On 02/06/2013 05:13 PM, Douglas Smith wrote:
> #2: data loading - wait didn't we do the source partitioning and loading
> also
> yet?  Um, looking at the Source.sql in the pt11 is looks like what you need
> is: "--theta-column 6 --phi-column 9".  The rest of the options should work
> fine.  Actually I think I looked that before, and the other columns
> would also
> work, it seems that there are repeated ra and dec fields in the table,and
> looking at the data quickly the values appear to be the same.
>
We talked about this on the call, but let's put it in writing so we're 
less likely to forget.

Looking at Source.sql in pt11, column 6 and 9 are NOT what you want, 
although it may accidentally work for now. Remember that we are not 
partitioning tables in isolation. For LSST catalogs, there is always one 
single table that is "in charge," so to speak. For PT11 and 
production-targeted data, this table is "Object". Object is partitioned 
according to a particular ra and decl, and every other table that is 
partitioned must be partitioned using the same values of ra and decl 
from the Object table. In Source, these are raObject and declObject. 
This is done so that Source can be joined with Object using objectId. In 
theory, you could leave the (raObject and declObject) columns out, but 
you would need to join Source with Object on objectId in order to get 
the right coordinates for partitioning, which is just way too expensive. 
Thus the same code that creates source rows populates raObject and 
declObject along with objectId.

For Summer12 and Winter13 (so far), the data is not production-targeted. 
The schema is different, and there was no attempt (as far as I know) to 
use a production-like schema. In these cases, you need to be told which 
is the table "in charge," and which columns in the other tables are the 
appropriate ra and decl columns to use.

-Daniel

########################################################################
Use REPLY-ALL to reply to list

To unsubscribe from the QSERV-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1