|Subject:||Re: table repair quirk
|Date:||Mon, 26 Nov 2018 11:27:05 +0000
|From:||Michael Kennedy <Info@KennedySoftware.ie>
Is it weird that 'Table-Repair' throws the error - surely that fellow
should just read the raw data, do basic checks on the fields, and build
all the indices, etc, and not worry about a data-record being in the
wrong 'sort-order'? ;-)
Any chance that the 'collating-sequences' / 'sort orders' are mixed up -
among the indices, or inconsistent between the actual data and the
When you split the table into sub-tables (as we mentioned previously),
did the issue:
#1: Disappear (therefore it might be a 'volume' issue, and, perhaps
also related to the 24 sec-indices), or
#2: Arise in just one split (therefore, seems like data-dependent,
and probably best to just repeatedly split the table, until the
offending record(s) are identified), or
#3: Arise in all splits (seems like a bug in the DB engine, or in
Table-repair, perhaps related to that huge number of sec-indices)!
If #2, and if you have a very small table with the error, I'll gladly
take a look at its 'insides' - if that might help.
On 26/11/2018 04:08, Kevin Zawicki wrote:
> May sound stupid, but can you move it to another PC and test?
> I would copy table:
> on copy delete first 12 indexes, test, if error, then delete first 6 of 12,
> Going in halves, try to find the index causing it..
> I have seen chimney sweep fix things rebuilds can't and vice versa.
> This does sound like you hit some limit.
> Is the data sound? any high ascii?
> "Steven Green" <firstname.lastname@example.org> wrote:
>> thanks.. been there, did all of those things.. assumed I'd find the
>> offending record and delete it.. then it started moving :-)
>> it's gotta be a bad pointer in one of the indexes, but with 24 to check
>> manually, it's not worth the effort, for now.. that's the biggest issue
>> these old apps.. all the eff-ing indexes.. it was so cutting-edge, so
>> simple.. make an index for everything.. 20 years later, you have 400 files
>> open at once, hundreds of thousands of records, and everyone beating on
>> of it, all day long
>> Steven Green
>> Myrtle Beach, South Carolina, USA
>> Collectibles and Memorabilia
>> Vintage Lego Sets and Parts
>> - and Paradox support, too
>> "Michael Kennedy" wrote in message
>> A few thoughts...
>> To me, 80,000 does not seem large, but 24 does ;-)
>> How about copying the table structure only, and then "adding" all the
>> current data into the new structure, etc...
>> Or... How about splitting the current table into, say, three sub-tables,
>> with about 27,000 records each (well under that darn 32,768 limit!), and
>> then running Repair on each sub-table. If each sub-table is error-free,
>> that might suggest that the data in the individual records is not the
>> problem, but perhaps the "volume" of records, and you might have hit a
>> variation of that darn 32,768 issue among all those secondary indices.
>> Also, if you have ChimneySweep handy, maybe it's worth a shot? (However,
>> I did try/stress it (a lot!), some years ago, with no success!).
>> - Mike
>> On 25/11/2018 14:29, Steven Green wrote:
>>> huge table.. 80,000 records, 24 sec indexes, a couple of memo fields..
>>> I didn't create this monster
>>> repair says record 28728 is out of sort order.. rebuild, still there..
>>> delete it, now the record BEFORE it is now out of sort order, not the
>>> record AFTER it
>>> stranger still, there's seemingly nothing wrong with the record that stops
>>> anything.. I can add it back to the table w/o an error, I can edit it
>>> leave the record w/o an error, so none of the lookup values is invalid
>>> Steven Green
>>> Myrtle Beach, South Carolina, USA
>>> Collectibles and Memorabilia
>>> Vintage Lego Sets and Parts
>>> - and Paradox support, too