#!/bin/env -S python3

from dataclasses import dataclass

import numpy
import pandas
import random
import math

def main():
    dataset = load_dataset("bin/iris/iris.data")
    # print("dataset", dataset)
    (dataset_training, dataset_test) = split_dataset(dataset, 0.8)
    # print("dataset_training", dataset_training, "dataset_test", dataset_test)
    centroid = KVMeans(3, 4)
    centroid.learn(dataset_training[["a", "b", "c", "d"]])



def load_dataset(path: str):
    data = pandas.read_csv(path, names=["a", "b", "c", "d", "category"])
    return data


def split_dataset(dataset: pandas.DataFrame, ratio: float):
    total_len = len(dataset)
    start_len = math.floor(total_len * ratio)
    (start, rest) = numpy.array_split(dataset, [start_len])
    return (start, rest)


@dataclass
class Centroid:
    dimension: int
    position: list[float]

    @staticmethod
    def new_random(dimension: int, rng: random.Random):
        return Centroid(dimension, [rng.random() for _ in range(dimension)])

class KVMeans:

    def __init__(self, mean_count: int, dimension: int):
        rng = random.Random(0)
        self.means = [Centroid.new_random(dimension, rng) for _ in range(mean_count)]
        self.dimension = dimension
        self.mean_count = mean_count

    def learn(self, dataset: pandas.DataFrame):
        assert dataset.columns.size == self.dimension
        for (_, line) in dataset.iterrows():
            for absc in range(self.dimension):
                pass

        # print([3 for _ in series])
        # print(series)
        # dataset.map(lambda e: print(e))

    def predict(self):
        pass

def distance_dim(a: list[float], b: list[float]):
    acc = 0
    for (a_, b_) in zip(a, b): acc += math.pow(a_ - b_, 2)
    return math.sqrt(acc)

if __name__ == "__main__": main()