Adding rewriter in HiveDriver should not be an issue. Will be different
jira than LENS-24.

Thanks
Amareshwari

On Wed, May 13, 2015 at 5:43 PM, Vikas Singh (Tech - VS) <
[email protected]> wrote:

> Hi Amareshwari
>
> Thanks for the reply.
> Just to set the context, According to the above patch the rewrite flow will
> be-
>
>             query  --> rewriteToCubeQL --> rewriteToHQL
>     Here rewriteToHQL does the cube rewrites which are similar to previous
> functionality. So the hook is added before any Cube rewrite is happening
> (before any actual fact table, storage table, candidate table resolution
> happens). In our use case we need rewrite to happen after all these
> resolutions happen. So having a rewriter at HQL Driver level, analogous to
> JDBC Driver will give flexibility to do rewrite at the CubeLevel as well as
> at the driver level (based on actual storage table, candidate tables).
>
>     So I propose adding a rewriter to HiveDriver as well. Please share your
> thoughts.
>
> Regards
> Vikas Singh
>
> On Thu, Apr 30, 2015 at 11:30 AM, amareshwarisr . <[email protected]>
> wrote:
>
> > Hello Vikas,
> >
> > Sorry missed replying this mail.
> >
> > From the requirement put up, i see query needs to be rewritten with where
> > clause appended. For plugging in any new rewriter, there is open issue
> > LENS-24 (https://issues.apache.org/jira/browse/LENS-24) and the changes
> > put
> > up at https://reviews.apache.org/r/29110/. The changes need a rebase
> with
> > master. One the pluggable rewriter is available, plugging in AuthRewriter
> > should be simple.
> >
> > Thanks
> > Amareshwari
> >
> >
> > On Thu, Apr 23, 2015 at 11:16 AM, Vikas Singh (Tech - VS) <
> > [email protected]> wrote:
> >
> > > Problem Statement: Implementing Authorisation filter in lens.
> > >
> > >     The way we do Row level authorisation is -
> > > For row level security, we have auth_db which stores map of significant
> > > columns(In which we are interested to do row level security) and
> allowed
> > > values per user. During run time we get all the Facts associated in
> > query,
> > > get their column set and do a intersection with set returned from
> > auth_db.
> > > Resulting set is appended as "IN Clause" or column!=value to the query
> > > accordingly.
> > >
> > >     My guess is the best place to do it will be driver ReWriter level
> as
> > it
> > > will handle CubeQl and bypassed HQL seamlessly. I was wondering if
> there
> > is
> > > any better place to do this, i.e. any hooks available.
> > >
> > > Regards
> > > Vikas Singh
> > >
> >
>

Reply via email to