Atan in micropython

[Redstonetill]

I currently use the pybricks micropython repo to create my own function inside of the pybricks firmware. But when trying to build the firmware I am getting an error: <artificial>:(.text.pb_module_tools_get_angle_error.lto_priv.0+0x1e2): undefined reference to atan'
(same is for fmod)

If anyone could help, I would be gratefull

#include <math.h>

[...]

static mp_obj_t pb_module_tools_get_angle_error(size_t n_args, const mp_obj_t *args) {
    // Stelle sicher, dass 5 Argumente übergeben werden
    if (n_args != 5) {
        mp_raise_ValueError(mp_obj_new_str("Expected 5 arguments", strlen("Expected 5 arguments")));
    }

    // Konvertiere die Argumente in C-Float-Werte
    double startX = mp_obj_get_float(args[0]);
    double startY = mp_obj_get_float(args[1]);
    double pointX = mp_obj_get_float(args[2]);
    double pointY = mp_obj_get_float(args[3]);
    double startAngle = mp_obj_get_float(args[4]);

    double angleToWaypoint = 0.0;

    if (startY == pointY) {angleToWaypoint = startX < pointX ? 0.0 : 180.0;}
    else {
        double dx = pointX - startX;
        double dy = startY - pointY;

        angleToWaypoint = atan(dx / dy) * 180 / (double) M_PI;

        if (dy < 0) {angleToWaypoint += 90;}
        if (dy > 0) {angleToWaypoint -= 90;}
    }

    // --- Get fastes way to turn ---
    double rightDistance = fmod(angleToWaypoint - startAngle + 720, 360.0);
    double leftDistance = fmod(startAngle - angleToWaypoint + 720, 360.0);
    
    if (rightDistance < leftDistance) {return mp_obj_new_float(rightDistance);}
    
    return mp_obj_new_float(-leftDistance);
}

static MP_DEFINE_CONST_FUN_OBJ_VAR_BETWEEN(pb_module_tools_get_angle_Error_obj, 5, 5, pb_module_tools_get_angle_error);

[dlech]

Which hub are you targeting?

[Redstonetill]

Sorry for not stating, I am currently using the Spike Prime Hub

[dlech]

Ah, I see the problem now. Pybricks hubs only use single precision floating point, not double

For portability,

double startX = mp_obj_get_float(args[0]);

should be

mp_float_t startX = mp_obj_get_float(args[0]);

or to force single-precision, you can do

float startX = mp_obj_get_float_to_f(args[0]);

(All other uses of double should be changed accordingly as well)

[laurensvalk]

And to explain:

undefined reference to atan

You can use atanf.

(Although you'd probably want to use atan2f since dividing dy is problematic when dy = startY - pointY = 0.)

[Redstonetill]

Thanks for the quick answer. Using atanf + fmodf and only float variables (single-precision) (mp_obj_get_float_to_f) seems to work just fine

[Redstonetill]

Fyi dy can never be zero since I am checking if the startY and pointY is the same, but still thanks for the tipp :)

[laurensvalk]

Glad it worked!

Fyi dy can never be zero since I am checking if the startY and pointY is the same, but still thanks for the tipp :)

Checking two doubles or floats for equality as startY == pointY is generally not reliable. They may be almost arbitrarily close which still means dividing by zero. It is safer to test something like abs(startY - pointY) < some_small_value