MySQL计数一个字段值出现在另一个表字段中的次数

Question

Given that i have 2 tables, how can i see how many distinct values of X is in distinct values of Y but is within 31 days (or a month) before date_X?

tb1
     date_X        X
    2015-05-01    cat
    2015-05-01    pig
    2015-05-01    mouse
    2015-04-01    dog
    2015-04-01    horse
tb2  
    date_Y         Y
    2015-04-30    cat
    2015-04-02    cat
    2015-04-05    mouse
    2015-04-15    rabbit
    2015-04-10    pig
    2015-03-20    dog
    2015-03-10    horse
    2015-03-09    frog

For example, i want:

date_period num_match count_y percent_match
2015-05-01   2            4        40
2014-04-01   2            3        67

date_period is unique(date_x)

num_match is the number of distinct(Y) that matches distinct(X) for up to 31 days before given date_period

count_y is the distinct(Y) for up to 31 days before given date_period.

percent_match is just num_match / count_y

This question is an extension to my earlier question here: join mysql on a date range

Answer 1

One way you can this is with an non-equijoin on date. Then you can count the distinct values of y either in the set or that match:

select x.date_x,
       count(distinct case when x.x = y.y then y.seqnum end) as nummatch,
       count(distinct y.seqnum) as count_y,
       (count(distinct case when x.x = y.y then y.seqnum end) /
        count(distinct y.seqnum) 
       ) as ratio
from x left join
     (select y.*, rownum as seqnum
      from y
     ) y
     on y.date_y between x.date_x - 31 and x.date_x
group by x.date_x;

EDIT:

The above treats the two "cat" lines in y as being different. I misread the desired results, so I think the appropriate query is:

select x.date_x,
       count(distinct case when x.x = y.y then y.y end) as nummatch,
       count(distinct y.y) as count_y,
       (count(distinct case when x.x = y.y then y.y end) /
        count(distinct y.y) 
       ) as ratio
from x left join
     y
     on y.date_y between x.date_x - 31 and x.date_x
group by x.date_x;