在子查询中使用UNION ALL分组

Question

I have two tables, one showing the search queries history, the other showing the locations.

The first table is for the search queries: searchTerm | date | userId | historyId The second is for the locations: userId | date | historyId | lat | lng

I am using UNION ALL to bring these together in a sub query with JOIN LEFT & JOIN RIGHT to create a FULL JOIN.

SELECT st, fa FROM
                 (SELECT webHistory.searchTerm AS st, locationHistory.fullAddress AS fa
                  FROM webHistory LEFT OUTER JOIN locationHistory
                  ON webHistory.historyId=locationHistory.historyId
                  UNION ALL
                  SELECT webHistory.searchTerm as st, locationHistory.fullAddress AS fa
                  FROM webHistory RIGHT OUTER JOIN locationHistory
                  ON webHistory.historyId=locationHistory.historyId) t

Every thing is working beautifully I'm getting results for just entries in webHistory table, and the same for the locationHistory table. The issue I have is when I'm getting rows that have been added to both. They show twice!

I can't just use union as it will JOIN more than I want it to. If two of the same searches are added, instead of showing two it will join them into one. So I need to use UNION ALL.

I have to use union as I'm JOINing from the LEFT and RIGHT and in mySql FULL JOIN isn't supported.

So I'm assuming I need to GROUP BY historyId as this id will always be the same on rows added at the same time (when a search is made by query and location). Problem is, i've tried and tested but I can't seem to get this to work. Am I missing something?

Lastly, while I'm asking a question and because I've not yet got to the point of testing this would the ORDER BY date be on the OUTER or SUBQUERIES as I want to show the most recent as the first result?

/* More Information */

I need to to output:

SearchTerm if there is no fullAddress. Full address if there is no searchTerm. SearchTerm & fullAddress if there is both.

Currently it outputs the first two fine, but the latter will output twice.

Answer 1

As far as understand, you want duplicates with the same historyId to be eliminated, so you should be able to use UNION as long as the historyId is included in the UNION so that duplicates with another historyId won't be eliminated;

SELECT st, fa FROM
 (SELECT webHistory.searchTerm AS st, 
         locationHistory.fullAddress AS fa, 
         webHistory.historyId
  FROM webHistory LEFT OUTER JOIN locationHistory
  ON webHistory.historyId=locationHistory.historyId
  UNION
  SELECT webHistory.searchTerm as st, 
         locationHistory.fullAddress AS fa,
         locationHistory.historyId
  FROM webHistory RIGHT OUTER JOIN locationHistory
  ON webHistory.historyId=locationHistory.historyId) t