On Tue, 12 Feb 2008, Patrick Mast wrote:
> Hello Przemek,
> Did you receive my private mail regarding "Eof == .T. on Phantom record"?
> Thanks!
It's normal Clipper behavior replicated in [x]Harbour:
proc main(rdd)
rddSetDefault(iif(empty(rdd),"DBFCDX",rdd))
? rddSetDefault()
dbCreate("_tst",{{"F1","C",1,0}})
use _tst
? bof(), eof(), lastrec(), recno()
dbSkip(-1)
? bof(), eof(), lastrec(), recno()
return
In the above code we are at phantom record but after dbSkip(-1)
the EOF flag is cleared. In such way work all CL52, SIX3 and CL53
RDDs. Also [x]Harbuor's DBF, DBFCDX, DBFNTX. Only our ADS RDD
do not always respect it (f.e. in this example) and this is side
effect of code which hides top phantom record.
I said many times in the past that EOF() checking to detect
phantom record is not enough even in standard Clipper RDDs.
Additionally some RDDs may not support phantom records at
all or can have both: top and bottom ones, f.e. ADS has both
(see above) though we are hiding this in our code. I suggested
to introduce new function like isPahntomRec() to detect such
situation. Of course we can also change current code to not
be Clipper compatible. It will be enough to change the
condition which exists at the end of SKIP methods:
/* Update Bof and Eof flags */
if( fForward )
pArea->fBof = FALSE;
else
pArea->fEof = FALSE;
and update only fBof flag but it may introduce some other problems
for code which strictly need Clipper compatibility though I do not
think if many people knows how exactly this flag are updated by Clipper.
best regards,
Przemek
_______________________________________________
Harbour mailing list
Harbour@harbour-project.org
http://lists.harbour-project.org/mailman/listinfo/harbour
