trade_on_close not followed when SL/TP hit in the same candle.

[admb4r]

Expected Behavior

When using the trade_on_close=False feature, trades will always occur on the following candle to which a trade was placed.

Actual Behavior

When a SL/TP is hit on the same candle a trade is placed, the logic does not cater for trade_on_close=False and instead places and closes the trade on the current candle.

I believe this is due to the check on line 908 of backtesting.py, which after a opening a trade with the correct time_index (the next candle in the backtest), instructs the backtest to reprocess orders if the trade is a market order and does not cater for the trade_on_close setting. This results in the SL/TP being checked for a trade that should not yet exist when trade_on_close=False.

backtesting.py/backtesting/backtesting.py

Lines 900 to 920 in 0107d4a

	# Open a new trade
	if need_size:
	self._open_trade(adjusted_price, need_size, order.sl, order.tp, time_index)
	# We need to reprocess the SL/TP orders newly added to the queue.
	# This allows e.g. SL hitting in the same bar the order was open.
	# See https://github.com/kernc/backtesting.py/issues/119
	if order.sl or order.tp:
	if is_market_order:
	reprocess_orders = True
	elif (low <= (order.sl or -np.inf) <= high or
	low <= (order.tp or -np.inf) <= high):
	warnings.warn(
	f"({data.index[-1]}) A contingent SL/TP order would execute in the "
	"same bar its parent stop/limit order was turned into a trade. "
	"Since we can't assert the precise intra-candle "
	"price movement, the affected SL/TP order will instead be executed on "
	"the next (matching) price/bar, making the result (of this trade) "
	"somewhat dubious. "
	"See https://github.com/kernc/backtesting.py/issues/119",
	UserWarning)

[kernc]

Hopefully fixed in 4aee89a.