Lost trying to use DISTINCT and GROUP BY - sql

Question

I'm having trouble with something that I thought would've been simple...
I have a simple model Statistic that stores a date (created_at), a user_fingerprint and a structure_id. From that, I'd like to create a graph to show #visitors per day.
So I did
#structure.statistics.order('DATE(created_at) ASC').group('DATE(created_at)').count
Which works and return what I expect:
=> {Sat, 18 May 2014=>50, Mon, 19 May 2014=>90}
Now I'd like the same, but I want to squeeze all rows with the same couple (created_at, user_fingerprint). For instance:
| created_at | user_fingerprint | structure_id |
|----------------------|------------------|--------------|
| Sat, 18 May 2014 2PM | '124512341' | 12 |
| Sat, 18 May 2014 4PM | '124512341' | 12 |
| Mon, 19 May 2014 6PM | '124512341' | 12 |
With this data, I would have:
=> {Sat, 18 May 2014=>1, Mon, 19 May 2014=>1}
# instead of
=> {Sat, 18 May 2014=>2, Mon, 19 May 2014=>1}
I would be able to do it in Ruby but I wondered if I could directly do it with SQL & Arel.
Solution regarding your answers
Here is what I did at the end:
#impressions = {}
# The following is to ensure I will have a key when there is no stat for a day.
(15.days.ago.to_date..Date.today).each { |date| #impressions[date] = 0 }
#structure.statistics.where( Statistic.arel_table[:created_at].gt(Date.today - 15.days) )
.order('DATE(created_at) ASC')
.group('DATE(created_at)')
.select('DATE(created_at) as created_at, COUNT(DISTINCT(user_fingerprint)) as user_count')
.each{ |stat| #impressions[stat.created_at] = stat.user_count }
I need to do a bit of Ruby though but that's good for me.

Answer 1

your query would look something like (Oracle dialect)
select trunc(created_at), user_fingerprint, count(distinct user_fingerprint)
from statistic
group by trunc(created_at), user_fingerprint
there is no SQL standard for getting date portion out of datetime data field.
oracle: trunc(dt_column)
sql server: cast(dt_column As Date)
mysql: DATE(dt_column)

Answer 2

#structure.statistics.order('DATE(created_at) ASC').group('DATE(created_at)').select('count(distinct(user_fingerprint)) as user_count').first.user_count