You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cassandra.apache.org by Chris Lohfink <cl...@gmail.com> on 2019/09/05 21:57:26 UTC

[DISCUSS] Server side CQL schema

In https://issues.apache.org/jira/browse/CASSANDRA-14825 its been discussed
on how a server side DDL can be provided to clients.

The original thought was to use a virtual table to make it less invasive of
a change while providing access to the TableCQLHelper output (and fixing
it).

The other is to add a DESCRIBE command to the cql protocol or query parser
so that the command would act same across all drivers as it does in cqlsh.

In both cases it would return a ResultSet so from drivers perspective it
wont be too much different.

Since the original thought was a virtual table the existing PR does that so
I lean in that direction. The virtual table is only 300 or so lines of code
too so its also smaller incremental step than changing protocol or grammar
(which can still happen, as mentioned in ticket "why not both").

There is no consensus in ticket after nearly a year, so to get the review
moving on Id like to open up here for further/final discussion and or vote.

Re: [DISCUSS] Server side CQL schema

Posted by Chris Lohfink <cl...@gmail.com>.
It’s not that complicated and can make any work. There is just no consensus
on which of the three options and don’t wanna go back and forth as
different people review it.

Chris

On Thursday, September 5, 2019, Dinesh Joshi <dj...@apache.org> wrote:

> Reading the ticket, I see 3 options being discussed.
>
> 1. Implement Server Side DESCRIBE
> 2. Virtual Table (current patch)
> 3. Go with Virtual Table now, implement Server Side DESCRIBE later (as it
> is invasive)
>
> Is it complicated to port your current patch to implement DESCRIBE instead
> of a virtual table?
>
> Dinesh
>
> > On Sep 5, 2019, at 2:57 PM, Chris Lohfink <cl...@gmail.com> wrote:
> >
> > In https://issues.apache.org/jira/browse/CASSANDRA-14825 its been
> discussed
> > on how a server side DDL can be provided to clients.
> >
> > The original thought was to use a virtual table to make it less invasive
> of
> > a change while providing access to the TableCQLHelper output (and fixing
> > it).
> >
> > The other is to add a DESCRIBE command to the cql protocol or query
> parser
> > so that the command would act same across all drivers as it does in
> cqlsh.
> >
> > In both cases it would return a ResultSet so from drivers perspective it
> > wont be too much different.
> >
> > Since the original thought was a virtual table the existing PR does that
> so
> > I lean in that direction. The virtual table is only 300 or so lines of
> code
> > too so its also smaller incremental step than changing protocol or
> grammar
> > (which can still happen, as mentioned in ticket "why not both").
> >
> > There is no consensus in ticket after nearly a year, so to get the review
> > moving on Id like to open up here for further/final discussion and or
> vote.
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
> For additional commands, e-mail: dev-help@cassandra.apache.org
>
>

Re: [DISCUSS] Server side CQL schema

Posted by Dinesh Joshi <dj...@apache.org>.
Reading the ticket, I see 3 options being discussed.

1. Implement Server Side DESCRIBE
2. Virtual Table (current patch)
3. Go with Virtual Table now, implement Server Side DESCRIBE later (as it is invasive)

Is it complicated to port your current patch to implement DESCRIBE instead of a virtual table?

Dinesh

> On Sep 5, 2019, at 2:57 PM, Chris Lohfink <cl...@gmail.com> wrote:
> 
> In https://issues.apache.org/jira/browse/CASSANDRA-14825 its been discussed
> on how a server side DDL can be provided to clients.
> 
> The original thought was to use a virtual table to make it less invasive of
> a change while providing access to the TableCQLHelper output (and fixing
> it).
> 
> The other is to add a DESCRIBE command to the cql protocol or query parser
> so that the command would act same across all drivers as it does in cqlsh.
> 
> In both cases it would return a ResultSet so from drivers perspective it
> wont be too much different.
> 
> Since the original thought was a virtual table the existing PR does that so
> I lean in that direction. The virtual table is only 300 or so lines of code
> too so its also smaller incremental step than changing protocol or grammar
> (which can still happen, as mentioned in ticket "why not both").
> 
> There is no consensus in ticket after nearly a year, so to get the review
> moving on Id like to open up here for further/final discussion and or vote.


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@cassandra.apache.org
For additional commands, e-mail: dev-help@cassandra.apache.org