KIAPBCI/kiap_bci/tests/test_prediction2.py

'''
description: testing prediction and speller matrix by writing freely on the console
author: Alessandro Tonin
date:   06.02.2020

Copyright (c) 2020 Alessandro Tonin.
All rights reserved.'''

import importlib
import logging
import sys
import time
from random import randint

import tkinter as tk
import yaml
# import pyaudio
from colorlog import ColoredFormatter

import aux
from aux import log
from paradigms import colorSpeller
import multiprocessing as mp
from multiprocessing import Array, Pipe

import readchar
import csv

import re

importlib.reload(colorSpeller)
importlib.reload(aux)

file_log = '/kiap/src/kiap_bci/tests/records_for_annotation_yml.yaml'
file_in = '/kiap/src/kiap_bci/tests/user_test.txt'
file_out = '/kiap/src/kiap_bci/tests/user_test_out_update_spcorpus_sp_us_gen.txt'

sp = colorSpeller.ColorSpeller()

update_corpus = True

# # process yaml file to correctly format the spelled sentences
# with open(file_log, 'r') as flog, open(file_in,'w') as fin:
# 	entries = yaml.load(flog,Loader=yaml.Loader)
# 	for k in entries.keys():
# 		phrase = entries[k]['phrase']
# 		init_phrase = entries[k]['phrase_start']
# 		phrase = phrase.strip() #just in case...
# 		if init_phrase and phrase.find(init_phrase) == 0:
# 			phrase = '['+phrase[0:len(init_phrase)]+']'+phrase[len(init_phrase):]
# 		fin.write(phrase+'\n')


# txt file with one speller selection in each line
# Read txt file line by line
with open(file_in,'r+') as fin, open(file_out, 'w') as fout:

	results_writer = csv.writer(fout, delimiter=',')
	results_writer.writerow(['num', 'target', 'string', 'vocab1', 'vocab2', 'vocab3', 'vocab4', 'vocab5', 'pred corr','pred wrong','saved char'])

	line = fin.readline()
	cnt = 1
	while line:
		# Check initial string
		init_str_start = line.find('[')
		init_str_end = line.find(']')
		if init_str_start != init_str_end:
			curr_str = line[init_str_start+1:init_str_end]
		else:
			curr_str = ''

		# Delete initial string identifiers
		for r in (('[',''),(']',''),('\n','')):
			line = line.replace(*r)
		log.warning(f'target: {line}')

		sp.set_current_string(curr_str)
		sp.external_updates()
		vocab = sp.get_vocabulary(5)
		# fill empty vocabulary with dash
		while len(vocab) < 5:
			vocab.append('#')

		log.info(f'string: {curr_str}')
		log.info(f'vocab: {vocab}')

		results_writer.writerow([cnt,line,curr_str]+vocab)
		
		# use while loop instead of for in order to skip iterations when prediction is correct
		idx_letter = 0
		pred_correct = 0
		pred_wrong = 0
		saved_char = 0
		while idx_letter < len(line):

			# find the actual word
			word_start = line.rfind(' ',0,idx_letter)
			word_end = line.find(' ',idx_letter)
			if word_end == -1: word_end = len(line) # because if not find result is -1
			curr_word = line[word_start+1:word_end]

			# Check if there is a probable word and if it is correct
			if len(vocab)>0 and vocab[0].startswith('*') and vocab[0][1:-1] == curr_word:
					# in case add the probable word to the string
				curr_str = curr_str[:word_start+1] + curr_word
				pred_correct += 1
				saved_char += word_end-idx_letter
				if word_end < len(line): 
					curr_str += ' '
					saved_char += 1
			elif len(vocab)>0 and vocab[0].startswith('*'):
				curr_str += line[idx_letter]
				pred_wrong += 1					

			else:
				# Add to speller the selection letter by letter
				curr_str += line[idx_letter]

						# Keep trace of predicted words
			

			sp.set_current_string(curr_str)
			sp.external_updates()
			# Check prediction after every new letter	
			vocab = sp.get_vocabulary(5)
			# fill empty vocabulary with dash
			while len(vocab) < 5:
				vocab.append('#')

			log.info(f'string: {curr_str}')
			log.info(f'vocab: {vocab}')
			
			idx_letter = len(curr_str)
			results_writer.writerow([cnt,line,curr_str]+vocab+[pred_correct,pred_wrong,saved_char])

		if update_corpus:
			sp._save_string()
			sp.speller_user_cfd = sp._createCFD(root=sp.corpora_path,corpus_name=sp.speller_user_corpus_name)
		line = fin.readline()
		cnt += 1