SQLite优化查询

Question

I'm having trouble with SQLite query optimization, it runs fine, but for large tables it takes too much time and I need some help with optimizing it.

Source table:

-------+----------+----------------
IdMain | IdParent | ColumnToUpdate
-------+----------+----------------
   1   |          |                
   2   |    1     |     999        <-- IdParent = 1 \
   3   |          |                                  \ 
   4   |    5     |     123                           > DISTINCT ITEMS COUNT = 1
   5   |          |                                  /  IdParent = 1
   6   |    1     |     999        <-- IdParent = 1 /   UPDATE Row with IdMain = IdParent
   7   |    4     |                
   8   |    3     |     456        
-------+----------+----------------

Query to optimize

UPDATE Table

SET ColumnToUpdate = (SELECT DISTINCT ColumnToUpdate
                        FROM Table T
                       WHERE T.ColumnToUpdate IS NOT NULL
                         AND T.IdParent = Table.IdMain)

WHERE Table.ColumnToUpdate IS NULL

AND (SELECT COUNT(*) FROM (SELECT DISTINCT ColumnToUpdate
                             FROM Table T2
                            WHERE T2.ColumnToUpdate IS NOT NULL
                              AND T2.IdParent = Table.IdMain)) = 1 ;

Expected table

-------+----------+----------------
IdMain | IdParent | ColumnToUpdate
-------+----------+----------------
   1   |          |     999         <-- UPDATE
   2   |    1     |     999        
   3   |          |                
   4   |    5     |     123        
   5   |          |                
   6   |    1     |     999        
   7   |    4     |                
   8   |    3     |     456        
-------+----------+----------------

Pseudo algorithm

FOR Row DO
BEGIN

  IF ColumnToUpdate = NULL THEN
  BEGIN

    // count distinct values in ColumnToUpdate
    X = COUNT(DISTINCT(ColumnToUpdate(WITH IdParent = IdMain))

    // update row ONLY when number of distinct count equals = 1
    IF X = 1 THEN
      UPDATE(ColumnToUpdate)

  END

END

I've tried to split it up within the source code (currently Delphi) but it works slow too. Is there any way to speed it up?

Answer 1

I wonder if this might speed things up:

UPDATE Table
    SET ColumnToUpdate = coalesce((SELECT ColumnToUpdate
                                   FROM Table T
                                   WHERE T.ColumnToUpdate IS NOT NULL AND
                                         T.IdParent = Table.IdMain
                                   GROUP BY ColumnToUpdate
                                   HAVING count(*) = 1),
                                  Table.ColumnToUpdate
                                 )
    WHERE Table.ColumnToUpdate IS NULL;

This only executes the subquery once instead of twice.

Also, an index on Table(IdParent, ColumnToUpdate) might also improve performance.