Stefan Manegold wrote:

On Tue, Sep 16, 2008 at 02:55:45PM +0000, Romulo Goncalves wrote:

...

Update of /cvsroot/monetdb/sql/src/server In directory sc8-pr-cvs16.sourceforge.net:/tmp/cvs-serv12591

Modified Files: sql_parser.mx sql_updates.mx Log Message: The number of records should be type lng to allow the load of billions of records in a single copy command.

[...]

...

U sql_updates.mx Index: sql_updates.mx =================================================================== RCS file: /cvsroot/monetdb/sql/src/server/sql_updates.mx,v retrieving revision 1.145 retrieving revision 1.146 diff -u -d -r1.145 -r1.146 --- sql_updates.mx 7 Sep 2008 14:28:00 -0000 1.145 +++ sql_updates.mx 16 Sep 2008 14:55:42 -0000 1.146 @@ -1411,11 +1411,11 @@ char *ssep = (seps->h->next->next)?seps->h->next->next->data.sval:"\""; char *ns = (null_string)?null_string:"null"; list *slist; - int nr = (nr_offset)?nr_offset->h->data.i_val:-1; - int offset = (nr_offset)?nr_offset->h->next->data.i_val:0; + lng nr = (nr_offset)?nr_offset->h->data.i_val:-1; ^^^ ^ + lng offset = (nr_offset)?nr_offset->h->next->data.i_val:0; ^^^ ^ shouldn't these be "l_val", now? Yes the cleaning is coming...

For the moment I will use type lng (l_val) in all the places. However, we should use wrd and do the check if the long fits or not (32bits or 64 bits) to have a perfect job. I need to know where is the best place. In my opinion it should be in the parser, but I do not know yet how to do the check there. Niels never did it there. I need to talk with him to see where is the best place for the check.... Thanks for the check ;) Romulo

^

...

- assert(!nr_offset || nr_offset->h->type == type_int); - assert(!nr_offset || nr_offset->h->next->type == type_int); + assert(!nr_offset || nr_offset->h->type == type_lng);

^^^

...

+ assert(!nr_offset || nr_offset->h->next->type == type_lng); ^^^ if (sname && !(s=mvc_bind_schema(sql, sname))) { (void) sql_error(sql, 02, "COPY INTO: no such schema '%s'", sname); return NULL;