SQLite - 3 different but connected selection in one query

[Iczer]

basically i want to get 2d array with rows consist of : [unique title, count of items where IsRead = 'false' under this title, total items under this title]

but as it seems fail to deliver selected in first SELECT statement unique title to consequent SELECTions

Query:

$sQuery = "SELECT DISTINCT Title AS fft FROM SRDB UNION ALL SELECT count(Item) FROM SRDB WHERE Title = fft AND IsRead = 'false' UNION ALL SELECT count(Item) FROM SRDB WHERE Title = fft GROUP BY Title;"

and after "_SQLite_GetTable2d()" receive Error:    no such column: fft

my table is:

$sSqlFill_Table = "(Title, Item, IsRead)"

Title_01, 'aaa', 'true'
Title_01, 'bbb', 'false'
Title_01, 'ccc', 'true'
Title_02, 'ddd', 'false'
Title_02, 'eee', 'true'
Title_02, 'fff', 'false'
Title_03, 'ggg', 'false'
Title_03, 'hhh', 'true'
Title_03, 'iii', 'true'

another case - is to get those 2 counts() for any predetermined title

so - how i can correctly make connected selections in one query?

[Miloud]

hi!

 

Maybe there is a mistake in the query, maybe you should quote fft

 

Please post a detailed example!

[jchd]

@Iczer

You need a different type of query:

select
      Title,
      (select count(*) from SRDB S WHERE S.Title = T.Title AND IsRead = 'f') as "Not read",
      count(*) "Total"
FROM SRDB T
GROUP BY Title
order by Title

 

[Iczer]

Thanks! Its working!

The only concern is - on real table it take  about  9,5 seconds for this query to get results. I have individual indexes on Title, Item and IsRead columns. Is there a way to speed up this query?

[jchd]

Using SQLite Expert I ran this:

explain query plan
select
      Title,
      (select count(*) from SRDB S WHERE S.Title = T.Title AND IsRead = 'f') as "Not read",
      count(*) "Total"
FROM SRDB T
GROUP BY Title
order by Title

on the test table created like this:

CREATE TABLE "srdb" (
  "Title" CHAR, 
  "Item" CHAR, 
  "Isread" CHAR);

CREATE INDEX "ixtitle" ON "srdb" ("Title");

CREATE INDEX "ixread" ON "srdb" ("Isread");

SQLite returns the following explanation for the above query:

selectid    order   from    detail
0   0   0   SCAN TABLE SRDB AS T USING COVERING INDEX ixtitle
0   0   0   EXECUTE CORRELATED SCALAR SUBQUERY 1
1   0   0   SEARCH TABLE SRDB AS S USING INDEX ixread (Isread=?)

This shows that the engine correctly uses the available simple indices. Of course the correlated subquery is inevitably slowing things down, but not knowing your data makes providing good hints difficult. If the IsRead column only contains boolean values, you can save some cycles by making it INT and storing 0/1 there (as integers). Also, the ORDER BY clause I used forces a final sort on he resultset, which again eats some time.

Now if you create a compound index like this:

CREATE INDEX "ixMixed" ON "srdb" ("Title", "Isread");

you discover that the engine now makes a better use of the mixed index:

selectid    order   from    detail
0   0   0   SCAN TABLE SRDB AS T USING COVERING INDEX ixtitle
0   0   0   EXECUTE CORRELATED SCALAR SUBQUERY 1
1   0   0   SEARCH TABLE SRDB AS S USING COVERING INDEX ixMixed (Title=? AND Isread=?)

How many rows does your table contain? Do you have a large number of rows having the same title on average?