def handle(self, *args, **options):
if settings.CENTRAL_SERVER:
raise CommandError("Run this command on the distributed server only.")
# Load videos
video_sizes = softload_json(REMOTE_VIDEO_SIZE_FILEPATH, logger=logging.debug)
# Query current files
all_video_filepaths = glob.glob(os.path.join(settings.CONTENT_ROOT, "*.mp4"))
logging.info("Querying sizes for %d video(s)." % len(all_video_filepaths))
# Get all current sizes
for video_filepath in all_video_filepaths:
youtube_id = os.path.splitext(os.path.basename(video_filepath))[0]
# Set to max, so that local compressed videos will not affect things.
video_sizes[youtube_id] = max(video_sizes.get(youtube_id, 0), os.path.getsize(video_filepath))
# Sort results
video_sizes = OrderedDict([(key, video_sizes[key]) for key in sorted(video_sizes.keys())])
logging.info("Saving results to disk.")
ensure_dir(os.path.dirname(REMOTE_VIDEO_SIZE_FILEPATH))
with open(REMOTE_VIDEO_SIZE_FILEPATH, "w") as fp:
json.dump(video_sizes, fp, indent=2)
def compute_data(data_types, who, where):
"""
Compute the data in "data_types" for each user in "who", for the topics selected by "where"
who: list of users
where: topic_path
data_types can include:
pct_mastery
effort
attempts
"""
# None indicates that the data hasn't been queried yet.
# We'll query it on demand, for efficiency
topics = None
exercises = None
videos = None
# Initialize an empty dictionary of data, video logs, exercise logs, for each user
data = OrderedDict(zip([w.id for w in who], [dict() for i in range(len(who))])) # maintain the order of the users
vid_logs = dict(zip([w.id for w in who], [[] for i in range(len(who))]))
ex_logs = dict(zip([w.id for w in who], [[] for i in range(len(who))]))
if UserLog.is_enabled():
activity_logs = dict(zip([w.id for w in who], [[] for i in range(len(who))]))
# Set up queries (but don't run them), so we have really easy aliases.
# Only do them if they haven't been done yet (tell this by passing in a value to the lambda function)
# Topics: topics.
# Exercises: names (ids for ExerciseLog objects)
# Videos: video_id (ids for VideoLog objects)
# This lambda partial creates a function to return all items with a particular path from the NODE_CACHE.
search_fun_single_path = partial(lambda t, p: t["path"].startswith(p), p=tuple(where))
# This lambda partial creates a function to return all items with paths matching a list of paths from NODE_CACHE.
search_fun_multi_path = partial(lambda ts, p: any([t["path"].startswith(p) for t in ts]), p=tuple(where))
# Functions that use the functions defined above to return topics, exercises, and videos based on paths.
query_topics = partial(lambda t, sf: t if t is not None else [t[0]["id"] for t in filter(sf, get_node_cache('Topic').values())], sf=search_fun_single_path)
query_exercises = partial(lambda e, sf: e if e is not None else [ex[0]["id"] for ex in filter(sf, get_node_cache('Exercise').values())], sf=search_fun_multi_path)
query_videos = partial(lambda v, sf: v if v is not None else [vid[0]["id"] for vid in filter(sf, get_node_cache('Video').values())], sf=search_fun_multi_path)
# No users, don't bother.
if len(who) > 0:
# Query out all exercises, videos, exercise logs, and video logs before looping to limit requests.
# This means we could pull data for n-dimensional coach report displays with the same number of requests!
# Note: User activity is polled inside the loop, to prevent possible slowdown for exercise and video reports.
exercises = query_exercises(exercises)
videos = query_videos(videos)
if exercises:
ex_logs = query_logs(data.keys(), exercises, "exercise", ex_logs)
if videos:
vid_logs = query_logs(data.keys(), videos, "video", vid_logs)
for data_type in (data_types if not hasattr(data_types, "lower") else [data_types]): # convert list from string, if necessary
if data_type in data[data.keys()[0]]: # if the first user has it, then all do; no need to calc again.
continue
#
# These are summary stats: you only get one per user
#
if data_type == "pct_mastery":
# Efficient query out, spread out to dict
for user in data.keys():
data[user][data_type] = 0 if not ex_logs[user] else 100. * sum([el['complete'] for el in ex_logs[user]]) / float(len(exercises))
elif data_type == "effort":
if "ex:attempts" in data[data.keys()[0]] and "vid:total_seconds_watched" in data[data.keys()[0]]:
# exercises and videos would be initialized already
for user in data.keys():
avg_attempts = 0 if len(exercises) == 0 else sum(data[user]["ex:attempts"].values()) / float(len(exercises))
avg_video_points = 0 if len(videos) == 0 else sum(data[user]["vid:total_seconds_watched"].values()) / float(len(videos))
data[user][data_type] = 100. * (0.5 * avg_attempts / 10. + 0.5 * avg_video_points / 750.)
else:
data_types += ["ex:attempts", "vid:total_seconds_watched", "effort"]
#
# These are detail stats: you get many per user
#
# Just querying out data directly: Video
elif data_type.startswith("vid:") and data_type[4:] in [f.name for f in VideoLog._meta.fields]:
for user in data.keys():
data[user][data_type] = OrderedDict([(v['video_id'], v[data_type[4:]]) for v in vid_logs[user]])
# Just querying out data directly: Exercise
elif data_type.startswith("ex:") and data_type[3:] in [f.name for f in ExerciseLog._meta.fields]:
for user in data.keys():
data[user][data_type] = OrderedDict([(el['exercise_id'], el[data_type[3:]]) for el in ex_logs[user]])
# User Log Queries
elif data_type.startswith("user:"******"", "activity", activity_logs)
for user in data.keys():
data[user][data_type] = [log[data_type[5:]] for log in activity_logs[user]]
# User Summary Queries
elif data_type.startswith("usersum:") and data_type[8:] in [f.name for f in UserLogSummary._meta.fields] and UserLog.is_enabled():
activity_logs = query_logs(data.keys(), "", "summaryactivity", activity_logs)
for user in data.keys():
data[user][data_type] = sum([log[data_type[8:]] for log in activity_logs[user]])
# Unknown requested quantity
else:
raise Exception("Unknown type: '%s' not in %s" % (data_type, str([f.name for f in ExerciseLog._meta.fields])))
# Returning empty list instead of None allows javascript on client
# side to read 'length' property without error.
exercises = exercises or []
videos = videos or []
return {
"data": data,
"topics": topics,
"exercises": exercises,
"videos": videos,
}
def compute_data(data_types, who, where):
"""
Compute the data in "data_types" for each user in "who", for the topics selected by "where"
who: list of users
where: topic_path
data_types can include:
pct_mastery
effort
attempts
"""
# None indicates that the data hasn't been queried yet.
# We'll query it on demand, for efficiency
topics = None
exercises = None
videos = None
# Initialize an empty dictionary of data, video logs, exercise logs, for each user
data = OrderedDict(zip([w.id for w in who], [dict() for i in range(len(who))])) # maintain the order of the users
vid_logs = dict(zip([w.id for w in who], [[] for i in range(len(who))]))
ex_logs = dict(zip([w.id for w in who], [[] for i in range(len(who))]))
if UserLog.is_enabled():
activity_logs = dict(zip([w.id for w in who], [[] for i in range(len(who))]))
# Set up queries (but don't run them), so we have really easy aliases.
# Only do them if they haven't been done yet (tell this by passing in a value to the lambda function)
# Topics: topics.
# Exercises: names (ids for ExerciseLog objects)
# Videos: video_id (ids for VideoLog objects)
# This lambda partial creates a function to return all items with a particular path from the NODE_CACHE.
search_fun_single_path = partial(lambda t, p: t["path"].startswith(p), p=tuple(where))
# This lambda partial creates a function to return all items with paths matching a list of paths from NODE_CACHE.
search_fun_multi_path = partial(lambda ts, p: any([t["path"].startswith(p) for t in ts]), p=tuple(where))
# Functions that use the functions defined above to return topics, exercises, and videos based on paths.
query_topics = partial(lambda t, sf: t if t is not None else [t[0]["id"] for t in filter(sf, get_node_cache('Topic').values())], sf=search_fun_single_path)
query_exercises = partial(lambda e, sf: e if e is not None else [ex[0]["id"] for ex in filter(sf, get_node_cache('Exercise').values())], sf=search_fun_multi_path)
query_videos = partial(lambda v, sf: v if v is not None else [vid[0]["id"] for vid in filter(sf, get_node_cache('Video').values())], sf=search_fun_multi_path)
# No users, don't bother.
if len(who) > 0:
# Query out all exercises, videos, exercise logs, and video logs before looping to limit requests.
# This means we could pull data for n-dimensional coach report displays with the same number of requests!
# Note: User activity is polled inside the loop, to prevent possible slowdown for exercise and video reports.
exercises = query_exercises(exercises)
videos = query_videos(videos)
if exercises:
ex_logs = query_logs(data.keys(), exercises, "exercise", ex_logs)
if videos:
vid_logs = query_logs(data.keys(), videos, "video", vid_logs)
for data_type in (data_types if not hasattr(data_types, "lower") else [data_types]): # convert list from string, if necessary
if data_type in data[data.keys()[0]]: # if the first user has it, then all do; no need to calc again.
continue
#
# These are summary stats: you only get one per user
#
if data_type == "pct_mastery":
# Efficient query out, spread out to dict
for user in data.keys():
data[user][data_type] = 0 if not ex_logs[user] else 100. * sum([el['complete'] for el in ex_logs[user]]) / float(len(exercises))
elif data_type == "effort":
if "ex:attempts" in data[data.keys()[0]] and "vid:total_seconds_watched" in data[data.keys()[0]]:
# exercises and videos would be initialized already
for user in data.keys():
avg_attempts = 0 if len(exercises) == 0 else sum(data[user]["ex:attempts"].values()) / float(len(exercises))
avg_video_points = 0 if len(videos) == 0 else sum(data[user]["vid:total_seconds_watched"].values()) / float(len(videos))
data[user][data_type] = 100. * (0.5 * avg_attempts / 10. + 0.5 * avg_video_points / 750.)
else:
data_types += ["ex:attempts", "vid:total_seconds_watched", "effort"]
#
# These are detail stats: you get many per user
#
# Just querying out data directly: Video
elif data_type.startswith("vid:") and data_type[4:] in [f.name for f in VideoLog._meta.fields]:
for user in data.keys():
data[user][data_type] = OrderedDict([(v['video_id'], v[data_type[4:]]) for v in vid_logs[user]])
# Just querying out data directly: Exercise
elif data_type.startswith("ex:") and data_type[3:] in [f.name for f in ExerciseLog._meta.fields]:
for user in data.keys():
data[user][data_type] = OrderedDict([(el['exercise_id'], el[data_type[3:]]) for el in ex_logs[user]])
# User Log Queries
elif data_type.startswith("user:"******"", "activity", activity_logs)
for user in data.keys():
data[user][data_type] = [log[data_type[5:]] for log in activity_logs[user]]
# User Summary Queries
elif data_type.startswith("usersum:") and data_type[8:] in [f.name for f in UserLogSummary._meta.fields] and UserLog.is_enabled():
activity_logs = query_logs(data.keys(), "", "summaryactivity", activity_logs)
for user in data.keys():
data[user][data_type] = sum([log[data_type[8:]] for log in activity_logs[user]])
# Unknown requested quantity
else:
raise Exception("Unknown type: '%s' not in %s" % (data_type, str([f.name for f in ExerciseLog._meta.fields])))
# Returning empty list instead of None allows javascript on client
# side to read 'length' property without error.
exercises = exercises or []
videos = videos or []
return {
"data": data,
"topics": topics,
"exercises": exercises,
"videos": videos,
}
