def test_registry():
def test_fn():
pass
r = Registry('test')
r.register(test_fn)
assert r['test_fn'] is not None
def setUp(self):
yield initDB(self)
self.user = yield User(first_name="First", last_name="Last", age=10).save()
self.avatar = yield Avatar(name="an avatar name", user_id=self.user.id).save()
self.picture = yield Picture(name="a pic", size=10, user_id=self.user.id).save()
self.favcolor = yield FavoriteColor(name="blue").save()
self.boy = yield Boy(name="Robert").save()
self.girl = yield Girl(name="Susan").save()
self.config = Registry.getConfig()
def setUp(self):
yield initDB(self)
self.user = yield User(first_name="First", last_name="Last",
age=10).save()
self.avatar = yield Avatar(name="an avatar name",
user_id=self.user.id).save()
self.picture = yield Picture(name="a pic",
size=10,
user_id=self.user.id).save()
self.favcolor = yield FavoriteColor(name="blue").save()
self.boy = yield Boy(name="Robert").save()
self.girl = yield Girl(name="Susan").save()
self.config = Registry.getConfig()
def setUp(self):
yield initDB(self)
self.config = Registry.getConfig()
import pandas as pd
from assets.mapping import colmap, local_identifiers
from .standardize import standardize
from .view import View
from db import query_db, get_db
from db.common import get_table_columns
from utils import Registry
from config import IDVIEWTYPE
import logging
preplogger = logging.getLogger(__name__)
# Creating simple config registry for function assignment
registry_idview = {}
view_registry = Registry(name='views')
## Prepare identifying information for matching ##
def create_data_view(tablename: str) -> pd.DataFrame:
raw_query = f"SELECT * FROM {tablename}"
dview = View(pd.read_sql_query(raw_query, get_db()), context={'partner': tablename.split('_')[0]})
preplogger.info(f'Data View created with statement:\n{raw_query}')
preplogger.info(f'Data View created with columns:\n{dview.columns}\nlength: {len(dview)}')
dview.standardized = standardize(dview.subset).data
return dview
################################################
class LogParser:
handler = Registry()
def _update_course(self, item):
self.course_name = (get_item(item, 'context.course_id').split(
':', 1)[-1] or self.course_name)
@handler.add(event_type=['load_video', 'edx.video.loaded'])
def _load_video(self, item):
self._update_course(item)
video_id = get_item(json.loads(get_item(item, 'event')), 'id')
page = get_item(item, 'page')
self.content.add_content('video', video_id)
self.modules.add_content(page, video_id)
@handler.add(event_type=['play_video', 'edx.video.played'])
def _play_video(self, item):
self._update_course(item)
user_id = get_item(item, 'context.user_id')
video_id = get_item(json.loads(get_item(item, 'event')), 'id')
self.users.view_content(user_id, video_id)
@handler.add(event_type='problem_check', event_source='server')
def _problem_check_server(self, item):
self._update_course(item)
(problem_id, user_id,
time) = get_items(item,
['event.problem_id', 'context.user_id', 'time'])
subtasks = get_item(item, 'event.submission', type_=dict)
for (subtask_id, subtask) in subtasks.items():
(question, task_type) = get_items(subtask,
['question', 'response_type'])
correct = get_item(subtask, 'correct', type_=bool)
self.tasks.add_task(problem_id, subtask_id, question, task_type)
self.users.score_task(user_id, problem_id, subtask_id, correct,
time)
@handler.add(event_type='edx.grades.problem.submitted')
def _problem_submitted(self, item):
self._update_course(item)
(user_id, problem_id, page, time) = get_items(
item, ['context.user_id', 'event.problem_id', 'referer', 'time'])
self.modules.add_task(page, problem_id)
self.users.post_solution(user_id, problem_id, convert_datetime(time))
@handler.add(event_type='openassessmentblock.create_submission')
def _create_submission(self, item):
self._update_course(item)
(submission_id, task_id, user_id, name,
page) = get_items(item, [
'event.submission_uuid', 'context.module.usage_key',
'context.user_id', 'context.module.display_name', 'referer'
])
self.users.create_submission(submission_id, user_id, task_id)
self.modules.add_task(page, task_id)
self.tasks.add_assessment(task_id, name)
@handler.add(event_type=[
'openassessmentblock.self_assess', 'openassessmentblock.peer_assess',
'openassessmentblock.staff_assess'
])
def _assess_submission(self, item):
self._update_course(item)
(submission_id,
user_id) = get_items(item,
['event.submission_uuid', 'context.user_id'])
scores = get_item(item, 'event.parts', type_=list)
points = sum(
get_item(score, 'option.points', type_=int) for score in scores)
max_points = sum(
get_item(score, 'criterion.points_possible', type_=int)
for score in scores)
self.users.assess(submission_id, user_id, points, max_points)
def __init__(self, log, course, answers, courses):
self.course_name = ''
self.users = Users()
self.tasks = Tasks()
self.modules = Modules()
self.content = Content()
self._parse(log)
for item in (self.users, self.tasks, self.modules, self.content):
item.update_data(course, answers)
self.course_long_name = courses[self.course_name]
def _parse(self, log):
for (i, line) in enumerate(log):
try:
item = json.loads(line.split(':', maxsplit=1)[-1])
LogParser.handler(self, item)
except Exception as e:
logging.warning('Error on process entry, line %d: %s', i, e)
def get_course_info(self):
return {
'short_name': self.course_name,
'long_name': self.course_long_name
}
def get_student_solutions(self, user_id=None):
if user_id is None:
for userid in self.users.submits:
yield self.get_student_solutions(userid)
else:
submits = self.users.submits[user_id]
for (taskid, tries) in submits.items():
for (time, correct) in tries:
yield (user_id, taskid, correct, time)
def get_student_content(self, user_id=None):
if user_id is None:
for userid in self.users.viewed_content:
yield self.get_student_content(userid)
else:
viewed = self.users.viewed_content[user_id]
for (_, content) in self.content.content.items():
for content_id in content:
if self.modules.get_content_module(content_id):
yield (user_id, content_id, int(content_id in viewed))
def get_assessments(self):
for submission_id in self.users.pr_submits:
(user_id, problem_id) = self.users.pr_submits[submission_id]
problem_id = get_id(problem_id)
assessments = self.users.assessments[submission_id]
for (reviewer, score, max_score) in assessments:
yield (user_id, problem_id, reviewer, score, max_score)
def get_tasks(self, task_id=None):
if task_id is None:
task_ids = set(self.tasks.tasks) | set(self.tasks.assessments)
for taskid in task_ids:
yield self.get_tasks(taskid)
else:
module = self.modules.get_task_module(task_id)
if not module:
return
if task_id in self.tasks.tasks:
for subtask in self.tasks.tasks[task_id]:
text = self.tasks.subtask_text.get(subtask) or 'NA'
yield (subtask, self.tasks.subtask_type[subtask], text,
module)
if task_id in self.tasks.assessments:
name = self.tasks.assessments[task_id] or 'NA'
yield (get_id(task_id), 'openassessment', name, module)
def get_content(self):
for (content_type, content) in self.content.content.items():
for content_id in content:
module = self.modules.get_content_module(content_id)
if module:
yield (content_id, content_type, 'NA', module)
from utils import Registry
DATASETS = Registry('dataset')
PIPELINES = Registry('pipeline')
DATA_LOADER = Registry('data_loader')
from utils import Registry
dataset_default_cfg = {'type': 'ModelNet'}
dataloader_default_cfg = {'type': 'DataListLoader'}
DATASETS = Registry(default_cfg=dataset_default_cfg)
DATALOADERS = Registry(default_cfg=dataloader_default_cfg)
from utils import Registry
DATASETS = Registry('dataset')
DATA_LOADER = Registry('data_loader')
from utils import Registry
DATASETS = Registry('dataset')
PIPELINES = Registry('pipeline')
from utils import Registry, OptimizerRegistry
from torch.optim import Adam, SGD
from torch.optim.lr_scheduler import StepLR, MultiStepLR, CosineAnnealingLR
optimizer_default_cfg = {'type': 'Adam', 'lr': 0.001}
lr_scheduler_default_cfg = None
OPTIMIZERS = OptimizerRegistry(default_cfg=optimizer_default_cfg)
OPTIMIZERS.register(Adam)
OPTIMIZERS.register(SGD)
LR_SCHEDULER = Registry(default_cfg=lr_scheduler_default_cfg)
LR_SCHEDULER.register(StepLR)
LR_SCHEDULER.register(CosineAnnealingLR)
LR_SCHEDULER.register(MultiStepLR)
from utils import Registry MODELS = Registry() ENCODERS = Registry() GENERATORS = Registry() MAPPINGS = Registry() DISCRIMINATORS = Registry()
def validate_registry(registry: Registry, mapping=colmap) -> bool:
for val in mapping.values():
assert val in registry.keys(
), f'{val} missing from registry {registry.name}'
# pandas standards.py
from utils import Registry
from assets.mapping import colmap
from mpi.prepare.view import View
from recordlinkage.preprocessing import clean
pandas_standards_registry = Registry('pandas_standards')
# Standardization Functions
def ssn_pool(view: View, colname='ssn_pool'):
def _test_ssn(x):
try:
tx = str(x)
assert len(tx) == 9
assert tx[0:3] != '000'
assert tx[0:3] != '666'
assert int(tx[0:3]) < 900
assert int(tx[3:5]) > 0
assert int(tx[-4:]) > 0
return int(x)
except:
return None
return view[colname].apply(_test_ssn)
pandas_standards_registry.register(ssn_pool)
# CMD
import torch
from utils import Registry
OPTIMIZER_REGISTRY = Registry("optimizer")
LR_SCHEDULER_REGISTRY = Registry("lr_scheduler")
def build_optimizer(cfg, model) -> torch.optim.Optimizer:
optimizer = OPTIMIZER_REGISTRY.get(cfg.schedule.optimizer.pop("name"))(model.parameters(), **cfg.schedule.optimizer)
return optimizer
def build_lr_scheduler(cfg, optimizer):
lr_scheduler = LR_SCHEDULER_REGISTRY.get(cfg.schedule.lr_scheduler.pop("name"))(optimizer, **cfg.schedule.lr_scheduler)
return lr_scheduler
from utils import Registry
model_default_cfg = {'type': 'PointNet2'}
MODELS = Registry(default_cfg=model_default_cfg)
# view_pandas.py
from utils import Registry, get_column_intersect
import pandas as pd
from assets.mapping import colmap, local_identifiers
df_registry = Registry('df_view')
## Access/View functions
def dfcolumns(data: pd.DataFrame) -> list:
return data.columns.tolist()
df_registry.register(dfcolumns, name='columns')
def dflen(data: pd.DataFrame) -> int:
return len(data)
df_registry.register(dflen, name='dlen')
def getitem(data: pd.DataFrame, idx) -> pd.Series:
return data[idx]
df_registry.register(getitem)
def head(data: pd.DataFrame, nrows:int) -> pd.DataFrame:
return data.head(nrows)
df_registry.register(head)
from utils import Registry
MODELS = Registry('model')
