Are there any model analytics gems? - ruby-on-rails

Question

I'm working on allowing clients to view analytics per day, week, month, in a period of time, grouped by hours or days or months, etc... All of that is based on the created_at attribute.
Is there any gem out there that already does this? Something like:
Posts.analytics(:by => :day, :period => :this_week, :column => :created_at)
Would return:
{
'2012-06-19' => 14,
'2012-06-20' => 0, // Empty rows padding support*
'2012-06-21' => 3
}
I'm trying to make it from scratch but it seems like a lot of unecessary work if there's already a gem to do the job.
Update
I tried to make an analytics module that gets included into all models for easy analytics generation, But it's really unreliable, Sometimed i get more days than i need, and it's really messy, Could anyone collaborate and rewrite/improve on this:
# Usage:
# include Analytics::Timeline
# Model.timeline(:period => :last_24_hours, :time_by => :hour)
module Analytics
module Timeline
def self.included(base)
base.class_eval {
def self.timeline(*filters)
filters = filters[0]
period = filters[:period] || :this_week
time_by = filters[:time_by] || :days
date_column = filters[:date_column] || :created_at
# Named periods conventions
period_range = case period
when :last_12_hours
[Time.now-12.hours, Time.now]
when :last_24_hours
[Time.now-24.hours, Time.now]
when :last_7_days
[Time.now-7.days, Time.now]
when :last_30_days
[Time.now-30.days, Time.now]
when :this_week
[Time.now.beginning_of_week, Time.now.end_of_week]
when :past_week
[(Time.now - 1.week).beginning_of_week, (Time.now - 1.week).end_of_week]
when :this_month
[Time.now.beginning_of_month, Time.now.end_of_month]
when :past_month
[(Time.now-1.month).beginning_of_month, (Time.now - 1.month).end_of_month]
when :this_year
[Time.now.beginning_of_year, Time.now.end_of_year]
end
period_range = period if period.kind_of?(Array)
period_range = [period, Time.now] if period.is_a?(String)
# determine the SQL group method
group_column = case time_by
when :months
time_suffix = "-01 00:00:00"
records = where("#{table_name}.#{date_column} > ? AND #{table_name}.#{date_column} <= ?", period_range[0].to_date, period_range[1].to_date)
"DATE_FORMAT(#{table_name}.#{date_column.to_s}, '%Y-%m')"
when :days
time_suffix = " 00:00:00"
records = where("#{table_name}.#{date_column} > ? AND #{table_name}.#{date_column} <= ?", period_range[0].to_date, period_range[1].to_date)
"DATE(#{table_name}.#{date_column.to_s})"
when :hours
time_suffix = ":00:00"
records = where("#{table_name}.#{date_column} > ? AND #{table_name}.#{date_column} <= ?", period_range[0], period_range[1])
"DATE_FORMAT(#{table_name}.#{date_column.to_s}, '%Y-%m-%d %H')"
when :minutes
time_suffix = ":00"
records = where("#{table_name}.#{date_column} > ? AND #{table_name}.#{date_column} <= ?", period_range[0], period_range[1])
"DATE_FORMAT(#{table_name}.#{date_column.to_s}, '%Y-%m-%d %H:%M')"
end
# Get counts per cycle
records = records.group(group_column).select("*, count(*) AS series_count, #{group_column} AS series_time")
series = {}
# Generate placeholder series
time_table = { :days => 60*60*24, :hours => 60*60, :minutes => 60, :seconds => 0 }
if time_by == :months
ticks = 12 * (period_range[1].year - period_range[0].year) + (period_range[1].month + 1) - period_range[0].month
else
ticks = (period_range[1] - period_range[0] + 1) / time_table[time_by]
end
ticks.to_i.times do |i|
time = period_range[1]-i.send(time_by)
time = case time_by
when :minutes
time.change(:sec => 0)
when :hours
time.change(:min => 0)
when :days
time.change(:hour => 0)
when :months
time.change(:day => 1, :hour => 0)
end
series[time.to_s(:db)] = 0
end
# Merge real counts with placeholder series
to_merge = {}
records.each do |r|
to_merge[r.series_time.to_s+time_suffix] = r.series_count
end
series.merge!(to_merge)
end
}
end
end
end

Answer 1

The ActiveRecord statistics gem seems like it could be really useful to you.
If the statistics gem doesn't help, the admin_data gem has some analytics built in. Check out the demo. Use of the entire admin system might be overkill but you could at least try to browse the source to mimic the analytics feature.