Delete midrange of records with DBU - Clipper

Donald,

>Is there a way to delete a range of corrupt records in the middle of a dbf
>file with DBU or a similar utility?

Make a copy of the dbf to experiment with. If the records are
corrupted in the sense that there are garbage characters that Clipper
chokes on when the record is read into variables, then I would try to
find out which records are affected and write a little program that
deletes them based on the record number. Then copy to a new dbf for
!deleted().

>A chunk of these records in the same DBF are duplicated - any ideas of how
>this corruption could occur?

This sounds like either a PACK attempt that crashed in the middle of
the operation (IMO this is the prime suspect and the exact reason why
PACK should never, never ever be used) or a hardware related problem.
I have seen examples of the latter kind where whatever happened to be
in a cache was written to disk on top of an existing dbf.

Regards,
Klas

-------
klas dot engwall at engwall dot com

http://www.engwall.com/clipper/

The LFN Library for Clipper
The LanMan Library for Clipper
The NFPAT1A Timeslice release patch for the Nanforum Toolkit

Thanks Klas

What would the core of this program look like? Forcing physical deletion?
Pack?!!!

Donald.

Donald

// This presumes you know which records are corrupt
// you'll need a tool like BRO (maybe DBU will do) to get the record numbers
// ie you'll have to do something by hand as there is no automatic way of
identifying corrupt records

aArray := { 1, 2, 33, 55, 78, 99 }
for i := 1 to len( aArray )
x->( DBGoto( aArray[1] )
x->( DBDelete() )
next
PACK

What I used to do was
COPY TO xyz
and after it failed checked the last record written to the xyz.dbf.
Load up the crook dbf in BRO and delete that record and any I thought was
corrupt around that record (+50-100)
Repeat the COPY TO until it completed successfully.

You'll find it easier to delete any indexes involved as well.

HTH
Steve

Donald,

>What would the core of this program look like? Forcing physical deletion?

That depends on the type of corruption, and you haven't said much
about that yet. If it is just the duplicate records you mentioned
yesterday, then all you have to do is identify them and delete() them
plus copy to a new file for !deleted() as the final step.

If there is garbage in the records you will have to try different
solutions until you find one that works. The ones Steve mentioned, for
example, and if nothing else helps maybe calculating the offsets of
the corrupted records in the file and fwrite()ing space characters as
replacement for the garbage. It is diffifult to say how much brute
force you will have to use without really knowing what the problem is.

But in any case, first you have to identify the record numbers of the
corrupted records. And as I said yesterday, make a copy of the file
before you do anything at all!!

>Pack?!!!

You're kidding, right? :-)

Regards,
Klas

-------
klas dot engwall at engwall dot com

http://www.engwall.com/clipper/

The LFN Library for Clipper
The LanMan Library for Clipper
The NFPAT1A Timeslice release patch for the Nanforum Toolkit

Steve,

>aArray := { 1, 2, 33, 55, 78, 99 }
>for i := 1 to len( aArray )
> x->( DBGoto( aArray[1] )

perhaps x->( DBGoto( aArray[i] ) would be more efficient :-)

> x->( DBDelete() )
>next
>PACK

Oh no, Steve, you said the "P" word right after I tried to push Donald
away from that and show him copy for !deleted() instead :-)

The duplicate records he said that he has in the middle of the file
sounds suspiciously like there has been a PACK crash already.

Regards,
Klas

-------
klas dot engwall at engwall dot com

http://www.engwall.com/clipper/

The LFN Library for Clipper
The LanMan Library for Clipper
The NFPAT1A Timeslice release patch for the Nanforum Toolkit

Klas

> Oh no, Steve, you said the "P" word right after I tried to push Donald
> away from that and show him copy for !deleted() instead :-)
I've never had a problem using pack on my machine for maintenance purposes <g>
- never used it in an app though IIRC as I used to recycle deleted records

> The duplicate records he said that he has in the middle of the file
> sounds suspiciously like there has been a PACK crash already.
That may have been an OS/Machine/Disk space/Network problem while 'packing'
though.

CYA
Steve

"Donald Leask" <vet@svh.co.za> wrote in message
news:e9o278$k2i$1@ctb-nnrp2.saix.net...
> Is there a way to delete a range of corrupt records in the middle of a dbf
> file with DBU or a similar utility?
>
> A chunk of these records in the same DBF are duplicated - any ideas of how
> this corruption could occur?
>
> Thanks,
>
> Donald Leask
> Johannesburg.
>
>

to clean dbf with dbu i use:

a] with dbu detect first/last recno of corrupted area
b] then, copy to temporary file [without any index], with clause

recno()<first_corrupted .or. recno()>last_corrupted

if you have few corrupted areas, repeat steps a] and b], you have to copy in
chain, always using previously copied temp file.

at the end, backup orig dbf, and replace it with last temp file, and
recreate indices

Steve,

>> Oh no, Steve, you said the "P" word right after I tried to push Donald
>> away from that and show him copy for !deleted() instead :-)
>I've never had a problem using pack on my machine for maintenance purposes <g>
> - never used it in an app though IIRC as I used to recycle deleted records
>
>> The duplicate records he said that he has in the middle of the file
>> sounds suspiciously like there has been a PACK crash already.
>That may have been an OS/Machine/Disk space/Network problem while 'packing'
>though.

Or ...
.... user's knee hitting the reset button under the table
.... user's foot getting entangled in the power cord
.... user's co-worker blowing the fuse
.... user turning off the computer to go home
.... non-thinking admin rebooting the server

The list goes on and on <g>

Regards,
Klas

-------
klas dot engwall at engwall dot com

http://www.engwall.com/clipper/

The LFN Library for Clipper
The LanMan Library for Clipper
The NFPAT1A Timeslice release patch for the Nanforum Toolkit

This works just fine - thanks!

BTW what would a DO WHILE loop to delete a range of records look like....
with variables nRangeStart and nRangeEnd?

Thanks,
Donald.

"sali" <sali@tel.net.ba> wrote in message
news:e9u5o0$b5j$1@ls5.tel.net.ba...
> "Donald Leask" <vet@svh.co.za> wrote in message
> news:e9o278$k2i$1@ctb-nnrp2.saix.net...
>> Is there a way to delete a range of corrupt records in the middle of a
>> dbf
>> file with DBU or a similar utility?
>>
>> A chunk of these records in the same DBF are duplicated - any ideas of
>> how
>> this corruption could occur?
>>
>> Thanks,
>>
>> Donald Leask
>> Johannesburg.
>>
>>
>
> to clean dbf with dbu i use:
>
> a] with dbu detect first/last recno of corrupted area
> b] then, copy to temporary file [without any index], with clause
>
> recno()<first_corrupted .or. recno()>last_corrupted
>
> if you have few corrupted areas, repeat steps a] and b], you have to copy
> in
> chain, always using previously copied temp file.
>
> at the end, backup orig dbf, and replace it with last temp file, and
> recreate indices
>
>
>
>
>