Exporters

Exporters generate platform-specific code from a fitted DecisionTree. They are used internally by Converter.convert(target=...) but can also be used directly.

create_exporter() Factory

from blackbox2c.exporters import create_exporter

exporter = create_exporter(format, **kwargs)

format value	Exporter class
'cpp' or 'c++'	CppExporter
'arduino'	ArduinoExporter
'micropython' or 'python'	MicroPythonExporter

---

CppExporter

Generates a C++11 class inside an ml namespace.

from blackbox2c.exporters import CppExporter

exporter = CppExporter(
    function_name='predict',
    class_name='Predictor',
    use_namespace=True,
    namespace='ml',
    use_fixed_point=False,
    precision=8,
)
code = exporter.generate(tree, feature_names=['sl', 'sw', 'pl', 'pw'])

Output structure:

namespace ml {
class Predictor {
public:
    static uint8_t predict(float features[4]) { ... }
};
} // namespace ml

---

ArduinoExporter

Generates an Arduino .h file with optional PROGMEM constants.

from blackbox2c.exporters import ArduinoExporter

exporter = ArduinoExporter(
    function_name='predict',
    use_progmem=True,
    use_fixed_point=False,
    precision=8,
)
code = exporter.generate(tree, feature_names=['temp', 'humidity'])

Include in your sketch:

#include "predict.h"

void loop() {
    float features[2] = { readTemp(), readHumidity() };
    uint8_t result = predict(features);
}

---

MicroPythonExporter

Generates a Python module for MicroPython-enabled boards (ESP32, Pico, etc.).

from blackbox2c.exporters import MicroPythonExporter

exporter = MicroPythonExporter(
    function_name='predict',
    class_name='Predictor',
    use_const=True,
)
code = exporter.generate(tree, feature_names=['x', 'y'])

Usage on device:

from predictor import Predictor
result = Predictor.predict([0.5, 1.2])

---

Direct Usage (after Converter)

converter = Converter(config)
converter.convert(model, X_train)  # runs pipeline, builds surrogate_tree_

exporter = create_exporter('arduino')
code = exporter.generate(
    converter.surrogate_tree_,
    feature_names=converter.feature_names_,
    class_names=converter.class_names_,
)