http://www.doctrine-project.org/jira This file is an XML representation of an issue en-us 5.2.7 850 21-02-2013 http://www.doctrine-project.org/jira/browse/DC-81 Doctrine 1 <p>When I try to limit left joined records with an additional WITH clause, it seems to be bugged when also specifying a LIMIT.</p> <p>My DQL:</p> <p>FROM Transporter t LEFT JOIN t.Profile p LEFT JOIN t.Requests r WITH (r.distributor_id = ?) WHERE t.is_active = ? ORDER BY p.company_country, p.company_name LIMIT 20</p> <p>This returns me 0 results, while removing the LIMIT (or the WITH) will return me 2 results.</p> Mac OS X 10.6.1 DC-81

Using WITH in combinations with LIMIT returns unexpected results

Bug Major Reopened Unresolved Jonathan H. Wage Gerry Vandermaesen Tue, 6 Oct 2009 07:59:08 +0000 Tue, 15 Jun 2010 10:15:33 +0000 1.2.0-ALPHA1 0 1 <p>Sorry this is just not enough information to produce the problem. I ran some basic tests looking for what you pointed out but I didn't see any issues. Plus our tests cover this functionality so I imagine that if it were broke we'd get some failures. However, I could be wrong so if you could re-open and provide a failing test case for us that would help with getting it fixed.</p> <p>I actually ran into this problem once again in another problem.</p> <p>Schema:</p> <p>Story:<br/> columns:<br/> id:<br/> type: integer(4)<br/> unsigned: true<br/> primary: true<br/> autoincrement: true</p> <p>Picture:<br/> columns:<br/> id:<br/> type: integer(4)<br/> unsigned: true<br/> primary: true<br/> autoincrement: true<br/> story_id:<br/> type: integer(4)<br/> unsigned: true<br/> notnull: true<br/> is_selected:<br/> type: boolean<br/> notnull: true<br/> default: false<br/> relations:<br/> Story:<br/> foreignAlias: Pictures<br/> local: story_id<br/> foreign: id<br/> type: one<br/> foreignType: many<br/> onDelete: CASCADE</p> <p>Query:</p> <p>Doctrine_Query::create()<br/> -&gt;from('Story s')<br/> -&gt;leftJoin('s.Pictures p WITH p.is_selected = ?', true)<br/> -&gt;where('s.id = ?', 5)<br/> -&gt;limit(10);</p> <p>Outputted SQL:</p> <p>SELECT "s"."id" AS "s_<em>id", "s"."first_name" AS "s</em><em>first_name", "s"."last_name" AS "s</em><em>last_name", "s"."country" AS "s</em><em>country", "s"."email" AS "s</em><em>email", "s"."content" AS "s</em><em>content", "s"."title" AS "s</em><em>title", "s"."summary" AS "s</em><em>summary", "s"."is_published" AS "s</em><em>is_published", "s"."is_rejected" AS "s</em><em>is_rejected", "s"."is_promoted" AS "s</em><em>is_promoted", "s"."published_at" AS "s</em><em>published_at", "s"."created_at" AS "s</em><em>created_at", "s"."updated_at" AS "s</em><em>updated_at", "p"."id" AS "p</em><em>id", "p"."story_id" AS "p</em><em>story_id", "p"."filename_original" AS "p</em><em>filename_original", "p"."filename_large" AS "p</em><em>filename_large", "p"."filename_thumb" AS "p</em><em>filename_thumb", "p"."mime_type" AS "p</em><em>mime_type", "p"."is_selected" AS "p</em><em>is_selected", "p"."created_at" AS "p</em><em>created_at", "p"."updated_at" AS "p</em>_updated_at" FROM "story" "s" LEFT JOIN "picture" "p" ON "s"."id" = "p"."story_id" AND ("p"."is_selected" = '1') WHERE "s"."id" IN (SELECT DISTINCT "s2"."id" FROM "story" "s2" LEFT JOIN "picture" "p2" ON "s2"."id" = "p2"."story_id" AND ("p2"."is_selected" = '5') WHERE "s2"."id" = '1' ORDER BY "s2"."published_at" DESC LIMIT 10) AND ("s"."id" = '5') ORDER BY "s"."published_at" DESC</p> <p>Notice the WHERE "s2"."id" = '1' !</p> <p>Can you provide a test case we can run?</p> <p>I could not figure out what test to write but I did narrow the problem down.</p> <p>It seems that the order of the generated SQL clauses does not match the order of the passed parameters, so in the final SQL query the values are mixed up. You can actually see that happening in the SQL query above. I also noticed there a special cases hardcoded in the code for MySQL and PgSQL as far as these "limit subqueries" go, so that might be another reason why you havent bumped into problem before. I was using an SQLite driver.</p> <p>I suppose something goes wrong in the SQL generation, but I don't know the core good enough to find out where exactly it does go wrong, so I'm afraid I can't attach a test case, but it should be easy to reproduce now anyway.</p>