交叉表中的Group_Concat

Question

With a table like:

Name   Event     Time
Steve  Start    19:53
Steve  Event1   19:51
Steve  Stop     19:45
Steve  Start    18:13
Steve  Event2   18:01
Steve  Event1   17:51
Steve  Stop     17:45
Bob    Start    19:47
Bob    Event2   19:31
Bob    Stop     17:57

representing events that occur between start and stops times. I'd like to turn this into:

Name    Start   Stop   Event1    Event2
Steve  19:53    19:45    1         0
Steve  18:13    17:45    1         1
Bob    19:47    17:57    0         1

My attempt:

select
    name,
    (CASE event WHEN 'start' THEN time ELSE NULL END) AS Start,
    (CASE event WHEN 'stop' THEN time ELSE NULL END) AS Stop,
    (CASE event WHEN 'event1' THEN 1 ELSE 0 END) AS Event1,
    (CASE event WHEN 'event2' THEN 1 ELSE 0 END) AS Event2
from atable

which results in:

Name    Start   Stop    Event1  Event2
Steve   19:53   0   0   0
Steve   0   0   1   0
Steve   0   19:45   0   0
Steve   18:13   0   0   0
Steve   0   0   0   1
Steve   0   0   1   0
Steve   0   17:45   0   0
Bob 19:47   0   0   0
Bob 0   0   1   0
Bob 0   17:57   0   0

How do I get the records in a single row?

I did try a Group_CONCAT per:

select
    name,
    GROUP_CONCAT((CASE event WHEN 'start' THEN time ELSE 0 END)) AS Start,
    GROUP_CONCAT((CASE event WHEN 'stop' THEN time ELSE 0 END)) AS Stop,
    GROUP_CONCAT((CASE event WHEN 'event1' THEN 1 ELSE 0 END)) AS Event1,
    GROUP_CONCAT((CASE event WHEN 'event2' THEN 1 ELSE 0 END)) AS Event2
from atable
group by name, event

But ended up with:

Name    Start   Stop    Event1  Event2
Bob 0   0   1   0
Bob 19:47   0   0   0
Bob 0   17:57   0   0
Steve   0,0 0,0 1,1 0,0
Steve   0   0   0   1
Steve   19:53,18:13 0,0 0,0 0,0
Steve   0,0 19:45,17:45 0,0 0,0

Answer 1

And now tested: Working SQL FIDDLE

How do I get the records in a single row?

Max or min would combine the rows, but since name isn't unique enough we need a 2nd way to identify the group within the group which would result in a 2nd row in your output. We can do this using user variables on the event start and a counter.

This assumes:

1. Ordering by time desc will retain the needed order
2. Multiple occurrences of 'start' in an event for the same user denote a new row needed in result.
3. Start will always have the lowest time for each set to become a row.

.

***DDL***
CREATE table atable (
  name varchar(10),
  `event` varchar(10),
  `time` time);

INSERT into atable values 
('Steve',  'Start',    '19:53'),
('Steve',  'Event1',   '19:51'),
('Steve',  'Stop',     '19:45'),
('Steve',  'Start',    '18:13'),
('Steve',  'Event2',   '18:01'),
('Steve',  'Event1',   '17:51'),
('Steve',  'Stop',     '17:45'),
('Bob',    'Start',    '19:47'),
('Bob',    'Event2',   '19:31'),
('Bob',    'Stop',     '17:57');

***DML***
SELECT
    name,
    max(CASE event WHEN 'start' THEN time ELSE NULL END) AS Start,
    max(CASE event WHEN 'stop' THEN time ELSE NULL END) AS Stop,
    max(CASE event WHEN 'event1' THEN 1 ELSE 0 END) AS Event1,
    max(CASE event WHEN 'event2' THEN 1 ELSE 0 END) AS Event2
FROM (SELECT A.*, CASE WHEN @Name=Name and Event = 'Start' THEN @GRP:=@GRP+1
                       WHEN @Name <> Name THEN @GRP:=@GRP+1
                       ELSE @GRP
                  end as mGROUP,
                  CASE WHEN @Name <> NAME then 
                       @Name:=Name
                  END  as Name2
      FROM  atable A
      CROSS JOIN (Select @GRP:=0) x
      CROSS JOIN (Select @Name:='') y
      ORDER BY Name, Time Desc) z
GROUP BY Name, mGROUP

How this works: The inner query "Z" adds two columns, mGROUP and Name2 to your base data set "atable". The Database engine initializes variables @GRP and @NAME via a cross join. We need @GRP so we can group each sub set of data together from atable to make it easier to combine the data. This column represents the missing group by we need in order to use min/max to combine the rows in your pivot.

The @Name variable is used as a control break method so we know when to increment the @grp counter. This value is reset in the Name2 column when the names vary. There are two times the @grp counter needs to be incremented. Once is when a 'Start' event is encountered, once is when @Name changes. The mGROUP column does this increment for us and when no increment is needed, the same value is output as it's part of the same group. (To see this work just run the inner query "Z" and evaluate the results)

Now that we have the needed grouping mechanism to combine the groups using max, we can use simple aggregation to combine the rows which is done via the outer query.