Mailing-List: contact dev-help@hawq.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hawq.incubator.apache.org
MIME-Version: 1.0
In-Reply-To: <CA+-yw=KyxcAh0H0AqKQES_B9XzyhcuF7CKCTpbcRAYBtrMDHcg@mail.gmail.com>
References: <CA+-yw=KyxcAh0H0AqKQES_B9XzyhcuF7CKCTpbcRAYBtrMDHcg@mail.gmail.com>
From: Hong Wu <xunzhangthu@gmail.com>
Date: Sat, 10 Dec 2016 12:06:22 +0800
Message-ID: <CAOPnHqdg70Cufd02s5KDUY14G9Tekm4PRXujp7Va0ZzGae=N=A@mail.gmail.com>
Subject: Re: Serialization/deserialization mechanism unification in HAWQ
To: dev@hawq.incubator.apache.org
Content-Type: multipart/alternative; boundary=001a113cc9fe5fd35c054345fe89
archived-at: Sat, 10 Dec 2016 04:06:33 -0000

--001a113cc9fe5fd35c054345fe89
Content-Type: text/plain; charset=UTF-8

Hi Alex,

It is reasonable to unify the serialization/deserialization process inside
HAWQ. We need additional effort to do that refactor and further evaluation.

Using protobuf-c or thrift library is one choice, but I think they are
using IDL to solve serialization/deserialization problem between different
languages. In other words, we need to do compile IDL to C code before
compiling HAWQ. This is not good. For HAWQ,  we just need a
unified protocol to do serialization/deserialization between C and
JAVA(at pxf end) codes.

Another choice is using an existing protocol to do
serialization/deserialization.

Best,
Hong

2016-12-09 8:46 GMT+08:00 Alex (Oleksandr) Diachenko <odiachenko@pivotal.io>
:

> Hi,
>
> I have noticed that we have multiple places where we
> serialize/deserialize data to/from a string.
>
> Couple examples:
>
>    - Filters strings in pxffilters.c.
>    - Information about data fragments in hd_work_mgr.c, pxfuriparser.c.
>
> I was thinking to use some library(like protobuf-c or similar) instead of
> having plenty of boilerplate code.
> Any thoughts?
>
> Regards, Alex.
>

--001a113cc9fe5fd35c054345fe89--