使用GROUP_CONCAT

Question

I have three tables. TB_Main is a table of Entities. TB_BoardMembers is a table of People. TB_BoardMembersLINK is a bridging table which references the other two by ids and also has start and end dates for when a Person was on the board of an Entity. These dates are often incomplete.

I have been asked to export as part of a report a CSV with one row per Entity per year in which I have a list of board members for that year with their occupations in a single field delimited by newlines.

I don't need bml.Entity in the result but added it to try to debug. I'm getting one row where I expect 85. Tried with and without GROUP BY and the fact that the result is the same suggests I am misusing GROUP_CONCAT. How should I construct this to get the result they want?

SELECT 
GROUP_CONCAT(
DISTINCT CONCAT(bm.First, ' ', bm.Last, 
IF (bm.Occupation != '', ' - ', ''),
bm.Occupation)  SEPARATOR "\n") as Board,
bml.Entity
FROM  
TB_Main arfe,
TB_BoardMembers  bm,
TB_BoardMembersLINK  bml
WHERE YEAR(bml.start) <= 2011 
AND (YEAR(bml.end) >= 2011 OR bml.end IS NULL)
AND bml.start > 0 
AND bml.Entity = arfe.ID
GROUP BY bml.Entity
ORDER BY Board

Answer 1

There are a few issues with this query. The main issue appears to be that you are missing a condition to link board members to the link table, so you have a cross join, ie you will be returning every broadband member regardless of their start/end dates, and assuming you have 85 rows where the criteria matches, you will actually be returning each board member 85 times. This highlights a very good reason to switch from the ANSI 89 implicit joins you are using, to the ANSI 92 explicit join syntax. This article highlights some very good reasons to make the switch.

So your query would become (I've had to guess at your field names):

SELECT  *
FROM    TB_Main arfe
        INNER JOIN TB_BoardMembersLINK  bml
            ON bml.Entity = arfe.ID
        INNER JOIN TB_BoardMembers  bm
            ON bm.ID = bml.BoardMemberID

The next thing I noticed about your query is that using functions in the where clause is not very efficient at all, so because of this:

WHERE   YEAR(bml.start) <= 2011 
AND     (YEAR(bml.end) >= 2011 OR bml.end IS NULL)

You are operating the YEAR function twice for every row, and removing any possible chance of using an index on bml.Start or bml.End (if any exist). Yet again Aaron Bertrand has written a nice article highlighting good practises when querying date ranges, it is target at SQL-Server, but the principles are still the same, so your where clause would become:

WHERE   bml.Start <= '20110101'
AND     (bml.End >= '20110101' OR bml.End IS NULL)
AND     bml.start > 0 

Your final query should then be:

SELECT  bml.Entity,
        GROUP_CONCAT(DISTINCT CONCAT(bm.First, ' ', bm.Last, 
            IF (bm.Occupation != '', ' - ', ''), bm.Occupation) 
            SEPARATOR "\n") as Board
FROM    TB_Main arfe
        INNER JOIN TB_BoardMembersLINK  bml
            ON bml.Entity = arfe.ID
        INNER JOIN TB_BoardMembers  bm
            ON bm.ID = bml.BoardMemberID
WHERE   bml.Start <= '20110101'
AND     (bml.End >= '20110101' OR bml.End IS NULL)
AND     bml.start > 0
GROUP BY bml.Entity
ORDER BY Board;

Example on SQL Fiddle

Answer 2

If you read up on Group_Concat

"This function returns a string result with the concatenated non-NULL values from a group."

Here in this case, the group seems to be just one group, as you say there is only one entity? I am not sure if that is the case from your description. Why dont you also group by firstname, lastname and Occupation, this may give you all the members.

I am also not sure of your joins, without real data its tough to explain that part as every join works for some set of data properly, even though its not the best way to write a query