def apply_to_sheet(self, worksheet: gspread.Worksheet):
"""
Add a row to the specified Google Sheet worksheet corresponding to
this update's new Slack message.
"""
user_name = slack_utils.get_user_display(self.user_id)
row_data = [
self.message_id,
user_name,
self.message,
self.timestamp.isoformat()
]
# Inserts row into the spreadsheet with an offset of 2
# (After row 1 (header row))
worksheet.insert_row(row_data, 2)
logger.info(f"Message with id={self.message} by {user_name} added")
def _ensure_sheet_formatting(worksheet: gspread.Worksheet):
"""
Ensure that the specified worksheet has the correct header row.
Overwrites the header row if a header mismatch is found.
"""
worksheet_headers = list(worksheet.row_values(1))
expected_headers = list(ColumnHeaders.__members__.keys())
# Check if the current_headers line up with the updated header structure
if worksheet_headers != expected_headers:
logger.warning("Prexisting table, with improper formatting: Fixing")
# TODO: move all data, not just headers
worksheet.delete_row(1)
worksheet.insert_row(expected_headers, 1)
else:
# TODO: ensure the below is still necessary
if worksheet.row_count in (0, 1):
worksheet.insert_row([], 1)
worksheet.insert_row(expected_headers, 1)
worksheet.delete_row(3)
worksheet.delete_row(1)
def process_datasets(config: SettingConfig, sheet: Worksheet) -> None:
print("Started running RASA")
# Compute and write the title of the spreadsheet based on the loaded configurations
spreadsheet_title = [config.identifier]
sheet.insert_row(spreadsheet_title, 1)
# For each scenario folder
spreadsheet_row_index = SPREADSHEET_START_VERTICAL_OFFSET
for file in os.listdir(config.datasets_path):
# Compute the path for the scenario folder
folder_path = os.path.join(config.datasets_path, file)
if os.path.isdir(folder_path):
# Break if the directory does not contain the splits
if len(os.listdir(folder_path)) <= 2:
print(
"Directory only contains train and test files, but no splits. Breaking."
)
break
# Compute the scenario file name
file_path = os.path.join(folder_path, file)
# Compute the reports path and create the directory
scenario_reports_path = os.path.join(folder_path,
'scenario_reports')
os.mkdir(scenario_reports_path)
# Compute the intent and slot reports paths and create them
intent_reports_path = os.path.join(scenario_reports_path,
'intent_reports')
slot_reports_path = os.path.join(scenario_reports_path,
'slot_reports')
os.mkdir(intent_reports_path)
os.mkdir(slot_reports_path)
scenario_slot_results = [f'Slot - {file}']
scenario_intent_results = [f'Intent - {file}']
for split_id in range(config.splits):
# Compute the identifier, get the train split and test split
identifier = f" {file}, split {split_id}"
train_split = file_path + "_train_" + f"{split_id}" + ".json"
test_split = file_path + "_test_" + f"{split_id}" + ".json"
# Run the subprocess for RASA training and testing, and wait for its completion
command = [
config.rasa_script_path, train_split, test_split,
config.rasa_config_path
]
subprocess.Popen(command, shell=True).wait()
# Process the slot and intent errors & reports and save their return values
intent_f1 = process_intent_result(identifier,
scenario_reports_path,
config)
slot_f1 = process_slot_result(identifier,
scenario_reports_path, config)
# Move the confusion matrix to the results path
copy_confusion_matrix(identifier, config)
scenario_slot_results.append(float("{:0.4f}".format(slot_f1)))
scenario_intent_results.append(
float("{:0.4f}".format(intent_f1)))
print(f"Finished processing split {identifier}")
# Append the mean value to each list for the scenario
scenario_intent_results.append(
float("{:0.4f}".format(mean(scenario_intent_results[1:]))))
scenario_slot_results.append(
float("{:0.4f}".format(mean(scenario_slot_results[1:]))))
# Append the standard deviation to each list for the scenario
scenario_intent_results.append(
float("{:0.3f}".format(
stdev(scenario_intent_results[1:len(scenario_intent_results
) - 2]))))
scenario_slot_results.append(
float("{:0.3f}".format(
stdev(scenario_slot_results[1:len(scenario_slot_results) -
2]))))
# Append the line in the google doc:
sheet.insert_row(scenario_slot_results, spreadsheet_row_index)
sheet.insert_row(scenario_intent_results, spreadsheet_row_index)
spreadsheet_row_index += 3