscillator(7,14,28)`	`ta.uo(highs, lows, closes, fast=7, medium=14, slow=28)`	`talib.ULTOSC(highs, lows, closes, timeperiod1=7, timeperiod2=14, timeperiod3=28)`	Requires HLC
`ChaikinOsc(3, 10)`	`ta.adosc(highs, lows, closes, volumes, fast=3, slow=10)`	`talib.ADOSC(highs, lows, closes, volumes, fastperiod=3, slowperiod=10)`	Requires HLCV
`PriceOscillator(Fast, Slow)`	`ta.apo(closes, fast=Fast, slow=Slow)`	`talib.APO(closes, fastperiod=Fast, slowperiod=Slow)`	Absolute Price Oscillator
`DirMovement(N, ...)`	`ta.dm(highs, lows, length=N)` + `ta.adx(...)`	`talib.PLUS_DI(...)` + `talib.MINUS_DI(...)` + `talib.ADX(...)`	Multi-output: +DI, −DI, ADX
`Extremes(High, Low, N, oHH, oLL, oHHBar, oLLBar)`	Manual: rolling max/min + argmax/argmin	Manual: rolling max/min + argmax/argmin	Returns highest, lowest, and their bar offsets
`TrueHigh`	`np.maximum(highs, closes.shift(1))`	Same	`max(High, Close[1])`
`TrueLow`	`np.minimum(lows, closes.shift(1))`	Same	`min(Low, Close[1])`
`Range`	`highs - lows`	Same	Per-bar range; no period
`NthHighest(N, Close, Len)`	Manual: `closes.rolling(Len).apply(lambda x: np.sort(x)[-N])`	Manual: sort rolling window	Nth largest value in window
`NthLowest(N, Close, Len)`	Manual: `closes.rolling(Len).apply(lambda x: np.sort(x)[N-1])`	Manual: sort rolling window	Nth smallest value in window
`NthHighestBar(N, Close, Len)`	Manual: find bar offset of Nth highest	Manual: find bar offset of Nth highest	Returns bars ago of Nth highest
`NthLowestBar(N, Close, Len)`	Manual: find bar offset of Nth lowest	Manual: find bar offset of Nth lowest	Returns bars ago of Nth lowest
`SwingHighBar(Occur, Price, Strength, Length)`	Manual: pivot detection with bar offset	Manual: pivot detection with bar offset	Bars since Nth swing high; same arg order as `SwingHigh`
`SwingLowBar(Occur, Price, Strength, Length)`	Manual: pivot detection with bar offset	Manual: pivot detection with bar offset	Bars since Nth swing low
`LinearRegAngle(Close, N)`	`np.degrees(np.arctan(ta.linreg(closes, length=N, slope=True)))`	`talib.LINEARREG_ANGLE(closes, timeperiod=N)`	Slope converted to degrees
`Correlation(Close, Volume, N)`	`ta.correlation(closes, volumes, length=N)`	`talib.CORREL(closes, volumes, timeperiod=N)`	Pearson correlation
`RSquared(Close, N)`	`ta.correlation(closes, ..., length=N) ** 2`	`talib.CORREL(closes, ..., timeperiod=N) ** 2`	Square of correlation
`StdError(Close, N)`	Manual: fit `np.polyfit(np.arange(N), closes[-N:], 1)`, then `np.sqrt(((y - yhat) ** 2).sum() / (N - 2))`	Manual: same residual computation on the window	Standard error of the linear-regression residuals — NOT stdev/sqrt(N) (that is SE of the mean)
`Median(Close, N)`	`ta.median(closes, length=N)`	`closes.rolling(N).median()`	Rolling median
`ELDate(dt)`	Manual: `(y - 1900) * 10000 + m * 100 + d`	Same	EasyLanguage YYYMMDD date format
`MinutesToTime(mins)`	Manual: `(mins // 60) * 100 + mins % 60`	Same	Minutes since midnight to HHMM
`TimeToMinutes(hhmm)`	Manual: `(hhmm // 100) * 60 + hhmm % 100`	Same	HHMM to minutes since midnight
`AvgPrice`	`(opens + highs + lows + closes) / 4`	`talib.AVGPRICE(opens, highs, lows, closes)`	Average of OHLC
`MedianPrice`	`(highs + lows) / 2`	`talib.MEDPRICE(highs, lows)`	Median of HL
`TypicalPrice`	`(highs + lows + closes) / 3`	`talib.TYPPRICE(highs, lows, closes)`	Typical price HLC
`WeightedClose`	`(highs + lows + 2 * closes) / 4`	`talib.WCLPRICE(highs, lows, closes)`	Weighted close HLCC
`MRO(cond, N, 1)`	Manual: find Nth True in `cond_series[-N:]`	Manual: iterate lookback	Most Recent Occurrence; returns bars ago
`IFF(cond, trueVal, falseVal)`	`trueVal if cond else falseVal`	Same	Python ternary; or `np.where(cond, trueVal, falseVal)` for Series
`TriAverage(Close, N)`	`ta.trima(closes, length=N)`	`talib.TRIMA(closes, timeperiod=N)`	Triangular MA
`FastK(N)`	`ta.stoch(highs, lows, closes, k=N, d=1, smooth_k=1)['STOCHk_N_1_1']`	`talib.STOCHF(highs, lows, closes, fastk_period=N)`	Raw Fast %K
`FastD(N)`	`ta.stoch(highs, lows, closes, k=N, d=3, smooth_k=1)['STOCHd_N_3_1']`	`talib.STOCHF(highs, lows, closes, fastk_period=N, fastd_period=3)`	Smoothed Fast %D — set `smooth_k=1` so %D is the SMA(3) of RAW %K, not of slowed %K
`SlowK(N)`	`ta.stoch(highs, lows, closes, k=N, d=3)['STOCHk_N_3_3']`	`talib.STOCH(highs, lows, closes, fastk_period=N, slowk_period=3)`	Slow %K
`SlowD(N)`	`ta.stoch(highs, lows, closes, k=N, d=3)['STOCHd_N_3_3']`	`talib.STOCH(highs, lows, closes, fastk_period=N, slowk_period=3, slowd_period=3)`	Slow %D
`FastKCustom(H, L, C, N)`	`ta.stoch(H, L, C, k=N, d=1, smooth_k=1)['STOCHk_...']`	`talib.STOCHF(H, L, C, fastk_period=N)`	Custom prices
`FastDCustom(H, L, C, N)`	`ta.stoch(H, L, C, k=N, d=3)['STOCHd_...']`	`talib.STOCHF(H, L, C, fastk_period=N, fastd_period=3)`	Custom prices
`SlowKCustom(H, L, C, N)`	`ta.stoch(H, L, C, k=N, d=3)['STOCHk_...']`	`talib.STOCH(H, L, C, fastk_period=N, slowk_period=3)`	Custom prices
`SlowDCustom(H, L, C, N)`	`ta.stoch(H, L, C, k=N, d=3)["STOCHd_N_3_3"]`	`talib.STOCH(H, L, C, fastk_period=N)[2]`	Slow %D with custom prices
`StochasticExp(H, L, C, N, S1, S2, ...)`	Manual: compute FastK then apply EMA smoothing	Manual: compute with `talib.STOCHF` then `talib.EMA`	Exponential smoothing variant
`ADXR(N)`	`ta.adx(highs, lows, closes, length=N)` then `(adx + adx.shift(N)) / 2`	`talib.ADXR(highs, lows, closes, timeperiod=N)`	TA-Lib has direct ADXR
`ADXCustom(H, L, C, N)`	`ta.adx(H, L, C, length=N)`	`talib.ADX(H, L, C, timeperiod=N)`	Pass custom price Series
`DMI(N)`	`ta.adx(highs, lows, closes, length=N)`	`talib.ADX(highs, lows, closes, timeperiod=N)`	Wrapper; same as ADX
`DMIPlusCustom(H, L, C, N)`	`ta.dm(H, L, length=N)['DMP_N']`	`talib.PLUS_DI(H, L, C, timeperiod=N)`	+DI with custom prices
`DMIMinusCustom(H, L, C, N)`	`ta.dm(H, L, length=N)['DMN_N']`	`talib.MINUS_DI(H, L, C, timeperiod=N)`	−DI with custom prices
`ParabolicCustom(step, limit)`	`ta.psar(highs, lows, af0=step, max_af=limit)`	`talib.SAR(highs, lows, acceleration=step, maximum=limit)`	Parabolic SAR with custom limit
`TRIX(Close, N)`	`ta.trix(closes, length=N)`	`talib.TRIX(closes, timeperiod=N)`	Triple EMA ROC
`MassIndex(SmoothLen, SumLen)`	`ta.massi(highs, lows, fast=SmoothLen, slow=SumLen)`	Manual: EMA ratio sum	Mass Index
`EaseOfMovement`	`ta.eom(highs, lows, closes, volumes)`	Manual: distance moved / box ratio	Requires HLCV
`SwingIndex`	Manual: Wilder swing index formula	Manual: same	No library built-in
`AccumSwingIndex`	Manual: cumulative sum of SwingIndex	Manual: same	No library built-in
`Detrend(Close, N)`	Manual: `closes - closes.rolling(N).mean().shift(N // 2 + 1)`	Manual: offset SMA	Detrended price
`PercentChange(Close, N)`	`closes.pct_change(N) * 100`	`talib.ROC(closes, timeperiod=N)`	Percent change
`UlcerIndex(Close, N)`	`ta.ui(closes, length=N)`	Manual: RMS of drawdown pct	Downside volatility
`ParabolicSAR(step, limit, ...)`	`ta.psar(highs, lows, af0=step, max_af=limit)`	`talib.SAR(highs, lows, acceleration=step, maximum=limit)`	Multi-output; extract position from PSARl/PSARs columns
`LinearReg(Close, N, TgtBar, ...)`	`ta.linreg(closes, length=N)` + manual slope/angle	`talib.LINEARREG(closes, timeperiod=N)` + `talib.LINEARREG_SLOPE/ANGLE/INTERCEPT`	Multi-output; combine four TA-Lib calls
`TrueRangeCustom(H, L, C)`	`ta.true_range(H, L, C)`	`talib.TRANGE(H, L, C)`	Custom prices
`VolatilityStdDev(NumDays)`	`np.log(closes / closes.shift(1)).rolling(NumDays).std() * np.sqrt(252)`	Manual: annualized stdev of log returns	Historical volatility
`StandardDevAnnual(Close, N, DataType)`	`ta.stdev(closes, length=N) * np.sqrt(252)`	`talib.STDDEV(closes, timeperiod=N) * np.sqrt(252)`	Annualized stdev
`HighestFC(Close, N)`	`closes.rolling(N).max()`	`talib.MAX(closes, timeperiod=N)`	Same as Highest
`LowestFC(Close, N)`	`closes.rolling(N).min()`	`talib.MIN(closes, timeperiod=N)`	Same as Lowest
`PivotHighVS(Inst, Price, LStr, RStr, Len)`	Manual: scan for pivot high with L/R strength	Manual: same	Asymmetric left/right strength
`PivotLowVS(Inst, Price, LStr, RStr, Len)`	Manual: scan for pivot low with L/R strength	Manual: same	Same
`PivotHighVSBar(Inst, Price, LStr, RStr, Len)`	Manual: bars ago of pivot high	Manual: same	Returns offset
`PivotLowVSBar(Inst, Price, LStr, RStr, Len)`	Manual: bars ago of pivot low	Manual: same	Returns offset
`Divergence(P1, P2, Str, Len, HiLo)`	Manual: compare pivot highs/lows of two series	Manual: same	No library built-in
`TimeSeriesForecast(Close, N)`	`ta.tsf(closes, length=N)`	`talib.TSF(closes, timeperiod=N)`	Time Series Forecast
`SummationFC(Close, N)`	`closes.rolling(N).sum()`	`talib.SUM(closes, timeperiod=N)`	Same as Summation
`OpenD(N)`	`df.resample('D').first()['open'].iloc[-1-N]`	Manual: resample to daily	Daily open
`HighD(N)`	`df.resample('D').max()['high'].iloc[-1-N]`	Manual: resample	Daily high
`LowD(N)`	`df.resample('D').min()['low'].iloc[-1-N]`	Manual: resample	Daily low
`CloseD(N)`	`df.resample('D').last()['close'].iloc[-1-N]`	Manual: resample	Daily close
`OpenW(N)`	`df.resample('W').first()['open'].iloc[-1-N]`	Manual: resample	Weekly open
`HighW(N)`	`df.resample('W').max()['high'].iloc[-1-N]`	Manual: resample	Weekly high
`LowW(N)`	`df.resample('W').min()['low'].iloc[-1-N]`	Manual: resample	Weekly low
`CloseW(N)`	`df.resample('W').last()['close'].iloc[-1-N]`	Manual: resample	Weekly close
`OpenM(N)`	`df.resample('ME').first()['open'].iloc[-1-N]`	Manual: resample	Monthly open
`HighM(N)`	`df.resample('ME').max()['high'].iloc[-1-N]`	Manual: resample	Monthly high
`LowM(N)`	`df.resample('ME').min()['low'].iloc[-1-N]`	Manual: resample	Monthly low
`CloseM(N)`	`df.resample('ME').last()['close'].iloc[-1-N]`	Manual: resample	Monthly close
`OpenY(N)`	`df.resample('YE').first()['open'].iloc[-1-N]`	Manual: resample	Yearly open
`HighY(N)`	`df.resample('YE').max()['high'].iloc[-1-N]`	Manual: resample	Yearly high
`LowY(N)`	`df.resample('YE').min()['low'].iloc[-1-N]`	Manual: resample	Yearly low
`CloseY(N)`	`df.resample('YE').last()['close'].iloc[-1-N]`	Manual: resample	Yearly close
`LRO(cond, N, Inst)`	Manual: `cond_series[-N:].iloc[::-1]` find Nth True from end	Manual: same	Least Recent Occurrence
`SummationIf(cond, Price, N)`	`(Price * cond).rolling(N).sum()`	Manual: multiply then sum	Conditional rolling sum
`IFFString(cond, trueStr, falseStr)`	`trueStr if cond else falseStr`	Same	String ternary; or `np.where`
`OBV`	`ta.obv(closes, volumes)`	`talib.OBV(closes, volumes)`	On Balance Volume
`VolumeROC(N)`	`ta.roc(volumes, length=N)`	`talib.ROC(volumes, timeperiod=N)`	Volume rate of change
`VolumeOsc(ShortLen, LongLen)`	`ta.sma(volumes, length=ShortLen) - ta.sma(volumes, length=LongLen)`	`talib.SMA(volumes, ShortLen) - talib.SMA(volumes, LongLen)`	Volume oscillator
`PriceVolTrend`	`ta.pvt(closes, volumes)`	Manual: cumulative `pct_change * volume`	Price Volume Trend
`LWAccDis`	Manual: `((closes - opens) / (highs - lows) * volumes).cumsum()`	Manual: same	Larry Williams A/D
`Fisher(Price)`	Manual: normalize, then `0.5 * np.log((1 + norm) / (1 - norm))`	Manual: same	Fisher transformation
`FisherINV(Price)`	Manual: `(np.exp(2 * Price) - 1) / (np.exp(2 * Price) + 1)`	Manual: same	Inverse Fisher
`C_Doji(Pct)`	Manual: `abs(close - open) <= (high - low) * Pct / 100`	`talib.CDLDOJI(opens, highs, lows, closes)`	TA-Lib has CDL* family
`C_Hammer_HangingMan(Len, Factor, ...)`	Manual: body/shadow ratios	`talib.CDLHAMMER(...)` / `talib.CDLHANGINGMAN(...)`	Separate TA-Lib functions
`C_BullEng_BearEng(Len, ...)`	Manual: engulfing detection	`talib.CDLENGULFING(opens, highs, lows, closes)`	+100=bullish, -100=bearish
`C_BullHar_BearHar(Len, ...)`	Manual: harami detection	`talib.CDLHARAMI(opens, highs, lows, closes)`	+100=bullish, -100=bearish
`C_MornDoji_EveDoji(Len, Pct, ...)`	Manual: 3-bar doji star	`talib.CDLMORNINGDOJISTAR(...)` / `talib.CDLEVENINGDOJISTAR(...)`	Separate functions
`C_MornStar_EveStar(Len, ...)`	Manual: 3-bar star	`talib.CDLMORNINGSTAR(...)` / `talib.CDLEVENINGSTAR(...)`	Separate functions
`C_PierceLine_DkCloud(Len, ...)`	Manual: piercing/cloud	`talib.CDLPIERCING(...)` / `talib.CDLDARKCLOUDCOVER(...)`	Separate functions
`C_ShootingStar(Len, Factor)`	Manual: shooting star	`talib.CDLSHOOTINGSTAR(opens, highs, lows, closes)`	TA-Lib direct
`C_3WhSolds_3BlkCrows(Len, Factor, ...)`	Manual: 3-bar trend	`talib.CDL3WHITESOLDIERS(...)` / `talib.CDL3BLACKCROWS(...)`	Separate functions
Statistical extended
`AvgDeviation(Close, N)`	`ta.mad(closes, length=N)`	Manual: `(closes - closes.rolling(N).mean()).abs().rolling(N).mean()`	Mean absolute deviation
`Variance(Close, N)`	`ta.variance(closes, length=N)`	`talib.VAR(closes, timeperiod=N)`	Population variance
`Kurtosis(Close, N)`	`closes.rolling(N).kurt()`	Manual: 4th moment	Excess kurtosis
`Skew(Close, N)`	`closes.rolling(N).skew()`	Manual: 3rd moment	Skewness
`PercentRank(ValToRank, Price, N)`	`closes.rolling(N).apply(lambda w: stats.percentileofscore(w, w.iloc[-1]))`	Manual	Percent rank
`Covariance(P1, P2, N)`	`P1.rolling(N).cov(P2, ddof=0)`	Manual	Covariance — pandas defaults to SAMPLE covariance (ddof=1); pass `ddof=0` for the population form PL uses
`Quartile(Close, N, Q)`	`closes.rolling(N).quantile(Q*0.25)`	Manual	Quartile value
`TrimMean(Close, N, Pct)`	`closes.rolling(N).apply(lambda w: stats.trim_mean(w, Pct/100))`	Manual	Trimmed mean
`Mode(Close, N, Type)`	`closes.rolling(N).apply(lambda w: w.mode().iloc[0])`	Manual	Modal value
`HarmonicMean(Close, N)`	`closes.rolling(N).apply(stats.hmean)`	Manual	Harmonic mean
Moving averages extended
`SmoothedAverage(Close, N)`	`ta.rma(closes, length=N)`	Manual: Wilder smoothing	Wilder/RMA
Miscellaneous
`BarAnnualization`	Manual: compute from bar frequency	Manual	Bars-per-year factor
`LastBarOnChart`	`idx == len(df) - 1`	Manual	True on last bar
Custom functions
`StochRSI(Close, N, M)`	`pandas_ta.stochrsi(df["close"], length=N, rsi_length=N, k=M)`	`pandas_ta.stochrsi()`	Stochastic RSI
`supertrend(N, Mult)`	`pandas_ta.supertrend(df["high"], df["low"], df["close"], length=N, multiplier=Mult)`	`pandas_ta.supertrend()`	Supertrend
`NVI(Start)`	Manual: accumulate on volume-down bars	Manual	Negative Volume Index
`PVI(Start)`	Manual: accumulate on volume-up bars	Manual	Positive Volume Index
`Coppo(N1, N2, N3)`	`pandas_ta.coppock(df["close"], length=N3, fast=N2, slow=N1)`	`pandas_ta.coppock()`	Coppock Curve
`LWTI(Close, P, N)`	Manual: `(sma(close-close.shift(P), N) / sma(high-low, N)) * 50 + 50`	Manual	Larry Williams TI
`TVI(Close, Vol, Tick)`	Manual: cumulative directional volume	Manual	Trade Volume Index
`SharpeRatio(Period, Rate, Calc, Cap)`	Manual: `(returns.mean() - rf) / returns.std()`	Manual	Portfolio Sharpe
`WRSI(N, Close)`	`pandas_ta.rsi(df["close"], length=N)`	`pandas_ta.rsi()`	Wilder RSI (default)
`NewMA(Close, N)`	Manual: Heikin-Ashi + TEMA hybrid	Manual	Hybrid MA

Table 4 — Strategy Orders

PowerLanguage	Python	Notes
`Buy("label") next bar market`	`orders.append(Order(label="label", action=Action.ENTRY, side=Side.LONG, order_type=OrderType.MARKET))`	Append to orders list
`SellShort("label") next bar market`	`Order(action=Action.ENTRY, side=Side.SHORT, order_type=OrderType.MARKET)`	`Side.SHORT` for short entry
`Sell("label") next bar market`	Close long: `Order(action=Action.EXIT, side=Side.LONG, order_type=OrderType.MARKET)`	WARNING: PL `Sell` exits a long — do NOT map to `Side.SHORT`
`BuyToCover("label") next bar market`	Close short: `Order(action=Action.EXIT, side=Side.SHORT, order_type=OrderType.MARKET)`	PL `BuyToCover` exits an existing short
`Buy("label") next bar at price limit`	`Order(order_type=OrderType.LIMIT, price=price)`	Limit order
`Buy("label") next bar at price stop`	`Order(order_type=OrderType.STOP, price=price)`	Stop order
`SetStopLoss(dollars)`	`stop_price = entry_price - dollars / (qty * point_value)`	PL takes a dollar amount; Python must convert to price level
`SetProfitTarget(dollars)`	`target_price = entry_price + dollars / (qty * point_value)`	Same dollar-to-price conversion

Table 5 — Plotting

PowerLanguage	Python	Notes
`Plot1(value, "label")`	`print(f"{label}: {value}")` or log to file	No chart in Python; output to stdout, `logging` module, or results list
`SetPlotColor(1, Red)`	N/A	No visual output; skip or store as metadata
`SetPlotWidth(1, 2)`	N/A	Skip
`NoPlot(1)`	Skip output for that bar	Conditional: don't append to results

Table 6 — Control Flow

PowerLanguage	Python	Notes
`If cond Then Begin ... End;`	`if cond:`	Direct mapping
`If cond Then ... Else ...`	`if cond: ... else: ...`	Direct mapping
`For idx = 1 to n Begin ... End;`	`for i in range(1, n + 1):`	PL `For` is inclusive; Python `range` excludes upper bound. PL side uses `idx` because `i` is reserved (alias for OpenInterest)
`For idx = n DownTo 1 Begin ... End;`	`for i in range(n, 0, -1):`	PL DownTo; Python reversed range
`While cond Begin ... End;`	`while cond:`	Direct mapping
`Switch (expr) Begin Case 1: ... End;`	`match expr:` with `case 1: ...`	Python 3.10+ `match/case`; no fallthrough (matches PL); PL empty case body is a compile error — use `Value1 = Value1;` as no-op
`Once Begin ... End;`	`if not self._once_done: ... self._once_done = True`	Use a bool attribute; PL `Once` runs on first bar only

Table 7 — Other Built-ins and Features

PowerLanguage	Python	Notes
`MarketPosition`	`self.position` (int: −1/0/+1)	Track manually; update on order fills
`EntryPrice`	`self.entry_price` (float)	Track manually; set when position opens
`CurrentContracts`	`self.current_contracts` (float)	Track manually; absolute contract count
`BarsSinceEntry`	`self.bars_since_entry` (int)	Increment each bar while `position != 0`; reset on new entry
`Print("text")`	`print("text")`	Direct mapping
`#BeginCmtry ... #EndCmtry`	No equivalent	PL expert commentary; skip or log
`Alert("msg")`	`print("ALERT: msg", file=sys.stderr)`	No built-in alert; write to stderr or use callback

Part 2: Semantic Differences and Gotchas

Order fill timing — the #1 conversion bug

PL Buy next bar at market fills at the NEXT bar's OPEN, and next bar at X Stop / Limit fills intrabar on the next bar at the stop/limit price (or at the open if the bar gaps through it). A converted simulator that flips position on the signal bar (same-bar close fill) is one bar early and uses the wrong price.

The executor must queue orders produced on bar N and fill them against bar N+1:

for i in range(len(bars) - 1):
    orders: list[Order] = []
    strategy.on_bar(bars[: i + 1], orders)        # orders queued on bar i...
    nxt = bars[i + 1]
    for o in orders:                              # ...filled on bar i+1
        if o.order_type is OrderType.MARKET:
            fill(o, price=nxt.open)                          # next bar's OPEN
        elif o.order_type is OrderType.STOP and nxt.high >= o.price:  # buy stop
            fill(o, price=max(o.price, nxt.open))            # intrabar at the stop price
        # sell stop: nxt.low <= o.price; limit orders mirror with reversed comparisons

EMA seeding/warmup gotcha: PL XAverage seeds recursively from the first bar's price. pandas-ta ema defaults to an SMA seed (sma=True) — pass sma=False to match PL, or document the difference. TA-Lib TA_EMA seeds with the SMA of the first Length values. Early-history signals therefore differ across targets — discard roughly the first 3×Length bars before comparing backtests.

Sell does not mean "go short". In PowerLanguage, Sell exits an existing long position. The Python equivalent is closing the position (setting self.position = 0). Do not create a new short entry as a translation of Sell.
PL is implicitly bar-driven; Python requires an explicit loop. PowerLanguage executes the entire script top-to-bottom on each bar close automatically. In Python, you must write the for i in range(len(bars)) loop yourself and call on_bar() on each iteration. Forgetting the loop wrapper means the strategy logic runs only once.
Negative indexing is convenient but still needs bounds guarding. bars[-2].close is Pythonic for "previous bar" but raises IndexError when len(bars) < 2. Guard with if len(bars) > n before any lookback access. PL handles this automatically via MaxBarsBack; Python does not.
pandas-ta operates on full Series, not streaming. Each on_bar() call recomputes the indicator on the entire series up to the current bar. This is correct but O(n²) over a backtest. For performance, pre-compute indicators outside the loop or cache results.
NaN values in indicator output. pandas-ta returns NaN for the first length - 1 bars where the indicator has insufficient history. Always check pd.notna(value) or not math.isnan(value) before using the result. PL silently returns 0 for insufficient history; Python raises errors or produces silent NaN propagation.
Mutable default arguments are a classic Python trap. Never write def __init__(self, buf=[]) — use None and assign inside: self.buf = buf if buf is not None else []. This doesn't affect PL conversion directly, but catches new Python users.
Python match does not fall through. Unlike C++ switch, Python 3.10+ match/case has no fallthrough. This matches PL's Switch/Case behavior, so the mapping is direct.
Float equality. Use math.isclose(a, b) or abs(a - b) < 1e-10 instead of ==. PL uses tolerance-based comparison for = on float values; Python == on floats is exact bitwise comparison.
No Value1..Value99 or Condition1..Condition99. These PL pre-declared variables must be replaced with named instance attributes. Use descriptive identifiers like self.rsi_value and self.is_oversold.
Position state is not tracked automatically. PL provides MarketPosition, EntryPrice, CurrentContracts, and BarsSinceEntry as built-in read-only variables updated by the engine. In Python, you must maintain these as instance attributes and update them manually whenever the execution engine fills an order.
Dollar amounts versus price levels for stops. SetStopLoss and SetProfitTarget in PL accept a dollar (currency) amount. Python has no implicit conversion — you must compute the price level manually: stop_price = entry_price - stop_dollars / (qty * point_value).
Multi-data series requires explicit design. PL Close of Data2 accesses a second feed bound to the chart. In Python, you must pass a second bar list (bars2: list[Bar]) to on_bar() or store it as an attribute. There is no implicit secondary-feed mechanism.
No Portfolio Money Management (PMM) equivalent. PL's PMM allows a single script to govern position sizing across multiple instruments. Python has no built-in portfolio-level abstraction. You must implement cross-instrument logic as a separate orchestrator that calls individual strategy instances.
datetime.utcfromtimestamp is deprecated in Python 3.12+. Use datetime.fromtimestamp(ts, tz=timezone.utc) instead to avoid deprecation warnings.

Part 3: Conversion Checklists

PL → Python Pre-Conversion

List every Inputs: declaration → __init__ parameter with default
List every Variables: declaration → instance attribute in __init__
Identify Value1..99 / Condition1..99 → plan meaningful replacement names
Identify all indicator calls → confirm pandas-ta / TA-Lib equivalent exists
Locate Data2/Data3 references → plan multi-feed architecture
Locate SetStopLoss/SetProfitTarget → note point value for dollar→price conversion
Check IntraBarPersist → decide if tick-level behavior matters
Identify .elf function calls → plan Python function equivalents

PL → Python Post-Conversion

Python → PL Pre-Conversion

List all __init__ attributes → categorize: inputs (immutable) vs variables (mutable) vs indicator state
Identify all pandas-ta / TA-Lib calls → map each to PL built-in
Check for match statements with guard clauses → plan PL Switch or If/Else
Identify any libraries beyond pandas-ta/TA-Lib → plan PL equivalents
Check for explicit position-tracking attributes → confirm PL built-ins suffice

Python → PL Post-Conversion

Every short entry maps to SellShort, not Sell
Every long close maps to Sell, not SellShort
All bars[-1-n].close lookbacks converted to Close[n]
All manual crossover comparisons converted to PL Crosses Over / Crosses Under
All attributes categorized: immutable → Inputs:, mutable → Variables:
All pandas-ta/TA-Lib calls replaced with PL function calls
NaN handling removed (PL handles lookback alignment internally)
Price-level stops/targets converted to dollar amounts for SetStopLoss/SetProfitTarget
Script compiled in MultiCharts PowerEditor with zero errors

name	powerlanguage-python-conversion
description	Use when converting or porting code between MultiCharts PowerLanguage and Python in either direction — Strategy ABC scaffold, concept mapping tables, semantic differences, pandas-ta/TA-Lib indicators, conversion checklists.

powerlanguage-python-conversion

同仓库更多 Skills

PowerLanguage ↔ Python Conversion

How to use

Part 0: Structural Template

Bar dataclass

Order types

Strategy ABC

Strategy subclass skeleton

Alternative: TA-Lib indicator pattern

Main loop

Part 1: Concept Mapping

Table 1 — Declarations

Table 2 — Data Access

Table 3 — Technical Indicators (dual-library)

Table 4 — Strategy Orders

Table 5 — Plotting

Table 6 — Control Flow

Table 7 — Other Built-ins and Features

Part 2: Semantic Differences and Gotchas

Order fill timing — the #1 conversion bug

Part 3: Conversion Checklists

PL → Python Pre-Conversion

PL → Python Post-Conversion

Python → PL Pre-Conversion

Python → PL Post-Conversion

PowerLanguage ↔ Python Conversion

How to use

Part 0: Structural Template

Bar dataclass

Order types

Strategy ABC

Strategy subclass skeleton

Alternative: TA-Lib indicator pattern

Main loop

Part 1: Concept Mapping

Table 1 — Declarations

Table 2 — Data Access

Table 3 — Technical Indicators (dual-library)

Table 4 — Strategy Orders

Table 5 — Plotting

Table 6 — Control Flow

Table 7 — Other Built-ins and Features

Part 2: Semantic Differences and Gotchas

Order fill timing — the #1 conversion bug

Part 3: Conversion Checklists

PL → Python Pre-Conversion

PL → Python Post-Conversion

Python → PL Pre-Conversion

Python → PL Post-Conversion

同仓库更多 Skills