I am trying to limit scans to a subset of stored rows by setting ranges.
 The scan has an iterator that does some aggregation and transformation of
the result.  Specifically, it changes the rowid to be more meaningful but
at the same time outside of the specified range.  By instrumenting the
iterator with debug messages, I can verify that the expected rows are
actually processed.  But it appears that the range filter is applied to the
result of the transformation that the iterator is making.  Basically, I'm
trying to get something like a sql projection:

select myTransformation(id) from myTable where id between (start and end);

Is it possible to set the precedence of the range filter and the iterator
that I am using so that I can do this transformation without filtering all
the results?

Thanks,
Anthony

The last time I looked at that code, every resulting key was silently
removed by the TabletServerBatchReader(Iterator).

That being said, it was previously impossible to turn off that final range
check as you asked. I'd have to double check the code again, though.

Would it be possible in your application to place your aggregated
information in some serialized representation inside of the Value and
preserve the original Key?

On Monday, November 12, 2012, Anthony Fox <[EMAIL PROTECTED]> wrote:
> I am trying to limit scans to a subset of stored rows by setting ranges.
 The scan has an iterator that does some aggregation and transformation of
the result.  Specifically, it changes the rowid to be more meaningful but
at the same time outside of the specified range.  By instrumenting the
iterator with debug messages, I can verify that the expected rows are
actually processed.  But it appears that the range filter is applied to the
result of the transformation that the iterator is making.  Basically, I'm
trying to get something like a sql projection:
> select myTransformation(id) from myTable where id between (start and end);
> Is it possible to set the precedence of the range filter and the iterator
that I am using so that I can do this transformation without filtering all
the results?
>
> Thanks,
> Anthony

Yes, we can definitely work around it as you suggest, though the
transformed key/value fit the semantics of the domain a bit better.  Since
we're using the intersecting iterator, I'm trying to hide the random
partition row id from a consumer of this API.
On Mon, Nov 12, 2012 at 9:44 AM, Josh Elser <[EMAIL PROTECTED]> wrote:

> The last time I looked at that code, every resulting key was silently
> removed by the TabletServerBatchReader(Iterator).
>
> That being said, it was previously impossible to turn off that final range
> check as you asked. I'd have to double check the code again, though.
>
> Would it be possible in your application to place your aggregated
> information in some serialized representation inside of the Value and
> preserve the original Key?
>
> On Monday, November 12, 2012, Anthony Fox <[EMAIL PROTECTED]> wrote:
> > I am trying to limit scans to a subset of stored rows by setting ranges.
>  The scan has an iterator that does some aggregation and transformation of
> the result.  Specifically, it changes the rowid to be more meaningful but
> at the same time outside of the specified range.  By instrumenting the
> iterator with debug messages, I can verify that the expected rows are
> actually processed.  But it appears that the range filter is applied to the
> result of the transformation that the iterator is making.  Basically, I'm
> trying to get something like a sql projection:
> > select myTransformation(id) from myTable where id between (start and
> end);
> > Is it possible to set the precedence of the range filter and the
> iterator that I am using so that I can do this transformation without
> filtering all the results?
> >
> > Thanks,
> > Anthony
>

Alternatively you could strip the partition id off of the returned key
before the data is accessed via your client API.

On Mon, Nov 12, 2012 at 7:29 AM, Anthony Fox <[EMAIL PROTECTED]> wrote:

> Yes, we can definitely work around it as you suggest, though the
> transformed key/value fit the semantics of the domain a bit better.  Since
> we're using the intersecting iterator, I'm trying to hide the random
> partition row id from a consumer of this API.
>
>
> On Mon, Nov 12, 2012 at 9:44 AM, Josh Elser <[EMAIL PROTECTED]> wrote:
>
>> The last time I looked at that code, every resulting key was silently
>> removed by the TabletServerBatchReader(Iterator).
>>
>> That being said, it was previously impossible to turn off that final
>> range check as you asked. I'd have to double check the code again, though.
>>
>> Would it be possible in your application to place your aggregated
>> information in some serialized representation inside of the Value and
>> preserve the original Key?
>>
>> On Monday, November 12, 2012, Anthony Fox <[EMAIL PROTECTED]> wrote:
>> > I am trying to limit scans to a subset of stored rows by setting
>> ranges.  The scan has an iterator that does some aggregation and
>> transformation of the result.  Specifically, it changes the rowid to be
>> more meaningful but at the same time outside of the specified range.  By
>> instrumenting the iterator with debug messages, I can verify that the
>> expected rows are actually processed.  But it appears that the range filter
>> is applied to the result of the transformation that the iterator is making.
>>  Basically, I'm trying to get something like a sql projection:
>> > select myTransformation(id) from myTable where id between (start and
>> end);
>> > Is it possible to set the precedence of the range filter and the
>> iterator that I am using so that I can do this transformation without
>> filtering all the results?
>> >
>> > Thanks,
>> > Anthony
>>
>
>