The 0991 error seems to be quite common when loading statistics from an
external file. It just means that the value of the repetition factor in the
file doesn't quite tally with the value computed from the histogram counts.
It is a sanity check, but seems to often run afoul of floating point
roundoff. I guess we should fix it to be a little more selective before
The 095F is also a sanity check. The "unique" flag is set during the
original histogram build if 90% or more of the rows have a distinct value in
the column. If the value of sunique on the iistatistics catalog is manually
overriden, or again because possible floating point rounding problems change
the recomputed value, the flag in the file being loaded may be found to be
inconsistent with the data.
In both cases, the messages are warnings and the data is still stored as it
was in the external file. In particular, the unique flag is stored as
displayed in the file, even if the 095F warning is issued.
> -----Original Message-----
> Sent: Sunday, October 08, 2000 7:19 AM
> To: 'i_ml'
> Subject: optimizedb -i warning
> When I load optimizer statistics from a text file( optimizedb -i ), I get
> lot of warning messages. What does it mean? Is there anything wrong with
> file? It seems the values in the file are not compliant with the tables.
> the text file is just generated by "statdump" from the same database and I
> change nothing of these tables since the statdump.
> The messages are like:
> W_OP0991 The repetition factor value ` 10480000.00000e-007'
> forcolumn `ADRQ_REQUEST_DATETIME' in table `ADRQ_ADMISSION_REQUEST' is not
> sufficiently close to tuple count divided by number of unique values.
> W_OP095F The unique flag value `Y' specified for column
> in table `AN_CUMCUST' does not conform to optimizer rules.