Hi, Teryl,

Good points.  Thanks for your offer to help with testing.  I should be 
able to get around to implement something in a day or two.

Regarding the possiblity of using the tablex interface, my concern is 
not to make the table object extensible, but simply to write it. 
Seems like just having a write function is a little better.  The 
concern is that it might clash with a future class that derives from 
both table and tablex.  In this case, the solution might be to use a 
different function name, say, backup or dump, instead of write.  Any 
comments?

Regarding the column names, since a column name like 'sum(a)' is not 
usable in future queries -- it will be interpreted as a function on 
column named a.  We will have to rename them to something else.  If 
there is an alias for it already in the select clause, it will be 
used, otherwise, it will be up to us to rename it to something else. 
One suggestion is simply to change all characters that are not allowed 
in a SQL column name to be the underscore character _.  Anyone has a 
better suggestion?

John


On 4/7/2010 11:12 AM, Teryl Taylor wrote:
> Hi John,
>
> You might be able to create a constructor on the ibis::tablex object
> which takes an ibis::table and then just use the write function on
> tablex if you want to keep the read/write separated between the two
> objects.    But some sort of a write function like in tablex should work
> well.  How would you handle columns that are made up of an aggregate
> function...so for instance SUM(A).  Would the column on disk just be
> called SUM(A)?
>
> Let me know if/when you do implement something like this,  I can
> probably help test it for you.  I'm writing a little tool that does
> queries out of a set of partitions, and it would be great to be able to
> drop the results into a temporary partition and then read that partition
> or join as needed.
>
> Best Regards,
>
> Teryl
>
>
>
> On Tue, Apr 6, 2010 at 8:38 PM, K. John Wu <[email protected]
> <mailto:[email protected]>> wrote:
>
>     Hi, Teryl,
>
>     Good point.  The functionality of writing out an existing table is
>     not there and will be added shortly.  I am planning to use the same
>     function name and prototype as ibis::tablex::write.  Do you have any
>     comment or suggestion?
>
>     John
>
>
>
>     On 4/6/2010 12:09 PM, Teryl Taylor wrote:
>
>         Hi,
>
>         I am doing my queries on a list of partitions using the select
>         method on
>         an ibis::table object.   Is there an easy way to take the new table
>         returned from a select or groupby method, and write it to a
>         temporary
>         partition or set of partitions on disk?
>
>
>         Regards,
>
>         Teryl
>
>
>
>         _______________________________________________
>         FastBit-users mailing list
>         [email protected] <mailto:[email protected]>
>         https://hpcrdm.lbl.gov/cgi-bin/mailman/listinfo/fastbit-users
>
>
_______________________________________________
FastBit-users mailing list
[email protected]
https://hpcrdm.lbl.gov/cgi-bin/mailman/listinfo/fastbit-users

Reply via email to