tajo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Hyunsik Choi <hyun...@apache.org>
Subject Re: [GSoc2013] - Outer Join
Date Tue, 16 Jul 2013 09:21:49 GMT
That's good idea. However, they don't have 'write grant' to apache
repository since they are not committer yet :) Github may be enough for
GSoC projects.

Thanks,
Hyunsik Choi


On Tue, Jul 16, 2013 at 5:49 AM, Henry Saputra <henry.saputra@gmail.com>wrote:

> Hi Hyunsik,
>
> I think we should create a git remote branch for each Tajo GSoC project so
> in this case Camelia can submit patches for her work and community can
> start testing accepted patches to help stabilize the implementation.
>
>
> - Henry
>
>
> On Fri, Jul 12, 2013 at 9:28 PM, Hyunsik Choi <hyunsik@apache.org> wrote:
>
> > Hi Camellia,
> >
> > I leave inline comments on your questions.
> >
> >
> > On Fri, Jul 12, 2013 at 9:13 PM, camelia c <camelie_1985@yahoo.com>
> wrote:
> >
> > > Hello,
> > >
> > > Thank You very much for Your feedback!
> > >
> > > I completed the outer joins to inner joins rewriting part and I plan to
> > > follow Your advice and move the rewriting methods to LogicalOptimizer.
> > > The new processing is described in
> > > https://sites.google.com/site/gsoc2013tajo34/home/validation , where I
> > > also uploaded the source code as files.
> > >
> > >
> > Your work looks good. However, first of all, I would like to encourage
> you
> > to learn SCM like Git.
> >
> > Actually, your source code cannot be merged into the current Tajo source
> > code because Tajo source code has been changed by multiple developers. It
> > is very hard to manage Individual source code files against updating
> source
> > tree.
> >
> > The main objective of Google summer code is to encourage open source
> > participation. So, you need to learn an overall system of open source
> > development. Above all, you should learn SCM like Git.
> >
> >
> > > 1)  I think that the allTables data structure as well as the
> > > validateOuterJoin and recursiveWhere methods should remain in class
> > > QueryAnalyzer, as they belong to the stage where the query is analyzed
> > and
> > > validated.
> > > In my opinion, only methods rewriteOuterJoin,
> > > recursiveRewriteMultiNullSupplier, recursiveRewriteNullRestricted
> should
> > be
> > > moved to class LogicalOptimizer as they perform optimizations on the
> > > logical plan.
> > > What do You think about this?
> > >
> > >
> > Sounds great. Let's go ahead with that :)
> >
> >
> > > 2) I would like to kindly ask You how can I continually rebase my work
> on
> > > the latest Tajo version, "rebase continually your work on updated
> source
> > > code"?
> > > Usually I issue this command:
> > >
> > > mvn package -DskipTests -Pdist -Dtar
> > >
> > > What should I do before this?
> > >
> >
> > If you download the source code via git, just type as follows:
> >
> > $ git pull origin master
> >
> > Probably, you meet some conflicts. If you don't know git, you should
> learn
> > git in order to solve the conflicts. You can refer some manuals available
> > online. I would like to recommend this one (http://git-scm.com/book).
> >
> >
> > >
> > > 3) I read on the mailing lists that the Tajo Cli changed and was
> > improved.
> > > But besides the query acceptance, does this affect in any way the
> stages
> > of
> > > the query processing, after its parsing?
> > >
> > >
> > Tajo Cli change was in only client side. It does not affect the part in
> > which you have worked.
> >
> >
> > > 4) Also, I read some posts on the mailing list related to integration
> > > tests.
> > > Where can I find these and how should I use them in order to verify
> that
> > > my work integrates well with the rest of the source code?
> > >
> > >
> > The following command verifies unit tests and integration tests. It
> > verifies most parts of Tajo.
> >
> > $ mvn clean install
> >
> >
> > > My work so far only affects queries containing at least one outer join,
> > so
> > > for queries consisting only of inner joins no modification is made.
> >
> >
> > > As a final remark, it was easier to manage the recursion without
> > > EvalTreeUtil. Hope it's ok.
> > >
> >
> > That's great.
> >
> >
> > >
> > >
> > > Thank You in advance!
> > >
> > > Yours  sincerely,
> > > Camelia
> > >
> > >
> > >
> > >
> > >
> >
> > Best regards,
> > Hyunsik
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message