I've searched around for this, but all the similar questions and answers are just different enough not to work.
I have a table with the following fields: person, thing, purdate. A new record is entered when a person buys each new thing.
I want to count the consecutive months that a person bought any "thing" (thing01 or thing02, it doesn't mater). If there is a break in consecutive purdays, then the count should start over.
With the data enclosed, I want to end up with this:
| Person | Consec Days |
| person_01 | 3 |
| person_02 | 3 |
| person_02 | 2 |
create table records (
person varchar(32) not null,
thing varchar(32) not null,
purdate datetime not null
insert into records (person, thing, purdate) values
('person_01', 'thing01', '2014-01-02'),
('person_01', 'thing02', '2014-01-02'),
('person_01', 'thing02', '2014-02-27'),
('person_01', 'thing02', '2014-03-27'),
('person_02', 'thing02', '2014-01-28'),
('person_02', 'thing01', '2014-02-28'),
('person_02', 'thing02', '2014-03-28'),
('person_02', 'thing02', '2014-05-29'),
('person_02', 'thing02', '2014-06-29')
You can do this in MySQL using variables (or very complicated correlated subqueries). In other databases, you would use window/analytic functions.
The logic is:
Here is a query that has been tested on your SQL Fiddle:
select person, count(*) as numMonths from (select person, ym, @ym, @person, if(@person = person and @ym = ym - 1, @grp, @grp := @grp + 1) as grp, @person := person, @ym := ym from (select distinct person, year(purdate)*12+month(purdate) as ym from records r ) r cross join (select @person := '', @ym := 0, @grp := 0) const order by 1, 2 ) pym group by person, grp;