controversial=(desc, '_controversy'))
def db_sort(sort):
cls, col = db_sorts[sort]
return cls(col)
search_sort = dict(hot='hot desc',
new='date desc',
top='points desc',
controversial='controversy desc',
old='date asc')
db_times = dict(all=None,
hour=Thing.c._date >= timeago('1 hour'),
day=Thing.c._date >= timeago('1 day'),
week=Thing.c._date >= timeago('1 week'),
month=Thing.c._date >= timeago('1 month'),
year=Thing.c._date >= timeago('1 year'))
# batched_time_sorts/batched_time_times: top and controversial
# listings with a time-component are really expensive, and for the
# ones that span more than a day they don't change much (if at all)
# within that time. So we have some hacks to avoid re-running these
# queries against the precomputer except up to once per day
# * To get the results of the queries, we return the results of the
# (potentially stale) query, merged with the query by 'day' (see
# get_links)
# * When we are adding the special queries to the queue, we add them
# with a preflight check to determine if they are runnable and a
new = (desc, '_date'),
top = (desc, '_score'),
controversial = (desc, '_controversy'))
def db_sort(sort):
cls, col = db_sorts[sort]
return cls(col)
search_sort = dict(hot = 'hot desc',
new = 'date desc',
top = 'points desc',
controversial = 'controversy desc',
old = 'date asc')
db_times = dict(all = None,
hour = Thing.c._date >= timeago('1 hour'),
day = Thing.c._date >= timeago('1 day'),
week = Thing.c._date >= timeago('1 week'),
month = Thing.c._date >= timeago('1 month'),
year = Thing.c._date >= timeago('1 year'))
# sorts for which there can be a time filter (by day, by week,
# etc). All of these but 'all' are done in mr_top, who knows about the
# structure of the stored CachedResults (so changes here may warrant
# changes there)
time_filtered_sorts = set(('top', 'controversial'))
#we need to define the filter functions here so cachedresults can be pickled
def filter_identity(x):
return x
precompute_limit = 1000
db_sorts = dict(hot=(desc, "_hot"), new=(desc, "_date"), top=(desc, "_score"), controversial=(desc, "_controversy"))
def db_sort(sort):
cls, col = db_sorts[sort]
return cls(col)
search_sort = dict(hot="hot desc", new="date desc", top="points desc", controversial="controversy desc", old="date asc")
db_times = dict(
all=None,
hour=Thing.c._date >= timeago("1 hour"),
day=Thing.c._date >= timeago("1 day"),
week=Thing.c._date >= timeago("1 week"),
month=Thing.c._date >= timeago("1 month"),
year=Thing.c._date >= timeago("1 year"),
)
# batched_time_sorts/batched_time_times: top and controversial
# listings with a time-component are really expensive, and for the
# ones that span more than a day they don't change much (if at all)
# within that time. So we have some hacks to avoid re-running these
# queries against the precomputer except up to once per day
# * To get the results of the queries, we return the results of the
# (potentially stale) query, merged with the query by 'day' (see
# get_links)
# * When we are adding the special queries to the queue, we add them
