class DataUtil:
def __init__(self, data_dir):
self.phrase_util = Phrases("resources/comm2.dev")
self.plot_colors = ["#0000ff", "#4400ff", "#8800ff", "#cc00ff", "#ff00cc", "#ff0088", "#ff0044"]
self.click_data_files = []
self.click_context_files = []
self.preconfig_file = None
self.data_dir = data_dir
for path, dir, files in os.walk(data_dir):
for file in files:
if "click_time_log" in file:
self.click_data_files += [os.path.join(path, file)]
if "params_data_use_num" in file:
self.click_context_files += [os.path.join(path, file)]
if "preconfig" in file:
self.preconfig_file = os.path.join(path, file)
self.rel_click_data = []
self.abs_click_data = []
self.selection_data = []
self.speed_changes = []
self.kde_list = []
self.clicks_by_speed = {}
self.clicks_by_phrase = {}
self.phrase_stats = {}
self.corrected_clicks = None
def load_data(self):
# size_list = []
for data_file in self.click_data_files:
data_handel = PickleUtil(data_file)
click_dict = data_handel.safe_load()
# size_list += [len(click_dict["click time list"])]
self.rel_click_data += [click[0] for click in click_dict["click time list"]]
self.rel_click_data = np.array(self.rel_click_data)
# size_list_2 = []
for data_file in self.click_context_files:
data_handel = PickleUtil(data_file)
context_dict = data_handel.safe_load()
self.abs_click_data += context_dict["press"]
# size_list_2 += [len(flatten(context_dict["press"]))]
self.speed_changes += context_dict["speed"]
self.selection_data += context_dict["choice"]
self.abs_click_data = np.array(flatten(self.abs_click_data))
self.speed_changes.sort(key=lambda x: x[0])
if self.preconfig_file is not None:
preconfig_handel = PickleUtil(self.preconfig_file)
preconfig = preconfig_handel.safe_load()
self.kde_list = np.array(preconfig["li"])/np.sum(preconfig["li"])
if len(self.abs_click_data) != len(self.rel_click_data):
raise ValueError("Click data length does not match context data length!")
print("Loaded " + str(len(self.abs_click_data)) + " clicks")
def split_data_phrase(self):
self.phrases = []
self.phrase_times = {}
phrase_start = 0
uncorrected_error = 0
corrected_error = 0
session_num = 0
start_time_prev = 0
for selection_num, selection in enumerate(self.selection_data): # scan through data to find phrases & time ints
if "target" in selection:
phrase = selection["target"]
typed = selection["typed"]
is_backspace = selection["backspace"]
is_undo = selection["undo"]
if phrase not in self.phrases:
self.phrases.append(phrase)
phrase_start = selection["time"]
if is_backspace: # accumulate corrected errors
corrected_error += 1
if is_undo:
if selection_num > 0:
# prev_typed = self.selection_data[selection_num-1]["typed"]
# typed_difference = prev_typed[len(typed):]
corrected_error += 1
_, completed_phrase = self.phrase_util.compare(typed, phrase)
if completed_phrase and not is_undo:
self.phrase_times[phrase] = (phrase_start, selection["time"])
uncorrected_error = calc_MSD(typed, phrase)[0]
total_error = (uncorrected_error + corrected_error) / len(max(typed, phrase))
self.phrase_stats[phrase] = {"error_unc": uncorrected_error, "error_cor": corrected_error,
"error_tot": total_error}
uncorrected_error = 0
corrected_error = 0
self.phrases = list(self.phrase_times.keys())
self.phrases.sort(key=lambda x: self.phrase_times[x][0])
print("Data partitioned into " + str(len(self.phrases)) + " sets by phrase")
for phrase in self.phrases: # split data according to phrase times calculated above
phrase_start, phrase_end = self.phrase_times[phrase]
phrase_click_indices = np.where((self.abs_click_data >= phrase_start) & (self.abs_click_data <= phrase_end))
phrase_abs_clicks = self.abs_click_data[phrase_click_indices]
phrase_rel_clicks = self.rel_click_data[phrase_click_indices]
if len(phrase_abs_clicks) > 0:
first_click_time = min(phrase_abs_clicks)
self.clicks_by_phrase[phrase] = {"abs": phrase_abs_clicks, "rel": phrase_rel_clicks}
num_clicks = len(phrase_abs_clicks)
time_int = phrase_end - first_click_time
num_characters = len(phrase)
num_words = len(phrase.split(" "))
clicks_per_char = num_clicks / num_characters
clicks_per_word = num_clicks / num_words
chars_per_min = num_characters / time_int * 60
words_per_min = num_words / time_int * 60
stat_dict = self.phrase_stats[phrase]
stat_dict["clicks_char"] = clicks_per_char
stat_dict["clicks_word"] = clicks_per_word
stat_dict["chars_min"] = chars_per_min
stat_dict["words_min"] = words_per_min
if phrase_start - start_time_prev > 20 * 60:
session_num += 1
start_time_prev = phrase_start
stat_dict["session"] = session_num
stat_dict["start_time"] = phrase_start
else:
del self.phrase_stats[phrase]
self.phrases = list(self.phrase_stats.keys())
# print(self.phrase_stats)
def split_data_speed(self):
num_speed_changes = len(self.speed_changes)
for change_index in range(num_speed_changes):
time_min = self.speed_changes[change_index][0]
if change_index == num_speed_changes-1:
time_max = float("inf")
else:
time_max = self.speed_changes[change_index+1][0]
clicks_int = np.array(np.where((self.abs_click_data > time_min) & (self.abs_click_data < time_max))[0],
dtype='int64')
matrix_index = np.vectorize(lambda m_index: self.rel_click_data[m_index])
if clicks_int.size > 0:
clock_speed = round(self.speed_changes[change_index][2], 2)
clicks_rel_int = matrix_index(clicks_int)
if clock_speed in self.clicks_by_speed.keys():
prev_clicks = self.clicks_by_speed[clock_speed]
self.clicks_by_speed[clock_speed] = np.concatenate((prev_clicks, clicks_rel_int))
else:
self.clicks_by_speed[clock_speed] = clicks_rel_int
print("Data partitioned into " + str(len(self.clicks_by_speed.keys())) + " sets by clock rotation speed")
def correct_data_speed(self):
print(self.clicks_by_speed.keys())
if 0.6 not in self.clicks_by_speed.keys():
raise ValueError("Base rotation speed not in data!")
base_clicks = self.clicks_by_speed[0.6]
base_clicks_mean = np.mean(base_clicks)
base_clicks = base_clicks - base_clicks_mean
base_clicks_std = np.std(base_clicks)
clock_speeds = list(self.clicks_by_speed.keys())
self.corrected_clicks = []
for clock_speed in clock_speeds:
clicks = self.clicks_by_speed[clock_speed]
clicks_mean = np.mean(clicks)
clicks = clicks - clicks_mean
clicks_std = np.std(clicks)
clicks *= base_clicks_std/clicks_std
self.corrected_clicks += clicks.tolist()
self.corrected_clicks = np.array(self.corrected_clicks)
def make_data_frame(self):
for phrase_num, phrase in enumerate(self.phrases):
if phrase_num == 0:
DF = pd.DataFrame(pd.DataFrame([self.phrase_stats[phrase]]))
DF["phrase"] = phrase
else:
df = pd.DataFrame([self.phrase_stats[phrase]])
df["phrase"] = phrase
DF = DF.append(df, ignore_index=True)
# DF = DF.sort_values(by=['start_time'])
self.DF = DF
def plot_data(self):
fig = plt.figure()
ax = plt.subplot(111)
plot_num = 0
for clock_speed in self.clicks_by_speed.keys():
plot_color = self.plot_colors[plot_num]
clicks = self.clicks_by_speed[clock_speed]/clock_speed*80
clicks_mean = np.mean(clicks)
clicks_std = np.std(clicks)
plot_label = "speed: "+str(clock_speed)+" ("+str(len(clicks))+" points)"
ax.hist(clicks, 20, range=[0, 80], density=True, color=plot_color, alpha=0.3, label=plot_label)
kernel = stats.gaussian_kde(clicks)
res = 10
plt.plot(np.arange(80*res)/res, kernel(np.arange(80*res)/res), color=plot_color, linewidth=2)
plot_num += 1
plt.axvline(clicks_mean, color=plot_color, linestyle="--", alpha=0.8)
for i in [-1,1]:
plt.axvline(clicks_mean + i*clicks_std, color=plot_color, linestyle=":", alpha=0.6)
if self.corrected_clicks is not None:
kernel = stats.gaussian_kde(self.corrected_clicks)
res = 10
plot_color = self.plot_colors[plot_num]
plot_label = "speed_adj (" + str(len(self.corrected_clicks)) + " points)"
plt.plot(np.arange(80 * res) / res, kernel(np.arange(80 * res) / res - 40), linestyle="--", color="0000", linewidth=2, label=plot_label)
# ax.bar(np.arange(self.kde_list.size), self.kde_list, fill=False, edgecolor='black', label="KDE")
ax.legend()
ax.set_xlim(0,80)
plt.show()
def plot_phrase_stats(self):
ind_var_name = "session"
for data_label in ['error_tot', 'error_tot', 'clicks_char', 'clicks_word', 'words_min', 'chars_min']:
dep_var_name = data_label
# if data_label == 'words_min':
# dep_var_name = "Words per Minute"
# elif data_label == 'chars_min':
# dep_var_name = "Characters per Minute"
# elif data_label == 'clicks_char':
# dep_var_name = "Presses per Character"
# elif data_label == 'clicks_word':
# dep_var_name = "Presses per Word"
# elif data_label == 'error_tot':
# dep_var_name = "Error Rate (Errors/Selection)"
# else:
# raise ValueError("Data Attribute Unknown: " + data_label)
DF = self.DF
pd.set_option('display.max_columns', 500)
fig, ax = plt.subplots()
fig.set_size_inches(10, 8)
sns.set(font_scale=1.5, rc={"lines.linewidth": 3})
sns.set_style({'font.serif': 'Helvetica'})
sns.lineplot(x=ind_var_name, y=dep_var_name,
data=DF, ci="sd", ax=ax)
# sns.lineplot(x=ind_var_name, y=dep_var_name,
# data=DF, ax=ax)
# sns.lineplot(x=ind_var_name, y=dep_var_name, hue="Adjusted Scanning Delay",
# palette=sns.cubehelix_palette(2, start=3, rot=0.2, dark=.2, light=.7, reverse=True),
# data=DF, ci="sd", ax=ax)
plt.title("Row Column Scanner: " + dep_var_name + " vs. " + ind_var_name)
sns.axes_style("darkgrid")
plt.show()
def print_stat_avg(self):
errors = []
clicks_char = []
clicks_word = []
words_min = []
chars_min = []
for phrase in self.phrases:
data_dict = self.phrase_stats[phrase]
errors.append(data_dict["error_tot"])
clicks_char.append(data_dict["clicks_char"])
clicks_word.append(data_dict["clicks_word"])
words_min.append(data_dict["words_min"])
chars_min.append(data_dict["chars_min"])
print("\n")
print("Characters/Min: ", np.average(chars_min), "+/-", np.std(chars_min))
print("Words/Min: ", np.average(words_min), "+/-", np.std(words_min))
print("Clicks/Character: ", np.average(clicks_char), "+/-", np.std(clicks_char))
print("Clicks/Word: ", np.average(clicks_word), "+/-", np.std(clicks_word))
print("Error Rate (%): ", np.average(errors) * 100, "+/-", np.std(errors) * 100)
def gen_kernel(self):
kernel = stats.gaussian_kde(self.rel_click_data)
kernel_handle = PickleUtil("resources\\kde_kernel.p")
kernel_handle.safe_save(kernel)
class Keyboard(MainWindow):
def __init__(self, screen_res, app):
super(Keyboard, self).__init__(screen_res)
self.app = app
self.font_scale = 1
self.key_chars = kconfig.key_chars
self.key_chars_sorted = kconfig.key_chars_sorted
self.key_config = "sorted"
# self.key_config = "default"
self.num_words = 7
self.words_first = True
# self.words_first = False
self.sound_set = True
self.pause_set = True
self.lm_prefix = ""
self.left_context = ""
self.typed_versions = [""]
self.cwd = os.getcwd()
self.gen_data_handel()
self.up_handel = PickleUtil(
os.path.join(self.user_handel, 'user_preferences.p'))
user_preferences = self.up_handel.safe_load()
if user_preferences is None:
user_preferences = [
'sorted', config.default_rotate_ind, config.default_pause_ind,
True
]
self.up_handel.safe_save(user_preferences)
self.key_config, self.speed, self.pause_index, self.is_write_data = user_preferences
self.scanning_delay = config.period_li[self.speed]
self.extra_delay = config.pause_li[self.pause_index]
self.params_handle_dict = {
'speed': [],
'extra_delay': [],
'params': [],
'start': [],
'press': [],
'choice': []
}
self.num_presses = 0
self.last_press_time = time.time()
self.last_update_time = time.time()
self.next_frame_time = time.time()
self.params_handle_dict['params'].append(
[config.period_li[config.default_rotate_ind], config.theta0])
self.params_handle_dict['start'].append(time.time())
self.click_time_list = []
lm_path = os.path.join(os.path.join(self.cwd, 'resources'),
'lm_word_medium.kenlm')
vocab_path = os.path.join(os.path.join(self.cwd, 'resources'),
'vocab_100k')
self.lm = LanguageModel(lm_path, vocab_path)
self.phrase_prompts = False
if self.phrase_prompts:
self.phrases = Phrases("resources/comm2.dev")
else:
self.phrases = None
# determine keyboard positions
# set up file handle for printing useful stuff
# set up "typed" text
self.typed = ""
self.context = ""
self.old_context_li = [""]
self.last_add_li = [0]
# set up "talked" text
# check for speech
# talk_fid = open(self.talk_file, 'wb')
# write words
self.generate_layout()
self.draw_words()
self.wpm_data = []
self.decay_avg_wpm = 0
self.wpm_time = 0
self.error_data = []
self.decay_avg_error = 1
self.row_scan = True
self.row_scan_num = -2
self.col_scan = False
self.col_scan_num = -1
self.init_ui()
# animate
self.on_timer()
def gen_data_handel(self):
self.cwd = os.getcwd()
self.data_path = user_data_dir('data', 'RowCol')
if os.path.exists(self.data_path):
user_files = list(os.walk(self.data_path))
users = user_files[0][1]
else:
pathlib.Path(self.data_path).mkdir(parents=True, exist_ok=True)
# os.mkdir(data_path)
user_files = None
users = []
input_method = 'text'
if user_files is not None and len(users) != 0:
message = QtWidgets.QMessageBox(
QtWidgets.QMessageBox.Information, "Load User Data",
"You can either create a new user profile or "
"load an existing user profile.")
message.addButton(QtWidgets.QPushButton('Create New User'),
QtWidgets.QMessageBox.YesRole)
message.addButton(QtWidgets.QPushButton('Load Previous User'),
QtWidgets.QMessageBox.NoRole)
message.setDefaultButton(QtWidgets.QMessageBox.Yes)
response = message.exec_()
if response == 0:
input_method = 'text'
else:
input_method = 'list'
if input_method == 'text':
valid_user_id = False
input_text = "Please input a Number that will be used to save your user information"
while not valid_user_id:
num, ok = QtWidgets.QInputDialog.getInt(
self, "User ID Number Input", input_text)
if str(num) not in users:
valid_user_id = True
else:
input_text = "The user ID you inputed already exists! \n please input a valid user ID or press " \
"\"cancel\" to choose an existing one from a list"
if ok == 0:
input_method = 'list'
break
if input_method == 'text':
self.user_id = num
user_id_path = os.path.join(self.data_path, str(self.user_id))
os.mkdir(user_id_path)
if input_method == 'list':
item, ok = QtWidgets.QInputDialog.getItem(
self, "Select User ID", "List of save User IDs:", users, 0,
False)
self.user_id = item
self.user_handel = os.path.join(self.data_path, str(self.user_id))
user_id_files = list(os.walk(self.user_handel))
user_id_calibrations = user_id_files[0][1]
if len(user_id_calibrations) == 0:
self.data_handel = os.path.join(self.user_handel, 'cal0')
os.mkdir(self.data_handel)
user_id_cal_files = None
self.user_cal_num = 0
else:
user_id_cal_files = user_id_files[-1][2]
self.data_handel = user_id_files[-1][0]
self.user_cal_num = len(user_id_calibrations) - 1
if user_id_cal_files is not None:
self.use_num = sum([
1 if 'params_data' in file_name else 0
for file_name in user_id_cal_files
])
else:
self.use_num = 0
print(self.data_handel)
def generate_layout(self):
if self.key_config == "sorted":
self.keys_list = np.array(self.key_chars_sorted)
closest_square = int(
np.ceil(np.sqrt(len(self.keys_list) + self.num_words)))
self.key_freq_map = np.zeros((closest_square, closest_square))
for row_num, row in enumerate(self.key_freq_map):
for col_num, _ in enumerate(row):
self.key_freq_map[row_num][col_num] = row_num + col_num
else:
self.keys_list = np.array(self.key_chars)
closest_square = int(
np.ceil(np.sqrt(len(self.keys_list) + self.num_words)))
self.key_freq_map = np.zeros((closest_square, closest_square))
for row_num, row in enumerate(self.key_freq_map):
for col_num, _ in enumerate(row):
self.key_freq_map[row_num][
col_num] = row_num * closest_square + col_num
self.key_layout = np.empty((closest_square, closest_square), dtype=str)
for word in range(self.num_words): # fill words first
if self.words_first:
word_index = np.unravel_index(word,
(closest_square, closest_square))
else:
word_index = np.unravel_index(
closest_square**2 - self.num_words + word,
(closest_square, closest_square))
self.key_layout[word_index] = kconfig.word_char
self.key_freq_map[word_index] = float("inf")
sorted_indicies = []
for i in range(len(self.keys_list)):
arg_min_index = np.unravel_index(self.key_freq_map.argmin(),
self.key_freq_map.shape)
sorted_indicies += [arg_min_index]
self.key_freq_map[arg_min_index] = float("inf")
sorted_indicies.reverse()
for key in self.keys_list:
lowest_index = sorted_indicies.pop()
self.key_layout[lowest_index] = key
print(self.key_layout)
empty_row_counts = np.sum(np.where(self.key_layout != '', 1, 0),
axis=1).tolist()
if 0 in empty_row_counts:
empty_index = empty_row_counts.index(0)
self.key_layout = np.delete(self.key_layout, (empty_index), axis=0)
self.key_rows_num = len(self.key_layout)
self.key_cols_nums = np.sum(np.where(self.key_layout != '', 1, 0),
axis=1).tolist()
# shift empty cells
for row_num, row in enumerate(self.key_layout):
if "" in row:
row_list = row.tolist()
row_list.sort(key=lambda x: 2 if x == "" else 1)
self.key_layout[row_num] = np.array(row_list)
def keyPressEvent(self, e):
if e.key() == QtCore.Qt.Key_Space:
self.last_gap_time = time.time() - self.last_update_time
self.last_press_time = time.time()
self.save_click_time(self.last_press_time, self.last_gap_time,
(self.row_scan_num, self.col_scan_num))
self.on_press()
def change_speed(self, value):
old_rotate = config.period_li[self.speed]
self.speed = value
self.time_rotate = config.period_li[self.speed]
self.params_handle_dict['speed'].append(
[time.time(), old_rotate, self.scanning_delay])
def change_extra_delay(self, value):
old_pause_length = config.pause_li[self.pause_index]
self.pause_index = value
self.extra_delay = config.pause_li[self.pause_index]
self.params_handle_dict['extra_delay'].append(
[time.time(), old_pause_length, self.extra_delay])
def toggle_sound_button(self, value):
self.sound_set = value
self.mainWidget.sldLabel.setFocus()
def toggle_pause_button(self, value):
self.pause_set = value
self.mainWidget.sldLabel.setFocus()
def draw_words(self):
self.words_li = self.lm.get_words(self.left_context, self.lm_prefix,
self.num_words)
def on_timer(self):
if self.focusWidget() == self.mainWidget.text_box:
self.mainWidget.sldLabel.setFocus(
) # focus on not toggle-able widget to allow keypress event
cur_time = time.time()
if cur_time >= self.next_frame_time:
self.update_frame()
def update_frame(self):
self.last_update_time = time.time()
if self.row_scan:
self.row_scan_num += 1
if self.row_scan_num >= self.key_rows_num:
self.row_scan_num = 0
if self.row_scan_num == 0 and self.pause_set:
self.next_frame_time += (config.period_li[self.speed] +
self.extra_delay)
else:
self.next_frame_time += config.period_li[self.speed]
elif self.col_scan:
self.next_frame_time += config.period_li[self.speed]
self.col_scan_num += 1
if self.col_scan_num >= self.key_cols_nums[self.row_scan_num]:
self.col_scan_num = 0
self.mainWidget.highlight_grid()
def on_press(self):
if self.wpm_time == 0:
self.wpm_time = time.time()
if self.phrase_prompts:
self.mainWidget.speed_slider.setEnabled(False)
self.mainWidget.extra_delay_slider.setEnabled(False)
self.mainWidget.speed_slider_label.setStyleSheet(
'QLabel { color: grey }')
self.mainWidget.sldLabel.setStyleSheet('QLabel { color: grey }')
self.mainWidget.extra_delay_label.setStyleSheet(
'QLabel { color: grey }')
self.mainWidget.extra_sldLabel.setStyleSheet(
'QLabel { color: grey }')
if self.sound_set:
self.play()
if self.row_scan:
self.row_scan = False
self.col_scan = True
self.next_frame_time = time.time()
self.on_timer()
elif self.col_scan:
self.make_selection()
self.num_presses += 1
def save_click_time(self, last_press_time, last_gap_time, index):
self.params_handle_dict['press'].append([last_press_time])
self.click_time_list.append((last_gap_time, index))
def make_selection(self):
self.winner = self.key_layout[max(0, self.row_scan_num)][max(
0, self.col_scan_num)]
self.draw_typed()
self.draw_words()
self.mainWidget.update_grid()
print(self.winner)
self.col_scan = False
self.row_scan = True
self.row_scan_num = -1
self.col_scan_num = -1
self.next_frame_time = time.time()
self.on_timer()
def draw_typed(self):
if len(self.typed_versions) > 0:
previous_text = self.typed_versions[-1]
else:
previous_text = ""
if "_" in previous_text:
previous_text = previous_text.replace("_", " ")
if self.winner == kconfig.word_char:
new_text = self.mainWidget.labels_by_row[self.row_scan_num][
self.col_scan_num].text()[1:] + ' '
new_text = new_text[len(self.lm_prefix):]
else:
new_text = self.winner
if self.winner == kconfig.back_char:
# self.prefix = self.prefix[:-1]
if self.typed_versions[-1] != '':
self.typed_versions += [previous_text[:-1]]
input_text = "<span style='color:#000000;'>" + self.typed_versions[
-1] + "</span>"
self.mainWidget.text_box.setText(
"<span style='color:#000000;'>" + self.typed_versions[-1] +
"</span>")
else:
input_text = ""
elif self.winner == kconfig.mybad_char:
if len(self.typed_versions) > 1:
self.typed_versions = self.typed_versions[:-1]
input_text = "<span style='color:#000000;'>" + self.typed_versions[
-1] + "</span>"
self.mainWidget.text_box.setText(
"<span style='color:#000000;'>" + self.typed_versions[-1] +
"</span>")
else:
input_text = ""
elif self.winner == kconfig.clear_char:
if len(self.typed_versions) > 1:
self.typed_versions += [" "]
self.mainWidget.text_box.setText("")
input_text = ""
else:
self.typed_versions += [previous_text + new_text]
if new_text in kconfig.break_chars:
new_text = new_text + ' '
if previous_text[-1] == " ":
previous_text = previous_text[:-1]
if len(new_text) > 0 and new_text[-1] == " ":
new_text = new_text[:-1] + "_"
new_text = new_text[:-1] + "_"
input_text = "<span style='color:#000000;'>" + previous_text + "</span><span style='color:#0000dd;'>" \
+ new_text + "</span>"
self.mainWidget.text_box.setText(
"<span style='color:#000000;'>" + previous_text +
"</span><span style='color:#0000dd;'>" + new_text + "</span>")
self.mainWidget.text_box.update()
self.update_prefixes()
# write output
if self.is_write_data:
choice_dict = {
"time": time.time(),
"undo": self.winner == kconfig.mybad_char,
"backspace": self.winner == kconfig.back_char,
"typed": self.typed_versions[-1]
}
if self.phrase_prompts:
choice_dict["target"] = self.phrases.cur_phrase
self.params_handle_dict['choice'].append(choice_dict)
if self.phrase_prompts:
self.update_phrases(self.typed_versions[-1], input_text)
def text_stat_update(self, phrase, typed):
_, cur_error = calc_MSD(phrase, typed)
self.error_data = [cur_error] + self.error_data
decaying_weights = np.power(0.8, np.arange(len(self.error_data)))
decaying_weights /= np.sum(decaying_weights)
decay_avg_error = sum(np.array(self.error_data) * decaying_weights)
error_delta = decay_avg_error / (self.decay_avg_error + 0.000001)
self.decay_avg_error = decay_avg_error
self.wpm_data = [(len(typed.split(" ")) - 1) /
(time.time() - self.wpm_time) * 60] + self.wpm_data
self.wpm_time = 0
decaying_weights = np.power(0.8, np.arange(len(self.wpm_data)))
decaying_weights /= np.sum(decaying_weights)
decay_avg_wpm = sum(np.array(self.wpm_data) * decaying_weights)
wpm_delta = decay_avg_wpm / (self.decay_avg_wpm + 0.000001)
self.decay_avg_wpm = decay_avg_wpm
if error_delta > 1:
error_red = int(min(4, error_delta) * 63)
error_green = 0
else:
error_green = int(min(4, 1 / error_delta) * 63)
error_red = 0
if wpm_delta < 1:
wpm_red = int(min(4, wpm_delta) * 63)
wpm_green = 0
else:
wpm_green = int(min(4, 1 / wpm_delta) * 63)
wpm_red = 0
self.mainWidget.error_label.setStyleSheet("color: rgb(" +
str(error_red) + ", " +
str(error_green) + ", 0);")
self.mainWidget.wpm_label.setStyleSheet("color: rgb(" + str(wpm_red) +
", " + str(wpm_green) +
", 0);")
self.mainWidget.error_label.setText("Error Rate: " +
str(round(decay_avg_error, 2)))
self.mainWidget.wpm_label.setText("Words/Min: " +
str(round(decay_avg_wpm, 2)))
def reset_context(self):
self.left_context = ""
self.context = ""
self.typed = ""
self.lm_prefix = ""
def update_phrases(self, cur_text, input_text):
cur_phrase_typed, next_phrase = self.phrases.compare(cur_text)
cur_phrase_highlighted = self.phrases.highlight(cur_text)
if next_phrase:
self.text_stat_update(self.phrases.cur_phrase,
self.typed_versions[-1])
self.typed_versions = ['']
self.mainWidget.text_box.setText('')
self.mainWidget.speed_slider.setEnabled(True)
self.mainWidget.speed_slider_label.setStyleSheet(
'QLabel { color: blue }')
self.mainWidget.sldLabel.setStyleSheet('QLabel { color: blue }')
self.mainWidget.extra_delay_slider.setEnabled(True)
self.mainWidget.extra_delay_label.setStyleSheet(
'QLabel { color: blue }')
self.mainWidget.extra_sldLabel.setStyleSheet(
'QLabel { color: blue }')
self.clear_text = False
undo_text = 'Clear'
self.phrases.sample()
input_text = ""
cur_phrase_highlighted = self.phrases.highlight("")
self.reset_context()
if self.is_write_data:
choice_dict = {
"time": time.time(),
"undo": False,
"backspace": False,
"typed": "",
"target": self.phrases.cur_phrase
}
self.params_handle_dict['choice'].append(choice_dict)
self.mainWidget.text_box.setText("<p>" + cur_phrase_highlighted +
"<\p><p>" + input_text +
"</span><\p>")
def update_prefixes(self):
cur_text = self.typed_versions[-1]
for bc in kconfig.break_chars:
if bc in cur_text:
cur_text = cur_text.split(bc)[-1]
cur_text_words = cur_text.split(" ")
print(cur_text_words)
if cur_text_words[-1] == '' or cur_text_words[
-1] in kconfig.break_chars:
self.lm_prefix = ""
self.left_context = cur_text
else:
self.lm_prefix = cur_text_words[-1]
self.left_context = cur_text[:(-len(self.lm_prefix) - 1)]
def play(self):
sound_file = "icons/bell.wav"
QtMultimedia.QSound.play(sound_file)
def data_auto_save(self):
if len(self.click_time_list) > 0:
print("auto saving data")
self.save_data()
def closeEvent(self, event):
print("CLOSING THRU CLOSEEVENT")
self.quit(event)
# self.deleteLater()
def quit(self, event=None):
self.save_data()
self.close()
def save_data(self):
user_preferences = [
self.key_config, self.speed, self.pause_index, self.is_write_data
]
self.up_handel.safe_save(user_preferences)
self.click_data_path = os.path.join(
self.data_handel, 'click_time_log_' + str(self.use_num) + '.p')
self.params_data_path = os.path.join(
self.data_handel, 'params_data_use_num' + str(self.use_num) + '.p')
print(self.params_data_path)
PickleUtil(self.click_data_path).safe_save({
'user id': self.user_id,
'use_num': self.use_num,
'click time list': self.click_time_list,
'rotate index': self.speed
})
PickleUtil(self.params_data_path).safe_save(self.params_handle_dict)
class DataUtil:
def __init__(self, data_directory):
self.data_dir = data_directory
self.phrase_util = Phrases(
"resources/twitter-phrases/watch-combined.txt")
self.plot_colors = [
"#0000ff", "#4400ff", "#8800ff", "#cc00ff", "#ff00cc", "#ff0088",
"#ff0044"
]
self.click_data_files = []
self.click_context_files = []
self.preconfig_file = None
for path, dir, files in os.walk(self.data_dir):
for file in files:
if "click_time_log" in file:
self.click_data_files += [os.path.join(path, file)]
if "params_data_use_num" in file:
self.click_context_files += [os.path.join(path, file)]
if "preconfig" in file:
self.preconfig_file = os.path.join(path, file)
self.rel_click_data = []
self.abs_click_data = []
self.selection_data = []
self.speed_changes = []
self.kde_list = []
self.clicks_by_speed = {}
self.clicks_by_phrase = {}
self.phrase_stats = {}
self.corrected_clicks = None
def load_data(self):
# size_list = []
self.full_click_data = []
for data_file in self.click_data_files:
data_handel = PickleUtil(data_file)
click_dict = data_handel.safe_load()
# size_list += [len(click_dict["click time list"])]
self.full_click_data += click_dict["click time list"]
self.rel_click_data += [
click[0] for click in click_dict["click time list"]
]
self.rel_click_data = np.array(self.rel_click_data)
self.bin_click_data = np.array(
[click[1] for click in self.full_click_data])
# size_list_2 = []
for data_file in self.click_context_files:
data_handel = PickleUtil(data_file)
context_dict = data_handel.safe_load()
self.abs_click_data += context_dict["press"]
# size_list_2 += [len(flatten(context_dict["press"]))]
self.speed_changes += context_dict["speed"]
self.selection_data += context_dict["choice"]
self.abs_click_data = np.array(flatten(self.abs_click_data))
self.speed_changes.sort(key=lambda x: x[0])
if self.preconfig_file is not None:
preconfig_handel = PickleUtil(self.preconfig_file)
preconfig = preconfig_handel.safe_load()
self.kde_list = np.array(preconfig["li"]) / np.sum(preconfig["li"])
if len(self.abs_click_data) != len(self.rel_click_data):
raise ValueError(
"Click data length does not match context data length!")
print("Loaded " + str(len(self.abs_click_data)) + " clicks")
def plot_click_recovery(self):
click_locations = [click[1] for click in self.full_click_data]
# print(click_locations)
click_pairs = []
abs_click_pairs = []
index = 0
while index < len(click_locations):
cur_click = click_locations[index]
abs_cur_click = self.abs_click_data[index]
if cur_click[1] == -1: # row scan
if len(click_pairs) == 0:
click_pairs.append([cur_click])
abs_click_pairs.append([abs_cur_click])
elif len(click_pairs[-1]) == 1:
click_pairs = click_pairs[:-1]
abs_click_pairs = abs_click_pairs[:-1]
else:
click_pairs.append([cur_click])
abs_click_pairs.append([abs_cur_click])
else:
if len(click_pairs[-1]) == 1:
click_pairs[-1].append(cur_click)
abs_click_pairs[-1].append(abs_cur_click)
index += 1
click_diffs = np.array(
[pair[1] - pair[0] for pair in abs_click_pairs if len(pair) == 2])
pair_orders = np.array(
[pair[1][1] for pair in click_pairs if len(pair) == 2])
recovery_times = click_diffs[np.where(pair_orders == 0)]
recovery_times = recovery_times[np.where(recovery_times <= 3)]
plt.hist(recovery_times, bins=20)
PickleUtil("recovery_time_951.p").safe_save(recovery_times)
# plt.show()
# plt.hist(pair_orders, bins=8)
plt.xlabel("time (s)")
plt.ylabel("count")
plt.title("Recovery Time")
plt.show()
def split_data_phrase(self):
self.phrases = []
self.phrase_times = {}
phrase_start = 0
uncorrected_error = 0
corrected_error = 0
session_num = 0
start_time_prev = 0
phrase_selections = 0
for selection_num, selection in enumerate(
self.selection_data
): # scan through data to find phrases & time ints
if "target" in selection:
phrase = selection["target"]
typed = selection["typed"]
is_backspace = selection["backspace"]
is_undo = selection["undo"]
if phrase not in self.phrases:
self.phrases.append(phrase)
phrase_start = selection["time"]
phrase_selections += 1
# if is_backspace: # accumulate corrected errors
# corrected_error += 1
if is_undo:
if selection_num > 0:
# prev_typed = self.selection_data[selection_num-1]["typed"]
# typed_difference = prev_typed[len(typed):]
corrected_error += 1
_, completed_phrase = self.phrase_util.compare(typed, phrase)
if completed_phrase and not is_undo:
self.phrase_times[phrase] = (phrase_start,
selection["time"])
uncorrected_error = calc_MSD(typed, phrase)[0] / len(
max(typed, phrase)) * 100
self.phrase_stats[phrase] = {
"error_unc": uncorrected_error,
"error_cor": corrected_error / phrase_selections * 100,
"selections": phrase_selections
}
corrected_error = 0
phrase_selections = 0
self.phrases = list(self.phrase_times.keys())
self.phrases.sort(key=lambda x: self.phrase_times[x][0])
print("Data partitioned into " + str(len(self.phrases)) +
" sets by phrase")
for phrase in self.phrases: # split data according to phrase times calculated above
phrase_start, phrase_end = self.phrase_times[phrase]
phrase_click_indices = np.where(
(self.abs_click_data >= phrase_start)
& (self.abs_click_data <= phrase_end))
phrase_abs_clicks = self.abs_click_data[phrase_click_indices]
phrase_rel_clicks = self.rel_click_data[phrase_click_indices]
phrase_bin_clicks = self.bin_click_data[phrase_click_indices]
if len(phrase_abs_clicks) > 0:
first_click_time = min(phrase_abs_clicks)
self.clicks_by_phrase[phrase] = {
"abs": phrase_abs_clicks,
"rel": phrase_rel_clicks,
"bin": phrase_bin_clicks
}
stat_dict = self.phrase_stats[phrase]
num_clicks = len(phrase_abs_clicks)
time_int = phrase_end - first_click_time
num_characters = len(phrase)
num_words = len(phrase.split(" "))
num_sel = stat_dict["selections"]
clicks_per_char = num_clicks / num_characters
clicks_per_word = num_clicks / num_words
clicks_per_sel = num_sel / num_words
chars_per_min = num_characters / time_int * 60
words_per_min = num_words / time_int * 60
sel_per_min = num_sel / time_int * 60
click_dist_mean = np.mean(phrase_bin_clicks)
click_dist_var = np.var(phrase_bin_clicks)
stat_dict["clicks_char"] = clicks_per_char
stat_dict["clicks_word"] = clicks_per_word
stat_dict["clicks_sel"] = clicks_per_sel
stat_dict["chars_min"] = chars_per_min
stat_dict["words_min"] = words_per_min
stat_dict["sel_min"] = sel_per_min
stat_dict["click_mean"] = click_dist_mean
stat_dict["click_var"] = click_dist_var
if phrase_start - start_time_prev > 19 * 60:
session_num += 1
start_time_prev = phrase_start
stat_dict["session"] = session_num
stat_dict["start_time"] = phrase_start
else:
del self.phrase_stats[phrase]
self.phrases = list(self.phrase_stats.keys())
# print(self.phrase_stats)
def split_data_speed(self):
num_speed_changes = len(self.speed_changes)
for change_index in range(num_speed_changes):
time_min = self.speed_changes[change_index][0]
if change_index == num_speed_changes - 1:
time_max = float("inf")
else:
time_max = self.speed_changes[change_index + 1][0]
clicks_int = np.array(
np.where((self.abs_click_data > time_min)
& (self.abs_click_data < time_max))[0],
dtype='int64')
matrix_index = np.vectorize(
lambda m_index: self.rel_click_data[m_index])
if clicks_int.size > 0:
clock_speed = round(self.speed_changes[change_index][2], 2)
clicks_rel_int = matrix_index(clicks_int)
if clock_speed in self.clicks_by_speed.keys():
prev_clicks = self.clicks_by_speed[clock_speed]
self.clicks_by_speed[clock_speed] = np.concatenate(
(prev_clicks, clicks_rel_int))
else:
self.clicks_by_speed[clock_speed] = clicks_rel_int
print("Data partitioned into " +
str(len(self.clicks_by_speed.keys())) +
" sets by clock rotation speed")
def correct_data_speed(self):
if 1.44 in self.clicks_by_speed.keys():
base_index = 1.44
elif round(period_li[default_rotate_ind],
2) in self.clicks_by_speed.keys():
base_index = round(period_li[default_rotate_ind], 2)
else:
base_index = max(self.clicks_by_speed.keys())
base_clicks = self.clicks_by_speed[base_index]
base_clicks_mean = np.mean(base_clicks)
# base_clicks = base_clicks - base_clicks_mean
base_clicks_std = np.std(base_clicks)
clock_speeds = list(self.clicks_by_speed.keys())
self.corrected_clicks = []
for clock_speed in clock_speeds:
clicks = self.clicks_by_speed[clock_speed]
clicks_mean = np.mean(clicks)
clicks = clicks - clicks_mean
clicks_std = np.std(clicks)
clicks *= base_clicks_std / clicks_std
clicks += base_clicks_mean
self.corrected_clicks += clicks.tolist()
self.corrected_clicks = np.array(self.corrected_clicks)
def make_data_frame(self):
for phrase_num, phrase in enumerate(self.phrases):
if phrase_num == 0:
DF = pd.DataFrame(pd.DataFrame([self.phrase_stats[phrase]]))
DF["phrase"] = phrase
else:
df = pd.DataFrame([self.phrase_stats[phrase]])
df["phrase"] = phrase
DF = DF.append(df, ignore_index=True)
# DF = DF.sor _values(by=['start_time'])
self.DF = DF
def save_hist(self):
# kernel = stats.gaussian_kde(self.rel_click_data) # rel_click_data
hist = self.kde_list
time_rotate = self.speed_changes[-1][-1]
x_range = (np.arange(80) - 40) / 80 * time_rotate
print(os.path.join(self.data_dir, "user_histogram.p"))
PickleUtil(os.path.join(self.data_dir, "user_histogram.p")).safe_save(
(x_range, hist))
def plot_data(self):
fig = plt.figure()
ax = plt.subplot(111)
plot_num = 0
for clock_speed in self.clicks_by_speed.keys():
plot_color = self.plot_colors[plot_num]
clicks = self.clicks_by_speed[clock_speed]
clicks_mean = np.mean(clicks)
clicks_std = np.std(clicks)
# clicks = clicks - clicks_mean
plot_label = "rotation: " + str(clock_speed) + " (" + str(
len(clicks)) + " points)"
# ax.hist(clicks, 30, range=[0, 80], density=True, color=plot_color, alpha=0.3, label=plot_label)
kernel = stats.gaussian_kde(clicks)
res = 80
plt.plot(np.arange(2.5 * res) / res,
kernel(np.arange(2.5 * res) / res),
color=plot_color,
linewidth=2,
label=plot_label)
plot_num += 1
# plt.axvline(clicks_mean, color=plot_color, linestyle="--", alpha=0.8)
# for i in [-1,1]:
# plt.axvline(clicks_mean + i*clicks_std, color=plot_color, linestyle=":", alpha=0.6)
if self.corrected_clicks is not None:
kernel = stats.gaussian_kde(self.corrected_clicks)
res = 80
plot_label = "rotation_adj (" + str(len(
self.corrected_clicks)) + " points)"
plt.plot(np.arange(2.5 * res) / res,
kernel(np.arange(2.5 * res) / res),
linestyle="--",
color="0000",
linewidth=2,
label=plot_label)
# ax.bar(np.arange(self.kde_list.size), self.kde_list, fill=False, edgecolor='black', label="KDE")
ax.legend()
# ax.set_xlim(20,60)
plt.show()
def plot_phrase_stats(self, DF2=None):
ind_var_name = "session"
for data_label in [
'error_unc', 'error_unc', 'error_cor', 'clicks_char',
'chars_min'
]:
var_name_dict = {
'sel_min': "Selections/Min",
'words_min': "Words/Min",
'chars_min': "Characters/Min",
'clicks_char': "Clicks/Character",
'clicks_word': "Clicks/Word",
'clicks_sel': "Clicks/Selection",
'error_unc': "Uncorrected Error Rate (%)",
'error_cor': "Corrections/Selections (%)"
}
dep_var_name = var_name_dict[data_label]
DF = self.DF
pd.set_option('display.max_columns', 500)
fig, ax = plt.subplots()
fig.set_size_inches(10, 8)
sns.set(font_scale=1.4, rc={"lines.linewidth": 3})
sns.set_style({'font.serif': 'Helvetica'})
if DF2 is not None:
sns.lineplot(x=ind_var_name,
y=data_label,
color="rosybrown",
data=DF2,
ci="sd",
ax=ax)
sns.lineplot(x=ind_var_name,
y=data_label,
color="brown",
data=DF2,
ax=ax)
sns.lineplot(x=ind_var_name,
y=data_label,
color="cadetblue",
data=DF,
ci="sd",
ax=ax)
sns.lineplot(x=ind_var_name,
y=data_label,
color="darkslategrey",
data=DF,
ax=ax)
ax.set(xlabel=ind_var_name, ylabel=dep_var_name)
plt.legend(title='Key:',
labels=[
'Row Col (SD)', ' (95% CI)', 'Nomon (SD)',
' (95% CI)'
])
plt.title("Webcam Double Switch: " + dep_var_name + " vs. " +
ind_var_name)
sns.axes_style("darkgrid")
plt.show()
def plot_click_dist_phrase(self):
DF = self.DF
fig, ax = plt.subplots()
fig.set_size_inches(10, 8)
sns.set(font_scale=1.4, rc={"lines.linewidth": 3})
sns.set_style({'font.serif': 'Helvetica'})
sns.lineplot(x="session",
y="error_unc",
color="darkslategrey",
data=DF,
ax=ax)
plt.show()
fig, ax = plt.subplots()
fig.set_size_inches(10, 8)
sns.set(font_scale=1.4, rc={"lines.linewidth": 3})
sns.set_style({'font.serif': 'Helvetica'})
ind_var_name = "session"
sns.lineplot(x=ind_var_name,
y="click_var",
color="darkslateblue",
data=DF,
ci="sd",
ax=ax)
sns.lineplot(x=ind_var_name,
y="click_var",
color="slateblue",
data=DF,
ax=ax)
sns.lineplot(x=ind_var_name,
y="click_mean",
color="cadetblue",
data=DF,
ci="sd",
ax=ax)
sns.lineplot(x=ind_var_name,
y="click_mean",
color="darkslategrey",
data=DF,
ax=ax)
ax.set(xlabel=ind_var_name,
ylabel="Pressing Distribution Variance and Mean (Bins)")
plt.legend(title='Key:',
labels=[
'Click Dist Var (SD)', ' (95% CI)',
'Click Dist Mean (SD)', ' (95% CI)'
])
plt.title("Participant 951: Pressing Distribution vs. " + ind_var_name)
sns.axes_style("darkgrid")
plt.show()
def print_stat_avg(self):
errors = []
clicks_char = []
clicks_word = []
words_min = []
chars_min = []
for phrase in self.phrases:
data_dict = self.phrase_stats[phrase]
errors.append(data_dict["error_cor"])
clicks_char.append(data_dict["clicks_char"])
clicks_word.append(data_dict["clicks_word"])
words_min.append(data_dict["words_min"])
chars_min.append(data_dict["chars_min"])
print("\n")
print("Characters/Min: ", np.average(chars_min), "+/-",
np.std(chars_min))
print("Words/Min: ", np.average(words_min), "+/-",
np.std(words_min))
print("Clicks/Character: ", np.average(clicks_char), "+/-",
np.std(clicks_char))
print("Clicks/Word: ", np.average(clicks_word), "+/-",
np.std(clicks_word))
print("Correction Rate (%): ", np.average(errors), "+/-",
np.std(errors))
def test_significance(self, DF2=None):
var_name_dict = {
'words_min': "Words/Min: ",
'chars_min': "Characters/Min: ",
'clicks_char': "Clicks/Character: ",
'clicks_word': "Clicks/Word: ",
'error_unc': "Error Rate: ",
'sel_min': "Selections/Min",
'error_cor': "Correction Rate: "
}
if DF2 is None:
df_start = self.DF[self.DF["session"] == 1]
df_end = self.DF[self.DF["session"] == max(self.DF["session"])]
print("\n")
print("Difference in means from first and last session: \n")
for data_label in ['error_unc', 'clicks_char', 'chars_min']:
sub_data_start = df_start[data_label]
sub_data_end = df_end[data_label]
t_val, p_val = stats.ttest_ind(sub_data_start.values,
sub_data_end.values,
equal_var=False)
if p_val > 0.05:
print(var_name_dict[data_label] +
": Not statistically significant (p-value: " +
str(p_val) + ")")
else:
print(var_name_dict[data_label] +
": Statistically significant (p-value: " +
str(p_val) + ")")
else:
# df_self = self.DF[self.DF["session"].isin([max(self.DF["session"]), max(self.DF["session"])-1])]
# df_other = DF2[DF2["session"].isin([max(DF2["session"]), max(DF2["session"])-1])]
df_self = self.DF
df_other = DF2
print("\n")
print(
"Difference in means from Nomon and Row Col in last session: \n"
)
for data_label in ['error_cor', 'clicks_char', 'chars_min']:
sub_data_self = df_self[data_label]
sub_data_other = df_other[data_label]
t_val, p_val = stats.ttest_ind(sub_data_self.values,
sub_data_other.values,
equal_var=False)
# print(stats.f_oneway(sub_data_self.values, sub_data_other.values))
if p_val > 0.05:
print(var_name_dict[data_label] +
": Not statistically significant (p-value: " +
str(p_val) + ")")
else:
print(var_name_dict[data_label] +
": Statistically significant (p-value: " +
str(p_val) + ")")
class Keyboard(MainWindow):
def __init__(self, screen_res, app):
super(Keyboard, self).__init__(screen_res)
self.app = app
self.is_simulation = False
self.pretrain_window = False
# 2 is turn fully on, 1 is turn on but reduce, 0 is turn off
self.word_pred_on = 2
# Number of word clocks to display in case word prediction == 1 (reduced)
self.reduce_display = 5
# get user data before initialization
self.gen_data_handel()
self.up_handel = PickleUtil(
os.path.join(self.user_handel, 'user_preferences.p'))
user_preferences = self.up_handel.safe_load()
if user_preferences is None:
first_load = True
user_preferences = [
'default', 1, False, 'alpha', 'off', config.default_rotate_ind,
True
]
self.up_handel.safe_save(user_preferences)
else:
first_load = False
self.clock_type, self.font_scale, self.high_contrast, self.layout_preference, self.pf_preference, \
self.start_speed, self.is_write_data = user_preferences
self.phrase_prompts = False # set to true for data collection mode
if self.phrase_prompts:
self.phrases = Phrases(
"resources/twitter-phrases/watch-combined.txt")
else:
self.phrases = None
if self.layout_preference == 'alpha':
self.target_layout = kconfig.alpha_target_layout
self.key_chars = kconfig.key_chars
elif self.layout_preference == 'qwerty':
self.target_layout = kconfig.qwerty_target_layout
self.key_chars = kconfig.key_chars
elif self.layout_preference == 'emoji':
self.target_layout = kconfig.emoji_target_layout
self.key_chars = kconfig.emoji_keys
self.word_pred_on = 0
# set up dictionary tree
# splash = StartWindow(screen_res, True)
self.pause_animation = False
self.output_manager = outputManager()
self.is_output_text = False
self.lm_prefix = ""
self.left_context = ""
self.cwd = os.getcwd()
word_lm_path = os.path.join(
os.path.join(self.cwd, 'resources'),
'mix4_opt_min_lower_100k_4gram_2.5e-9_prob8_bo4_compress255.kenlm')
char_lm_path = os.path.join(
os.path.join(self.cwd, 'resources'),
'mix4_opt_min_lower_12gram_6e-9_prob9_bo4_compress255.kenlm')
vocab_path = os.path.join(os.path.join(self.cwd, 'resources'),
'vocab_lower_100k.txt')
char_path = os.path.join(os.path.join(self.cwd, 'resources'),
'char_set.txt')
self.lm = LanguageModel(word_lm_path, char_lm_path, vocab_path,
char_path)
# initialize pygame and joystick
if kconfig.target_evt is kconfig.joy_evt:
pygame.init()
if pygame.joystick.get_count() < 1:
# no joysticks found
print("Please connect a joystick.\n")
self.quit(None)
else:
# create a new joystick object from
# ---the first joystick in the list of joysticks
Joy0 = pygame.joystick.Joystick(0)
# tell pygame to record joystick events
Joy0.init()
# start looking for events
self.parent.after(0, self.find_events)
# not in selection pause
self.in_pause = False
# determine keyboard positions
self.init_locs()
# get old data if there is such
# Just for default. Loaded again when bc initializes
self.rotate_index = config.default_rotate_ind
# set up file handle for printing useful stuff
self.undefined = False
self.params_handle_dict = {
'speed': [],
'params': [],
'start': [],
'press': [],
'choice': []
}
self.num_presses = 0
self.params_handle_dict['params'].append(
[config.period_li[config.default_rotate_ind], config.theta0])
self.params_handle_dict['start'].append(time.time())
self.gen_scale()
self.pause_set = True
# set up "typed" text
self.typed = ""
self.btyped = ""
self.context = ""
self.old_context_li = [""]
self.last_add_li = [0]
# set up "talked" text
# self.talk_file = "talk.txt"
self.sound_set = True
self.press_lock = False
self.press_lock_status = False
# check for speech
# talk_fid = open(self.talk_file, 'wb')
# write words
self.init_words()
self.bars = kconfig.bars
self.bc_init = False
self.previous_undo_text = ''
self.previous_winner = 0
self.wpm_data = []
self.decay_avg_wpm = 0
self.wpm_time = 0
self.error_data = []
self.decay_avg_error = 1
self.clear_text = False
self.pretrain = False
self.emoji_box_highlight = [-1, -1]
self.init_ui()
self.mainWidget.clockgrid_widget.update_word_clocks(self.words_li)
sound_file = "icons/bell.wav"
self.sound_player = QtMultimedia.QSound(sound_file)
self.time_rotate = config.period_li[self.start_speed]
# get language model results
self.gen_word_prior(False)
self.clock_spaces = np.zeros((len(self.clock_centers), 2))
self.bc = BroderClocks(self)
self.mainWidget.change_value(self.start_speed)
self.bc.init_follow_up(self.word_score_prior)
# draw histogram
self.init_histogram()
self.save_environment()
self.consent = False
if first_load:
self.pretrain = True
self.welcome = Pretraining(self, screen_res)
# animate
# record to prevent double tap
self.last_key_press_time = time.time()
self.last_release_time = time.time()
self.update_radii = False
self.on_timer()
def gen_data_handel(self):
self.cwd = os.getcwd()
self.data_path = user_data_dir('data', 'Nomon')
if os.path.exists(self.data_path):
user_files = list(os.walk(self.data_path))
users = user_files[0][1]
else:
pathlib.Path(self.data_path).mkdir(parents=True, exist_ok=True)
user_files = None
users = []
input_method = 'text'
if user_files is not None and len(users) != 0:
message = QtWidgets.QMessageBox()
message.setText("You can either create a new user profile or "
"load an existing user profile.")
message.setWindowTitle("Load User")
message.addButton('Create New User', QtWidgets.QMessageBox.YesRole)
message.addButton('Load Previous User',
QtWidgets.QMessageBox.NoRole)
message.setDefaultButton(QtWidgets.QMessageBox.Yes)
response = message.exec_()
if response == 0:
input_method = 'text'
else:
input_method = 'list'
if input_method == 'text':
valid_user_id = False
input_text = "Please input a Number that will be used to save your user information"
while not valid_user_id:
num, ok = QtWidgets.QInputDialog.getInt(
self, "User ID Number Input", input_text)
if str(num) not in users:
valid_user_id = True
else:
input_text = "The user ID you inputed already exists! \n please input a valid user ID or press " \
"\"cancel\" to choose an existing one from a list"
if ok == 0:
input_method = 'list'
break
if input_method == 'text':
self.user_id = num
user_id_path = os.path.join(self.data_path, str(self.user_id))
os.mkdir(user_id_path)
if input_method == 'list':
item, ok = QtWidgets.QInputDialog.getItem(
self, "Select User ID", "List of save User IDs:", users, 0,
False)
self.user_id = item
self.user_handel = os.path.join(self.data_path, str(self.user_id))
user_id_files = list(os.walk(self.user_handel))
user_id_calibrations = user_id_files[0][1]
if len(user_id_calibrations) == 0:
self.data_handel = os.path.join(self.user_handel, 'cal0')
os.mkdir(self.data_handel)
user_id_cal_files = None
self.user_cal_num = 0
else:
user_id_cal_files = user_id_files[-1][2]
self.data_handel = user_id_files[-1][0]
self.user_cal_num = len(user_id_calibrations) - 1
if user_id_cal_files is not None:
self.use_num = sum([
1 if 'params_data' in file_name else 0
for file_name in user_id_cal_files
])
else:
self.use_num = 0
print(self.data_handel)
def save_environment(self):
self.rotate_index_past = self.rotate_index
self.window_size_past = self.frameGeometry()
self.clock_type_past = self.clock_type
self.layout_preference_past = self.layout_preference
self.high_contrast_past = self.high_contrast
def data_auto_save(self):
if len(self.bc.click_time_list) > 0:
print("auto saving data")
self.bc.save_when_quit()
def find_events(self):
# check everything in the queue of pygame events
events = pygame.event.get()
for event in events:
# event type for pressing any of the joystick buttons down
if event.type == pygame.JOYBUTTONDOWN:
# generate the event I've defined
self.canvas.event_generate(kconfig.joy_evt)
# return to check for more events in a moment
self.parent.after(20, self.find_events)
def keyPressEvent(self, e):
# if e.key() == QtCore.Qt.Key_Space:
# if self.press_lock:
# if not self.press_lock_status:
# self.on_press()
# self.press_lock_status = True
# else:
# self.on_press()
if e.key() == QtCore.Qt.Key_Control:
if self.press_lock:
if self.press_lock_status:
self.press_lock_status = False
if e.key() == QtCore.Qt.Key_Return:
if self.phrase_prompts:
if len(self.typed_versions) > 0:
self.update_phrases(self.typed_versions[-1],
"",
next_phrase=True)
else:
self.update_phrases("", "", next_phrase=True)
def init_locs(self):
# size of keyboard
key_chars = kconfig.key_chars
self.N_rows = len(key_chars)
self.N_keys_row = []
self.N_keys = 0
self.N_alpha_keys = 0
for row in range(self.N_rows):
n_keys = len(key_chars[row])
for col in range(n_keys):
if not isinstance(key_chars[row], list):
if (key_chars[row][col] in kconfig.main_chars
and (len(key_chars[row]) == 1)):
self.N_alpha_keys = self.N_alpha_keys + 1
elif ((key_chars[row] == kconfig.space_char)
and (len(key_chars[row]) == 1)):
self.N_alpha_keys = self.N_alpha_keys + 1
elif (key_chars[row] == kconfig.break_chars[1]
and (len(key_chars[row]) == 1)):
self.N_alpha_keys = self.N_alpha_keys + 1
self.N_keys_row.append(n_keys)
self.N_keys += n_keys
# print "NKEYS is " + str(self.N_keys)
# print "And N_alpha_keys is " + str(self.N_alpha_keys)
# width difference when include letter
word_clock_offset = 7 * kconfig.clock_rad
rect_offset = word_clock_offset - kconfig.clock_rad
word_offset = 8.5 * kconfig.clock_rad
rect_end = rect_offset + kconfig.word_w
# clock, key, word locations
self.clock_centers = []
self.win_diffs = []
self.word_locs = []
self.char_locs = []
self.rect_locs = []
self.keys_li = []
self.keys_ref = []
index = 0 # how far into the clock_centers matrix
word = 0 # word index
key = 0 # key index
# kconfig.N_pred = 2 # number of words per key
self.key_height = 6.5 * kconfig.clock_rad
self.w_canvas = 0
self.index_to_wk = [] # overall index to word or key index
for row in range(0, self.N_rows):
y = row * self.key_height
self.w_canvas = max(
self.w_canvas, self.N_keys_row[row] *
(6 * kconfig.clock_rad + kconfig.word_w))
for col in range(0, self.N_keys_row[row]):
x = col * (6 * kconfig.clock_rad + kconfig.word_w)
# predictive words
self.clock_centers.append(
[x + word_clock_offset, y + 1 * kconfig.clock_rad])
self.clock_centers.append(
[x + word_clock_offset, y + 3 * kconfig.clock_rad])
self.clock_centers.append(
[x + word_clock_offset, y + 5 * kconfig.clock_rad])
# win diffs
self.win_diffs.extend([
config.win_diff_base, config.win_diff_base,
config.win_diff_base
])
# word position
self.word_locs.append(
[x + word_offset, y + 1 * kconfig.clock_rad])
self.word_locs.append(
[x + word_offset, y + 3 * kconfig.clock_rad])
self.word_locs.append(
[x + word_offset, y + 5 * kconfig.clock_rad])
# rectangles
self.rect_locs.append([
x + rect_offset, y, x + rect_end, y + 2 * kconfig.clock_rad
])
self.rect_locs.append([
x + rect_offset, y + 2 * kconfig.clock_rad, x + rect_end,
y + 4 * kconfig.clock_rad
])
self.rect_locs.append([
x + rect_offset, y + 4 * kconfig.clock_rad, x + rect_end,
y + 6 * kconfig.clock_rad
])
# indices
self.index_to_wk.append(word)
self.index_to_wk.append(word + 1)
self.index_to_wk.append(word + 2)
index += 3
word += 3
## key character
# reference to index of key character
key_char = key_chars[row][col]
self.keys_li.append(key_chars[row][col])
self.keys_ref.append(index)
self.index_to_wk.append(key)
# key character position
self.char_locs.append(
[x + 2 * kconfig.clock_rad, y + 3 * kconfig.clock_rad])
# clock position for key character
self.clock_centers.append(
[x + 1 * kconfig.clock_rad, y + 3 * kconfig.clock_rad])
# rectangles
self.rect_locs.append(
[x, y, x + rect_offset, y + 6 * kconfig.clock_rad])
# win diffs
if (key_char == kconfig.mybad_char) or (
key_char == kconfig.yourbad_char) or (
key_char == kconfig.back_char
): # or (key_char == kconfig.break_char)
self.win_diffs.append(config.win_diff_high)
else:
self.win_diffs.append(config.win_diff_base)
index += 1
key += 1
def gen_scale(self):
scale_length = self.w_canvas / 2 # (len(kconfig.key_chars[0])-1)*kconfig.word_w
tick_int = int((len(config.period_li) - 1) * kconfig.word_pt * 3 /
(1.0 * scale_length)) + 1
self.time_rotate = config.period_li[self.rotate_index]
def toggle_pause_button(self, value):
self.pause_set = value
self.mainWidget.sldLabel.setFocus()
def toggle_sound_button(self, value):
self.sound_set = value
self.mainWidget.sldLabel.setFocus()
def toggle_talk_button(self, value):
self.talk_set = value
self.mainWidget.sldLabel.setFocus(
) # focus on not toggle-able widget to allow keypress event
def toggle_learn_button(self, value):
config.is_learning = value
self.mainWidget.sldLabel.setFocus(
) # focus on not toggle-able widget to allow keypress event
def change_speed(self, index):
# speed (as shown on scale)
speed_index = int(index)
# period (as stored in config.py)
self.rotate_index = speed_index
old_rotate = self.time_rotate
self.time_rotate = config.period_li[self.rotate_index]
self.bc.clock_inf.clock_util.change_period(self.time_rotate)
# note period change in log file
self.params_handle_dict['speed'].append(
[time.time(), old_rotate, self.time_rotate])
# update the histogram
self.draw_histogram()
def init_histogram(self):
# histogram
bars = self.bc.get_histogram()
self.bars = bars
# undo_text
self.undo_loc = [
(self.N_keys_row[self.N_rows - 1] - 1) *
(6 * kconfig.clock_rad + kconfig.word_w) - self.w_canvas / 2,
2 * kconfig.clock_rad
]
def draw_histogram(self, bars=None):
if bars == None:
bars = self.bc.get_histogram()
# bars = [1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
self.bars = bars
self.mainWidget.histogram.update()
def init_words(self):
if self.word_pred_on == 1:
num_words_total = 5
elif self.word_pred_on == 0:
num_words_total = 0
else:
num_words_total = kconfig.num_words_total
(self.words_li, self.word_freq_li,
self.key_freq_li) = self.lm.get_words(self.left_context,
self.context,
self.keys_li,
num_words_total=num_words_total)
if self.layout_preference == "emoji":
self.key_freq_li = np.array([
np.log(1 / len(self.key_chars))
for i in range(len(self.key_chars))
])
self.word_id = []
self.word_pair = []
word = 0
index = 0
windex = 0
self.words_on = []
self.words_off = []
self.word_list = []
# self.flag_args = []
temp_word_list = [
word_item for sublist in self.words_li for word_item in sublist
]
for word_item in temp_word_list:
if word_item != '':
self.word_list.append(word_item)
# print "TURNED ON AND WORD LIST IS" + str(self.word_list)
len_con = len(self.context)
for key in range(0, self.N_alpha_keys):
for pred in range(0, kconfig.N_pred):
word_str = self.words_li[key][pred]
len_word = len(word_str)
if (len_con > 1) and (len_word > kconfig.max_chars_display):
word_str = "+" + word_str[len_con:len_word]
self.word_pair.append((key, pred))
if word_str == '':
self.words_off.append(index)
else:
self.words_on.append(index)
windex += 1
word += 1
index += 1
self.words_on.append(index)
self.word_pair.append((key, ))
index += 1
for key in range(self.N_alpha_keys, self.N_keys):
for pred in range(0, kconfig.N_pred):
word_str = self.words_li[key][pred]
self.word_pair.append((key, pred))
self.words_off.append(index)
index += 1
self.words_on.append(index)
self.word_pair.append((key, ))
index += 1
self.typed_versions = ['']
def draw_words(self):
if self.word_pred_on == 1:
num_words_total = 5
elif self.word_pred_on == 0:
num_words_total = 0
else:
num_words_total = kconfig.num_words_total
(self.words_li, self.word_freq_li,
self.key_freq_li) = self.lm.get_words(self.left_context,
self.context,
self.keys_li,
num_words_total=num_words_total)
self.mainWidget.clockgrid_widget.update_word_clocks(self.words_li)
if self.layout_preference == "emoji":
self.key_freq_li = np.array([
np.log(1 / len(self.key_chars))
for i in range(len(self.key_chars))
])
word = 0
index = 0
self.words_on = []
self.words_off = []
self.word_list = []
temp_word_list = [
word_item for sublist in self.words_li for word_item in sublist
]
for word_item in temp_word_list:
if word_item != '' and all(c in "abcdefghijklmnopqrstuvwxyz\' "
for c in word_item):
self.word_list.append(word_item)
len_con = len(self.context)
windex = 0
for key in range(0, self.N_alpha_keys):
for pred in range(0, kconfig.N_pred):
word_str = self.words_li[key][pred]
len_word = len(word_str)
if len_con > 1 and len_word > kconfig.max_chars_display:
word_str = "+" + word_str[len_con:len_word]
if word_str == '':
self.words_off.append(index)
else:
self.words_on.append(index)
windex += 1
word += 1
index += 1
self.words_on.append(index)
self.word_pair.append((key, ))
index += 1
for key in range(self.N_alpha_keys, self.N_keys):
for pred in range(0, kconfig.N_pred):
self.word_pair.append((key, pred))
self.words_off.append(index)
index += 1
self.words_on.append(index)
self.word_pair.append((key, ))
index += 1
def gen_word_prior(self, is_undo):
self.word_score_prior = []
N_on = len(self.words_on)
if not is_undo:
for index in self.words_on:
pair = self.word_pair[index]
# word case
if len(pair) == 2:
(key, pred) = pair
prob = self.word_freq_li[key][pred] + np.log(
kconfig.rem_prob)
self.word_score_prior.append(prob)
else:
key = pair[0]
prob = self.key_freq_li[key]
prob = prob + np.log(kconfig.rem_prob)
if self.keys_li[key] == kconfig.mybad_char or self.keys_li[
key] == kconfig.yourbad_char:
prob = np.log(kconfig.undo_prob)
# if self.keys_li[key] in kconfig.break_chars:
# prob = np.log(kconfig.break_prob)
if self.keys_li[key] == kconfig.back_char:
prob = np.log(kconfig.back_prob)
if self.keys_li[key] == kconfig.clear_char:
prob = np.log(kconfig.undo_prob)
if self.keys_li[key] == kconfig.speak_char:
prob = np.log(kconfig.undo_prob)
self.word_score_prior.append(prob)
else:
for index in self.words_on:
pair = self.word_pair[index]
if len(pair) == 1:
key = pair[0]
if (self.keys_li[key] == kconfig.mybad_char) or (
self.keys_li[key] == kconfig.yourbad_char):
prob = kconfig.undo_prob
self.word_score_prior.append(np.log(prob))
else:
self.word_score_prior.append(-float("inf"))
else:
self.word_score_prior.append(-float("inf"))
def reset_context(self):
self.left_context = ""
self.context = ""
self.typed = ""
self.lm_prefix = ""
self.draw_words()
def update_phrases(self, cur_text, input_text, next_phrase=False):
cur_phrase_typed, _ = self.phrases.compare(cur_text)
cur_phrase_highlighted = self.phrases.highlight(cur_text)
if next_phrase:
if len(cur_text + input_text) > 0:
self.text_stat_update(self.phrases.cur_phrase,
self.typed_versions[-1])
self.typed_versions = ['']
self.mainWidget.text_box.setText('')
self.mainWidget.speed_slider.setEnabled(True)
self.mainWidget.speed_slider_label.setStyleSheet(
'QLabel { color: blue }')
self.mainWidget.sldLabel.setStyleSheet('QLabel { color: blue }')
self.clear_text = False
undo_text = 'Clear'
self.phrases.sample()
input_text = ""
cur_phrase_highlighted = self.phrases.highlight("")
self.reset_context()
if self.is_write_data:
choice_dict = {
"time": time.time(),
"undo": False,
"backspace": False,
"typed": "",
"target": self.phrases.cur_phrase
}
self.params_handle_dict['choice'].append(choice_dict)
self.mainWidget.text_box.setText("<p>" + cur_phrase_highlighted +
"<\p><p>" + input_text +
"</span><\p>")
if self.layout_preference == 'emoji':
if len(self.phrases.cur_phrase) > len(cur_text):
target_emoji = self.phrases.cur_phrase[len(cur_text)]
self.emoji_box_highlight = np.where(
np.array(self.target_layout[:-1]) == target_emoji)
else:
self.emoji_box_highlight = [-1, -1]
self.mainWidget.update()
# next_emoji_clock = self.mainWidget.clockgrid_widget.clocks[4*key_index+3]
def draw_typed(self):
new_text = ''
redraw_words = False
# phrases
delete = False
undo = False
undo_text = ""
if self.phrase_prompts:
previous_text = self.mainWidget.text_box.toPlainText().split(
"\n")[-1]
else:
previous_text = self.mainWidget.text_box.toPlainText()
if len(previous_text) > 0 and previous_text[-1] == "_":
previous_text = previous_text[:-1] + " "
if len(self.last_add_li) > 1:
last_add = self.last_add_li[-1]
if last_add > 0: # typed something
new_text = self.typed[-last_add:len(self.typed)]
undo_text = new_text
elif last_add == -1: # backspace
new_text = ''
delete = True
undo_text = kconfig.back_char
else:
new_text = ''
undo_text = new_text
if new_text in [". ", ", ", "? ", "! "]:
previous_text = previous_text[:-1]
redraw_words = True
index = self.previous_winner
if self.mainWidget.clockgrid_widget.clocks[index] != '':
if self.mainWidget.clockgrid_widget.clocks[index].text in [
kconfig.mybad_char, 'Undo'
]:
undo = True
delete = False
if self.typed_versions[-1] == '' and len(self.typed_versions) > 1:
undo_text = 'Clear'
if self.clear_text:
if self.is_output_text:
self.output_manager.remove_text(len(self.typed_versions[-1]))
self.typed_versions += ['']
input_text = ""
self.lm_prefix = ""
self.mainWidget.text_box.setText('')
self.clear_text = False
undo_text = 'Clear'
elif delete:
# self.prefix = self.prefix[:-1]
if self.typed_versions[-1] != '':
self.typed_versions += [previous_text[:-1]]
new_text = self.typed_versions[-1]
if len(new_text) > 0 and new_text[-1] == " ":
new_text = new_text[:-1] + "_"
input_text = "<span style='color:#000000;'>" + new_text + "</span>"
self.mainWidget.text_box.setText(
"<span style='color:#000000;'>" + new_text + "</span>")
if self.is_output_text:
self.output_manager.type_text(kconfig.back_char)
else:
input_text = ""
elif undo:
if len(self.typed_versions) > 1:
if self.is_output_text:
cur_length = len(self.typed_versions[-1])
prev_length = len(self.typed_versions[-2])
if prev_length > cur_length:
text_diff_length = prev_length - cur_length
text_diff = self.typed_versions[-2][-text_diff_length:]
self.output_manager.type_text(text_diff)
else:
if cur_length >= 2 and self.typed_versions[-1][
-2:] in ["? ", "! ", ", ", ". "]:
self.output_manager.type_text(kconfig.back_char +
kconfig.back_char +
" ")
else:
self.output_manager.remove_text(cur_length -
prev_length)
self.typed_versions = self.typed_versions[:-1]
new_text = self.typed_versions[-1]
if len(new_text) > 0 and new_text[-1] == " ":
new_text = new_text[:-1] + "_"
input_text = "<span style='color:#000000;'>" + new_text + "</span>"
self.mainWidget.text_box.setText(
"<span style='color:#000000;'>" + new_text + "</span>")
else:
input_text = ""
else:
self.typed_versions += [previous_text + new_text]
if self.is_output_text:
text_to_output = new_text
for break_char in kconfig.break_chars:
text_to_output = text_to_output.replace(
break_char + " ", kconfig.back_char + break_char + " ")
self.output_manager.type_text(text_to_output)
if len(new_text) > 0 and new_text[-1] == " ":
new_text = new_text[:-1] + "_"
input_text = "<span style='color:#000000;'>" + previous_text + "</span><span style='color:#0000dd;'>"\
+ new_text + "</span>"
self.mainWidget.text_box.setText(
"<span style='color:#000000;'>" + previous_text +
"</span><span style='color:#0000dd;'>" + new_text + "</span>")
if undo_text == kconfig.mybad_char:
undo_text = "Undo"
elif undo_text == kconfig.back_char:
undo_text = "Backspace"
elif undo_text == kconfig.clear_char:
undo_text = "Clear"
self.previous_undo_text = undo_text
self.mainWidget.undo_label.setText("<font color='green'>" + undo_text +
"</font>")
if self.phrase_prompts:
self.update_phrases(self.typed_versions[-1], input_text)
def text_stat_update(self, phrase, typed):
_, cur_error = calc_MSD(phrase, typed)
self.error_data = [cur_error] + self.error_data
decaying_weights = np.power(0.8, np.arange(len(self.error_data)))
decaying_weights /= np.sum(decaying_weights)
decay_avg_error = sum(np.array(self.error_data) * decaying_weights)
error_delta = decay_avg_error / (self.decay_avg_error + 0.000001)
self.decay_avg_error = decay_avg_error
self.wpm_data = [(len(typed.split(" ")) - 1) /
(time.time() - self.wpm_time) * 60] + self.wpm_data
self.wpm_time = 0
decaying_weights = np.power(0.8, np.arange(len(self.wpm_data)))
decaying_weights /= np.sum(decaying_weights)
decay_avg_wpm = sum(np.array(self.wpm_data) * decaying_weights)
wpm_delta = decay_avg_wpm / (self.decay_avg_wpm + 0.000001)
self.decay_avg_wpm = decay_avg_wpm
if error_delta > 1:
error_red = int(min(4, error_delta) * 63)
error_green = 0
else:
error_green = int(min(4, 1 / error_delta) * 63)
error_red = 0
if wpm_delta < 1:
wpm_red = int(min(4, wpm_delta) * 63)
wpm_green = 0
else:
wpm_green = int(min(4, 1 / wpm_delta) * 63)
wpm_red = 0
self.mainWidget.error_label.setStyleSheet("color: rgb(" +
str(error_red) + ", " +
str(error_green) + ", 0);")
self.mainWidget.wpm_label.setStyleSheet("color: rgb(" + str(wpm_red) +
", " + str(wpm_green) +
", 0);")
self.mainWidget.error_label.setText("Error Rate: " +
str(round(decay_avg_error, 2)))
self.mainWidget.wpm_label.setText("Words/Min: " +
str(round(decay_avg_wpm, 2)))
def on_pause(self):
self.mainWidget.pause_timer.start(kconfig.pause_length)
self.in_pause = True
self.mainWidget.keygrid_widget.highlight = True
self.mainWidget.keygrid_widget.update()
def end_pause(self):
self.mainWidget.pause_timer.stop()
self.in_pause = False
self.mainWidget.keygrid_widget.highlight = False
self.mainWidget.keygrid_widget.update()
self.on_timer()
def on_timer(self):
if self.focusWidget() == self.mainWidget.text_box:
self.mainWidget.sldLabel.setFocus(
) # focus on not toggle-able widget to allow keypress event
if self.bc_init:
self.bc.clock_inf.clock_util.increment(self.words_on)
is_input = win32api.GetAsyncKeyState(win32con.VK_SPACE)
if is_input:
if not self.output_manager.ignore_keyevent:
win32api.keybd_event(0x20, 0, win32con.KEYEVENTF_KEYUP, 0)
if self.is_output_text:
win32api.keybd_event(0x08, 0, 0, 0)
win32api.keybd_event(0x08, 0, win32con.KEYEVENTF_KEYUP, 0)
if not self.pretrain:
self.on_press()
else:
self.output_manager.ignore_keyevent = False
def on_press(self):
if time.time() - self.last_key_press_time > 1:
self.last_key_press_time = time.time()
# self.canvas.focus_set()
if self.wpm_time == 0:
self.wpm_time = time.time()
if self.sound_set:
self.play()
if self.phrase_prompts:
self.mainWidget.speed_slider.setEnabled(False)
self.mainWidget.speed_slider_label.setStyleSheet(
'QLabel { color: grey }')
self.mainWidget.sldLabel.setStyleSheet(
'QLabel { color: grey }')
if self.is_write_data:
self.num_presses += 1
self.bc.select()
def play(self):
self.sound_player.play()
def highlight_winner(self, index):
if self.mainWidget.clockgrid_widget.clocks[index] != '':
self.mainWidget.clockgrid_widget.clocks[index].selected = True
self.mainWidget.highlight_timer.start(kconfig.pause_length)
def end_highlight(self):
index = self.previous_winner
if self.mainWidget.clockgrid_widget.clocks[index] != '':
self.mainWidget.clockgrid_widget.clocks[index].selected = False
self.mainWidget.highlight_timer.stop()
def talk_winner(self, talk_string):
pass
def make_choice(self, index):
is_undo = False
is_equalize = False
is_backspace = False
# now pause (if desired)
if self.pause_set == 1:
self.on_pause()
self.mainWidget.pause_timer.start(kconfig.pause_length)
# highlight winner
self.previous_winner = index
self.highlight_winner(index)
# initialize talk string
talk_string = ""
# if selected a key
if (index - kconfig.N_pred) % (kconfig.N_pred + 1) == 0:
new_char = self.keys_li[self.index_to_wk[index]]
# special characters
if new_char == kconfig.speak_char:
new_char = ""
if len(self.typed_versions) > 0:
text_to_speak = self.typed_versions[-1]
text_to_speak = text_to_speak.replace("_", " ")
else:
text_to_speak = ""
self.output_manager.speak_text(text_to_speak)
self.old_context_li.append(self.context)
self.context = ""
self.last_add_li.append(0)
elif new_char == kconfig.space_char:
if len(self.context) > 1:
talk_string = self.context
else:
talk_string = "space"
new_char = ' '
self.old_context_li.append(self.context)
self.context = ""
self.last_add_li.append(1)
elif new_char == kconfig.mybad_char or new_char == kconfig.yourbad_char:
talk_string = new_char
# if added characters that turn
if len(self.last_add_li) > 1:
last_add = self.last_add_li.pop()
self.context = self.old_context_li.pop()
if last_add > 0: # if added text that turn
self.typed = self.typed[0:-last_add]
elif last_add == -1: # if backspaced that turn
letter = self.btyped[-1]
self.btyped = self.btyped[0:-1]
self.typed += letter
if new_char == kconfig.yourbad_char:
is_equalize = True
new_char = ''
is_undo = True
elif new_char == kconfig.back_char:
talk_string = new_char
is_backspace = True
# if delete the last character that turn
self.old_context_li.append(self.context)
print(self.context)
lt = len(self.typed)
if lt > 0: # typed anything yet?
self.btyped += self.typed[-1]
self.last_add_li.append(-1)
self.typed = self.typed[0:-1]
lt -= 1
if lt == 0:
self.context = ""
elif len(self.context) > 0:
self.context = self.context[0:-1]
elif not (self.typed[-1]).isalpha():
self.context = ""
else:
i = -1
while (i >= -lt) and (self.typed[i].isalpha()):
i -= 1
self.context = self.typed[i + 1:lt]
new_char = ''
elif new_char == kconfig.clear_char:
talk_string = 'clear'
new_char = '_'
self.old_context_li.append(self.context)
self.context = ""
self.last_add_li.append(1)
self.reset_context()
self.clear_text = True
elif new_char.isalpha(
) or new_char == "'" or new_char in kconfig.emoji_keys:
talk_string = new_char
self.old_context_li.append(self.context)
self.context += new_char
self.last_add_li.append(1)
if new_char in [".", ",", "?", "!"]:
talk_string = "Full stop"
self.old_context_li.append(self.context)
self.context = ""
self.typed += new_char
if " " + new_char in self.typed:
self.last_add_li.append(2)
self.typed = self.typed.replace(" " + new_char, new_char + " ")
else:
self.typed += new_char
# if selected a word
else:
key = self.index_to_wk[index] // kconfig.N_pred
pred = self.index_to_wk[index] % kconfig.N_pred
new_word = self.words_li[key][pred]
new_selection = new_word
length = len(self.context)
talk_string = new_word.rstrip(kconfig.space_char) # talk string
if length > 0:
self.typed = self.typed[0:-length]
self.typed += new_word
self.last_add_li.append(len(new_word) - len(self.context))
self.old_context_li.append(self.context)
self.context = ""
# update the screen
if self.context != "":
self.left_context = self.typed[:-len(self.context)]
else:
self.left_context = self.typed
# write output
if self.is_write_data:
choice_dict = {
"time": time.time(),
"undo": is_undo,
"backspace": is_backspace,
"typed": self.typed
}
if self.phrase_prompts:
choice_dict["target"] = self.phrases.cur_phrase
self.params_handle_dict['choice'].append(choice_dict)
self.draw_words()
self.draw_typed()
# update the word prior
self.gen_word_prior(is_undo)
# # talk the string
# if self.talk_set.get() == 1:
# self.talk_winner(talk_string)
return self.words_on, self.words_off, self.word_score_prior, is_undo, is_equalize
def present_choice(self):
self.draw_histogram()
# self.canvas.update_idletasks()
def closeEvent(self, event):
print("CLOSING THRU CLOSEEVENT")
self.quit(event)
# self.deleteLater()
def quit(self, event=None):
self.bc.quit_bc()
self.close()
def launch_help(self):
help_window = Pretraining(self,
self.mainWidget.screen_res,
help_screen=True)
help_window.show()
self.pretrain = True
self.mainWidget.repaint()
def launch_retrain(self):
retrain_window = Pretraining(self, self.mainWidget.screen_res)
retrain_window.screen_num = 3
retrain_window.next_screen()
retrain_window.show()
# retrain_window.setCentralWidget(retrain_window.mainWidget)
retrain_window.retrain = True
self.pretrain = True
self.mainWidget.update()
