accumulo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Devin Pinkston <devinfpinks...@gmail.com>
Subject Re: Column Scan / table metadata
Date Thu, 19 Sep 2013 11:09:31 GMT
Thanks for all of the input guys, I really appreciate it!


On Wed, Sep 18, 2013 at 11:16 PM, David Medinets
<david.medinets@gmail.com>wrote:

> How would you define 'modestly-sized tables'? Are you thinking of an
> absolute number like 100 Billion entries or some number of entries per
> tablet? Or perhaps a time estimate - like a map-reduce job takes 60 minutes
> to scan the table?
>
>
> On Wed, Sep 18, 2013 at 2:57 PM, Josh Elser <josh.elser@gmail.com> wrote:
>
> > There isn't a reliable way to ascertain the column set for a table via
> the
> > Accumulo API.
> >
> > Scanning all of the keys in a table would work; however, this quickly
> > becomes too costly to perform for modestly sized tables.
> >
> > An easy way to manage this is to build up the set of columns as part of
> > your "ingest" code and store them in Accumulo (a separate table is
> > easiest). By adding a quick cache to your ingest code, you can track a
> > column schema without much extra effort or cost.
> >
> >
> > On Wed, Sep 18, 2013 at 2:42 PM, Devin Pinkston <
> devinfpinkston@gmail.com
> > >wrote:
> >
> > > I have been looking through the Accumulo source to try and find the
> best
> > > way to derive the column structure/metadata of a table.  If I have a
> > table
> > > "sample", and I want to find all the column families/qualifiers, is
> > there a
> > > built-in facility in Accumulo to get a list of columns in that table?
>  Or
> > > would my best option be to scan() the entire table, and only put unique
> > > column families/qualifiers into a list and return to the user?
> > >
> > > I am imagining the user has no idea of what their columns are like in
> > this
> > > table, that is why I ask.
> > >
> > > Thanks!
> > >
> >
>



-- 
Devin Pinkston
(703) 727-4252
devinfpinkston@gmail.com

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message