import numpy as np

import argparse

import re

import os

import sys

parser = argparse.ArgumentParser()

# parser.add_argument('-d', '--datasets', nargs='+', type=str)

# parser.add_argument('-r', '--rate', type=float)

parser.add_argument('-t', '--tree', type=int)

# parser.add_argument('--is-retrain', action='store_true')

args = parser.parse_args()

def load_file(file_base_path, ratio: str, n_runs):

removing_time = [0 for _ in range(n_runs)]

training_time = [0 for _ in range(n_runs)]

for i in range(n_runs):

fn = f"{file_base_path}_deltaboost_{ratio}_{i}.out"

# print("Loading file: ", fn)

with open(fn, "r") as f:

for line in f:

if "removing time" in line:

removing_time[i] = float(re.findall("\d+\.\d+", line)[0])

if "training time" in line:

training_time[i] += float(re.findall("\d+\.\d+", line)[0])

if "Init booster time" in line:

training_time[i] += float(re.findall("\d+\.\d+", line)[0])

db_train_time_str = rf"{np.mean(training_time): .3f} \textpm {np.std(training_time):.3f}"

db_remove_time_str = rf"{np.mean(removing_time): .3f} \textpm {np.std(removing_time):.3f}"

fn_gbdt = f"{file_base_path}_gbdt_{ratio}.out"

gbdt_time = 0

# print("Loading file: ", fn_gbdt)

with open(fn_gbdt, "r") as f:

for line in f:

if "training time" in line:

gbdt_time += float(re.findall("\d+\.\d+", line)[0])

if "Init booster time" in line:

gbdt_time += float(re.findall("\d+\.\d+", line)[0])

gbdt_time_str = rf"{gbdt_time: .3f}"

speedup = gbdt_time / np.mean(removing_time)

print(f"{gbdt_time_str}\t& {db_train_time_str}\t & {db_remove_time_str}\t & {speedup:.2f}x \\\\")

if __name__ == '__main__':

datasets = ['codrna', 'covtype', 'gisette', 'cadata', 'msd']

print("Thunder\t& DB-Train\t& DB-Remove\t& Speedup (Thunder) \\\\")

for dataset in datasets:

load_file(f"../out/remove_test/tree{args.tree}/"+dataset, "1e-03", 100)

load_file(f"../out/remove_test/tree{args.tree}/"+dataset, "1e-02", 100)