Hi Jan, I just thought: how does your duplicate generating scjoins relate to the generalized scjoin I proposed in my TDM 2004-paper? http://db.cs.utwente.nl/Publications/PaperStore/db-utwente-0000003587.pdf Maurice. Jan Rittinger wrote:
Update of /cvsroot/monetdb/pathfinder/compiler/algebra/opt In directory sc8-pr-cvs7.sourceforge.net:/tmp/cvs-serv23019/algebra/opt
Modified Files: opt_set.c Log Message: -- Rewrite scjoin operators into duplicate generating ones as intermediate step. (This helps the join pushdown phase to push down more joins.)
Index: opt_set.c =================================================================== RCS file: /cvsroot/monetdb/pathfinder/compiler/algebra/opt/opt_set.c,v retrieving revision 1.1 retrieving revision 1.2 diff -u -d -r1.1 -r1.2 --- opt_set.c 13 Nov 2006 17:10:59 -0000 1.1 +++ opt_set.c 17 Nov 2006 14:03:41 -0000 1.2 @@ -50,6 +50,11 @@ #define L(p) ((p)->child[0]) /** starting from p, make a step right */ #define R(p) ((p)->child[1]) +/** and so on ... */ +#define LL(p) (L(L(p))) +#define LR(p) (R(L(p))) +#define RL(p) (L(R(p))) +#define RR(p) (R(R(p)))
#define SEEN(p) ((p)->bit_dag)
@@ -71,6 +76,56 @@
/* action code */ switch (p->kind) { + case la_eqjoin: + /* Rewrite scjoin into duplicate generating scjoins + (underneath join operators). This hopefully allows + the above join to be pushed down in a later phase. */ + if (PFprop_set (p->prop) && + L(p)->kind == la_scjoin && + PFprop_set (L(p)->prop)) { + PFalg_att_t item_res; + item_res = PFalg_ori_name ( + PFalg_unq_name (att_item, 0), + ~(L(p)->sem.scjoin.item | + L(p)->sem.scjoin.iter)); + *L(p) = *PFla_project ( + PFla_dup_scjoin ( + LL(p), + LR(p), + L(p)->sem.scjoin.axis, + L(p)->sem.scjoin.ty, + L(p)->sem.scjoin.item, + item_res), + PFalg_proj (L(p)->sem.scjoin.iter, + L(p)->sem.scjoin.iter), + PFalg_proj (L(p)->sem.scjoin.item, + item_res)); + break; + } + if (PFprop_set (p->prop) && + R(p)->kind == la_scjoin && + PFprop_set (R(p)->prop)) { + PFalg_att_t item_res; + item_res = PFalg_ori_name ( + PFalg_unq_name (att_item, 0), + ~(R(p)->sem.scjoin.item | + R(p)->sem.scjoin.iter)); + *R(p) = *PFla_project ( + PFla_dup_scjoin ( + RL(p), + RR(p), + R(p)->sem.scjoin.axis, + R(p)->sem.scjoin.ty, + R(p)->sem.scjoin.item, + item_res), + PFalg_proj (R(p)->sem.scjoin.iter, + R(p)->sem.scjoin.iter), + PFalg_proj (R(p)->sem.scjoin.item, + item_res)); + break; + } + break; + case la_distinct: if (PFprop_set (p->prop)) *p = *PFla_dummy (L(p));
------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Monetdb-pf-checkins mailing list Monetdb-pf-checkins@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-pf-checkins
-- ---------------------------------------------------------------------- Dr.Ir. M. van Keulen - Assistant Professor, Data Management Technology Univ. of Twente, Dept of EEMCS, POBox 217, 7500 AE Enschede, Netherlands Email: m.vankeulen@utwente.nl, Phone: +31 534893688, Fax: +31 534892927 Room: ZI 3039, WWW: http://www.cs.utwente.nl/~keulen