On Thu, 12 Feb 2009, Szak�ts Viktor wrote:
Hi Viktor,
> Pls find enclosed patch which implements:
> dbCreateTemp( <cAlias>, <aStruct>, <cRDD>, <cDelimArg>, <cCodePage>,
> <nConnection> ) -> <lSuccess>
> This creates and opens a new temporary table. The file automatically
> gets deleted on close, all indexes created are also handled as temporary.
> Please review and comment.
It needs some additional modification:
1. it does not touch memo files and memo RDD at all. I has to be fixed.
2. <cAlias> is obligatory in dbCreateTemp() so RT error should be generated
when it's missing though we can add support for empty alias ("") which
is also legal in Clipper. In such case alias is not allocated at all
and not generated from file name. So we have to choices when szAlias
is NULL. Assing "" to it or generate RT error. The RT error should be
made in dbCreateTemp() function. In the C function hb_rddCreateTableTemp()
we should have:
if( szAlias == NULL )
szAlias = "";
3. pInfo.fTemporary is not initialized in hb_rddOpenTable() and
hb_rddCreateTable() what will cause very serious bugs.
Rest should be OK though I haven't look at it very deeply but I see one
problem with adding new member to DBOPENINFO.
In practice it means breaking binary compatibility so any existing 3-rd
party RDDs will stop to work and will have to be recompiled for new code.
Such modifications should be made only if they are strictly necessary and
we should group them with other modifications we plan to introduce to not
force unnecessary upgrades or create problems when user will try to use
old binaries with new Harbour code. I do not know how many 3-rd party RDDs
exist for Harbour. For sure MEDIATOR. If you know some others then please
send some information here.
Jacek used to read this list so there is a chance that he will release
new MEDIATOR version for Harbour quite fast. Anyhow I also want to make
some other modifications in RDD structures (f.e. reorder some members to
make them alignment independent) and such modification should be done
together. Probably it should be the part of type modifications which also
may interact with this code so I suggest to not introduce this one now.
It's also possible to not change DBOPENINFO but pass information about
temporary flag using DBI_* or RDDI_* interface. Such method is already
used in some places, f.e. in SIX3 to pass encryption password when table
is open or even in the function you created (hb_rddCreateTableTemp()) to
set field delimiters by DBI_SETDELIMITER. BTW pDelim is used in practice
only by DELIM export RDD and I do not know if it's necessary to keep it
in hb_rddCreateTableTemp(). You can remove it and add code which will
pass temporary flag as DBI_ISTEMPORARY, f.e.:
errCode = SELF_INFO( pArea, DBI_ISTEMPORARY, pTrueItm );
and change in dbf1.c:
case DBI_ISTEMPORARY:
hb_itemPutL( pItem, pArea->fTemporary );
break;
to:
case DBI_ISTEMPORARY:
if( !pArea->pDataFile && pArea->pMemoFile && HB_IS_LOGICAL( pItem ) )
pArea->fTemporary = hb_itemGetL( pItem );
else
hb_itemPutL( pItem, pArea->fTemporary );
break;
and do not change DBOPENINFO structure.
In such case we can introduce it even now of course after resolving the few
problem above.
best regards,
Przemek
_______________________________________________
Harbour mailing list
[email protected]
http://lists.harbour-project.org/mailman/listinfo/harbour
