def __regenerate_models():
users = db(db.user_model.id > 0).select()
for u in users:
u.last_cat = None
u.count_pair = 0
u.num_ideas = 0
u.count_transition_pairs = 0
u.transition_graph = '[]'
u.category_matrix = '{}'
u.update_record()
# Get all ideas
ideas = db(db.idea.id > 0).select()
# For each idea, get tags and update model
for i in ideas:
# Get tags
tags = db((db.tag_idea.idea == i.id)
& (db.tag.id == db.tag_idea.tag)).select(
db.tag.tag, groupby=db.tag_idea.id)
tags = [t.tag for t in tags]
# Get contextual info
problem_id = i.problem
user_id = i.userId
# Get model and update it
model = user_models.UserModel(user_id, problem_id)
model.update(tags)
def download_ideas():
__check_auth()
problem_id = long(request.vars['problem'])
fields = [
'user', 'condition', 'id', 'idea', 'timestamp', 'origin', 'sources',
'tags'
]
# Get records
rows = db(db.idea.problem == problem_id).select()
records = []
for r in rows:
model = user_models.UserModel(r.userId, problem_id)
tags = db((db.tag_idea.idea == r.id)
& (db.tag.id == db.tag_idea.tag)).select()
tags = [t.tag.tag for t in tags]
records.append([
r.userId,
model.user_condition, # condition
r.id,
r.idea,
r.dateAdded,
r.origin,
r.sources,
'|'.join(tags) # tags
])
# Return
filename = 'ideas_%d_%s.csv' % (problem_id, datetime.date.today())
return __prepare_csv_response(fields, records, filename, problem_id)
def get_inferred_categories(user_id, problem_id, db):
'''
I intended this class to not have to access DB at all, but right now I can't see a better way to do this.
'''
user_model = user_models.UserModel(user_id, problem_id)
all_users = db((db.idea.pool == True) & (db.idea.problem == problem_id)
& (db.idea.userId == db.user_info.id)
& (db.idea.userId <> user_id)).select(db.idea.userId,
db.user_info.userId,
distinct=True)
models = []
for u in all_users:
models.append(user_models.UserModel(u.idea.userId, problem_id))
nearest_neighbors = find_n_nearest(user_model, models, 5, db)
# Infer categories from nearest neighbors
inferred = infer_categories(user_model, nearest_neighbors, True)
return inferred
def usermodel():
__check_auth()
user_id = request.vars['user']
problem_id = request.vars['problem']
if not user_id or not problem_id:
redirect(URL('stats', 'index'))
# retrieve contextual info
user = db(db.user_info.id == user_id).select().first()
problem = db(db.problem.id == problem_id).select().first()
if not user or not problem:
redirect(URL('stats', 'index'))
# Retrieve user model
user_model = user_models.UserModel(user.id, problem.id)
# Get all user models in problem_id
all_users = __get_users_in_problem(problem_id, [user_id])
models = []
for u in all_users:
models.append(user_models.UserModel(u.idea.userId, problem_id))
nearest_neighbors = collab_filter.find_n_nearest(user_model, models, 5, db)
# Infer categories from nearest neighbors
inferred = collab_filter.infer_categories(user_model, nearest_neighbors,
True)
# Get logs
logs = db((db.action_log.problem == problem_id)
& (db.action_log.userId == user_id)).select()
return dict(user=user,
problem=problem,
model=user_model,
nearest_neighbors=nearest_neighbors,
inferred=inferred,
inferred_json=json.dumps(inferred),
logs=logs,
adjacent=', '.join(
__handle_unicode(
user_model.transition_graph.get_adjacent(
user_model.last_cat))),
inspiration_cats=', '.join(
__handle_unicode(
user_model.get_inspiration_categories(3))),
ordered_tags=', '.join(
__handle_unicode(user_model.get_ordered_tags())))
def __get_data(problem_id):
users = __get_users_in_problem(problem_id)
# Define fields
fields = [
'problem_id',
'user_id',
'public_id',
'initial_login',
'condition',
'num_ideas',
'num_refined',
'num_combined',
'num_inspirations',
'num_clicks_sp',
'breadth',
'depth_avg',
'depth_max',
'breadth_1',
'breadth_2',
'avg_click_i',
'sp_fulfill',
'insp_fulfill',
'category_switch_ratio',
'model_url',
]
# TODO get data
records = []
for u in users:
user = db(db.user_info.id == u.idea.userId).select().first()
public_id = user.userId
initial_login = user.initialLogin
model = user_models.UserModel(u.idea.userId, problem_id)
# Get the number of inspiration requests
num_inspirations = db(
(db.action_log.problem == problem_id)
& (db.action_log.userId == u.idea.userId)
& (db.action_log.actionName == 'get_available_tasks')).count()
# Get the number of clicks on a solution space cell
num_clicks_sp = db((db.action_log.problem == problem_id)
& (db.action_log.userId == u.idea.userId)
& (db.action_log.actionName == 'get_ideas')).select(
).find(lambda r: 'tags' in r.extraInfo)
# Get the number of refined and combined ideas
refined = db((db.idea.problem == problem_id)
& (db.idea.userId == u.idea.userId)
& (db.idea.origin == 'refinement')).count()
combined = db((db.idea.problem == problem_id)
& (db.idea.userId == u.idea.userId)
& (db.idea.origin == 'combine')).count()
breadth_1, breadth_2 = __get_breadth_per_half(user, problem_id)
# Get fulfillment and related metrics
fulfill = __get_fulfillment_metrics(user, problem_id)
# Create record
user_record = [
problem_id,
u.idea.userId,
public_id,
initial_login,
model.user_condition,
model.get_num_ideas(),
refined,
combined,
num_inspirations,
len(num_clicks_sp),
model.get_breadth(),
model.get_depth_avg(),
model.get_depth_max(),
breadth_1,
breadth_2,
fulfill['avg_click_i'],
fulfill['sp_fulfill'],
fulfill['insp_fulfill'],
model.category_switch_ratio,
'http://' + request.env.http_host +
URL('stats', 'usermodel?problem=%d&user=%d' %
(problem_id, u.idea.userId)),
]
# Add to array
records.append(user_record)
return fields, records
def __update_user_model(user_id, problem_id, tags, tag_names):
model = user_models.UserModel(user_id, problem_id)
tags_names = [tag_names[t] for t in tags]
model.update(tags_names)
def __get_tasks(user_id, problem_id):
# Get tags
model = user_models.UserModel(user_id, problem_id)
inspiration_categories, all_inferred = model.get_inspiration_categories(
NUM_TASKS)
# Remove duplicates
inspiration_categories = list(set(inspiration_categories))
inspiration_categories.extend(all_inferred)
# Get idea ids that the user already has used as inspiration
completed_tasks = db((db.task.completed_by == user_id)
& (db.task.task_type == 'RatingTask')
& (db.task.problem == problem_id)).select(
db.task.idea)
completed_tasks = [row.idea for row in completed_tasks]
if len(inspiration_categories) < NUM_TASKS:
# there were not enough inspiration categories. Randomily add more
tags = [
t.tag for t in db(db.tag.problem == problem_id).select(
orderby='<random>')
]
inspiration_categories.extend(tags)
tasks = []
for t in inspiration_categories:
try:
# Get tag id
t_id = db((db.tag.problem == problem_id)
& (db.tag.tag == t)).select(db.tag.id).first().id
# Get an idea id
if t_id:
idea_id = db((db.tag_idea.tag == t_id)
& (~db.tag_idea.idea.belongs(completed_tasks))
& (db.tag_idea.idea == db.idea.id)
& (db.idea.userId != user_id)
& (db.idea.pool == True)).select(
db.tag_idea.idea,
orderby='<random>').first().idea
if idea_id:
# Get a task for this idea and add it to list
task = db((db.task.idea == idea_id)
& (db.task.idea == db.idea.id)
& (db.task.idea == db.tag_idea.idea) &
(db.tag_idea.tag == db.tag.id)).select().first()
if task:
tasks.append(task)
except Exception:
pass
if len(tasks) == NUM_TASKS:
break
# Add favorites
favorites = __get_favorites(user_id)
for t in tasks:
if t.idea.id in favorites:
t.idea.favorite = True
else:
t.idea.favorite = False
# Filter to max number of tasks and return
return tasks
def get_versioning_structure():
'''
Return the tree structure for the versioning panel
Structure:
[
{
id:31,
type: 'combination',
tags: ['tag1', 'tag2'],
children: [
{
id: 01,
...,
children: ...
},
{
id: 91,
...,
children: ...
}
]
}
]
'''
problem_id = util.get_problem_id(request)
user_id = session.user_id
cache_type = 'versioning'
timestamp = datetime.datetime.min
complete_idea_map = dict()
filtered_idea_list = []
ids = []
# Retrieve latest cache
'''
TODO enable cache again
Cache is temporarily disabled due to a bug where refinements (and probably mergers) wouldn't be accurately reflected in the updated model.
E.g. if an idea was refined, the verisoning view would still show the pre-refined idea as the latest version, alongside with the refined idea.
The solution should involve updating the cached model when you detect a new idea that's the result of of merge or refinement
'''
cache = None # db((db.visualization_cache.problem == problem_id) & (db.visualization_cache.type == cache_type)).select().first()
if cache:
json_cache = json.loads(cache.cache, cls=ClassDecoder)
timestamp = cache.timestamp
complete_idea_map = json_cache['complete_idea_map']
filtered_idea_list = json_cache['filtered_idea_list']
ids = json_cache['ids']
# get all ideas
query = ((db.idea.id == db.tag_idea.idea) & (db.tag.id == db.tag_idea.tag)
& (db.idea.problem == problem_id) &
(db.idea.dateAdded > timestamp))
results = db(query).select(
orderby=db.idea.dateAdded,
groupby=db.idea.id) # Ordered from older to newer
# Get user model and ordered tag sequence
model = user_models.UserModel(user_id, problem_id)
ordered_tags = model.get_ordered_tags()
# build idea map
for i, r in enumerate(results):
tags = [t.tag.tag for t in r.idea.tag_idea.select()]
complete_idea_map[str(r.idea.id)] = Node(tags=tags,
type=r.idea.origin,
id=r.idea.id,
i=i)
children = []
if r.idea.sources:
children = [complete_idea_map[str(c)] for c in r.idea.sources]
# TODO update previous ideas to reflect their replacement
complete_idea_map[str(r.idea.id)].children = children
if not r.idea.replacedBy:
filtered_idea_list.append(complete_idea_map[str(r.idea.id)])
ids.append(r.idea.id)
# Trim nodes to avoid redundancy
for i in filtered_idea_list:
__trim_node(i, ids)
# Update cache
key = (db.visualization_cache.problem
== problem_id) & (db.visualization_cache.type == cache_type)
db.visualization_cache.update_or_insert(
key,
problem=problem_id,
type=cache_type,
cache=json.dumps(dict(complete_idea_map=complete_idea_map,
filtered_idea_list=filtered_idea_list,
ids=ids),
cls=ClassEncoder),
timestamp=datetime.datetime.now())
# sort ideas according to user model
filtered_idea_list.sort(key=lambda idea: (min(
[ordered_tags.index(t) for t in idea.tags]), idea.i))
# Log
log_action(user_id, problem_id, 'get_versioning_structure',
{'cache': (cache != None)})
# return
response.headers['Content-Type'] = 'application/json'
return json.dumps(filtered_idea_list, cls=ClassEncoder)
def get_solution_space():
user_id = session.user_id
problem_id = util.get_problem_id(request)
user_model = user_models.UserModel(user_id, problem_id)
connections = dict()
timestamp = datetime.datetime.min
max_n = 0
# Retrieve latest cache
# TODO Cache is temporarily disabled since I change the handling of an idea's pool flag to run the studies. That means each user will have their own cache.
cache = __get_cache(problem_id)
if not cache:
# If no cache, build it
cache = __build_cache(problem_id, user_model)
# Get info from cache
json_cache = json.loads(cache.cache)
connections = json_cache['connections']
all_tags = json_cache['tags']
timestamp = cache.timestamp
max_n = json_cache['max_n']
# Get tags ordered by the user model
tags = user_model.get_ordered_tags()
# get user's ideas with respective tags
ideas = db((db.idea.id == db.tag_idea.idea)
& (db.tag.id == db.tag_idea.tag)
& (db.idea.problem == problem_id)).select(orderby=~db.idea.id,
groupby=db.idea.id)
# extract tags
for idea in ideas:
idea_tags = list()
for tag in idea.idea.tag_idea.select():
tag = tag.tag.tag.lower()
idea_tags.append(tag)
all_tags.append(tag)
idea_tags.sort() # this contains a sorted array of tags for idea
# insert into data structure
key = '|'.join(idea_tags)
if key not in connections.keys():
connections[key] = dict(tags=idea_tags, n=0)
n = connections[key]['n'] + 1
connections[key]['n'] = n
if n > max_n:
max_n = n
# tags = tags[:SOLUTION_SPACE_MAX_TAGS]
# since another user may have added another tag, and since "tags" holds ALL tags, we need to remove those that are not in "ideas"
final_tags = tags # start with ordered array of tags
for t in all_tags: # Go over all tags
if t not in final_tags: # If not already in the array, add it
final_tags.append(t)
# Create minimap overview and generate outcome dict
overview = __generate_birdseye_solutionspace(final_tags,
connections,
max_n=max_n)
outcome = json.dumps(
dict(tags=final_tags,
connections=connections,
max_n=max_n,
overview=overview))
# Log
log_action(user_id, problem_id, 'get_solution_space', {
'cache': (cache != None),
'tags': final_tags
})
return outcome
def __update_user_model(user_id, problem_id, tags):
model = user_models.UserModel(user_id, problem_id)
model.update(tags)
