Integrate Calcite and JDBC

[Julian Feinauer <j....@pragmaticminds.de>]

Hi all,

there has been quite a bit of discussion this at the beginning but its become quiet about that.
I think it would be great to move the JDBC Implementation to Calcite for multiple reasons.

First, you get support for (nearly) everything in the SQL Standard.
Second, you don’t have to maintain that.
Third, we could even consider adding indices or dialects later on to improve the performance drastically.

But the first step, for me, would be to revisit how JDBC is currently implemented and how the syntax maps SQL to the IoTDB data model.
Then, I will, if my time allows it, happily help with the implementation.

Julian

AW: Integrate Calcite and JDBC

[Julian Feinauer <j....@pragmaticminds.de>]

Hey,

I agree although I think long term we should move to calcite like nearly all apache projects did.
You can even have your special syntax there.

But in the meantime is fine to have two implementations.

Julian

Von meinem Mobiltelefon gesendet


-------- Ursprüngliche Nachricht --------
Betreff: Re: Integrate Calcite and JDBC
Von: Xiangdong Huang
An: dev@iotdb.apache.org
Cc:

Hi,

I think it is a good idea.
Current JDBC and service-rpc module should be improved in my opinion..
For example, the SQL parser takes about 40% time cost for an "insert"
command.. (I remember that there is an issue mentioned that..).
Second, PreparedStatement is not implemented in current version (even in
the master branch).

Using Calcite is a kind of good choice, and for achieving that we need to
define how to map current SQL syntax to the standard SQL syntax.

By the way, I think it is also fine to keep maintaining the two JDBC and
SQL implementations (one for the current SQL and the other for the standard
SQL) for a while.
The implementation which has the best performance will be the final first
class citizen (Or, different users can use choose different implementation,
if someone prefer to the performance while others prefer the standard sql).


Best,
-----------------------------------
Xiangdong Huang
School of Software, Tsinghua University

 黄向东
清华大学 软件学院


Julian Feinauer <j....@pragmaticminds.de> 于2019年8月6日周二 上午5:59写道：

> Hi all,
>
> there has been quite a bit of discussion this at the beginning but its
> become quiet about that.
> I think it would be great to move the JDBC Implementation to Calcite for
> multiple reasons.
>
> First, you get support for (nearly) everything in the SQL Standard.
> Second, you don’t have to maintain that.
> Third, we could even consider adding indices or dialects later on to
> improve the performance drastically.
>
> But the first step, for me, would be to revisit how JDBC is currently
> implemented and how the syntax maps SQL to the IoTDB data model.
> Then, I will, if my time allows it, happily help with the implementation.
>
> Julian
>

Re: Integrate Calcite and JDBC

[Xiangdong Huang <sa...@gmail.com>]

Hi,

I think it is a good idea.
Current JDBC and service-rpc module should be improved in my opinion..
For example, the SQL parser takes about 40% time cost for an "insert"
command.. (I remember that there is an issue mentioned that..).
Second, PreparedStatement is not implemented in current version (even in
the master branch).

Using Calcite is a kind of good choice, and for achieving that we need to
define how to map current SQL syntax to the standard SQL syntax.

By the way, I think it is also fine to keep maintaining the two JDBC and
SQL implementations (one for the current SQL and the other for the standard
SQL) for a while.
The implementation which has the best performance will be the final first
class citizen (Or, different users can use choose different implementation,
if someone prefer to the performance while others prefer the standard sql).


Best,
-----------------------------------
Xiangdong Huang
School of Software, Tsinghua University

 黄向东
清华大学 软件学院


Julian Feinauer <j....@pragmaticminds.de> 于2019年8月6日周二 上午5:59写道：

> Hi all,
>
> there has been quite a bit of discussion this at the beginning but its
> become quiet about that.
> I think it would be great to move the JDBC Implementation to Calcite for
> multiple reasons.
>
> First, you get support for (nearly) everything in the SQL Standard.
> Second, you don’t have to maintain that.
> Third, we could even consider adding indices or dialects later on to
> improve the performance drastically.
>
> But the first step, for me, would be to revisit how JDBC is currently
> implemented and how the syntax maps SQL to the IoTDB data model.
> Then, I will, if my time allows it, happily help with the implementation.
>
> Julian
>