hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jonathan Gray <jg...@facebook.com>
Subject RE: Fetching only Rowkey by scanning
Date Fri, 21 May 2010 15:21:14 GMT
There is no concept of a row/rowkey outside the context of a KeyValue.  A row exists when a
column exists in the row, therefore there is no way to just get each rowkey.  It would be
possible to add a new API but underneath it would be doing the same thing as the FirstKeyOnlyFilter,
so the only savings would be the network cost of the column/ts/value in the single KeyValue
per row, which for most is negligible.

> -----Original Message-----
> From: Patrick Datko [mailto:patrick.datko@ymc.ch]
> Sent: Friday, May 21, 2010 5:24 AM
> To: user@hbase.apache.org
> Subject: Re: Fetching only Rowkey by scanning
> 
> Hey Bob,
> 
> i was looking through the HBase API and didn't find anything that gives
> only the rowkey back. But why you need only the rowkey???
> I think the performance won't be bad if you got a KeyValue back instead
> of only the Key.
> 
> Patrick
> 
> Am Freitag, den 21.05.2010, 13:09 +0200 schrieb Bob Schulze:
> > > ..
> > > Am Freitag, den 21.05.2010, 10:54 +0200 schrieb Bob Schulze:
> > >> What is the best option to just return the rowkeys when scanning a
> table
> > >> range?
> > >> Is it enough, not to set any family or column in the Scan()?
> > >>
> > >>
> > Am 21.05.2010 11:13, schrieb Patrick Datko:
> > > Hey Bob,
> > >
> > > i think you can use the 'FirstKeyOnlyFilter', this should probably
> work
> > > and commit only the First KV of a row back.
> > >
> > Thx, Patrick,
> >
> > 	As I am looking for the RK only, isn't this still returning more
> than I
> > need? The KV does have also timestamp and other properties...
> >
> > Bob


Mime
View raw message