Shouldn't this have gone to the Dec2011 branch? On 2012-01-04 22:30, Martin Kersten wrote: > Changeset: fa4034576f02 for MonetDB > URL: http://dev.monetdb.org/hg/MonetDB?cmd=changeset;node=fa4034576f02 > Modified Files: > monetdb5/optimizer/opt_joinpath.mx > Branch: default > Log Message: > > Fix for bug 2965 > Typechecking should be included as extra defense line. > Furthermore, we need to remove reference counts during > error handling. > > > diffs (74 lines): > > diff --git a/monetdb5/optimizer/opt_joinpath.mx > b/monetdb5/optimizer/opt_joinpath.mx > --- a/monetdb5/optimizer/opt_joinpath.mx > +++ b/monetdb5/optimizer/opt_joinpath.mx > @@ -580,7 +580,7 @@ ALGjoinPathBody(Client cntxt, int top, B > * BEWARE. you may not use a size estimation, because it > * may fire a BATproperty check in a few cases. > * In case a join fails, we may try another order first before > - * abandoning the task. > + * abandoning the task. It can handle cases where a Cartesian > product emerges. > * > * A left-join sequence only requires the result to be sorted > * against the first operand. For all others operand pairs, the > cheapest join suffice. > @@ -599,12 +599,20 @@ ALGjoinPathBody(Client cntxt, int top, B > b = BATsemijoin(joins[j], joins[j + 1]); > } > if (b==NULL){ > + if ( postpone[j] && postpone[j+1]){ > + for( --top; top>=0; top--) > + BBPreleaseref(joins[top]->batCacheid); > + GDKfree(postpone); > + return NULL; > + } > postpone[j] = TRUE; > postpone[j+1] = TRUE; > postponed = 0; > for( k=0; k<top; k++) > postponed += postpone[k]== TRUE; > if ( postponed == top){ > + for( --top; top>=0; top--) > + BBPreleaseref(joins[top]->batCacheid); > GDKfree(postpone); > return NULL; > } > @@ -624,6 +632,12 @@ ALGjoinPathBody(Client cntxt, int top, B > mnstr_printf(cntxt->fdout, "#joinPath %d:= > join(%d,%d)\n", > b->batCacheid, joins[j]->batCacheid, > joins[j + 1]->batCacheid); > > + if ( b == 0 ){ > + for( --top; top>=0; top--) > + BBPreleaseref(joins[top]->batCacheid); > + GDKfree(postpone); > + return 0; > + } > BBPdecref(joins[j]->batCacheid, FALSE); > BBPdecref(joins[j+1]->batCacheid, FALSE); > joins[j] = b; > @@ -645,17 +659,25 @@ ALGjoinPath(Client cntxt, MalBlkPtr mb, > BAT *b, **joins = (BAT**)GDKmalloc(pci->argc*sizeof(BAT*)); > str joinPathRef = putName("joinPath",8); > str leftjoinPathRef = putName("leftjoinPath",12); > + int error = 0; > > if ( joins == NULL) > throw(MAL, "algebra.joinPath", MAL_MALLOC_FAIL); > (void)mb; > for (i = pci->retc; i < pci->argc; i++) { > bid = (int *) getArgReference(stk, pci, i); > - if ((b = BATdescriptor(*bid)) == NULL) { > + b = BATdescriptor(*bid); > + if ( b && top ) { > + if (TYPEerror(joins[top-1]->ttype, b->htype) ) { > + b= NULL; > + error = 1; > + } > + } > + if ( b == NULL) { > for( --top; top>=0; top--) > BBPreleaseref(joins[top]->batCacheid); > GDKfree(joins); > - throw(MAL, "algebra.joinPath", INTERNAL_BAT_ACCESS); > + throw(MAL, "algebra.joinPath", error? > SEMANTIC_TYPE_MISMATCH: INTERNAL_BAT_ACCESS); > } > joins[top++] = b; > } > _______________________________________________ > Checkin-list mailing list > checkin-l...@monetdb.org > http://mail.monetdb.org/mailman/listinfo/checkin-list
-- Sjoerd Mullender
signature.asc
Description: OpenPGP digital signature
------------------------------------------------------------------------------ Ridiculously easy VDI. With Citrix VDI-in-a-Box, you don't need a complex infrastructure or vast IT resources to deliver seamless, secure access to virtual desktops. With this all-in-one solution, easily deploy virtual desktops for less than the cost of PCs and save 60% on VDI infrastructure costs. Try it free! http://p.sf.net/sfu/Citrix-VDIinabox
_______________________________________________ Monetdb-developers mailing list Monetdb-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-developers
