Numba compilation error I don't understand

[Kayne88]

I hope someone can point out my mistake, I wasn't able to resolve or google it myself.

Here's my code:

@njit
def pre_sim_func_nb(c, ohlc_5m, ohlc_30m, ema_1h, swing_change, cancellation_threshold):
    swing_hls = ind_swing_hls(ohlc_30m, swing_change, cancellation_threshold)
    sl = np.full(c.target_shape[1], np.nan, dtype=np.float_)
    tp = np.full(c.target_shape[1], np.nan, dtype=np.float_)
    return (swing_hls, sl, tp, ohlc_5m, ema_1h,)

@njit
def order_func_nb(c, swing_hls, sl, tp, ohlc_5m, ema_1h, size, size_type, fees, supp_res_threshold, max_margin):
     ...

@njit(debug=True)
def post_order_func_nb(c, swing_hls, sl, tp):
      if c.order_result.status == OrderStatus.Filled:
      if c.position_now == 0:
          sl[c.col] = np.nan
          tp[c.col] = np.nan
      elif c.order_result.side == OrderSide.Buy and c.position_before < 0 and c.position_now > 0:
          sl[c.col] = swing_hls[0, c.i]
          tp[c.col] = swing_hls[1, c.i]
      elif c.order_result.side == OrderSide.Sell and c.position_before > 0 and c.position_now < 0:
          sl[c.col] = swing_hls[0, c.i]
          tp[c.col] = swing_hls[1, c.i]

def simulate(ohlc_5m, ohlc_30m, ema_1h, swing_change, cancellation_threshold, supp_res_threshold, size, size_type, fees, max_margin):
    return vbt.Portfolio.from_order_func(
        ohlc_5m[:,3],
        order_func_nb,
        vbt.Rep('size'), vbt.Rep('size_type'), vbt.Rep('fees'), vbt.Rep("supp_res_threshold"), vbt.Rep("max_margin"),  # order_args
        pre_sim_func_nb=pre_sim_func_nb,
        pre_sim_args=(ohlc_5m, ohlc_30m, ema_1h, vbt.Rep('swing_change'), vbt.Rep('cancellation_threshold'),),
        post_order_func_nb=post_order_func_nb,
        broadcast_named_args=dict(
            swing_change=swing_change,
            cancellation_threshold=cancellation_threshold,
            supp_res_threshold=supp_res_threshold,
            size=size,
            size_type=size_type,
            fees=fees,
            max_margin=max_margin
        ),
        freq="5min"
    )

pf = simulate(
    ohlc_5m=backtest_data["chart_5m"].to_numpy(),
    ohlc_30m=backtest_data["chart_30m"].to_numpy(),
    ema_1h=backtest_data["ema_1h"].to_numpy(),
    swing_change=swing_change,
    cancellation_threshold=cancellation_threshold,
    supp_res_threshold=supp_res_threshold,
    size=trade_size,
    size_type=size_type,
    fees=fees,
    max_margin=max_margin
)

I get the following compilation error:

TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Internal error at <numba.core.typeinfer.CallConstraint object at 0x0000024301A1A128>.
too many positional arguments
During: resolving callee type: type(CPUDispatcher(<function post_order_func_nb at 0x000002430C7A3488>))
During: typing of call at C:\Users\User\Anaconda3\lib\site-packages\vectorbt\portfolio\nb.py (3142)
Enable logging at debug level for details.
File "......\Anaconda3\lib\site-packages\vectorbt\portfolio\nb.py", line 3142:
... def post_sim_func_nb(c):

)
post_order_func_nb(post_order_ctx, *pre_segment_out, *post_order_args)
^

I couldn't spot, where I put too many positional arguments. Anybody sees it? Btw the code compiled correctly, when the post_order_func_nb argument is ommited.

Cheers

[polakowo]

@Kayne88 please format your code, it's difficult to see what it does.

[Kayne88]

@polakowo done

[polakowo]

Shouldn't the arguments of post_order_func_nb be c, swing_hls, sl, tp, ohlc_5m, ema_1h?

[Kayne88]

I dont need ohlc_5m and ema_1h in that particular function, so I thought I ommit them. However, is it required by the framework to always give all the attributes passed down the stack into each subsequent function? @polakowo

[polakowo]

@Kayne88 yes. If you don't need those arguments you can do c, swing_hls, sl, tp, _, _, but you always need to accept the exact same number of arguments that was passed by the function up the stack.

[Kayne88]

@polakowo Thanks a lot, that was the issue. As always very helpful.