def create_individual_dict_from_csv(project_short_name = None, user_ids_to_include = None, user_ids_to_exclude = None):
""" Create an individual dictionary for each segment
Args:
project_short_name: Input the project short name
user_ids_to_include: Input the user ids to include
user_ids_to_exclude: Input the user ids to exclude, i.e. test user ids
Returns:
task_data dictionary keyed on task_id, row, col, values - link to image, answers for that particular segment or csv file (uncomment code)
"""
# pull data on tasks (images) and task_runs (classifications)
task_run_items = pd.read_csv(project_short_name + '_task_run.csv')
tasks = pd.read_csv(project_short_name + '_task.csv')
# create output dict
task_data = dict()
# create dictionary for skip data (if needed)
skips = dict()
# iterate through all tasks (images) - fastest way to create dictionaries for each image segment
for i, task in tasks.iterrows():
task_id = task['task__id'] # find image ID
#iterate through all sections of image
for row in range(6):
for col in range(6):
# call function from pybossautils for unique identifier for each segment
key = create_key(task_id, row, col)
# construct dictionary of segment information
# could be a dataframe
task_data[key] = {"task_id": task_id,
### Uncomment if doing matlab visual representation
# "img_num": id_only,
"row": row,
"col": col,
"link":task["taskinfo__url_b"],
'user_answers': dict()
}
# create space for skip button info, only needed if skip button present
# built as a separate dictionary to avoid storing unneeded data
skips[task_id] = {
'count_users':0,
'blank':{'id':[], 'reverse':[]},
'qual':{'id':[], 'reverse':[]},
'diff':{'id':[], 'reverse':[]},
}
# iterate through all task_runs (classifications) and add data to appropriate section of segment dictionary
for i, task_run in task_run_items.iterrows():
task_id = task_run['task_run__task_id']
# links task run to task info, not currently needed
# candidate_tasks = [t for j, t in tasks.iterrows() if unicode(t['task__id']) == unicode(task_id)]
# task = candidate_tasks[0]
# pull user ID and check if its in a list to include/exclude
user_id = task_run['task_run__user_id']
if (user_ids_to_include is None and user_ids_to_exclude is None) \
or (user_ids_to_include is not None and user_id in user_ids_to_include) \
or (user_ids_to_exclude is not None and user_id not in user_ids_to_exclude) \
: # If we passed in a list of user ids then check that this user_id is in it
# pull info, including answers, must be read through JSON enterpreter
task_answers = json.loads(task_run['task_run__info'])
# iterate through all segments in this task run and add answers
segments = task_answers['squares']
# determine if answers should be skipped due to user action in tasks with skip button
isSkipped = 0
if project_short_name in skip_versions:
skip_answers = task_answers['skipSlide']
# only skip if user skipped and did not change mind
if (skip_answers['reason']) and not (skip_answers['changeOfMind']):
isSkipped = 1
if isSkipped == 0:
for seg in segments:
row = seg['row']
col = seg["column"]
answer = seg['answer']
key = create_key(task_id, row, col)
# pull info from task dictionary
task_data_item = task_data.get(key)
# answers stored in full in nested dictionary
#user_answers:UID:answer
task_data_item['user_answers'][user_id] = {"answer": answer}
# if skip button included then extract skip usage info
if project_short_name in skip_versions:
# iterate through skips dictionary keys created above
# if key is part of the string that is the reason
# skip_answers = task_answers['skipSlide']
for i in skips[task_id].keys():
if skip_answers['reason'] != None: # reason is none when not skipped
if i in skip_answers['reason']:
skips[task_id][i]['id'].append(user_id) # create list of users who skip
if skip_answers['changeOfMind']:
skips[task_id][i]['reverse'].append(user_id) # create list of users who change minds
skips[task_id]['count_users']+=1 # counts how many users have seen the image
# print skip dictionary to main section dictionary if needed
if project_short_name in skip_versions:
for i, task in tasks.iterrows():
task_id = task['task__id']
for r in range(6):
for c in range(6):
key = create_key(task_id, r, c)
task_data[key]['skip_ans'] = skips[task_id]
### Return the task data
return task_data
tasks = pd.read_csv('tb2-r1.2_task.csv')
with open('results_mvp2_1.2.csv','wb') as f:
# create file structure and write protocol
fieldnamz = ['task_id', 'qual_skip', 'q_reverse', 'q_total', 'blank_skip', 'b_reverse', 'b_total', 'diff_skip', 'd_reverse', 'd_total', 'count']
id_write = csv.DictWriter(f, fieldnames=fieldnamz, lineterminator = '\n')
id_write.writeheader()
w = csv.writer(f)
# iterate through tasks
for i, task in tasks.iterrows():
task_id = task['task__id']
row = 0
col = 0
key = create_key(task_id, row, col)
# poor quality skip data
qual = task_data[key]['skip_ans']['qual']
# calculate number of poor quality skips where user didn't change mind
qual_diff = len(qual['id'])-len(qual['reverse'])
# same for blank skips
blank = task_data[key]['skip_ans']['blank']
blank_diff = len(blank['id'])-len(blank['reverse'])
# same for too dfficult skips
diff = task_data[key]['skip_ans']['diff']
diff_diff = len(diff['id'])-len(diff['reverse'])
# compile the above into a line and write to file
dat = [task['taskinfo__original_name'], len(qual['id']), len(qual['reverse']), qual_diff, len(blank['id']), len(blank['reverse']), blank_diff, len(diff['id']), len(diff['reverse']), diff_diff, task_data[key]['skip_ans']['count_users']]
w.writerow(dat)
