On 12/06/2011 03:28 PM, Max Vlasov wrote:
Hi, Noticed a strange regression after 3.7.7 (in 3.7.8 and inherited in 3.7.9)I have a Russian morphology database and different queries working with it. I narrowed it to the following case and populated with a couple of English words (just to make sense) The following database CREATE TABLE [Beginnings] ([Id] INTEGER PRIMARY KEY AUTOINCREMENT,[Title] TEXT, [EndingId] INTEGER); CREATE TABLE [Endings] (Id INT,Title TEXT,EndingId INT); INSERT INTO Beginnings (Id, Title, EndingId) VALUES (1, 'FACTOR', 18); INSERT INTO Beginnings (Id, Title, EndingId) VALUES (2, 'SWIMM', 18); INSERT INTO Endings (Id, Title, EndingId) VALUES (1, 'ING', 18); There's a query that searches for primary form of a united list of some words (here FACTORING and SWIMMING): SELECT SrcWord, Beginnings.Title FROM (SELECT 'FACTORING' AS SrcWord UNION SELECT 'SWIMMING' AS SrcWord ) LEFT JOIN Beginnings WHERE Beginnings.Id= (SELECT BeginningId FROM (SELECT SrcWord, B.Id as BeginningId, B.Title || E.Title As Connected FROM Beginnings B LEFT JOIN Endings E ON B.EndingId=E.EndingId WHERE Connected=SrcWord LIMIT 1))
There is a candidate fix for this in fossil now. Alternatively, a workaround with 3.7.9 is to remove the "LIMIT 1" from the sub-query. Or replace the "LEFT JOIN" in the subquery with a regular JOIN. The query then seems to work as expected. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
