def update_prediction_data():
    """Fetch prediction data and write it to cassandra."""
    min_daily_by_sr = _min_daily_pageviews_by_sr(NDAYS_TO_QUERY)

    # combine front page values (sometimes frontpage gets '' for its name)
    if '' in min_daily_by_sr:
        fp = DefaultSR.name.lower()
        min_daily_by_sr[fp] = min_daily_by_sr.get(fp, 0) + min_daily_by_sr['']
        del min_daily_by_sr['']

    filtered = {sr_name: num for sr_name, num in min_daily_by_sr.iteritems()
                if num > 100}
    PromoMetrics.set(MIN_DAILY_CASS_KEY, filtered)

def get_predicted_pageviews(srs, start, end):
    srs, is_single = tup(srs, ret_is_single=True)
    sr_names = [sr.name for sr in srs]

    # default subreddits require a different inventory factor
    content_langs = [g.site_lang]
    default_srids = Subreddit.top_lang_srs(content_langs,
                                           limit=g.num_default_reddits,
                                           filter_allow_top=True, over18=False,
                                           ids=True)

    # prediction does not vary by date
    daily_inventory = PromoMetrics.get(MIN_DAILY_CASS_KEY, sr_names=sr_names)
    dates = get_date_range(start, end)
    ret = {}
    for sr in srs:
        if not isinstance(sr, FakeSubreddit) and sr._id in default_srids:
            factor = DEFAULT_INVENTORY_FACTOR
        else:
            factor = INVENTORY_FACTOR
        sr_daily_inventory = daily_inventory.get(sr.name, 0) * factor
        sr_daily_inventory = int(sr_daily_inventory)
        ret[sr.name] = dict.fromkeys(dates, sr_daily_inventory)

    if is_single:
        return ret[srs[0].name]
    else:
        return ret

def get_predicted_by_date(sr_name, start, stop=None):
    """Return dict mapping datetime objects to predicted pageviews."""
    # lowest pageviews any day the last 2 weeks
    min_daily = PromoMetrics.get(MIN_DAILY_CASS_KEY, sr_name).get(sr_name, 0)
    # expand out to the requested range of dates
    ndays = (stop - start).days if stop else 1  # default is one day
    predicted = OrderedDict()
    for i in range(ndays):
        date = start + timedelta(i)
        predicted[date] = min_daily
    return predicted

def get_predicted_by_date(sr_name, start, stop=None):
    """
    For now, use lowest pageviews in the subreddit any day the last two weeks
    as a simple heuristic.
    """
    # lowest pageviews any day the last 2 weeks
    min_daily = PromoMetrics.get(CassKeys.MIN_DAILY, sr_name).get(sr_name, 0)
    # expand out to the requested range of dates
    ndays = (stop - start).days if stop else 1  # default is one day
    predicted = OrderedDict()
    for i in range(ndays):
        date = start + timedelta(i)
        predicted[date] = min_daily
    return predicted

def get_predicted_pageviews(srs, start, end):
    srs, is_single = tup(srs, ret_is_single=True)
    sr_names = [sr.name for sr in srs]

    # prediction does not vary by date
    daily_inventory = PromoMetrics.get(MIN_DAILY_CASS_KEY, sr_names=sr_names)
    dates = get_date_range(start, end)
    ret = {}
    for sr in srs:
        sr_daily_inventory = daily_inventory.get(sr.name, 0) * INVENTORY_FACTOR
        sr_daily_inventory = int(sr_daily_inventory)
        ret[sr.name] = dict.fromkeys(dates, sr_daily_inventory)

    if is_single:
        return ret[srs[0].name]
    else:
        return ret

def get_predicted_pageviews(srs, location=None):
    """
    Return predicted number of pageviews for sponsored headlines.

    Predicted geotargeted impressions are estimated as:

    geotargeted impressions = (predicted untargeted impressions) *
                                 (fp impressions for location / fp impressions)

    """

    srs, is_single = tup(srs, ret_is_single=True)
    sr_names = [sr.name for sr in srs]

    # default subreddits require a different inventory factor
    default_srids = LocalizedDefaultSubreddits.get_global_defaults()

    if location:
        no_location = Location(None)
        r = LocationPromoMetrics.get(DefaultSR, [no_location, location])
        location_pageviews = r[(DefaultSR, location)]
        all_pageviews = r[(DefaultSR, no_location)]
        if all_pageviews:
            location_factor = float(location_pageviews) / float(all_pageviews)
        else:
            location_factor = 0.
    else:
        location_factor = 1.0

    # prediction does not vary by date
    daily_inventory = PromoMetrics.get(MIN_DAILY_CASS_KEY, sr_names=sr_names)
    ret = {}
    for sr in srs:
        if not isinstance(sr, FakeSubreddit) and sr._id in default_srids:
            default_factor = DEFAULT_INVENTORY_FACTOR
        else:
            default_factor = INVENTORY_FACTOR
        base_pageviews = daily_inventory.get(sr.name, 0)
        ret[sr.name] = int(base_pageviews * default_factor * location_factor)

    if is_single:
        return ret[srs[0].name]
    else:
        return ret

def get_predicted_pageviews(srs):
    srs, is_single = tup(srs, ret_is_single=True)
    sr_names = [sr.name for sr in srs]

    # default subreddits require a different inventory factor
    default_srids = LocalizedDefaultSubreddits.get_global_defaults()

    # prediction does not vary by date
    daily_inventory = PromoMetrics.get(MIN_DAILY_CASS_KEY, sr_names=sr_names)
    ret = {}
    for sr in srs:
        if not isinstance(sr, FakeSubreddit) and sr._id in default_srids:
            factor = DEFAULT_INVENTORY_FACTOR
        else:
            factor = INVENTORY_FACTOR
        ret[sr.name] = int(daily_inventory.get(sr.name, 0) * factor)

    if is_single:
        return ret[srs[0].name]
    else:
        return ret

def get_available_pageviews(srs, start, end, datestr=False,
                            ignore=None):
    srs, is_single = tup(srs, ret_is_single=True)
    sr_names = [sr.name for sr in srs]
    daily_inventory = PromoMetrics.get(MIN_DAILY_CASS_KEY, sr_names=sr_names)
    sold_by_sr_by_date = get_sold_pageviews(srs, start, end, ignore)

    datekey = lambda dt: dt.strftime('%m/%d/%Y') if datestr else dt

    ret = {}
    for sr in srs:
        sold_by_date = sold_by_sr_by_date[sr.name]
        sr_ad_inventory = int(daily_inventory.get(sr.name, 0) * 1.00)
        ret[sr.name] = {
            datekey(date): max(0, sr_ad_inventory - sold)
            for date, sold in sold_by_date.iteritems()
        }

    if is_single:
        return ret[srs[0].name]
    else:
        return ret

def update_prediction_data():
    """
    Fetches prediction data and writes it to cassandra.
    """
    min_daily_by_sr = _min_daily_pageviews_by_sr(NDAYS_TO_QUERY)
    PromoMetrics.set(CassKeys.MIN_DAILY, min_daily_by_sr)

def update_prediction_data():
    """Fetch prediction data and write it to cassandra."""
    min_daily_by_sr = _min_daily_pageviews_by_sr(NDAYS_TO_QUERY)
    PromoMetrics.set(MIN_DAILY_CASS_KEY, min_daily_by_sr)