This is awesome, thank you Daniel. I agree that focusing on enough SQL for
TPCH queries would be a great idea and way to focus our efforts.

Subqueries may be the largest remaining outstanding item that I see -- I
have some ideas of how to implement them on the planner side if others are
interested in collaborating.

Andrew

On Wed, Feb 3, 2021 at 4:02 PM Andy Grove <andygrov...@gmail.com> wrote:

> Thanks for the update on this, Daniël. It is great to see the progress with
> this!
>
> Perhaps it is worth creating one JIRA issue per failing query detailing the
> errors and we can link these to the issues that are causing the failures?
>
> On Wed, Feb 3, 2021 at 1:57 PM Mike Seddon <seddo...@gmail.com> wrote:
>
> > Hi Daniël,
> >
> > I am working on 22 as part of https://github.com/apache/arrow/pull/9243
> >
> > We also need to convert all the Float64 schema types to Decimal(n).
> >
> > Cheers,
> > Mike
> >
> > On Thu, Feb 4, 2021 at 5:44 AM Daniël Heres <danielhe...@gmail.com>
> wrote:
> >
> > > Hey all,
> > >
> > > Quite some features have been added to DataFusion in the last couple of
> > > months.
> > >
> > > One test of the functionality we support this is the TPC-H benchmark.
> We
> > > now can run 7 out of 22 queries without errors.
> > > I think a nice goal would be having complete support for the full
> suite,
> > as
> > > it means a lot of functionality is included, helps optimization and
> helps
> > > us to test against other engines.
> > >
> > > These queries fail currently because of missing features or bugs:
> > >
> > > * 2 IN (Subquery) in (WHERE) expression
> > > * 4 Intervals https://github.com/apache/arrow/pull/9373
> > > * 7 Fails with error "Schema contains duplicate unqualified field name
> > > \'n_nationkey\'" https://issues.apache.org/jira/browse/ARROW-11432
> > > * 8 Fails with error "Schema contains duplicate unqualified field name
> > > \'n_nationkey\'" https://issues.apache.org/jira/browse/ARROW-11432
> > > * 9 Fails with error "Cartesian joins are not supported"
> > > * 11 HAVING support  https://github.com/apache/arrow/pull/9364 (but
> also
> > > requires IN (subquery) in expression)
> > > * 13 Filters in JOIN condition: "Unsupported expression \'NotLike\' in
> > JOIN
> > > condition"
> > > * 14 CASE WHEN expressions are not coerced yet, query fails with error
> > > "false_values downcast failed"
> > > * 15 VIEW/multiple statement support: "The context currently only
> > supports
> > > a single SQL statement"
> > > * 16 IN (Subquery) in (WHERE) expression
> > > * 17 Subquery in (WHERE) expression
> > > * 18 IN (Subquery) in (WHERE) expression
> > > * 19 Fails with error "Cartesian joins are not supported"
> > > * 20 IN (Subquery) in (WHERE) expression
> > > * 21 Compound identifier not supported: "Unsupported compound
> identifier
> > > \'[\"l1\", \"l_suppkey\"]"
> > > * 22 Fails with parser error for the syntax SUBSTRING(col FROM 1)
> > >
> > > FOR 2)
> > > Other functionality not causing failures now, but needed:
> > >
> > > * EXTRACT https://github.com/apache/arrow/pull/9359
> > > * EXISTS
> > >
> > > Am I missing any JIRA issues / PRs or features in this list? I would
> like
> > > to create some issues on JIRA so we can tackle this.
> > >
> > > Best regards,
> > >
> > > Daniël
> > >
> >
>

Reply via email to