def timeseries_grade_percentiles(c, assignment_name, num_points=40):
"""
Returns a timeseries of grades with percentiles. Here is an example:
[["2015-07-17 19:00:36-0700", 0.0, 0.0, 0.0, ... 0.0, 0.0],
["2015-07-17 19:10:36-0700", 0.0, 0.0, 0.0, ... 1.0, 2.0],
["2015-07-17 19:20:36-0700", 0.0, 0.0, 0.0, ... 3.0, 4.0],
["2015-07-17 19:30:36-0700", 0.0, 0.0, 0.5, ... 5.0, 6.0],
["2015-07-17 19:40:36-0700", 0.0, 0.0, 1.0, ... 7.0, 8.0]]
"""
data_keys = range(0, 105, 5)
assignment = get_assignment_by_name(assignment_name)
if not assignment:
return
# There is a slight problem that because of DST, ordering by "started" may not always
# produce the correct result. When the timezone changes, lexicographical order does not
# match the actual order of the times. However, this only happens once a year in the middle
# of the night, so f**k it.
c.execute(
"""SELECT source, score, started FROM builds WHERE job = ? AND status = ?
ORDER BY started""",
[assignment_name, SUCCESS],
)
# XXX: There is no easy way to exclude builds started by staff ("super") groups.
# But because this graph is to show the general trend, it's usually fine if staff builds
# are included. Plus, the graph only shows up in the admin interface anyway.
builds = [(source, score, parse_time(started)) for source, score, started in c.fetchall()]
if not builds:
return []
source_set = map(lambda b: b[0], builds)
started_time_set = map(lambda b: b[2], builds)
min_started = min(started_time_set)
max_started = max(started_time_set)
assignment_min_started = parse_time(assignment.not_visible_before)
assignment_max_started = parse_time(assignment.due_date)
data_min = min(min_started, assignment_min_started)
data_max = max(max_started, assignment_max_started)
data_points = []
best_scores_so_far = {source: 0 for source in source_set}
time_delta = (data_max - data_min) / (num_points - 1)
current_time = data_min
for source, score, started_time in builds:
while current_time < started_time:
percentiles = np.percentile(best_scores_so_far.values(), data_keys)
data_points.append([format_js_compatible_time(current_time)] + list(percentiles))
current_time += time_delta
if score is not None:
best_scores_so_far[source] = max(score, best_scores_so_far[source])
percentiles = list(np.percentile(best_scores_so_far.values(), data_keys))
now_time = now()
while current_time - (time_delta / 2) < data_max:
data_points.append([format_js_compatible_time(current_time)] + percentiles)
if current_time >= now_time:
percentiles = [None] * len(percentiles)
current_time += time_delta
return data_points
def timeseries_grade_percentiles(c, assignment_name, num_points=40):
"""
Returns a timeseries of grades with percentiles. Here is an example:
[["2015-07-17 19:00:36-0700", 0.0, 0.0, 0.0, ... 0.0, 0.0],
["2015-07-17 19:10:36-0700", 0.0, 0.0, 0.0, ... 1.0, 2.0],
["2015-07-17 19:20:36-0700", 0.0, 0.0, 0.0, ... 3.0, 4.0],
["2015-07-17 19:30:36-0700", 0.0, 0.0, 0.5, ... 5.0, 6.0],
["2015-07-17 19:40:36-0700", 0.0, 0.0, 1.0, ... 7.0, 8.0]]
"""
data_keys = range(0, 105, 5)
assignment = get_assignment_by_name(assignment_name)
if not assignment:
return
# There is a slight problem that because of DST, ordering by "started" may not always
# produce the correct result. When the timezone changes, lexicographical order does not
# match the actual order of the times. However, this only happens once a year in the middle
# of the night, so f**k it.
c.execute(
'''SELECT source, score, started FROM builds WHERE job = ? AND status = ?
ORDER BY started''', [assignment_name, SUCCESS])
# XXX: There is no easy way to exclude builds started by staff ("super") groups.
# But because this graph is to show the general trend, it's usually fine if staff builds
# are included. Plus, the graph only shows up in the admin interface anyway.
builds = [(source, score, parse_time(started))
for source, score, started in c.fetchall()]
if not builds:
return []
source_set = tuple(map(lambda b: b[0], builds))
started_time_set = tuple(map(lambda b: b[2], builds))
min_started = min(started_time_set)
max_started = max(started_time_set)
assignment_min_started = parse_time(assignment.not_visible_before)
assignment_max_started = parse_time(assignment.due_date)
data_min = min(min_started, assignment_min_started)
data_max = max(max_started, assignment_max_started)
data_points = []
best_scores_so_far = {source: 0 for source in source_set}
time_delta = (data_max - data_min) / (num_points - 1)
current_time = data_min
for source, score, started_time in builds:
while current_time < started_time:
percentiles = np.percentile(tuple(best_scores_so_far.values()),
data_keys)
data_points.append([format_js_compatible_time(current_time)] +
list(percentiles))
current_time += time_delta
if score is not None:
best_scores_so_far[source] = max(score,
best_scores_so_far[source])
percentiles = list(
np.percentile(tuple(best_scores_so_far.values()), data_keys))
now_time = now()
while current_time - (time_delta / 2) < data_max:
data_points.append([format_js_compatible_time(current_time)] +
percentiles)
if current_time >= now_time:
percentiles = [None] * len(percentiles)
current_time += time_delta
return data_points
def test_time_functions(self):
timestamp_str = now_str()
timestamp_obj = parse_time(timestamp_str)
self.assertEqual(timestamp_str, format_time(timestamp_obj))
def test_time_functions(self):
timestamp_str = now_str()
timestamp_obj = parse_time(timestamp_str)
self.assertEqual(timestamp_str, format_time(timestamp_obj))